From patchwork Fri Mar 10 08:19:37 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124952 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 474F041D3D; Fri, 10 Mar 2023 09:20:59 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id F1BE542D0B; Fri, 10 Mar 2023 09:20:32 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id E562A4114B for ; Fri, 10 Mar 2023 09:20:26 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId4021875; Fri, 10 Mar 2023 00:20:26 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=Hw8e6d9i5PQkxHJ5jSMZs5Wxc3uCrv+XLmykk/MWBIQ=; b=PFNDrS0q80atvQZmNlEBK8PeBIXX0QGRYBUtrjwhAaWAvxyi2X0peuEW4x5dIIgNNQfx +MGrpDK67ay8nu4DZx4t+crNjMAU4uV2xt+8rXq/pL9u/xxgT0X97QPRpY4Wpc4iL7CH iRCTRIoeaxi1yYNTGLpqT645YQJ0+JylyN0rbnQXLqZqSNrYshLcGxewH8LRNFRJkT5j jJVjLl9r8Usv6OYiQ5ivLrYgSPk53xUBqY0iIhPNIw4AAPqW8lYf5alsAZDO3/hrt5qk sfw7IeO5ya36b+qSq42fmMz1JBWUt4v3ioIjLG1uCYDS+iEyhAjWBMFi4TInWPGHtXwa zA== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-3 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Fri, 10 Mar 2023 00:20:25 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:20 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 33FBE5C68EB; Fri, 10 Mar 2023 00:20:18 -0800 (PST) From: Srikanth Yalavarthi To: Thomas Monjalon , Srikanth Yalavarthi , Nithin Dabilpuram , "Kiran Kumar K" , Sunil Kumar Kori , Satha Rao CC: , , , , , Subject: [PATCH v6 01/39] common/cnxk: add ML headers and ROC code for cnxk Date: Fri, 10 Mar 2023 00:19:37 -0800 Message-ID: <20230310082015.20200-2-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: d1HmGfzlQFqmc7i9HSiuw9voYJ4O97Gs X-Proofpoint-ORIG-GUID: d1HmGfzlQFqmc7i9HSiuw9voYJ4O97Gs X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added ML cnxk headers for register, structure definitions and ROC layer. Implemented ROC functions, registered logtype for ML module with the name pmd.ml.cnxk and defined ML hardware ID. Signed-off-by: Srikanth Yalavarthi --- Depends-on: series-27324 ("Implementation of mldev test application") MAINTAINERS | 9 + drivers/common/cnxk/hw/ml.h | 170 ++++++++ drivers/common/cnxk/meson.build | 1 + drivers/common/cnxk/roc_api.h | 4 + drivers/common/cnxk/roc_constants.h | 2 + drivers/common/cnxk/roc_dev_priv.h | 1 + drivers/common/cnxk/roc_ml.c | 626 ++++++++++++++++++++++++++++ drivers/common/cnxk/roc_ml.h | 152 +++++++ drivers/common/cnxk/roc_ml_priv.h | 24 ++ drivers/common/cnxk/roc_platform.c | 1 + drivers/common/cnxk/roc_platform.h | 2 + drivers/common/cnxk/roc_priv.h | 3 + drivers/common/cnxk/version.map | 29 ++ 13 files changed, 1024 insertions(+) create mode 100644 drivers/common/cnxk/hw/ml.h create mode 100644 drivers/common/cnxk/roc_ml.c create mode 100644 drivers/common/cnxk/roc_ml.h create mode 100644 drivers/common/cnxk/roc_ml_priv.h -- 2.17.1 diff --git a/MAINTAINERS b/MAINTAINERS index 320842e13f..d58df9197c 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1442,6 +1442,15 @@ F: drivers/raw/dpaa2_cmdif/ F: doc/guides/rawdevs/dpaa2_cmdif.rst +ML Device Drivers +------------------------ + +Marvell ML CNXK +M: Srikanth Yalavarthi +F: drivers/common/cnxk/hw/ml.h +F: drivers/common/cnxk/roc_ml* + + Packet processing ----------------- diff --git a/drivers/common/cnxk/hw/ml.h b/drivers/common/cnxk/hw/ml.h new file mode 100644 index 0000000000..3ead42b807 --- /dev/null +++ b/drivers/common/cnxk/hw/ml.h @@ -0,0 +1,170 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef __ML_HW_H__ +#define __ML_HW_H__ + +#include + +/* Constants */ +#define ML_ANBX_NR 0x3 + +/* Base offsets */ +#define ML_MLAB_BLK_OFFSET 0x20000000 /* CNF10KB */ +#define ML_AXI_START_ADDR 0x800000000 + +/* MLW register offsets / ML_PF_BAR0 */ +#define ML_CFG 0x10000 +#define ML_MLR_BASE 0x10008 +#define ML_AXI_BRIDGE_CTRL(a) (0x10020 | (uint64_t)(a) << 3) +#define ML_JOB_MGR_CTRL 0x10060 +#define ML_CORE_INT_LO 0x10140 +#define ML_CORE_INT_HI 0x10160 +#define ML_JCMDQ_IN(a) (0x11000 | (uint64_t)(a) << 3) /* CN10KA */ +#define ML_JCMDQ_STATUS 0x11010 /* CN10KA */ +#define ML_STGX_STATUS(a) (0x11020 | (uint64_t)(a) << 3) /* CNF10KB */ +#define ML_STG_CONTROL 0x11100 /* CNF10KB */ +#define ML_PNB_CMD_TYPE 0x113a0 /* CNF10KB */ +#define ML_SCRATCH(a) (0x14000 | (uint64_t)(a) << 3) +#define ML_ANBX_BACKP_DISABLE(a) (0x18000 | (uint64_t)(a) << 12) /* CN10KA */ +#define ML_ANBX_NCBI_P_OVR(a) (0x18010 | (uint64_t)(a) << 12) /* CN10KA */ +#define ML_ANBX_NCBI_NP_OVR(a) (0x18020 | (uint64_t)(a) << 12) /* CN10KA */ + +/* MLIP configuration register offsets / ML_PF_BAR0 */ +#define ML_SW_RST_CTRL 0x12084000 +#define ML_A35_0_RST_VECTOR_BASE_W(a) (0x12084014 + (a) * (0x04)) +#define ML_A35_1_RST_VECTOR_BASE_W(a) (0x1208401c + (a) * (0x04)) + +/* MLW scratch register offsets */ +#define ML_SCRATCH_WORK_PTR (ML_SCRATCH(0)) +#define ML_SCRATCH_FW_CTRL (ML_SCRATCH(1)) +#define ML_SCRATCH_DBG_BUFFER_HEAD_C0 (ML_SCRATCH(2)) +#define ML_SCRATCH_DBG_BUFFER_TAIL_C0 (ML_SCRATCH(3)) +#define ML_SCRATCH_DBG_BUFFER_HEAD_C1 (ML_SCRATCH(4)) +#define ML_SCRATCH_DBG_BUFFER_TAIL_C1 (ML_SCRATCH(5)) +#define ML_SCRATCH_EXCEPTION_SP_C0 (ML_SCRATCH(6)) +#define ML_SCRATCH_EXCEPTION_SP_C1 (ML_SCRATCH(7)) + +/* ML job completion structure */ +struct ml_jce_s { + /* WORD 0 */ + union ml_jce_w0 { + struct { + uint64_t rsvd_0_3 : 4; + + /* Reserved for future architecture */ + uint64_t ggrp_h : 2; + + /* Tag type */ + uint64_t ttype : 2; + + /* Physical function number */ + uint64_t pf_func : 16; + + /* Unused [7] + Guest Group [6:0] */ + uint64_t ggrp : 8; + + /* Tag */ + uint64_t tag : 32; + } s; + uint64_t u64; + } w0; + + /* WORD 1 */ + union ml_jce_w1 { + struct { + /* Work queue pointer */ + uint64_t wqp : 53; + uint64_t rsvd_53_63 : 11; + + } s; + uint64_t u64; + } w1; +}; + +/* ML job command structure */ +struct ml_job_cmd_s { + /* WORD 0 */ + union ml_job_cmd_w0 { + struct { + uint64_t rsvd_0_63; + } s; + uint64_t u64; + } w0; + + /* WORD 1 */ + union ml_job_cmd_w1 { + struct { + /* Job pointer */ + uint64_t jobptr : 53; + uint64_t rsvd_53_63 : 11; + } s; + uint64_t u64; + } w1; +}; + +/* ML A35 0 RST vector base structure */ +union ml_a35_0_rst_vector_base_s { + struct { + /* Base address */ + uint64_t addr : 37; + uint64_t rsvd_37_63 : 27; + } s; + + struct { + /* WORD 0 */ + uint32_t w0; + + /* WORD 1 */ + uint32_t w1; + } w; + + uint64_t u64; +}; + +/* ML A35 1 RST vector base structure */ +union ml_a35_1_rst_vector_base_s { + struct { + /* Base address */ + uint64_t addr : 37; + uint64_t rsvd_37_63 : 27; + } s; + + struct { + /* WORD 0 */ + uint32_t w0; + + /* WORD 1 */ + uint32_t w1; + } w; + + uint64_t u64; +}; + +/* Work pointer scratch register */ +union ml_scratch_work_ptr_s { + struct { + /* Work pointer */ + uint64_t work_ptr : 37; + uint64_t rsvd_37_63 : 27; + } s; + uint64_t u64; +}; + +/* Firmware control scratch register */ +union ml_scratch_fw_ctrl_s { + struct { + uint64_t rsvd_0_15 : 16; + + /* Valid job bit */ + uint64_t valid : 1; + + /* Done status bit */ + uint64_t done : 1; + uint64_t rsvd_18_63 : 46; + } s; + uint64_t u64; +}; + +#endif /* __ML_HW_H__ */ diff --git a/drivers/common/cnxk/meson.build b/drivers/common/cnxk/meson.build index 849735921c..b4aa0a050c 100644 --- a/drivers/common/cnxk/meson.build +++ b/drivers/common/cnxk/meson.build @@ -26,6 +26,7 @@ sources = files( 'roc_irq.c', 'roc_ie_ot.c', 'roc_mbox.c', + 'roc_ml.c', 'roc_model.c', 'roc_nix.c', 'roc_nix_bpf.c', diff --git a/drivers/common/cnxk/roc_api.h b/drivers/common/cnxk/roc_api.h index 993a2f7a68..bbc94ab48e 100644 --- a/drivers/common/cnxk/roc_api.h +++ b/drivers/common/cnxk/roc_api.h @@ -31,6 +31,7 @@ /* HW structure definition */ #include "hw/cpt.h" #include "hw/dpi.h" +#include "hw/ml.h" #include "hw/nix.h" #include "hw/npa.h" #include "hw/npc.h" @@ -110,4 +111,7 @@ #include "roc_nix_inl_dp.h" #include "roc_nix_inl.h" +/* ML */ +#include "roc_ml.h" + #endif /* _ROC_API_H_ */ diff --git a/drivers/common/cnxk/roc_constants.h b/drivers/common/cnxk/roc_constants.h index c94916db6d..291b6a4bc9 100644 --- a/drivers/common/cnxk/roc_constants.h +++ b/drivers/common/cnxk/roc_constants.h @@ -52,6 +52,8 @@ #define PCI_DEVID_CN10K_RVU_CPT_PF 0xA0F2 #define PCI_DEVID_CN10K_RVU_CPT_VF 0xA0F3 +#define PCI_DEVID_CN10K_ML_PF 0xA092 + #define PCI_SUBSYSTEM_DEVID_CN10KA 0xB900 #define PCI_SUBSYSTEM_DEVID_CN10KAS 0xB900 #define PCI_SUBSYSTEM_DEVID_CNF10KA 0xBA00 diff --git a/drivers/common/cnxk/roc_dev_priv.h b/drivers/common/cnxk/roc_dev_priv.h index 27bf68fddb..feda173ce5 100644 --- a/drivers/common/cnxk/roc_dev_priv.h +++ b/drivers/common/cnxk/roc_dev_priv.h @@ -94,6 +94,7 @@ struct dev { void *roc_nix; void *roc_cpt; void *roc_tim; + void *roc_ml; bool disable_shared_lmt; /* false(default): shared lmt mode enabled */ const struct plt_memzone *lmt_mz; } __plt_cache_aligned; diff --git a/drivers/common/cnxk/roc_ml.c b/drivers/common/cnxk/roc_ml.c new file mode 100644 index 0000000000..7390697b1d --- /dev/null +++ b/drivers/common/cnxk/roc_ml.c @@ -0,0 +1,626 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#include "roc_api.h" +#include "roc_priv.h" + +#define TIME_SEC_IN_MS 1000 + +static int +roc_ml_reg_wait_to_clear(struct roc_ml *roc_ml, uint64_t offset, uint64_t mask) +{ + uint64_t start_cycle; + uint64_t wait_cycles; + uint64_t reg_val; + + wait_cycles = (ROC_ML_TIMEOUT_MS * plt_tsc_hz()) / TIME_SEC_IN_MS; + start_cycle = plt_tsc_cycles(); + do { + reg_val = roc_ml_reg_read64(roc_ml, offset); + + if (!(reg_val & mask)) + return 0; + } while (plt_tsc_cycles() - start_cycle < wait_cycles); + + return -ETIME; +} + +uint64_t +roc_ml_reg_read64(struct roc_ml *roc_ml, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + return plt_read64(PLT_PTR_ADD(ml->ml_reg_addr, offset)); +} + +void +roc_ml_reg_write64(struct roc_ml *roc_ml, uint64_t val, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + plt_write64(val, PLT_PTR_ADD(ml->ml_reg_addr, offset)); +} + +uint32_t +roc_ml_reg_read32(struct roc_ml *roc_ml, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + return plt_read32(PLT_PTR_ADD(ml->ml_reg_addr, offset)); +} + +void +roc_ml_reg_write32(struct roc_ml *roc_ml, uint32_t val, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + plt_write32(val, PLT_PTR_ADD(ml->ml_reg_addr, offset)); +} + +void +roc_ml_reg_save(struct roc_ml *roc_ml, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + if (offset == ML_MLR_BASE) { + ml->ml_mlr_base = + FIELD_GET(ROC_ML_MLR_BASE_BASE, roc_ml_reg_read64(roc_ml, offset)); + ml->ml_mlr_base_saved = true; + } +} + +void * +roc_ml_addr_ap2mlip(struct roc_ml *roc_ml, void *addr) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + uint64_t ml_mlr_base; + + ml_mlr_base = (ml->ml_mlr_base_saved) ? ml->ml_mlr_base : + FIELD_GET(ROC_ML_MLR_BASE_BASE, + roc_ml_reg_read64(roc_ml, ML_MLR_BASE)); + return PLT_PTR_ADD(addr, ML_AXI_START_ADDR - ml_mlr_base); +} + +void * +roc_ml_addr_mlip2ap(struct roc_ml *roc_ml, void *addr) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + uint64_t ml_mlr_base; + + ml_mlr_base = (ml->ml_mlr_base_saved) ? ml->ml_mlr_base : + FIELD_GET(ROC_ML_MLR_BASE_BASE, + roc_ml_reg_read64(roc_ml, ML_MLR_BASE)); + return PLT_PTR_ADD(addr, ml_mlr_base - ML_AXI_START_ADDR); +} + +uint64_t +roc_ml_addr_pa_to_offset(struct roc_ml *roc_ml, uint64_t phys_addr) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + if (roc_model_is_cn10ka()) + return phys_addr - ml->pci_dev->mem_resource[0].phys_addr; + else + return phys_addr - ml->pci_dev->mem_resource[0].phys_addr - ML_MLAB_BLK_OFFSET; +} + +uint64_t +roc_ml_addr_offset_to_pa(struct roc_ml *roc_ml, uint64_t offset) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + if (roc_model_is_cn10ka()) + return ml->pci_dev->mem_resource[0].phys_addr + offset; + else + return ml->pci_dev->mem_resource[0].phys_addr + ML_MLAB_BLK_OFFSET + offset; +} + +void +roc_ml_scratch_write_job(struct roc_ml *roc_ml, void *work_ptr) +{ + union ml_scratch_work_ptr_s reg_work_ptr; + union ml_scratch_fw_ctrl_s reg_fw_ctrl; + + reg_work_ptr.u64 = 0; + reg_work_ptr.s.work_ptr = PLT_U64_CAST(roc_ml_addr_ap2mlip(roc_ml, work_ptr)); + + reg_fw_ctrl.u64 = 0; + reg_fw_ctrl.s.valid = 1; + + roc_ml_reg_write64(roc_ml, reg_work_ptr.u64, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(roc_ml, reg_fw_ctrl.u64, ML_SCRATCH_FW_CTRL); +} + +bool +roc_ml_scratch_is_valid_bit_set(struct roc_ml *roc_ml) +{ + union ml_scratch_fw_ctrl_s reg_fw_ctrl; + + reg_fw_ctrl.u64 = roc_ml_reg_read64(roc_ml, ML_SCRATCH_FW_CTRL); + + if (reg_fw_ctrl.s.valid == 1) + return true; + + return false; +} + +bool +roc_ml_scratch_is_done_bit_set(struct roc_ml *roc_ml) +{ + union ml_scratch_fw_ctrl_s reg_fw_ctrl; + + reg_fw_ctrl.u64 = roc_ml_reg_read64(roc_ml, ML_SCRATCH_FW_CTRL); + + if (reg_fw_ctrl.s.done == 1) + return true; + + return false; +} + +bool +roc_ml_scratch_enqueue(struct roc_ml *roc_ml, void *work_ptr) +{ + union ml_scratch_work_ptr_s reg_work_ptr; + union ml_scratch_fw_ctrl_s reg_fw_ctrl; + bool ret = false; + + reg_work_ptr.u64 = 0; + reg_work_ptr.s.work_ptr = PLT_U64_CAST(roc_ml_addr_ap2mlip(roc_ml, work_ptr)); + + reg_fw_ctrl.u64 = 0; + reg_fw_ctrl.s.valid = 1; + + if (plt_spinlock_trylock(&roc_ml->sp_spinlock) != 0) { + bool valid = roc_ml_scratch_is_valid_bit_set(roc_ml); + bool done = roc_ml_scratch_is_done_bit_set(roc_ml); + + if (valid == done) { + roc_ml_clk_force_on(roc_ml); + roc_ml_dma_stall_off(roc_ml); + + roc_ml_reg_write64(roc_ml, reg_work_ptr.u64, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(roc_ml, reg_fw_ctrl.u64, ML_SCRATCH_FW_CTRL); + + ret = true; + } + plt_spinlock_unlock(&roc_ml->sp_spinlock); + } + + return ret; +} + +bool +roc_ml_scratch_dequeue(struct roc_ml *roc_ml, void *work_ptr) +{ + union ml_scratch_work_ptr_s reg_work_ptr; + bool ret = false; + + if (plt_spinlock_trylock(&roc_ml->sp_spinlock) != 0) { + bool valid = roc_ml_scratch_is_valid_bit_set(roc_ml); + bool done = roc_ml_scratch_is_done_bit_set(roc_ml); + + if (valid && done) { + reg_work_ptr.u64 = roc_ml_reg_read64(roc_ml, ML_SCRATCH_WORK_PTR); + if (work_ptr == + roc_ml_addr_mlip2ap(roc_ml, PLT_PTR_CAST(reg_work_ptr.u64))) { + roc_ml_dma_stall_on(roc_ml); + roc_ml_clk_force_off(roc_ml); + + roc_ml_reg_write64(roc_ml, 0, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(roc_ml, 0, ML_SCRATCH_FW_CTRL); + ret = true; + } + } + plt_spinlock_unlock(&roc_ml->sp_spinlock); + } + + return ret; +} + +void +roc_ml_scratch_queue_reset(struct roc_ml *roc_ml) +{ + if (plt_spinlock_trylock(&roc_ml->sp_spinlock) != 0) { + roc_ml_dma_stall_on(roc_ml); + roc_ml_clk_force_off(roc_ml); + roc_ml_reg_write64(roc_ml, 0, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(roc_ml, 0, ML_SCRATCH_FW_CTRL); + plt_spinlock_unlock(&roc_ml->sp_spinlock); + } +} + +bool +roc_ml_jcmdq_enqueue_lf(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd) +{ + bool ret = false; + + if (FIELD_GET(ROC_ML_JCMDQ_STATUS_AVAIL_COUNT, + roc_ml_reg_read64(roc_ml, ML_JCMDQ_STATUS)) != 0) { + roc_ml_reg_write64(roc_ml, job_cmd->w0.u64, ML_JCMDQ_IN(0)); + roc_ml_reg_write64(roc_ml, job_cmd->w1.u64, ML_JCMDQ_IN(1)); + ret = true; + } + + return ret; +} + +bool +roc_ml_jcmdq_enqueue_sl(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd) +{ + bool ret = false; + + if (plt_spinlock_trylock(&roc_ml->fp_spinlock) != 0) { + if (FIELD_GET(ROC_ML_JCMDQ_STATUS_AVAIL_COUNT, + roc_ml_reg_read64(roc_ml, ML_JCMDQ_STATUS)) != 0) { + roc_ml_reg_write64(roc_ml, job_cmd->w0.u64, ML_JCMDQ_IN(0)); + roc_ml_reg_write64(roc_ml, job_cmd->w1.u64, ML_JCMDQ_IN(1)); + ret = true; + } + plt_spinlock_unlock(&roc_ml->fp_spinlock); + } + + return ret; +} + +void +roc_ml_clk_force_on(struct roc_ml *roc_ml) +{ + uint64_t reg_val = 0; + + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val |= ROC_ML_CFG_MLIP_CLK_FORCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); +} + +void +roc_ml_clk_force_off(struct roc_ml *roc_ml) +{ + uint64_t reg_val = 0; + + roc_ml_reg_write64(roc_ml, 0, ML_SCRATCH_WORK_PTR); + + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_MLIP_CLK_FORCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); +} + +void +roc_ml_dma_stall_on(struct roc_ml *roc_ml) +{ + uint64_t reg_val = 0; + + reg_val = roc_ml_reg_read64(roc_ml, ML_JOB_MGR_CTRL); + reg_val |= ROC_ML_JOB_MGR_CTRL_STALL_ON_IDLE; + roc_ml_reg_write64(roc_ml, reg_val, ML_JOB_MGR_CTRL); +} + +void +roc_ml_dma_stall_off(struct roc_ml *roc_ml) +{ + uint64_t reg_val = 0; + + reg_val = roc_ml_reg_read64(roc_ml, ML_JOB_MGR_CTRL); + reg_val &= ~ROC_ML_JOB_MGR_CTRL_STALL_ON_IDLE; + roc_ml_reg_write64(roc_ml, reg_val, ML_JOB_MGR_CTRL); +} + +bool +roc_ml_mlip_is_enabled(struct roc_ml *roc_ml) +{ + uint64_t reg_val; + + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + + if ((reg_val & ROC_ML_CFG_MLIP_ENA) != 0) + return true; + + return false; +} + +int +roc_ml_mlip_reset(struct roc_ml *roc_ml, bool force) +{ + uint64_t reg_val; + + /* Force reset */ + if (force) { + /* Set ML(0)_CFG[ENA] = 0. */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + + /* Set ML(0)_CFG[MLIP_ENA] = 0. */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_MLIP_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + + /* Clear ML_MLR_BASE */ + roc_ml_reg_write64(roc_ml, 0, ML_MLR_BASE); + } + + if (roc_model_is_cn10ka()) { + /* Wait for all active jobs to finish. + * ML_CFG[ENA] : When set, MLW will accept job commands. This + * bit can be cleared at any time. If [BUSY] is set, software + * must wait until [BUSY] == 0 before setting this bit. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_CFG, ROC_ML_CFG_BUSY); + + /* (1) Set ML(0)_AXI_BRIDGE_CTRL(0..1)[FENCE] = 1 to instruct + * the AXI bridge not to accept any new transactions from MLIP. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(1)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + + /* (2) Wait until ML(0)_AXI_BRIDGE_CTRL(0..1)[BUSY] = 0 which + * indicates that there is no outstanding transactions on + * AXI-NCB paths. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_AXI_BRIDGE_CTRL(0), + ROC_ML_AXI_BRIDGE_CTRL_BUSY); + roc_ml_reg_wait_to_clear(roc_ml, ML_AXI_BRIDGE_CTRL(1), + ROC_ML_AXI_BRIDGE_CTRL_BUSY); + + /* (3) Wait until ML(0)_JOB_MGR_CTRL[BUSY] = 0 which indicates + * that there are no pending jobs in the MLW's job manager. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_JOB_MGR_CTRL, ROC_ML_JOB_MGR_CTRL_BUSY); + + /* (4) Set ML(0)_CFG[ENA] = 0. */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + + /* (5) Set ML(0)_CFG[MLIP_ENA] = 0. */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_MLIP_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + + /* (6) Set ML(0)_AXI_BRIDGE_CTRL(0..1)[FENCE] = 0.*/ + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val &= ~ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + } + + if (roc_model_is_cnf10kb()) { + /* (1) Clear MLAB(0)_CFG[ENA]. Any new jobs will bypass the job + * execution stages and their completions will be returned to + * PSM. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + + /* (2) Quiesce the ACC and DMA AXI interfaces: For each of the + * two MLAB(0)_AXI_BRIDGE_CTRL(0..1) registers: + * + * (a) Set MLAB(0)_AXI_BRIDGE_CTRL(0..1)[FENCE] to block new AXI + * commands from MLIP. + * + * (b) Poll MLAB(0)_AXI_BRIDGE_CTRL(0..1)[BUSY] == 0. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + + roc_ml_reg_wait_to_clear(roc_ml, ML_AXI_BRIDGE_CTRL(0), + ROC_ML_AXI_BRIDGE_CTRL_BUSY); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(1)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + + roc_ml_reg_wait_to_clear(roc_ml, ML_AXI_BRIDGE_CTRL(1), + ROC_ML_AXI_BRIDGE_CTRL_BUSY); + + /* (3) Clear MLAB(0)_CFG[MLIP_ENA] to reset MLIP. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_CFG); + reg_val &= ~ROC_ML_CFG_MLIP_ENA; + roc_ml_reg_write64(roc_ml, reg_val, ML_CFG); + +cnf10kb_mlip_reset_stage_4a: + /* (4) Flush any outstanding jobs in MLAB's job execution + * stages: + * + * (a) Wait for completion stage to clear: + * - Poll MLAB(0)_STG(0..2)_STATUS[VALID] == 0. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_STGX_STATUS(0), ROC_ML_STG_STATUS_VALID); + roc_ml_reg_wait_to_clear(roc_ml, ML_STGX_STATUS(1), ROC_ML_STG_STATUS_VALID); + roc_ml_reg_wait_to_clear(roc_ml, ML_STGX_STATUS(2), ROC_ML_STG_STATUS_VALID); + +cnf10kb_mlip_reset_stage_4b: + /* (4b) Clear job run stage: Poll + * MLAB(0)_STG_CONTROL[RUN_TO_COMP] == 0. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_STG_CONTROL, ROC_ML_STG_CONTROL_RUN_TO_COMP); + + /* (4b) Clear job run stage: If MLAB(0)_STG(1)_STATUS[VALID] == + * 1: + * - Set MLAB(0)_STG_CONTROL[RUN_TO_COMP]. + * - Poll MLAB(0)_STG_CONTROL[RUN_TO_COMP] == 0. + * - Repeat step (a) to clear job completion stage. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_STGX_STATUS(1)); + if (reg_val & ROC_ML_STG_STATUS_VALID) { + reg_val = roc_ml_reg_read64(roc_ml, ML_STG_CONTROL); + reg_val |= ROC_ML_STG_CONTROL_RUN_TO_COMP; + roc_ml_reg_write64(roc_ml, reg_val, ML_STG_CONTROL); + + roc_ml_reg_wait_to_clear(roc_ml, ML_STG_CONTROL, + ROC_ML_STG_CONTROL_RUN_TO_COMP); + + goto cnf10kb_mlip_reset_stage_4a; + } + + /* (4c) Clear job fetch stage: Poll + * MLAB(0)_STG_CONTROL[FETCH_TO_RUN] == 0. + */ + roc_ml_reg_wait_to_clear(roc_ml, ML_STG_CONTROL, ROC_ML_STG_CONTROL_FETCH_TO_RUN); + + /* (4c) Clear job fetch stage: If + * MLAB(0)_STG(0..2)_STATUS[VALID] == 1: + * - Set MLAB(0)_STG_CONTROL[FETCH_TO_RUN]. + * - Poll MLAB(0)_STG_CONTROL[FETCH_TO_RUN] == 0. + * - Repeat step (b) to clear job run and completion stages. + */ + reg_val = (roc_ml_reg_read64(roc_ml, ML_STGX_STATUS(0)) | + roc_ml_reg_read64(roc_ml, ML_STGX_STATUS(1)) | + roc_ml_reg_read64(roc_ml, ML_STGX_STATUS(2))); + + if (reg_val & ROC_ML_STG_STATUS_VALID) { + reg_val = roc_ml_reg_read64(roc_ml, ML_STG_CONTROL); + reg_val |= ROC_ML_STG_CONTROL_RUN_TO_COMP; + roc_ml_reg_write64(roc_ml, reg_val, ML_STG_CONTROL); + + roc_ml_reg_wait_to_clear(roc_ml, ML_STG_CONTROL, + ROC_ML_STG_CONTROL_RUN_TO_COMP); + + goto cnf10kb_mlip_reset_stage_4b; + } + + /* (5) Reset the ACC and DMA AXI interfaces: For each of the two + * MLAB(0)_AXI_BRIDGE_CTRL(0..1) registers: + * + * (5a) Set and then clear + * MLAB(0)_AXI_BRIDGE_CTRL(0..1)[FLUSH_WRITE_DATA]. + * + * (5b) Clear MLAB(0)_AXI_BRIDGE_CTRL(0..1)[FENCE]. + */ + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FLUSH_WRITE_DATA; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val &= ~ROC_ML_AXI_BRIDGE_CTRL_FLUSH_WRITE_DATA; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(0)); + reg_val &= ~ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(0)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(1)); + reg_val |= ROC_ML_AXI_BRIDGE_CTRL_FLUSH_WRITE_DATA; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(1)); + reg_val &= ~ROC_ML_AXI_BRIDGE_CTRL_FLUSH_WRITE_DATA; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + + reg_val = roc_ml_reg_read64(roc_ml, ML_AXI_BRIDGE_CTRL(1)); + reg_val &= ~ROC_ML_AXI_BRIDGE_CTRL_FENCE; + roc_ml_reg_write64(roc_ml, reg_val, ML_AXI_BRIDGE_CTRL(1)); + } + + return 0; +} + +int +roc_ml_dev_init(struct roc_ml *roc_ml) +{ + struct plt_pci_device *pci_dev; + struct dev *dev; + struct ml *ml; + + if (roc_ml == NULL || roc_ml->pci_dev == NULL) + return -EINVAL; + + PLT_STATIC_ASSERT(sizeof(struct ml) <= ROC_ML_MEM_SZ); + + ml = roc_ml_to_ml_priv(roc_ml); + memset(ml, 0, sizeof(*ml)); + pci_dev = roc_ml->pci_dev; + dev = &ml->dev; + + ml->pci_dev = pci_dev; + dev->roc_ml = roc_ml; + + ml->ml_reg_addr = ml->pci_dev->mem_resource[0].addr; + ml->ml_mlr_base = 0; + ml->ml_mlr_base_saved = false; + + plt_ml_dbg("ML: PCI Physical Address : 0x%016lx", ml->pci_dev->mem_resource[0].phys_addr); + plt_ml_dbg("ML: PCI Virtual Address : 0x%016lx", + PLT_U64_CAST(ml->pci_dev->mem_resource[0].addr)); + + plt_spinlock_init(&roc_ml->sp_spinlock); + plt_spinlock_init(&roc_ml->fp_spinlock); + + return 0; +} + +int +roc_ml_dev_fini(struct roc_ml *roc_ml) +{ + struct ml *ml = roc_ml_to_ml_priv(roc_ml); + + if (ml == NULL) + return -EINVAL; + + return 0; +} + +int +roc_ml_blk_init(struct roc_bphy *roc_bphy, struct roc_ml *roc_ml) +{ + struct dev *dev; + struct ml *ml; + + if ((roc_ml == NULL) || (roc_bphy == NULL)) + return -EINVAL; + + PLT_STATIC_ASSERT(sizeof(struct ml) <= ROC_ML_MEM_SZ); + + ml = roc_ml_to_ml_priv(roc_ml); + memset(ml, 0, sizeof(*ml)); + + dev = &ml->dev; + + ml->pci_dev = roc_bphy->pci_dev; + dev->roc_ml = roc_ml; + + plt_ml_dbg( + "MLAB: Physical Address : 0x%016lx", + PLT_PTR_ADD_U64_CAST(ml->pci_dev->mem_resource[0].phys_addr, ML_MLAB_BLK_OFFSET)); + plt_ml_dbg("MLAB: Virtual Address : 0x%016lx", + PLT_PTR_ADD_U64_CAST(ml->pci_dev->mem_resource[0].addr, ML_MLAB_BLK_OFFSET)); + + ml->ml_reg_addr = PLT_PTR_ADD(ml->pci_dev->mem_resource[0].addr, ML_MLAB_BLK_OFFSET); + ml->ml_mlr_base = 0; + ml->ml_mlr_base_saved = false; + + plt_spinlock_init(&roc_ml->sp_spinlock); + plt_spinlock_init(&roc_ml->fp_spinlock); + + return 0; +} + +int +roc_ml_blk_fini(struct roc_bphy *roc_bphy, struct roc_ml *roc_ml) +{ + struct ml *ml; + + if ((roc_ml == NULL) || (roc_bphy == NULL)) + return -EINVAL; + + ml = roc_ml_to_ml_priv(roc_ml); + + if (ml == NULL) + return -EINVAL; + + return 0; +} + +uint16_t +roc_ml_sso_pf_func_get(void) +{ + return idev_sso_pffunc_get(); +} diff --git a/drivers/common/cnxk/roc_ml.h b/drivers/common/cnxk/roc_ml.h new file mode 100644 index 0000000000..3cd82be6a6 --- /dev/null +++ b/drivers/common/cnxk/roc_ml.h @@ -0,0 +1,152 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _ROC_ML_H_ +#define _ROC_ML_H_ + +#include "roc_api.h" + +#define ROC_ML_MEM_SZ (6 * 1024) +#define ROC_ML_TIMEOUT_MS 10000 + +/* ML_CFG */ +#define ROC_ML_CFG_JD_SIZE GENMASK_ULL(1, 0) +#define ROC_ML_CFG_MLIP_ENA BIT_ULL(2) +#define ROC_ML_CFG_BUSY BIT_ULL(3) +#define ROC_ML_CFG_WRAP_CLK_FORCE BIT_ULL(4) +#define ROC_ML_CFG_MLIP_CLK_FORCE BIT_ULL(5) +#define ROC_ML_CFG_ENA BIT_ULL(6) + +/* ML_MLR_BASE */ +#define ROC_ML_MLR_BASE_BASE GENMASK_ULL(51, 0) + +/* ML_STG_STATUS */ +#define ROC_ML_STG_STATUS_VALID BIT_ULL(0) +#define ROC_ML_STG_STATUS_ADDR_ERR BIT_ULL(1) +#define ROC_ML_STG_STATUS_DMA_ERR BIT_ULL(2) +#define ROC_ML_STG_STATUS_TIMEOUT BIT_ULL(3) +#define ROC_ML_STG_STATUS_NFAT_ERR BIT_ULL(4) +#define ROC_ML_STG_STATUS_JOB_ERR BIT_ULL(5) +#define ROC_ML_STG_STATUS_ELAPSED_TICKS GENMASK_ULL(47, 6) + +/* ML_STG_CONTROL */ +#define ROC_ML_STG_CONTROL_FETCH_TO_RUN BIT_ULL(0) +#define ROC_ML_STG_CONTROL_RUN_TO_COMP BIT_ULL(1) + +/* ML_AXI_BRIDGE */ +#define ROC_ML_AXI_BRIDGE_CTRL_AXI_RESP_CTRL BIT_ULL(0) +#define ROC_ML_AXI_BRIDGE_CTRL_BRIDGE_CTRL_MODE BIT_ULL(1) +#define ROC_ML_AXI_BRIDGE_CTRL_FORCE_AXI_ID GENMASK_ULL(11, 2) +#define ROC_ML_AXI_BRIDGE_CTRL_CSR_WR_BLK BIT_ULL(13) +#define ROC_ML_AXI_BRIDGE_CTRL_NCB_WR_BLK BIT_ULL(14) +#define ROC_ML_AXI_BRIDGE_CTRL_CSR_RD_BLK BIT_ULL(15) +#define ROC_ML_AXI_BRIDGE_CTRL_NCB_RD_BLK BIT_ULL(16) +#define ROC_ML_AXI_BRIDGE_CTRL_FENCE BIT_ULL(17) +#define ROC_ML_AXI_BRIDGE_CTRL_BUSY BIT_ULL(18) +#define ROC_ML_AXI_BRIDGE_CTRL_FORCE_WRESP_OK BIT_ULL(19) +#define ROC_ML_AXI_BRIDGE_CTRL_FORCE_RRESP_OK BIT_ULL(20) +#define ROC_ML_AXI_BRIDGE_CTRL_CSR_FORCE_CMPLT BIT_ULL(21) +#define ROC_ML_AXI_BRIDGE_CTRL_WR_CNT_GEAR GENMASK_ULL(25, 22) +#define ROC_ML_AXI_BRIDGE_CTRL_RD_GEAR GENMASK_ULL(28, 26) +#define ROC_ML_AXI_BRIDGE_CTRL_CSR_CUTTHROUGH_MODE BIT_ULL(29) +#define ROC_ML_AXI_BRIDGE_CTRL_GAA_WRITE_CREDITS GENMASK_ULL(33, 30) +#define ROC_ML_AXI_BRIDGE_CTRL_GAA_READ_CREDITS GENMASK_ULL(37, 34) +#define ROC_ML_AXI_BRIDGE_CTRL_GAA_LOAD_WRITE_CREDITS BIT_ULL(38) +#define ROC_ML_AXI_BRIDGE_CTRL_GAA_LOAD_READ_CREDITS BIT_ULL(39) +#define ROC_ML_AXI_BRIDGE_CTRL_FLUSH_WRITE_DATA BIT_ULL(40) + +/* ML_JOB_MGR_CTRL */ +#define ROC_ML_JOB_MGR_CTRL_STALL_ON_ERR BIT_ULL(0) +#define ROC_ML_JOB_MGR_CTRL_PF_OVERRIDE BIT_ULL(1) +#define ROC_ML_JOB_MGR_CTRL_PF_FUNC_OVERRIDE GENMASK_ULL(19, 4) +#define ROC_ML_JOB_MGR_CTRL_BUSY BIT_ULL(20) +#define ROC_ML_JOB_MGR_CTRL_STALL_ON_IDLE BIT_ULL(21) + +/* ML_JCMDQ_STATUS */ +#define ROC_ML_JCMDQ_STATUS_AVAIL_COUNT GENMASK_ULL(4, 0) + +/* ML_ANBX_BACKP_DISABLE */ +#define ROC_ML_ANBX_BACKP_DISABLE_EXTMSTR_B_BACKP_DISABLE BIT_ULL(0) +#define ROC_ML_ANBX_BACKP_DISABLE_EXTMSTR_R_BACKP_DISABLE BIT_ULL(1) + +/* ML_ANBX_NCBI_P_OVR */ +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MSH_DST_OVR_VLD BIT_ULL(0) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MSH_DST_OVR GENMASK_ULL(11, 1) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_NS_OVR_VLD BIT_ULL(12) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_NS_OVR BIT_ULL(13) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_PADDR_OVR_VLD BIT_ULL(14) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_PADDR_OVR BIT_ULL(15) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_RO_OVR_VLD BIT_ULL(16) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_RO_OVR BIT_ULL(17) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MPADID_VAL_OVR_VLD BIT_ULL(18) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MPADID_VAL_OVR BIT_ULL(19) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MPAMDID_OVR_VLD BIT_ULL(20) +#define ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_MPAMDID_OVR BIT_ULL(21) + +/* ML_ANBX_NCBI_NP_OVR */ +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MSH_DST_OVR_VLD BIT_ULL(0) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MSH_DST_OVR GENMASK_ULL(11, 1) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_NS_OVR_VLD BIT_ULL(12) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_NS_OVR BIT_ULL(13) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_PADDR_OVR_VLD BIT_ULL(14) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_PADDR_OVR BIT_ULL(15) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_RO_OVR_VLD BIT_ULL(16) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_RO_OVR BIT_ULL(17) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MPADID_VAL_OVR_VLD BIT_ULL(18) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MPADID_VAL_OVR BIT_ULL(19) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MPAMDID_OVR_VLD BIT_ULL(20) +#define ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_MPAMDID_OVR BIT_ULL(21) + +/* ML_SW_RST_CTRL */ +#define ROC_ML_SW_RST_CTRL_ACC_RST BIT_ULL(0) +#define ROC_ML_SW_RST_CTRL_CMPC_RST BIT_ULL(1) + +struct roc_ml { + struct plt_pci_device *pci_dev; + plt_spinlock_t sp_spinlock; + plt_spinlock_t fp_spinlock; + uint8_t reserved[ROC_ML_MEM_SZ] __plt_cache_aligned; +} __plt_cache_aligned; + +/* Register read and write functions */ +uint64_t __roc_api roc_ml_reg_read64(struct roc_ml *roc_ml, uint64_t offset); +void __roc_api roc_ml_reg_write64(struct roc_ml *roc_ml, uint64_t val, uint64_t offset); +uint32_t __roc_api roc_ml_reg_read32(struct roc_ml *roc_ml, uint64_t offset); +void __roc_api roc_ml_reg_write32(struct roc_ml *roc_ml, uint32_t val, uint64_t offset); +void __roc_api roc_ml_reg_save(struct roc_ml *roc_ml, uint64_t offset); + +/* Address translation functions */ +uint64_t __roc_api roc_ml_addr_pa_to_offset(struct roc_ml *roc_ml, uint64_t phys_addr); +uint64_t __roc_api roc_ml_addr_offset_to_pa(struct roc_ml *roc_ml, uint64_t offset); +void *__roc_api roc_ml_addr_ap2mlip(struct roc_ml *roc_ml, void *addr); +void *__roc_api roc_ml_addr_mlip2ap(struct roc_ml *roc_ml, void *addr); + +/* Scratch and JCMDQ functions */ +void __roc_api roc_ml_scratch_write_job(struct roc_ml *roc_ml, void *jd); +bool __roc_api roc_ml_scratch_is_valid_bit_set(struct roc_ml *roc_ml); +bool __roc_api roc_ml_scratch_is_done_bit_set(struct roc_ml *roc_ml); +bool __roc_api roc_ml_scratch_enqueue(struct roc_ml *roc_ml, void *work_ptr); +bool __roc_api roc_ml_scratch_dequeue(struct roc_ml *roc_ml, void *work_ptr); +void __roc_api roc_ml_scratch_queue_reset(struct roc_ml *roc_ml); +bool __roc_api roc_ml_jcmdq_enqueue_lf(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd); +bool __roc_api roc_ml_jcmdq_enqueue_sl(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd); + +/* Device management functions */ +void __roc_api roc_ml_clk_force_on(struct roc_ml *roc_ml); +void __roc_api roc_ml_clk_force_off(struct roc_ml *roc_ml); +void __roc_api roc_ml_dma_stall_on(struct roc_ml *roc_ml); +void __roc_api roc_ml_dma_stall_off(struct roc_ml *roc_ml); +bool __roc_api roc_ml_mlip_is_enabled(struct roc_ml *roc_ml); +int __roc_api roc_ml_mlip_reset(struct roc_ml *roc_ml, bool force); + +/* Device / block functions */ +int __roc_api roc_ml_dev_init(struct roc_ml *roc_ml); +int __roc_api roc_ml_dev_fini(struct roc_ml *roc_ml); +int __roc_api roc_ml_blk_init(struct roc_bphy *roc_bphy, struct roc_ml *roc_ml); +int __roc_api roc_ml_blk_fini(struct roc_bphy *roc_bphy, struct roc_ml *roc_ml); + +/* Utility functions */ +uint16_t __roc_api roc_ml_sso_pf_func_get(void); + +#endif /*_ROC_ML_H_*/ diff --git a/drivers/common/cnxk/roc_ml_priv.h b/drivers/common/cnxk/roc_ml_priv.h new file mode 100644 index 0000000000..ad5fe90bab --- /dev/null +++ b/drivers/common/cnxk/roc_ml_priv.h @@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _ROC_ML_PRIV_H_ +#define _ROC_ML_PRIV_H_ + +#include "roc_api.h" + +struct ml { + struct plt_pci_device *pci_dev; + struct dev dev; + uint8_t *ml_reg_addr; + uint64_t ml_mlr_base; + bool ml_mlr_base_saved; +} __plt_cache_aligned; + +static inline struct ml * +roc_ml_to_ml_priv(struct roc_ml *roc_ml) +{ + return (struct ml *)&roc_ml->reserved[0]; +} + +#endif /* _ROC_ML_PRIV_H_ */ diff --git a/drivers/common/cnxk/roc_platform.c b/drivers/common/cnxk/roc_platform.c index ce0f9b870c..f91b95ceab 100644 --- a/drivers/common/cnxk/roc_platform.c +++ b/drivers/common/cnxk/roc_platform.c @@ -63,6 +63,7 @@ roc_plt_init(void) RTE_LOG_REGISTER(cnxk_logtype_base, pmd.cnxk.base, NOTICE); RTE_LOG_REGISTER(cnxk_logtype_mbox, pmd.cnxk.mbox, NOTICE); RTE_LOG_REGISTER(cnxk_logtype_cpt, pmd.crypto.cnxk, NOTICE); +RTE_LOG_REGISTER(cnxk_logtype_ml, pmd.ml.cnxk, NOTICE); RTE_LOG_REGISTER(cnxk_logtype_npa, pmd.mempool.cnxk, NOTICE); RTE_LOG_REGISTER(cnxk_logtype_nix, pmd.net.cnxk, NOTICE); RTE_LOG_REGISTER(cnxk_logtype_npc, pmd.net.cnxk.flow, NOTICE); diff --git a/drivers/common/cnxk/roc_platform.h b/drivers/common/cnxk/roc_platform.h index ed562a82bf..c50fcd6c6d 100644 --- a/drivers/common/cnxk/roc_platform.h +++ b/drivers/common/cnxk/roc_platform.h @@ -234,6 +234,7 @@ extern int cnxk_logtype_base; extern int cnxk_logtype_mbox; extern int cnxk_logtype_cpt; +extern int cnxk_logtype_ml; extern int cnxk_logtype_npa; extern int cnxk_logtype_nix; extern int cnxk_logtype_npc; @@ -261,6 +262,7 @@ extern int cnxk_logtype_ree; #define plt_base_dbg(fmt, ...) plt_dbg(base, fmt, ##__VA_ARGS__) #define plt_cpt_dbg(fmt, ...) plt_dbg(cpt, fmt, ##__VA_ARGS__) #define plt_mbox_dbg(fmt, ...) plt_dbg(mbox, fmt, ##__VA_ARGS__) +#define plt_ml_dbg(fmt, ...) plt_dbg(ml, fmt, ##__VA_ARGS__) #define plt_npa_dbg(fmt, ...) plt_dbg(npa, fmt, ##__VA_ARGS__) #define plt_nix_dbg(fmt, ...) plt_dbg(nix, fmt, ##__VA_ARGS__) #define plt_npc_dbg(fmt, ...) plt_dbg(npc, fmt, ##__VA_ARGS__) diff --git a/drivers/common/cnxk/roc_priv.h b/drivers/common/cnxk/roc_priv.h index 122d411fe7..14fe2e452a 100644 --- a/drivers/common/cnxk/roc_priv.h +++ b/drivers/common/cnxk/roc_priv.h @@ -47,4 +47,7 @@ /* REE */ #include "roc_ree_priv.h" +/* ML */ +#include "roc_ml_priv.h" + #endif /* _ROC_PRIV_H_ */ diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map index 4bc14901a7..b298a21b84 100644 --- a/drivers/common/cnxk/version.map +++ b/drivers/common/cnxk/version.map @@ -8,6 +8,7 @@ INTERNAL { cnxk_logtype_base; cnxk_logtype_cpt; cnxk_logtype_mbox; + cnxk_logtype_ml; cnxk_logtype_nix; cnxk_logtype_npa; cnxk_logtype_npc; @@ -98,6 +99,34 @@ INTERNAL { roc_idev_npa_nix_get; roc_idev_num_lmtlines_get; roc_idev_nix_inl_meta_aura_get; + roc_ml_reg_read64; + roc_ml_reg_write64; + roc_ml_reg_read32; + roc_ml_reg_write32; + roc_ml_reg_save; + roc_ml_addr_ap2mlip; + roc_ml_addr_mlip2ap; + roc_ml_addr_pa_to_offset; + roc_ml_addr_offset_to_pa; + roc_ml_scratch_write_job; + roc_ml_scratch_is_valid_bit_set; + roc_ml_scratch_is_done_bit_set; + roc_ml_scratch_enqueue; + roc_ml_scratch_dequeue; + roc_ml_scratch_queue_reset; + roc_ml_jcmdq_enqueue_lf; + roc_ml_jcmdq_enqueue_sl; + roc_ml_clk_force_on; + roc_ml_clk_force_off; + roc_ml_dma_stall_on; + roc_ml_dma_stall_off; + roc_ml_mlip_is_enabled; + roc_ml_mlip_reset; + roc_ml_dev_init; + roc_ml_dev_fini; + roc_ml_blk_init; + roc_ml_blk_fini; + roc_ml_sso_pf_func_get; roc_model; roc_se_auth_key_set; roc_se_ciph_key_set; From patchwork Fri Mar 10 08:19:38 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124947 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 3C5AC41D3D; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id D68E641140; Fri, 10 Mar 2023 09:20:26 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id A1BD540141 for ; Fri, 10 Mar 2023 09:20:24 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV16009730; Fri, 10 Mar 2023 00:20:23 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=rs1V8FbY5Am1k1FJ+0fmZHKLoi8aAM6KUuxxo1h9iYQ=; b=joWaEDTULPrtgBj6/3gcQUAhS40lYdxwIZX1Oi7Fr+HNb7wPJHsXJsO2YiL6MICui+Iy cQBcFjO5bVM5JjDbfVLj5pceogt/muF3anaAnSCTPRskAerHMrZQ6newXjYc/A/zmdLF JLkrhJNB1e1rWExCsIjAhFzi4okV+RvHd1D8+Y2obLYy8o+ib8FMGFCL9n3Fb4VgF5LS ZGJtT95wLmvdzkrlgj86XRsjVmCEMt7+4wKxHo71AcjvPQy/qkwxAOACilbxkIV7iqFt /WFZ3B4abc7ONXeMnR9Nbnis4tvpO/XS2Y50We3HwRED8NOWw+o02xzW7bTGuuBwMrim Xg== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Fri, 10 Mar 2023 00:20:23 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:21 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 6E3CE5C68E8; Fri, 10 Mar 2023 00:20:18 -0800 (PST) From: Srikanth Yalavarthi To: Thomas Monjalon , Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 02/39] ml/cnxk: add skeleton for ML cnxk driver Date: Fri, 10 Mar 2023 00:19:38 -0800 Message-ID: <20230310082015.20200-3-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: bW1vaGp96r1FB9laAWBLNN_M2gyahZo1 X-Proofpoint-ORIG-GUID: bW1vaGp96r1FB9laAWBLNN_M2gyahZo1 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added initial source files and build files for ML cnxk driver. Signed-off-by: Srikanth Yalavarthi --- MAINTAINERS | 1 + doc/guides/rel_notes/release_23_03.rst | 7 +++++++ drivers/meson.build | 1 + drivers/ml/cnxk/cn10k_ml_dev.c | 8 ++++++++ drivers/ml/cnxk/cn10k_ml_dev.h | 8 ++++++++ drivers/ml/cnxk/meson.build | 26 ++++++++++++++++++++++++++ drivers/ml/meson.build | 8 ++++++++ 7 files changed, 59 insertions(+) create mode 100644 drivers/ml/cnxk/cn10k_ml_dev.c create mode 100644 drivers/ml/cnxk/cn10k_ml_dev.h create mode 100644 drivers/ml/cnxk/meson.build create mode 100644 drivers/ml/meson.build diff --git a/MAINTAINERS b/MAINTAINERS index d58df9197c..8f695516c7 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1449,6 +1449,7 @@ Marvell ML CNXK M: Srikanth Yalavarthi F: drivers/common/cnxk/hw/ml.h F: drivers/common/cnxk/roc_ml* +F: drivers/ml/cnxk/ Packet processing diff --git a/doc/guides/rel_notes/release_23_03.rst b/doc/guides/rel_notes/release_23_03.rst index 8186545082..851c41e4e0 100644 --- a/doc/guides/rel_notes/release_23_03.rst +++ b/doc/guides/rel_notes/release_23_03.rst @@ -224,6 +224,13 @@ New Features * Test case for inferences from multiple models in ordered mode. * Test case for inferences from multiple models.in interleaving mode. +* **Implementation of Marvell CNXK machine learning driver. ** + + * Added ml/cnxk driver which provides support for machine learning inference + operations on Marvell's CN10K series of SoC's. + * Added ML ROC code for ml/cnxk driver to common/cnxk. + * Added implementation with support for all rte_ml APIs. + Removed Items ------------- diff --git a/drivers/meson.build b/drivers/meson.build index 0618c31a69..31924823e1 100644 --- a/drivers/meson.build +++ b/drivers/meson.build @@ -14,6 +14,7 @@ subdirs = [ 'mempool', # depends on common and bus. 'dma', # depends on common and bus. 'net', # depends on common, bus, mempool + 'ml', # depends on common, bus, mempool 'raw', # depends on common, bus, dma and net. 'crypto', # depends on common, bus and mempool (net in future). 'compress', # depends on common, bus, mempool. diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c new file mode 100644 index 0000000000..cc96a7bdb3 --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -0,0 +1,8 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#include +#include + +#include "cn10k_ml_dev.h" diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h new file mode 100644 index 0000000000..049ac13fcd --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -0,0 +1,8 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _CN10K_ML_DEV_H_ +#define _CN10K_ML_DEV_H_ + +#endif /* _CN10K_ML_DEV_H_ */ diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build new file mode 100644 index 0000000000..2ec6a88e3f --- /dev/null +++ b/drivers/ml/cnxk/meson.build @@ -0,0 +1,26 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright (c) 2022 Marvell. + +if not is_linux or not dpdk_conf.get('RTE_ARCH_64') + build = false + reason = 'only supported on 64-bit Linux' + subdir_done() +endif + +driver_sdk_headers = files( + 'cn10k_ml_dev.h', +) + +sources = files( + 'cn10k_ml_dev.c', +) + +deps += ['mldev', 'common_cnxk'] + +if get_option('buildtype').contains('debug') + cflags += [ '-DCNXK_ML_DEV_DEBUG' ] +else + cflags += [ '-UCNXK_ML_DEV_DEBUG' ] +endif + +pmd_supports_disable_iova_as_pa = true diff --git a/drivers/ml/meson.build b/drivers/ml/meson.build new file mode 100644 index 0000000000..54bc394c47 --- /dev/null +++ b/drivers/ml/meson.build @@ -0,0 +1,8 @@ +# SPDX-License-Identifier: BSD-3-Clause +# Copyright (c) 2022 Marvell. + +drivers = [ + 'cnxk', +] + +std_deps = ['mldev'] From patchwork Fri Mar 10 08:19:39 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124954 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 9F76941D3D; Fri, 10 Mar 2023 09:21:12 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 61DA242D29; Fri, 10 Mar 2023 09:20:35 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 206124161A for ; Fri, 10 Mar 2023 09:20:27 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId5021875; Fri, 10 Mar 2023 00:20:26 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=2SFZ1LQtLSDaUp9R90Y+TQ7Qvd4L640wiK/wG5NVntc=; b=eqGrfdk6qdLw0dqEwq91PQwx+/4jm+zMD+AUMBCNwQ9mj+ivM8oF0TQv5opmGymmxW40 f8tuN/B+LP2Sv5yWv4mL2CeuDWiHLH/L1IQcIrKN/YTdR9bD6+wfllr8Rj8QtB5Gt7/a 4M/zKaH+WkFeTyU5ODVVafqe0OwXuJCBm+c6RFo43V5D3kmDcYDW6MPWEilNUmOi3R26 tUdiEE8FW77p8xLpaHKK1Lhk+RXMmJXz/vaL8vHeAZnL3sesgxE1faADldTKDDUxl/iG 00Ufs51z1VNsClYWoS1WgpbCgSxG+U1UxzQuz0N2ILSzAEuIRXjdiqkT8aM95WeLN0dZ kQ== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-4 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Fri, 10 Mar 2023 00:20:26 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:21 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 9E6C05C68F2; Fri, 10 Mar 2023 00:20:18 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi , Anatoly Burakov CC: , , , , , Subject: [PATCH v6 03/39] ml/cnxk: enable probe and remove of ML device Date: Fri, 10 Mar 2023 00:19:39 -0800 Message-ID: <20230310082015.20200-4-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: wx20eizcrrLYLoHs7TLkuq-CLVkvWJ-i X-Proofpoint-ORIG-GUID: wx20eizcrrLYLoHs7TLkuq-CLVkvWJ-i X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org ML inference engine on cn10k platform is a PCI based device. Added driver support to probe and remove the device for cn10k poll mode driver. The device is named by the PMD as "ml_cn10k". Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 114 +++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_dev.h | 11 ++++ drivers/ml/cnxk/cn10k_ml_ops.c | 10 +++ drivers/ml/cnxk/cn10k_ml_ops.h | 11 ++++ drivers/ml/cnxk/meson.build | 2 + 5 files changed, 148 insertions(+) create mode 100644 drivers/ml/cnxk/cn10k_ml_ops.c create mode 100644 drivers/ml/cnxk/cn10k_ml_ops.h diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index cc96a7bdb3..c2e93c9a1a 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -2,7 +2,121 @@ * Copyright (c) 2022 Marvell. */ +#include +#include #include #include +#include + +#include #include "cn10k_ml_dev.h" +#include "cn10k_ml_ops.h" + +/* Dummy operations for ML device */ +struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; + +static int +cn10k_ml_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) +{ + struct rte_ml_dev_pmd_init_params init_params; + struct cn10k_ml_dev *mldev; + char name[RTE_ML_STR_MAX]; + struct rte_ml_dev *dev; + int ret; + + PLT_SET_USED(pci_drv); + + init_params = (struct rte_ml_dev_pmd_init_params){ + .socket_id = rte_socket_id(), .private_data_size = sizeof(struct cn10k_ml_dev)}; + + ret = roc_plt_init(); + if (ret < 0) { + plt_err("Failed to initialize platform model"); + return ret; + } + + rte_pci_device_name(&pci_dev->addr, name, sizeof(name)); + dev = rte_ml_dev_pmd_create(name, &pci_dev->device, &init_params); + if (dev == NULL) { + ret = -ENODEV; + goto error_exit; + } + + /* Get private data space allocated */ + mldev = dev->data->dev_private; + + if (rte_eal_process_type() == RTE_PROC_PRIMARY) { + mldev->roc.pci_dev = pci_dev; + + ret = roc_ml_dev_init(&mldev->roc); + if (ret) { + plt_err("Failed to initialize ML ROC, ret = %d", ret); + goto pmd_destroy; + } + + dev->dev_ops = &cn10k_ml_ops; + } else { + plt_err("CN10K ML Ops are not supported on secondary process"); + dev->dev_ops = &ml_dev_dummy_ops; + } + + dev->enqueue_burst = NULL; + dev->dequeue_burst = NULL; + dev->op_error_get = NULL; + + return 0; + +pmd_destroy: + rte_ml_dev_pmd_destroy(dev); + +error_exit: + plt_err("Could not create device (vendor_id: 0x%x device_id: 0x%x)", pci_dev->id.vendor_id, + pci_dev->id.device_id); + + return ret; +} + +static int +cn10k_ml_pci_remove(struct rte_pci_device *pci_dev) +{ + struct cn10k_ml_dev *mldev; + char name[RTE_ML_STR_MAX]; + struct rte_ml_dev *dev; + int ret; + + if (pci_dev == NULL) + return -EINVAL; + + rte_pci_device_name(&pci_dev->addr, name, sizeof(name)); + + dev = rte_ml_dev_pmd_get_named_dev(name); + if (dev == NULL) + return -ENODEV; + + if (rte_eal_process_type() == RTE_PROC_PRIMARY) { + mldev = dev->data->dev_private; + ret = roc_ml_dev_fini(&mldev->roc); + if (ret) + return ret; + } + + return rte_ml_dev_pmd_destroy(dev); +} + +static struct rte_pci_id pci_id_ml_table[] = { + {RTE_PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_CN10K_ML_PF)}, + /* sentinel */ + {}, +}; + +static struct rte_pci_driver cn10k_mldev_pmd = { + .id_table = pci_id_ml_table, + .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_NEED_IOVA_AS_VA, + .probe = cn10k_ml_pci_probe, + .remove = cn10k_ml_pci_remove, +}; + +RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); +RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); +RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 049ac13fcd..833a09791a 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -5,4 +5,15 @@ #ifndef _CN10K_ML_DEV_H_ #define _CN10K_ML_DEV_H_ +#include + +/* Marvell OCTEON CN10K ML PMD device name */ +#define MLDEV_NAME_CN10K_PMD ml_cn10k + +/* Device private data */ +struct cn10k_ml_dev { + /* Device ROC */ + struct roc_ml roc; +}; + #endif /* _CN10K_ML_DEV_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c new file mode 100644 index 0000000000..39843e3ee5 --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -0,0 +1,10 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#include +#include + +#include "cn10k_ml_ops.h" + +struct rte_ml_dev_ops cn10k_ml_ops = {0}; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h new file mode 100644 index 0000000000..b14221d02c --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _CN10K_ML_OPS_H_ +#define _CN10K_ML_OPS_H_ + +/* Device ops */ +extern struct rte_ml_dev_ops cn10k_ml_ops; + +#endif /* _CN10K_ML_OPS_H_ */ diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build index 2ec6a88e3f..caed62a9f3 100644 --- a/drivers/ml/cnxk/meson.build +++ b/drivers/ml/cnxk/meson.build @@ -9,10 +9,12 @@ endif driver_sdk_headers = files( 'cn10k_ml_dev.h', + 'cn10k_ml_ops.h', ) sources = files( 'cn10k_ml_dev.c', + 'cn10k_ml_ops.c', ) deps += ['mldev', 'common_cnxk'] From patchwork Fri Mar 10 08:19:40 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124948 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 20D0141D3D; Fri, 10 Mar 2023 09:20:35 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 16780427E9; Fri, 10 Mar 2023 09:20:28 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id EC38B40ED9 for ; Fri, 10 Mar 2023 09:20:24 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV17009730 for ; Fri, 10 Mar 2023 00:20:24 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=nRWowxXrv/j/E7IHjvGnHP75t00sN6kQlM76kytHbOo=; b=Hyxlwd5NTAlprCbTuVYfPgHajwwvf9SYqsFjVzfSGAzljISIKf31rq3TBI4kloSJ6GJ9 KtsMZYG437aTRrvBFQBVDTgNUzjN05/Q4GdZspG9AkZPLMplhoj+xYs/9vykx1QKTrWF fM0/Rnl2tG+egVTEBXR4mwUtPMR9XT2Ezg1Ff1nUf5E79UBT32LTQIpcFFz9VzgguEhr ErIvE2CqjXuAHR2zIOjD41MtWLdBpWlr+GgE6hTbUQvkUMawfIc1Ss3ERnJGZ/fyiip3 NBtGzSpQXJpX9L0ARo7tBiCa9S5ICWYDBRYyopNbkPh0gN9PBZLiHXxd0/ghz3htacFO mQ== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-3 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:23 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:21 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id C92AE5C68F6; Fri, 10 Mar 2023 00:20:18 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 04/39] ml/cnxk: add driver support to get device info Date: Fri, 10 Mar 2023 00:19:40 -0800 Message-ID: <20230310082015.20200-5-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: cVMDrW9X8Vah3K4JZK4R_FzJoRKZHOac X-Proofpoint-ORIG-GUID: cVMDrW9X8Vah3K4JZK4R_FzJoRKZHOac X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support to get the cn10k ML device information. This is a driver implementation for the RTE function rte_ml_dev_info_get. ML device on cn10k supports one queue-pair in lock-free mode and does not support segmented input output data. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 15 +++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 23 ++++++++++++++++++++++- 2 files changed, 37 insertions(+), 1 deletion(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 833a09791a..13d26373e4 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -10,6 +10,21 @@ /* Marvell OCTEON CN10K ML PMD device name */ #define MLDEV_NAME_CN10K_PMD ml_cn10k +/* Device alignment size */ +#define ML_CN10K_ALIGN_SIZE 128 + +/* Maximum number of models per device */ +#define ML_CN10K_MAX_MODELS 16 + +/* Maximum number of queue-pairs per device */ +#define ML_CN10K_MAX_QP_PER_DEVICE 1 + +/* Maximum number of descriptors per queue-pair */ +#define ML_CN10K_MAX_DESC_PER_QP 1024 + +/* Maximum number of segments for IO data */ +#define ML_CN10K_MAX_SEGMENTS 1 + /* Device private data */ struct cn10k_ml_dev { /* Device ROC */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 39843e3ee5..bad5ad4713 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -5,6 +5,27 @@ #include #include +#include "cn10k_ml_dev.h" #include "cn10k_ml_ops.h" -struct rte_ml_dev_ops cn10k_ml_ops = {0}; +static int +cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) +{ + if (dev_info == NULL) + return -EINVAL; + + memset(dev_info, 0, sizeof(struct rte_ml_dev_info)); + dev_info->driver_name = dev->device->driver->name; + dev_info->max_models = ML_CN10K_MAX_MODELS; + dev_info->max_queue_pairs = ML_CN10K_MAX_QP_PER_DEVICE; + dev_info->max_desc = ML_CN10K_MAX_DESC_PER_QP; + dev_info->max_segments = ML_CN10K_MAX_SEGMENTS; + dev_info->min_align_size = ML_CN10K_ALIGN_SIZE; + + return 0; +} + +struct rte_ml_dev_ops cn10k_ml_ops = { + /* Device control ops */ + .dev_info_get = cn10k_ml_dev_info_get, +}; From patchwork Fri Mar 10 08:19:41 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124949 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 2E21441D3D; Fri, 10 Mar 2023 09:20:40 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 3CE6F42B71; Fri, 10 Mar 2023 09:20:29 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 6C5E640141 for ; Fri, 10 Mar 2023 09:20:25 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV18009730 for ; Fri, 10 Mar 2023 00:20:24 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=2RFu/gl/NqbNXc5+dVPUy7n0x0C9JfRPZ42jF/U228E=; b=LRJY4xU9cxlFWmc8K+kezme66vkV2b/8JoaGFRve6gzdc9/D0WnPXEVW1WoYFqCW80rN 0awNfMzK8uxH5uWXac82iC1jInnv6cl1KQeW1QK+HuZ0rZZhW2mDfgkKDywI5YAiaE2C cPPn50blAlc8Qb00VqvjVKmhkyWs+7152kFE+67GIUVvG9xZw0A50UuvWZN+Fo83hj7f 0dirD38iEGxzDTL0Le1BVhfR/pl/x7t8rwjsD/3LEVvW6VL7nI0XgWoFscZv/KPdstdm 3YLkNDhd0NnW8RqltNhf2Du66FvT9idn73oSAwWwBFhrx4jTQ8OJVNyU4/UPXKhYr1M2 GA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-4 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:24 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:21 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id F40145C68F8; Fri, 10 Mar 2023 00:20:18 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 05/39] ml/cnxk: add support for configure and close Date: Fri, 10 Mar 2023 00:19:41 -0800 Message-ID: <20230310082015.20200-6-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: H_NblFZe60Yi8QracijIL2fyLsOW7uVm X-Proofpoint-ORIG-GUID: H_NblFZe60Yi8QracijIL2fyLsOW7uVm X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Implemented driver functions to configure and close ML devices. Added skeleton code and support to reconfigure ML device. PCI device remove is enabled in device close. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 2 ++ drivers/ml/cnxk/cn10k_ml_dev.h | 21 ++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 60 ++++++++++++++++++++++++++++++++++ 3 files changed, 83 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index c2e93c9a1a..fd45226add 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -65,6 +65,8 @@ cn10k_ml_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_de dev->dequeue_burst = NULL; dev->op_error_get = NULL; + mldev->state = ML_CN10K_DEV_STATE_PROBED; + return 0; pmd_destroy: diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 13d26373e4..e7fb5fc2e2 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -25,10 +25,31 @@ /* Maximum number of segments for IO data */ #define ML_CN10K_MAX_SEGMENTS 1 +/* ML command timeout in seconds */ +#define ML_CN10K_CMD_TIMEOUT 5 + +/* Device configuration state enum */ +enum cn10k_ml_dev_state { + /* Probed and not configured */ + ML_CN10K_DEV_STATE_PROBED = 0, + + /* Configured */ + ML_CN10K_DEV_STATE_CONFIGURED, + + /* Started */ + ML_CN10K_DEV_STATE_STARTED, + + /* Closed */ + ML_CN10K_DEV_STATE_CLOSED +}; + /* Device private data */ struct cn10k_ml_dev { /* Device ROC */ struct roc_ml roc; + + /* Configuration state */ + enum cn10k_ml_dev_state state; }; #endif /* _CN10K_ML_DEV_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index bad5ad4713..3a78d8c816 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -25,7 +25,67 @@ cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) return 0; } +static int +cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *conf) +{ + struct rte_ml_dev_info dev_info; + struct cn10k_ml_dev *mldev; + + if (dev == NULL || conf == NULL) + return -EINVAL; + + /* Get CN10K device handle */ + mldev = dev->data->dev_private; + + cn10k_ml_dev_info_get(dev, &dev_info); + if (conf->nb_models > dev_info.max_models) { + plt_err("Invalid device config, nb_models > %u\n", dev_info.max_models); + return -EINVAL; + } + + if (conf->nb_queue_pairs > dev_info.max_queue_pairs) { + plt_err("Invalid device config, nb_queue_pairs > %u\n", dev_info.max_queue_pairs); + return -EINVAL; + } + + if (mldev->state == ML_CN10K_DEV_STATE_PROBED) { + plt_ml_dbg("Configuring ML device, nb_queue_pairs = %u, nb_models = %u", + conf->nb_queue_pairs, conf->nb_models); + } else if (mldev->state == ML_CN10K_DEV_STATE_CONFIGURED) { + plt_ml_dbg("Re-configuring ML device, nb_queue_pairs = %u, nb_models = %u", + conf->nb_queue_pairs, conf->nb_models); + } else if (mldev->state == ML_CN10K_DEV_STATE_STARTED) { + plt_err("Device can't be reconfigured in started state\n"); + return -ENOTSUP; + } else if (mldev->state == ML_CN10K_DEV_STATE_CLOSED) { + plt_err("Device can't be reconfigured after close\n"); + return -ENOTSUP; + } + + mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; + + return 0; +} + +static int +cn10k_ml_dev_close(struct rte_ml_dev *dev) +{ + struct cn10k_ml_dev *mldev; + + if (dev == NULL) + return -EINVAL; + + mldev = dev->data->dev_private; + + mldev->state = ML_CN10K_DEV_STATE_CLOSED; + + /* Remove PCI device */ + return rte_dev_remove(dev->device); +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, + .dev_configure = cn10k_ml_dev_configure, + .dev_close = cn10k_ml_dev_close, }; From patchwork Fri Mar 10 08:19:42 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124956 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 3FC7E41D3D; Fri, 10 Mar 2023 09:21:25 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id B40D642D38; Fri, 10 Mar 2023 09:20:37 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 8F00441151 for ; Fri, 10 Mar 2023 09:20:27 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId6021875 for ; Fri, 10 Mar 2023 00:20:27 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=h+nuKyOStfUr7Wzp0+ID0qetCKzgh30YE4DUhFIe5g4=; b=kmRfxg7O5zp2ymL+pWgf/2SpZx15QVRz/Pxf3zpCFm189e7Th+BP+QcER5N6mTg+j1Hp 9LfVkHlR4WuQAIkkvUV5zJKmypoNGmbyqt2R1YgV1RJIRBplm0QAmBSgJicrzqdGepQ5 XcWcV9httid1M40wS4UZMw7bl/r7VXXcZ2FpkkD9Gzd53xN/yicg7Z3TGWJBrvLOMioI 4GdxpjwOjhwy+fEk2fNXBxfqho7yNfKxJZ+Al+tS9CnBK1WdyLjwWp5mPeukoqKWa2kI ViLi9mmIagiwLNTdSpAjR5dk3ClY2PGw375D96iQHmSq/GXmAWS3p3EfFRIEGkUn6M3V eg== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-5 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:26 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:21 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 28DE35C68FB; Fri, 10 Mar 2023 00:20:19 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 06/39] ml/cnxk: parse ML firmware path from device args Date: Fri, 10 Mar 2023 00:19:42 -0800 Message-ID: <20230310082015.20200-7-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: QJeuuY-VDghVMkEtRSY-kjpKXGGTDN-K X-Proofpoint-ORIG-GUID: QJeuuY-VDghVMkEtRSY-kjpKXGGTDN-K X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled parsing ML firmware path for cn10k. Default path is set as "/lib/firmware/mlip-fw.bin", when args are not provided. Added internal structures for ML firmware. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 71 ++++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_dev.h | 12 ++++++ drivers/ml/cnxk/meson.build | 2 +- 3 files changed, 84 insertions(+), 1 deletion(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index fd45226add..117cac43aa 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -4,6 +4,8 @@ #include #include +#include +#include #include #include #include @@ -13,9 +15,70 @@ #include "cn10k_ml_dev.h" #include "cn10k_ml_ops.h" +#define CN10K_ML_FW_PATH "fw_path" + +#define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" + +static const char *const valid_args[] = {CN10K_ML_FW_PATH, NULL}; + /* Dummy operations for ML device */ struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; +static int +parse_string_arg(const char *key __rte_unused, const char *value, void *extra_args) +{ + if (value == NULL || extra_args == NULL) + return -EINVAL; + + *(char **)extra_args = strdup(value); + + if (!*(char **)extra_args) + return -ENOMEM; + + return 0; +} + +static int +cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mldev) +{ + struct rte_kvargs *kvlist = NULL; + bool fw_path_set = false; + char *fw_path = NULL; + int ret = 0; + + if (devargs == NULL) + goto check_args; + + kvlist = rte_kvargs_parse(devargs->args, valid_args); + if (kvlist == NULL) { + plt_err("Error parsing devargs\n"); + return -EINVAL; + } + + if (rte_kvargs_count(kvlist, CN10K_ML_FW_PATH) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_FW_PATH, &parse_string_arg, &fw_path); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", CN10K_ML_FW_PATH); + ret = -EINVAL; + goto exit; + } + fw_path_set = true; + } + +check_args: + if (!fw_path_set) + mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; + else + mldev->fw.path = fw_path; + plt_info("ML: %s = %s", CN10K_ML_FW_PATH, mldev->fw.path); + +exit: + if (kvlist) + rte_kvargs_free(kvlist); + + return ret; +} + static int cn10k_ml_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) { @@ -49,6 +112,12 @@ cn10k_ml_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_de if (rte_eal_process_type() == RTE_PROC_PRIMARY) { mldev->roc.pci_dev = pci_dev; + ret = cn10k_mldev_parse_devargs(dev->device->devargs, mldev); + if (ret) { + plt_err("Failed to parse devargs ret = %d", ret); + goto pmd_destroy; + } + ret = roc_ml_dev_init(&mldev->roc); if (ret) { plt_err("Failed to initialize ML ROC, ret = %d", ret); @@ -122,3 +191,5 @@ static struct rte_pci_driver cn10k_mldev_pmd = { RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); + +RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH "="); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index e7fb5fc2e2..5333566cff 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -43,6 +43,15 @@ enum cn10k_ml_dev_state { ML_CN10K_DEV_STATE_CLOSED }; +/* ML firmware structure */ +struct cn10k_ml_fw { + /* Device reference */ + struct cn10k_ml_dev *mldev; + + /* Firmware file path */ + const char *path; +}; + /* Device private data */ struct cn10k_ml_dev { /* Device ROC */ @@ -50,6 +59,9 @@ struct cn10k_ml_dev { /* Configuration state */ enum cn10k_ml_dev_state state; + + /* Firmware */ + struct cn10k_ml_fw fw; }; #endif /* _CN10K_ML_DEV_H_ */ diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build index caed62a9f3..7dc8a29a80 100644 --- a/drivers/ml/cnxk/meson.build +++ b/drivers/ml/cnxk/meson.build @@ -17,7 +17,7 @@ sources = files( 'cn10k_ml_ops.c', ) -deps += ['mldev', 'common_cnxk'] +deps += ['mldev', 'common_cnxk', 'kvargs'] if get_option('buildtype').contains('debug') cflags += [ '-DCNXK_ML_DEV_DEBUG' ] From patchwork Fri Mar 10 08:19:43 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124950 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 3D40A41D3D; Fri, 10 Mar 2023 09:20:46 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 6321142BC9; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 04EF040141 for ; Fri, 10 Mar 2023 09:20:25 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV19009730 for ; Fri, 10 Mar 2023 00:20:25 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=nQ6iYw4LCw+Fs1foBg17pdZEbqXiWITP7G3zayNHCG0=; b=G36shqWgyvJfyD3Z+app0UU1+QDyK9YzhHtSqRZNONrCCiHksV2uxDrGtJvrzK0/BRRK c0OMcRIALpc+P8k7uPCNyicNXABxNb5a74T+zNaPhBDwz4ahPrLP4wStTD5NdKC9gD0L 2HfbyOFwej2/3nuSAJcHMpIE42BFQ8FzatFkPrS50lDgM/Y9ClfvWxuNRvrYlIf7IFH0 6tJksXcY9QyutLmwHsnAQnbKBfzKfncwftMXxCBKR6msoe3NV2P1xefOCrLfwvO7ea0U 6k0FoPVGe72foJT80gD9e0P3JwLQwRKbxxWOTWDDF6fhAozfTyrmElW3D4yOMu2FWUtG Tg== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-5 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:24 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:21 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 532275C68FD; Fri, 10 Mar 2023 00:20:19 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 07/39] ml/cnxk: enable firmware load and device reset Date: Fri, 10 Mar 2023 00:19:43 -0800 Message-ID: <20230310082015.20200-8-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: COh_fonzrEwCtL83xi90ulS188Zn00ty X-Proofpoint-ORIG-GUID: COh_fonzrEwCtL83xi90ulS188Zn00ty X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support to load ML firmware on cn10ka ROC model. Reset MLIP device during dev_close driver operation. Device can't be reconfigured after a call to close. Job execution is disabled after firmware load, execution is enabled in device start state. Added internal request structure. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 327 +++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_dev.h | 156 ++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 21 +++ drivers/ml/cnxk/cn10k_ml_ops.h | 14 ++ 4 files changed, 518 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 117cac43aa..90fca45ddd 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -12,6 +12,8 @@ #include +#include + #include "cn10k_ml_dev.h" #include "cn10k_ml_ops.h" @@ -19,6 +21,15 @@ #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" +/* ML firmware macros */ +#define FW_MEMZONE_NAME "ml_cn10k_fw_mz" +#define FW_STACK_BUFFER_SIZE 0x40000 +#define FW_DEBUG_BUFFER_SIZE (2 * 0x20000) +#define FW_EXCEPTION_BUFFER_SIZE 0x400 +#define FW_LINKER_OFFSET 0x80000 +#define FW_WAIT_CYCLES 100 +#define FW_LOAD_FLAGS 0x1 + static const char *const valid_args[] = {CN10K_ML_FW_PATH, NULL}; /* Dummy operations for ML device */ @@ -175,6 +186,322 @@ cn10k_ml_pci_remove(struct rte_pci_device *pci_dev) return rte_ml_dev_pmd_destroy(dev); } +static void +cn10k_ml_fw_print_info(struct cn10k_ml_fw *fw) +{ + plt_info("ML Firmware Version = %s", fw->req->jd.fw_load.version); + + plt_ml_dbg("Firmware capabilities = 0x%016lx", fw->req->jd.fw_load.cap.u64); + plt_ml_dbg("Version = %s", fw->req->jd.fw_load.version); + plt_ml_dbg("core0_debug_ptr = 0x%016lx", fw->req->jd.fw_load.debug.core0_debug_ptr); + plt_ml_dbg("core1_debug_ptr = 0x%016lx", fw->req->jd.fw_load.debug.core1_debug_ptr); + plt_ml_dbg("debug_buffer_size = %u bytes", fw->req->jd.fw_load.debug.debug_buffer_size); + plt_ml_dbg("core0_exception_buffer = 0x%016lx", + fw->req->jd.fw_load.debug.core0_exception_buffer); + plt_ml_dbg("core1_exception_buffer = 0x%016lx", + fw->req->jd.fw_load.debug.core1_exception_buffer); + plt_ml_dbg("exception_state_size = %u bytes", + fw->req->jd.fw_load.debug.exception_state_size); + plt_ml_dbg("flags = 0x%016lx", fw->req->jd.fw_load.flags); +} + +uint64_t +cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw) +{ + PLT_SET_USED(fw); + + return FW_LOAD_FLAGS; +} + +static int +cn10k_ml_fw_load_cn10ka(struct cn10k_ml_fw *fw, void *buffer, uint64_t size) +{ + union ml_a35_0_rst_vector_base_s a35_0_rst_vector_base; + union ml_a35_0_rst_vector_base_s a35_1_rst_vector_base; + struct cn10k_ml_dev *mldev; + uint64_t timeout_cycle; + uint64_t reg_val64; + uint32_t reg_val32; + uint64_t offset; + bool timeout; + int ret = 0; + uint8_t i; + + mldev = fw->mldev; + + /* Reset HEAD and TAIL debug pointer registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C1); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C1); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_EXCEPTION_SP_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_EXCEPTION_SP_C1); + + /* (1) Write firmware images for ACC's two A35 cores to the ML region in LLC / DRAM. */ + rte_memcpy(PLT_PTR_ADD(fw->data, FW_LINKER_OFFSET), buffer, size); + + /* (2) Set ML(0)_MLR_BASE = Base IOVA of the ML region in LLC/DRAM. */ + reg_val64 = PLT_PTR_SUB_U64_CAST(fw->data, rte_eal_get_baseaddr()); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_MLR_BASE); + plt_ml_dbg("ML_MLR_BASE => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_MLR_BASE)); + roc_ml_reg_save(&mldev->roc, ML_MLR_BASE); + + /* (3) Set ML(0)_AXI_BRIDGE_CTRL(1) = 0x184003 to remove back-pressure check on DMA AXI + * bridge. + */ + reg_val64 = (ROC_ML_AXI_BRIDGE_CTRL_AXI_RESP_CTRL | + ROC_ML_AXI_BRIDGE_CTRL_BRIDGE_CTRL_MODE | ROC_ML_AXI_BRIDGE_CTRL_NCB_WR_BLK | + ROC_ML_AXI_BRIDGE_CTRL_FORCE_WRESP_OK | ROC_ML_AXI_BRIDGE_CTRL_FORCE_RRESP_OK); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_AXI_BRIDGE_CTRL(1)); + plt_ml_dbg("ML_AXI_BRIDGE_CTRL(1) => 0x%016lx", + roc_ml_reg_read64(&mldev->roc, ML_AXI_BRIDGE_CTRL(1))); + + /* (4) Set ML(0)_ANB(0..2)_BACKP_DISABLE = 0x3 to remove back-pressure on the AXI to NCB + * bridges. + */ + for (i = 0; i < ML_ANBX_NR; i++) { + reg_val64 = (ROC_ML_ANBX_BACKP_DISABLE_EXTMSTR_B_BACKP_DISABLE | + ROC_ML_ANBX_BACKP_DISABLE_EXTMSTR_R_BACKP_DISABLE); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_ANBX_BACKP_DISABLE(i)); + plt_ml_dbg("ML_ANBX_BACKP_DISABLE(%u) => 0x%016lx", i, + roc_ml_reg_read64(&mldev->roc, ML_ANBX_BACKP_DISABLE(i))); + } + + /* (5) Set ML(0)_ANB(0..2)_NCBI_P_OVR = 0x3000 and ML(0)_ANB(0..2)_NCBI_NP_OVR = 0x3000 to + * signal all ML transactions as non-secure. + */ + for (i = 0; i < ML_ANBX_NR; i++) { + reg_val64 = (ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_NS_OVR | + ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_NS_OVR_VLD); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_ANBX_NCBI_P_OVR(i)); + plt_ml_dbg("ML_ANBX_NCBI_P_OVR(%u) => 0x%016lx", i, + roc_ml_reg_read64(&mldev->roc, ML_ANBX_NCBI_P_OVR(i))); + + reg_val64 |= (ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_NS_OVR | + ML_ANBX_NCBI_NP_OVR_ANB_NCBI_NP_NS_OVR_VLD); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_ANBX_NCBI_NP_OVR(i)); + plt_ml_dbg("ML_ANBX_NCBI_NP_OVR(%u) => 0x%016lx", i, + roc_ml_reg_read64(&mldev->roc, ML_ANBX_NCBI_NP_OVR(i))); + } + + /* (6) Set ML(0)_CFG[MLIP_CLK_FORCE] = 1, to force turning on the MLIP clock. */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 |= ROC_ML_CFG_MLIP_CLK_FORCE; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + /* (7) Set ML(0)_JOB_MGR_CTRL[STALL_ON_IDLE] = 0, to make sure the boot request is accepted + * when there is no job in the command queue. + */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_JOB_MGR_CTRL); + reg_val64 &= ~ROC_ML_JOB_MGR_CTRL_STALL_ON_IDLE; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_JOB_MGR_CTRL); + plt_ml_dbg("ML_JOB_MGR_CTRL => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_JOB_MGR_CTRL)); + + /* (8) Set ML(0)_CFG[ENA] = 0 and ML(0)_CFG[MLIP_ENA] = 1 to bring MLIP out of reset while + * keeping the job manager disabled. + */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 |= ROC_ML_CFG_MLIP_ENA; + reg_val64 &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + /* (9) Wait at least 70 coprocessor clock cycles. */ + plt_delay_us(FW_WAIT_CYCLES); + + /* (10) Write ML outbound addresses pointing to the firmware images written in step 1 to the + * following registers: ML(0)_A35_0_RST_VECTOR_BASE_W(0..1) for core 0, + * ML(0)_A35_1_RST_VECTOR_BASE_W(0..1) for core 1. The value written to each register is the + * AXI outbound address divided by 4. Read after write. + */ + offset = PLT_PTR_ADD_U64_CAST( + fw->data, FW_LINKER_OFFSET - roc_ml_reg_read64(&mldev->roc, ML_MLR_BASE)); + a35_0_rst_vector_base.s.addr = (offset + ML_AXI_START_ADDR) / 4; + a35_1_rst_vector_base.s.addr = (offset + ML_AXI_START_ADDR) / 4; + + roc_ml_reg_write32(&mldev->roc, a35_0_rst_vector_base.w.w0, ML_A35_0_RST_VECTOR_BASE_W(0)); + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_A35_0_RST_VECTOR_BASE_W(0)); + plt_ml_dbg("ML_A35_0_RST_VECTOR_BASE_W(0) => 0x%08x", reg_val32); + + roc_ml_reg_write32(&mldev->roc, a35_0_rst_vector_base.w.w1, ML_A35_0_RST_VECTOR_BASE_W(1)); + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_A35_0_RST_VECTOR_BASE_W(1)); + plt_ml_dbg("ML_A35_0_RST_VECTOR_BASE_W(1) => 0x%08x", reg_val32); + + roc_ml_reg_write32(&mldev->roc, a35_1_rst_vector_base.w.w0, ML_A35_1_RST_VECTOR_BASE_W(0)); + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_A35_1_RST_VECTOR_BASE_W(0)); + plt_ml_dbg("ML_A35_1_RST_VECTOR_BASE_W(0) => 0x%08x", reg_val32); + + roc_ml_reg_write32(&mldev->roc, a35_1_rst_vector_base.w.w1, ML_A35_1_RST_VECTOR_BASE_W(1)); + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_A35_1_RST_VECTOR_BASE_W(1)); + plt_ml_dbg("ML_A35_1_RST_VECTOR_BASE_W(1) => 0x%08x", reg_val32); + + /* (11) Clear MLIP's ML(0)_SW_RST_CTRL[ACC_RST]. This will bring the ACC cores and other + * MLIP components out of reset. The cores will execute firmware from the ML region as + * written in step 1. + */ + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_SW_RST_CTRL); + reg_val32 &= ~ROC_ML_SW_RST_CTRL_ACC_RST; + roc_ml_reg_write32(&mldev->roc, reg_val32, ML_SW_RST_CTRL); + reg_val32 = roc_ml_reg_read32(&mldev->roc, ML_SW_RST_CTRL); + plt_ml_dbg("ML_SW_RST_CTRL => 0x%08x", reg_val32); + + /* (12) Wait for notification from firmware that ML is ready for job execution. */ + fw->req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&fw->req->status); + fw->req->jd.hdr.job_type = ML_CN10K_JOB_TYPE_FIRMWARE_LOAD; + fw->req->jd.hdr.result = roc_ml_addr_ap2mlip(&mldev->roc, &fw->req->result); + fw->req->jd.fw_load.flags = cn10k_ml_fw_flags_get(fw); + plt_write64(ML_CN10K_POLL_JOB_START, &fw->req->status); + plt_wmb(); + + /* Enqueue FW load through scratch registers */ + timeout = true; + timeout_cycle = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + roc_ml_scratch_enqueue(&mldev->roc, &fw->req->jd); + + plt_rmb(); + do { + if (roc_ml_scratch_is_done_bit_set(&mldev->roc) && + (plt_read64(&fw->req->status) == ML_CN10K_POLL_JOB_FINISH)) { + timeout = false; + break; + } + } while (plt_tsc_cycles() < timeout_cycle); + + /* Check firmware load status, clean-up and exit on failure. */ + if ((!timeout) && (fw->req->result.error_code == 0)) { + cn10k_ml_fw_print_info(fw); + } else { + /* Set ML to disable new jobs */ + reg_val64 = (ROC_ML_CFG_JD_SIZE | ROC_ML_CFG_MLIP_ENA); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + + /* Clear scratch registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_FW_CTRL); + + if (timeout) { + plt_err("Firmware load timeout"); + ret = -ETIME; + } else { + plt_err("Firmware load failed"); + ret = -1; + } + + return ret; + } + + /* (13) Set ML(0)_JOB_MGR_CTRL[STALL_ON_IDLE] = 0x1; this is needed to shut down the MLIP + * clock when there are no more jobs to process. + */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_JOB_MGR_CTRL); + reg_val64 |= ROC_ML_JOB_MGR_CTRL_STALL_ON_IDLE; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_JOB_MGR_CTRL); + plt_ml_dbg("ML_JOB_MGR_CTRL => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_JOB_MGR_CTRL)); + + /* (14) Set ML(0)_CFG[MLIP_CLK_FORCE] = 0; the MLIP clock will be turned on/off based on job + * activities. + */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 &= ~ROC_ML_CFG_MLIP_CLK_FORCE; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + /* (15) Set ML(0)_CFG[ENA] to enable ML job execution. */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 |= ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + /* Reset scratch registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_FW_CTRL); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_WORK_PTR); + + /* Disable job execution, to be enabled in start */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + /* Additional fixes: Set RO bit to fix O2D DMA bandwidth issue on cn10ka */ + for (i = 0; i < ML_ANBX_NR; i++) { + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_ANBX_NCBI_P_OVR(i)); + reg_val64 |= (ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_RO_OVR | + ML_ANBX_NCBI_P_OVR_ANB_NCBI_P_RO_OVR_VLD); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_ANBX_NCBI_P_OVR(i)); + } + + return ret; +} + +int +cn10k_ml_fw_load(struct cn10k_ml_dev *mldev) +{ + const struct plt_memzone *mz; + struct cn10k_ml_fw *fw; + void *fw_buffer = NULL; + uint64_t mz_size = 0; + uint64_t fw_size = 0; + int ret = 0; + + fw = &mldev->fw; + fw->mldev = mldev; + + /* Read firmware image to a buffer */ + ret = rte_firmware_read(fw->path, &fw_buffer, &fw_size); + if (ret < 0) { + plt_err("Can't read firmware data: %s\n", fw->path); + return ret; + } + + /* Reserve memzone for firmware load completion and data */ + mz_size = sizeof(struct cn10k_ml_req) + fw_size + FW_STACK_BUFFER_SIZE + + FW_DEBUG_BUFFER_SIZE + FW_EXCEPTION_BUFFER_SIZE; + mz = plt_memzone_reserve_aligned(FW_MEMZONE_NAME, mz_size, 0, ML_CN10K_ALIGN_SIZE); + if (mz == NULL) { + plt_err("plt_memzone_reserve failed : %s", FW_MEMZONE_NAME); + if (fw_buffer != NULL) + free(fw_buffer); + return -ENOMEM; + } + fw->req = mz->addr; + + /* Reset firmware load completion structure */ + memset(&fw->req->jd, 0, sizeof(struct cn10k_ml_jd)); + memset(&fw->req->jd.fw_load.version[0], '\0', MLDEV_FIRMWARE_VERSION_LENGTH); + + /* Reset device, if in active state */ + if (roc_ml_mlip_is_enabled(&mldev->roc)) + roc_ml_mlip_reset(&mldev->roc, true); + + /* Load firmware */ + fw->data = PLT_PTR_ADD(mz->addr, sizeof(struct cn10k_ml_req)); + ret = cn10k_ml_fw_load_cn10ka(fw, fw_buffer, fw_size); + if (fw_buffer != NULL) + free(fw_buffer); + if (ret < 0) + cn10k_ml_fw_unload(mldev); + + return ret; +} + +void +cn10k_ml_fw_unload(struct cn10k_ml_dev *mldev) +{ + const struct plt_memzone *mz; + uint64_t reg_val; + + /* Disable and reset device */ + reg_val = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val &= ~ROC_ML_CFG_MLIP_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val, ML_CFG); + roc_ml_mlip_reset(&mldev->roc, true); + + mz = plt_memzone_lookup(FW_MEMZONE_NAME); + if (mz != NULL) + plt_memzone_free(mz); +} + static struct rte_pci_id pci_id_ml_table[] = { {RTE_PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_CN10K_ML_PF)}, /* sentinel */ diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 5333566cff..00d23eb3ca 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -10,6 +10,9 @@ /* Marvell OCTEON CN10K ML PMD device name */ #define MLDEV_NAME_CN10K_PMD ml_cn10k +/* Firmware version string length */ +#define MLDEV_FIRMWARE_VERSION_LENGTH 32 + /* Device alignment size */ #define ML_CN10K_ALIGN_SIZE 128 @@ -28,6 +31,19 @@ /* ML command timeout in seconds */ #define ML_CN10K_CMD_TIMEOUT 5 +/* Poll mode job state */ +#define ML_CN10K_POLL_JOB_START 0 +#define ML_CN10K_POLL_JOB_FINISH 1 + +/* Job types */ +enum cn10k_ml_job_type { + ML_CN10K_JOB_TYPE_MODEL_RUN = 0, + ML_CN10K_JOB_TYPE_MODEL_STOP, + ML_CN10K_JOB_TYPE_MODEL_START, + ML_CN10K_JOB_TYPE_FIRMWARE_LOAD, + ML_CN10K_JOB_TYPE_FIRMWARE_SELFTEST, +}; + /* Device configuration state enum */ enum cn10k_ml_dev_state { /* Probed and not configured */ @@ -43,6 +59,136 @@ enum cn10k_ml_dev_state { ML_CN10K_DEV_STATE_CLOSED }; +/* Firmware stats */ +struct cn10k_ml_fw_stats { + /* Firmware start cycle */ + uint64_t fw_start; + + /* Firmware end cycle */ + uint64_t fw_end; + + /* Hardware start cycle */ + uint64_t hw_start; + + /* Hardware end cycle */ + uint64_t hw_end; +}; + +/* Result structure */ +struct cn10k_ml_result { + /* Job error code */ + uint64_t error_code; + + /* Firmware stats */ + struct cn10k_ml_fw_stats stats; + + /* User context pointer */ + void *user_ptr; +}; + +/* Firmware capability structure */ +union cn10k_ml_fw_cap { + uint64_t u64; + + struct { + /* CMPC completion support */ + uint64_t cmpc_completions : 1; + + /* Poll mode completion support */ + uint64_t poll_completions : 1; + + /* SSO completion support */ + uint64_t sso_completions : 1; + + /* Support for model side loading */ + uint64_t side_load_model : 1; + + /* Batch execution */ + uint64_t batch_run : 1; + + /* Max number of models to be loaded in parallel */ + uint64_t max_models : 8; + + /* Firmware statistics */ + uint64_t fw_stats : 1; + + /* Hardware statistics */ + uint64_t hw_stats : 1; + + /* Max number of batches */ + uint64_t max_num_batches : 16; + + uint64_t rsvd : 33; + } s; +}; + +/* Firmware debug info structure */ +struct cn10k_ml_fw_debug { + /* ACC core 0 debug buffer */ + uint64_t core0_debug_ptr; + + /* ACC core 1 debug buffer */ + uint64_t core1_debug_ptr; + + /* ACC core 0 exception state buffer */ + uint64_t core0_exception_buffer; + + /* ACC core 1 exception state buffer */ + uint64_t core1_exception_buffer; + + /* Debug buffer size per core */ + uint32_t debug_buffer_size; + + /* Exception state dump size */ + uint32_t exception_state_size; +}; + +/* Job descriptor header (32 bytes) */ +struct cn10k_ml_jd_header { + /* Job completion structure */ + struct ml_jce_s jce; + + /* Model ID */ + uint64_t model_id : 8; + + /* Job type */ + uint64_t job_type : 8; + + /* Flags for fast-path jobs */ + uint64_t fp_flags : 16; + + /* Flags for slow-path jobs */ + uint64_t sp_flags : 16; + uint64_t rsvd : 16; + + /* Job result pointer */ + uint64_t *result; +}; + +/* Job descriptor structure */ +struct cn10k_ml_jd { + /* Job descriptor header (32 bytes) */ + struct cn10k_ml_jd_header hdr; + + union { + struct cn10k_ml_jd_section_fw_load { + /* Firmware capability structure (8 bytes) */ + union cn10k_ml_fw_cap cap; + + /* Firmware version (32 bytes) */ + uint8_t version[MLDEV_FIRMWARE_VERSION_LENGTH]; + + /* Debug capability structure (40 bytes) */ + struct cn10k_ml_fw_debug debug; + + /* Flags to control error handling */ + uint64_t flags; + + uint8_t rsvd[8]; + } fw_load; + }; +}; + /* ML firmware structure */ struct cn10k_ml_fw { /* Device reference */ @@ -50,6 +196,12 @@ struct cn10k_ml_fw { /* Firmware file path */ const char *path; + + /* Data buffer */ + uint8_t *data; + + /* Firmware load / handshake request structure */ + struct cn10k_ml_req *req; }; /* Device private data */ @@ -64,4 +216,8 @@ struct cn10k_ml_dev { struct cn10k_ml_fw fw; }; +uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); +int cn10k_ml_fw_load(struct cn10k_ml_dev *mldev); +void cn10k_ml_fw_unload(struct cn10k_ml_dev *mldev); + #endif /* _CN10K_ML_DEV_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 3a78d8c816..3df1254dca 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -30,6 +30,7 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c { struct rte_ml_dev_info dev_info; struct cn10k_ml_dev *mldev; + int ret; if (dev == NULL || conf == NULL) return -EINVAL; @@ -51,6 +52,11 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c if (mldev->state == ML_CN10K_DEV_STATE_PROBED) { plt_ml_dbg("Configuring ML device, nb_queue_pairs = %u, nb_models = %u", conf->nb_queue_pairs, conf->nb_models); + + /* Load firmware */ + ret = cn10k_ml_fw_load(mldev); + if (ret != 0) + return ret; } else if (mldev->state == ML_CN10K_DEV_STATE_CONFIGURED) { plt_ml_dbg("Re-configuring ML device, nb_queue_pairs = %u, nb_models = %u", conf->nb_queue_pairs, conf->nb_models); @@ -77,6 +83,21 @@ cn10k_ml_dev_close(struct rte_ml_dev *dev) mldev = dev->data->dev_private; + /* Unload firmware */ + cn10k_ml_fw_unload(mldev); + + /* Clear scratch registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_FW_CTRL); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C1); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C1); + + /* Reset ML_MLR_BASE */ + roc_ml_reg_write64(&mldev->roc, 0, ML_MLR_BASE); + plt_ml_dbg("ML_MLR_BASE = 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_MLR_BASE)); + mldev->state = ML_CN10K_DEV_STATE_CLOSED; /* Remove PCI device */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index b14221d02c..fe18730aca 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -5,6 +5,20 @@ #ifndef _CN10K_ML_OPS_H_ #define _CN10K_ML_OPS_H_ +#include "cn10k_ml_dev.h" + +/* ML request */ +struct cn10k_ml_req { + /* Job descriptor */ + struct cn10k_ml_jd jd; + + /* Job result */ + struct cn10k_ml_result result; + + /* Status field for poll mode requests */ + volatile uint64_t status; +} __rte_aligned(ROC_ALIGN); + /* Device ops */ extern struct rte_ml_dev_ops cn10k_ml_ops; From patchwork Fri Mar 10 08:19:44 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124953 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 2AA6741D3D; Fri, 10 Mar 2023 09:21:06 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 3382542D13; Fri, 10 Mar 2023 09:20:34 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id EED6C41151 for ; Fri, 10 Mar 2023 09:20:26 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1B009730 for ; Fri, 10 Mar 2023 00:20:26 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=UxghtxWNIoa/urab9LWIQBwNFxc60gxPtTINtsrmkoM=; b=WXVrkAWpyvAqLH+7+Sj6k2IcrHVpSVjmcsQwe3Hqskv/LaiaIwvHFh0dmfnnVCj8d1MD MmEF62mUYs0omhisjWL7NI9kCQdnY1OGPiqOh+Da4ir6rmI80OLtZW96PjJf8EzlJIN5 nozXPkdxvT8cNW/VGAXPIJNiMeMSD6y/cUtpaAbALGUehO3d4pjxk1ruqB7FGndUM5Gn oi+rmFlLOE5vzeRg3uFx6c3UP9N7GXDCLnwAtJx3PmmOl6yu/D2nLre/Ww2PJGUMbuDY RsjTiXU9YiBtLaCr7wg+XqC+1WisxoKk9mVOekWnoKXJ/AVQYgPfJPuVeKOsYyTuL169 zA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-7 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:26 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 7D4095C68F3; Fri, 10 Mar 2023 00:20:19 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 08/39] ml/cnxk: enable support for simulator environment Date: Fri, 10 Mar 2023 00:19:44 -0800 Message-ID: <20230310082015.20200-9-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: gJjArUoFv4ebGv74Ber1ItXceiJMnVTv X-Proofpoint-ORIG-GUID: gJjArUoFv4ebGv74Ber1ItXceiJMnVTv X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled device initialization and firmware load on simulator platform. Firmware load stage on simulator would involve launching a firmware handshake request only. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 119 +++++++++++++++++++++++++++++---- 1 file changed, 107 insertions(+), 12 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 90fca45ddd..837f006bf0 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -213,6 +213,89 @@ cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw) return FW_LOAD_FLAGS; } +static int +cn10k_ml_fw_load_asim(struct cn10k_ml_fw *fw) +{ + struct cn10k_ml_dev *mldev; + uint64_t timeout_cycle; + uint64_t reg_val64; + bool timeout; + int ret = 0; + + mldev = fw->mldev; + + /* Reset HEAD and TAIL debug pointer registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_HEAD_C1); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_DBG_BUFFER_TAIL_C1); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_EXCEPTION_SP_C0); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_EXCEPTION_SP_C1); + + /* Set ML_MLR_BASE to base IOVA of the ML region in LLC/DRAM. */ + reg_val64 = rte_eal_get_baseaddr(); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_MLR_BASE); + plt_ml_dbg("ML_MLR_BASE = 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_MLR_BASE)); + roc_ml_reg_save(&mldev->roc, ML_MLR_BASE); + + /* Update FW load completion structure */ + fw->req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&fw->req->status); + fw->req->jd.hdr.job_type = ML_CN10K_JOB_TYPE_FIRMWARE_LOAD; + fw->req->jd.hdr.result = roc_ml_addr_ap2mlip(&mldev->roc, &fw->req->result); + fw->req->jd.fw_load.flags = cn10k_ml_fw_flags_get(fw); + plt_write64(ML_CN10K_POLL_JOB_START, &fw->req->status); + plt_wmb(); + + /* Enqueue FW load through scratch registers */ + timeout = true; + timeout_cycle = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + roc_ml_scratch_enqueue(&mldev->roc, &fw->req->jd); + + plt_rmb(); + do { + if (roc_ml_scratch_is_done_bit_set(&mldev->roc) && + (plt_read64(&fw->req->status) == ML_CN10K_POLL_JOB_FINISH)) { + timeout = false; + break; + } + } while (plt_tsc_cycles() < timeout_cycle); + + /* Check firmware load status, clean-up and exit on failure. */ + if ((!timeout) && (fw->req->result.error_code == 0)) { + cn10k_ml_fw_print_info(fw); + } else { + /* Set ML to disable new jobs */ + reg_val64 = (ROC_ML_CFG_JD_SIZE | ROC_ML_CFG_MLIP_ENA); + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + + /* Clear scratch registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_WORK_PTR); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_FW_CTRL); + + if (timeout) { + plt_err("Firmware load timeout"); + ret = -ETIME; + } else { + plt_err("Firmware load failed"); + ret = -1; + } + + return ret; + } + + /* Reset scratch registers */ + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_FW_CTRL); + roc_ml_reg_write64(&mldev->roc, 0, ML_SCRATCH_WORK_PTR); + + /* Disable job execution, to be enabled in start */ + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + return ret; +} + static int cn10k_ml_fw_load_cn10ka(struct cn10k_ml_fw *fw, void *buffer, uint64_t size) { @@ -447,16 +530,22 @@ cn10k_ml_fw_load(struct cn10k_ml_dev *mldev) fw = &mldev->fw; fw->mldev = mldev; - /* Read firmware image to a buffer */ - ret = rte_firmware_read(fw->path, &fw_buffer, &fw_size); - if (ret < 0) { - plt_err("Can't read firmware data: %s\n", fw->path); - return ret; + if (roc_env_is_emulator() || roc_env_is_hw()) { + /* Read firmware image to a buffer */ + ret = rte_firmware_read(fw->path, &fw_buffer, &fw_size); + if (ret < 0) { + plt_err("Can't read firmware data: %s\n", fw->path); + return ret; + } + + /* Reserve memzone for firmware load completion and data */ + mz_size = sizeof(struct cn10k_ml_req) + fw_size + FW_STACK_BUFFER_SIZE + + FW_DEBUG_BUFFER_SIZE + FW_EXCEPTION_BUFFER_SIZE; + } else if (roc_env_is_asim()) { + /* Reserve memzone for firmware load completion */ + mz_size = sizeof(struct cn10k_ml_req); } - /* Reserve memzone for firmware load completion and data */ - mz_size = sizeof(struct cn10k_ml_req) + fw_size + FW_STACK_BUFFER_SIZE + - FW_DEBUG_BUFFER_SIZE + FW_EXCEPTION_BUFFER_SIZE; mz = plt_memzone_reserve_aligned(FW_MEMZONE_NAME, mz_size, 0, ML_CN10K_ALIGN_SIZE); if (mz == NULL) { plt_err("plt_memzone_reserve failed : %s", FW_MEMZONE_NAME); @@ -475,10 +564,16 @@ cn10k_ml_fw_load(struct cn10k_ml_dev *mldev) roc_ml_mlip_reset(&mldev->roc, true); /* Load firmware */ - fw->data = PLT_PTR_ADD(mz->addr, sizeof(struct cn10k_ml_req)); - ret = cn10k_ml_fw_load_cn10ka(fw, fw_buffer, fw_size); - if (fw_buffer != NULL) - free(fw_buffer); + if (roc_env_is_emulator() || roc_env_is_hw()) { + fw->data = PLT_PTR_ADD(mz->addr, sizeof(struct cn10k_ml_req)); + ret = cn10k_ml_fw_load_cn10ka(fw, fw_buffer, fw_size); + if (fw_buffer != NULL) + free(fw_buffer); + } else if (roc_env_is_asim()) { + fw->data = NULL; + ret = cn10k_ml_fw_load_asim(fw); + } + if (ret < 0) cn10k_ml_fw_unload(mldev); From patchwork Fri Mar 10 08:19:45 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124955 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 1C18B41D3D; Fri, 10 Mar 2023 09:21:19 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 7DD3942D2F; Fri, 10 Mar 2023 09:20:36 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 578E84111C for ; Fri, 10 Mar 2023 09:20:27 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1C009730 for ; Fri, 10 Mar 2023 00:20:26 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=JG6jXO4k0Prwv+SBur/RV9/8mZnzG5/3LyHHY9iJ7vc=; b=aj1em0nrzWwaNQh9Tz0WhlMYWeRMKsVR30dPl+OXT8pcfhXT58mTge9Yus8l6u+gapt6 jLjt3f8AT31DNhRrRYtu8NJN2N9gZ+xk10NQ1xlvstdnsRf2nxkYNMQ15eW8gw7CwVyM R+j24jpSNQ0OHfjRBNcfdV82Ss03odljpmsTK2VJ9n+RAzu/8/EnLMBPSiCT5kIHMb9O ce7o8NAIarrADXJljsqp6no8NC0RfrP+5/QembLVzyblVMjsYtaUcN2V7XpZhpMppws0 hQ4fb3my2EIA04rx1HGP00amZn4XGojD2jZPIjfFf2IOSnYq5dtQT2Erf2RuX2ETXz1u DA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-8 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:26 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id A48733F708E; Fri, 10 Mar 2023 00:20:19 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 09/39] ml/cnxk: enable support for device start and stop Date: Fri, 10 Mar 2023 00:19:45 -0800 Message-ID: <20230310082015.20200-10-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: -_qu115Wl4yhO88xDViUiTBMi0DgSkVF X-Proofpoint-ORIG-GUID: -_qu115Wl4yhO88xDViUiTBMi0DgSkVF X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Implemented ML driver functions to start and stop ML device. Start / Stop would enable or disable ML device to accept inference requests. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 42 +++++++++++++++++++++++++++++++--- 1 file changed, 39 insertions(+), 3 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 3df1254dca..a9f14fe4c5 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -104,9 +104,45 @@ cn10k_ml_dev_close(struct rte_ml_dev *dev) return rte_dev_remove(dev->device); } +static int +cn10k_ml_dev_start(struct rte_ml_dev *dev) +{ + struct cn10k_ml_dev *mldev; + uint64_t reg_val64; + + mldev = dev->data->dev_private; + + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 |= ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + mldev->state = ML_CN10K_DEV_STATE_STARTED; + + return 0; +} + +static int +cn10k_ml_dev_stop(struct rte_ml_dev *dev) +{ + struct cn10k_ml_dev *mldev; + uint64_t reg_val64; + + mldev = dev->data->dev_private; + + reg_val64 = roc_ml_reg_read64(&mldev->roc, ML_CFG); + reg_val64 &= ~ROC_ML_CFG_ENA; + roc_ml_reg_write64(&mldev->roc, reg_val64, ML_CFG); + plt_ml_dbg("ML_CFG => 0x%016lx", roc_ml_reg_read64(&mldev->roc, ML_CFG)); + + mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ - .dev_info_get = cn10k_ml_dev_info_get, - .dev_configure = cn10k_ml_dev_configure, - .dev_close = cn10k_ml_dev_close, + .dev_info_get = cn10k_ml_dev_info_get, .dev_configure = cn10k_ml_dev_configure, + .dev_close = cn10k_ml_dev_close, .dev_start = cn10k_ml_dev_start, + .dev_stop = cn10k_ml_dev_stop, }; From patchwork Fri Mar 10 08:19:46 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124951 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 4299941D3D; Fri, 10 Mar 2023 09:20:53 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 9FAC642C76; Fri, 10 Mar 2023 09:20:31 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 7AC0C4111C for ; Fri, 10 Mar 2023 09:20:26 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1A009730 for ; Fri, 10 Mar 2023 00:20:25 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=RYpDoya02thFYmDn17LxXYEPWQh4uGg+c5B0gjY+ebo=; b=ZsmV+rnK8OU6dnvFBIGfaBG/5OFDm/Rpcq8fkjnG5xhL4xQJZSGz8zWABWTN9d3RLWFU kNwmDYY+kdeUmagolrrTfH/LLdLiakc4A6StGbaMSKqUp9CHoX6+rHjIxzpg7DoU94ew TW9hw0qKOBMUXpCiXbDiisu2OlScs6Vqik8aDHRB7JWnJYkYQsPFyIY9DILFJamrHKHD fCrsMMXNQZ2Mxuw9BMBRPQitkgfvuxmNtoYA3M3/0Oi4qO2QSDfwzLXJGryFK4IQ43jf iiwtgJpI0X0I1SEk8kzN8UxppkfVrqA92TO4wFCCDQA/3BARtqfVnLkn8+puFJIBh9mT 7g== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-6 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:25 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id CE8F25C6904; Fri, 10 Mar 2023 00:20:19 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 10/39] ml/cnxk: add support to create device queue-pairs Date: Fri, 10 Mar 2023 00:19:46 -0800 Message-ID: <20230310082015.20200-11-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: 6gQB8DE8eh46OO9qNYopdC3qVntdMLyj X-Proofpoint-ORIG-GUID: 6gQB8DE8eh46OO9qNYopdC3qVntdMLyj X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support to create and destroy device queue-pairs. Updated configure stage to create array to store queue-pair handles. Added internal structure for queue-pair, queue and ML inference requests. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 207 ++++++++++++++++++++++++++++++++- drivers/ml/cnxk/cn10k_ml_ops.h | 33 +++++- 2 files changed, 237 insertions(+), 3 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index a9f14fe4c5..82670330d1 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -8,6 +8,97 @@ #include "cn10k_ml_dev.h" #include "cn10k_ml_ops.h" +static void +qp_memzone_name_get(char *name, int size, int dev_id, int qp_id) +{ + snprintf(name, size, "cn10k_ml_qp_mem_%u:%u", dev_id, qp_id); +} + +static int +cn10k_ml_qp_destroy(const struct rte_ml_dev *dev, struct cn10k_ml_qp *qp) +{ + const struct rte_memzone *qp_mem; + char name[RTE_MEMZONE_NAMESIZE]; + int ret; + + qp_memzone_name_get(name, RTE_MEMZONE_NAMESIZE, dev->data->dev_id, qp->id); + qp_mem = rte_memzone_lookup(name); + ret = rte_memzone_free(qp_mem); + if (ret) + return ret; + + rte_free(qp); + + return 0; +} + +static int +cn10k_ml_dev_queue_pair_release(struct rte_ml_dev *dev, uint16_t queue_pair_id) +{ + struct cn10k_ml_qp *qp; + int ret; + + qp = dev->data->queue_pairs[queue_pair_id]; + if (qp == NULL) + return -EINVAL; + + ret = cn10k_ml_qp_destroy(dev, qp); + if (ret) { + plt_err("Could not destroy queue pair %u", queue_pair_id); + return ret; + } + + dev->data->queue_pairs[queue_pair_id] = NULL; + + return 0; +} + +static struct cn10k_ml_qp * +cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_desc, int socket_id) +{ + const struct rte_memzone *qp_mem; + char name[RTE_MEMZONE_NAMESIZE]; + struct cn10k_ml_qp *qp; + uint32_t len; + uint8_t *va; + + /* Allocate queue pair */ + qp = rte_zmalloc_socket("cn10k_ml_pmd_queue_pair", sizeof(struct cn10k_ml_qp), ROC_ALIGN, + socket_id); + if (qp == NULL) { + plt_err("Could not allocate queue pair"); + return NULL; + } + + /* For request queue */ + len = nb_desc * sizeof(struct cn10k_ml_req); + qp_memzone_name_get(name, RTE_MEMZONE_NAMESIZE, dev->data->dev_id, qp_id); + qp_mem = rte_memzone_reserve_aligned( + name, len, socket_id, RTE_MEMZONE_SIZE_HINT_ONLY | RTE_MEMZONE_256MB, ROC_ALIGN); + if (qp_mem == NULL) { + plt_err("Could not reserve memzone: %s", name); + goto qp_free; + } + + va = qp_mem->addr; + memset(va, 0, len); + + /* Initialize Request queue */ + qp->id = qp_id; + qp->queue.reqs = (struct cn10k_ml_req *)va; + qp->queue.head = 0; + qp->queue.tail = 0; + qp->queue.wait_cycles = ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + qp->nb_desc = nb_desc; + + return qp; + +qp_free: + rte_free(qp); + + return NULL; +} + static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { @@ -30,6 +121,9 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c { struct rte_ml_dev_info dev_info; struct cn10k_ml_dev *mldev; + struct cn10k_ml_qp *qp; + uint32_t mz_size; + uint16_t qp_id; int ret; if (dev == NULL || conf == NULL) @@ -68,21 +162,83 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c return -ENOTSUP; } + /* Configure queue-pairs */ + if (dev->data->queue_pairs == NULL) { + mz_size = sizeof(dev->data->queue_pairs[0]) * conf->nb_queue_pairs; + dev->data->queue_pairs = + rte_zmalloc("cn10k_mldev_queue_pairs", mz_size, RTE_CACHE_LINE_SIZE); + if (dev->data->queue_pairs == NULL) { + dev->data->nb_queue_pairs = 0; + plt_err("Failed to get memory for queue_pairs, nb_queue_pairs %u", + conf->nb_queue_pairs); + return -ENOMEM; + } + } else { /* Re-configure */ + void **queue_pairs; + + /* Release all queue pairs as ML spec doesn't support queue_pair_destroy. */ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { + qp = dev->data->queue_pairs[qp_id]; + if (qp != NULL) { + ret = cn10k_ml_dev_queue_pair_release(dev, qp_id); + if (ret < 0) + return ret; + } + } + + queue_pairs = dev->data->queue_pairs; + queue_pairs = + rte_realloc(queue_pairs, sizeof(queue_pairs[0]) * conf->nb_queue_pairs, + RTE_CACHE_LINE_SIZE); + if (queue_pairs == NULL) { + dev->data->nb_queue_pairs = 0; + plt_err("Failed to realloc queue_pairs, nb_queue_pairs = %u", + conf->nb_queue_pairs); + ret = -ENOMEM; + goto error; + } + + memset(queue_pairs, 0, sizeof(queue_pairs[0]) * conf->nb_queue_pairs); + dev->data->queue_pairs = queue_pairs; + } + dev->data->nb_queue_pairs = conf->nb_queue_pairs; + mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; return 0; + +error: + if (dev->data->queue_pairs != NULL) + rte_free(dev->data->queue_pairs); + + return ret; } static int cn10k_ml_dev_close(struct rte_ml_dev *dev) { struct cn10k_ml_dev *mldev; + struct cn10k_ml_qp *qp; + uint16_t qp_id; if (dev == NULL) return -EINVAL; mldev = dev->data->dev_private; + /* Destroy all queue pairs */ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { + qp = dev->data->queue_pairs[qp_id]; + if (qp != NULL) { + if (cn10k_ml_qp_destroy(dev, qp) != 0) + plt_err("Could not destroy queue pair %u", qp_id); + dev->data->queue_pairs[qp_id] = NULL; + } + } + + if (dev->data->queue_pairs) + rte_free(dev->data->queue_pairs); + /* Unload firmware */ cn10k_ml_fw_unload(mldev); @@ -140,9 +296,56 @@ cn10k_ml_dev_stop(struct rte_ml_dev *dev) return 0; } +static int +cn10k_ml_dev_queue_pair_setup(struct rte_ml_dev *dev, uint16_t queue_pair_id, + const struct rte_ml_dev_qp_conf *qp_conf, int socket_id) +{ + struct rte_ml_dev_info dev_info; + struct cn10k_ml_qp *qp; + uint32_t nb_desc; + + if (queue_pair_id >= dev->data->nb_queue_pairs) { + plt_err("Queue-pair id = %u (>= max queue pairs supported, %u)\n", queue_pair_id, + dev->data->nb_queue_pairs); + return -EINVAL; + } + + if (dev->data->queue_pairs[queue_pair_id] != NULL) + cn10k_ml_dev_queue_pair_release(dev, queue_pair_id); + + cn10k_ml_dev_info_get(dev, &dev_info); + if ((qp_conf->nb_desc > dev_info.max_desc) || (qp_conf->nb_desc == 0)) { + plt_err("Could not setup queue pair for %u descriptors", qp_conf->nb_desc); + return -EINVAL; + } + plt_ml_dbg("Creating queue-pair, queue_pair_id = %u, nb_desc = %u", queue_pair_id, + qp_conf->nb_desc); + + /* As the number of usable descriptors is 1 less than the queue size being created, we + * increment the size of queue by 1 than the requested size, except when the requested size + * is equal to the maximum possible size. + */ + nb_desc = + (qp_conf->nb_desc == dev_info.max_desc) ? dev_info.max_desc : qp_conf->nb_desc + 1; + qp = cn10k_ml_qp_create(dev, queue_pair_id, nb_desc, socket_id); + if (qp == NULL) { + plt_err("Could not create queue pair %u", queue_pair_id); + return -ENOMEM; + } + dev->data->queue_pairs[queue_pair_id] = qp; + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ - .dev_info_get = cn10k_ml_dev_info_get, .dev_configure = cn10k_ml_dev_configure, - .dev_close = cn10k_ml_dev_close, .dev_start = cn10k_ml_dev_start, + .dev_info_get = cn10k_ml_dev_info_get, + .dev_configure = cn10k_ml_dev_configure, + .dev_close = cn10k_ml_dev_close, + .dev_start = cn10k_ml_dev_start, .dev_stop = cn10k_ml_dev_stop, + + /* Queue-pair handling ops */ + .dev_queue_pair_setup = cn10k_ml_dev_queue_pair_setup, + .dev_queue_pair_release = cn10k_ml_dev_queue_pair_release, }; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index fe18730aca..289c7c5587 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -5,9 +5,13 @@ #ifndef _CN10K_ML_OPS_H_ #define _CN10K_ML_OPS_H_ +#include + +#include + #include "cn10k_ml_dev.h" -/* ML request */ +/* Request structure */ struct cn10k_ml_req { /* Job descriptor */ struct cn10k_ml_jd jd; @@ -19,6 +23,33 @@ struct cn10k_ml_req { volatile uint64_t status; } __rte_aligned(ROC_ALIGN); +/* Request queue */ +struct cn10k_ml_queue { + /* Array of requests */ + struct cn10k_ml_req *reqs; + + /* Head of the queue, used for enqueue */ + uint64_t head; + + /* Tail of the queue, used for dequeue */ + uint64_t tail; + + /* Wait cycles before timeout */ + uint64_t wait_cycles; +}; + +/* Queue-pair structure */ +struct cn10k_ml_qp { + /* ID */ + uint32_t id; + + /* Number of descriptors */ + uint64_t nb_desc; + + /* Request queue */ + struct cn10k_ml_queue queue; +}; + /* Device ops */ extern struct rte_ml_dev_ops cn10k_ml_ops; From patchwork Fri Mar 10 08:19:47 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124960 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id BFE4541D3D; Fri, 10 Mar 2023 09:21:50 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id C554C42D46; Fri, 10 Mar 2023 09:20:42 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 94A3740685 for ; Fri, 10 Mar 2023 09:20:28 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId8021875 for ; Fri, 10 Mar 2023 00:20:28 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=I0o7HE4dEvwmgtfFpIvw6QX8bKVmXCBaCVCOnFP9nZA=; b=FldAoTvJmfxrFTQ14yicbeQNWJa+zQ70EX78OG+oxltWHDpCwDQOtFfSpW8E/l6KScGW BUBy/Mz8uBgc91Slmv8+9wHGKuQJIezwbD8CTyiq8I9cIwAeeUTJgC/crIrajhk+hO32 IQoVJV1PsPX1GyFcmh3Olnkg2rSXzIXNdxcJ9TpSVFUH7a8VV6/PLL2Hjh5wqQXjv76j pyNhW2UQ4Cm6ovUIxg8NMajjAtA9sRXdNSz1E2jKTMoU8uCA4awyd3ot+xXuB5f4nhB6 a1XqnWP4WXTBFAhFLkMUM0RqEPbkxeYisGvGwHgoPOTx54QaA2B63qvuXT72ZxRVWov4 2g== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-7 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:27 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 0BCE15C6906; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 11/39] ml/cnxk: add functions to load and unload models Date: Fri, 10 Mar 2023 00:19:47 -0800 Message-ID: <20230310082015.20200-12-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: lRPtgXr40rSg1hUliyOP8FovqW0lMB1g X-Proofpoint-ORIG-GUID: lRPtgXr40rSg1hUliyOP8FovqW0lMB1g X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added cnxk driver implementations to load and unload ML models. Enabled support in configure stage to allocate model handles array. Assign model ID and allocate resources per each model during load stage and release resources during model unload. Added internal structures to handle ML models. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 3 + drivers/ml/cnxk/cn10k_ml_model.c | 5 + drivers/ml/cnxk/cn10k_ml_model.h | 40 ++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 154 +++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.h | 5 + drivers/ml/cnxk/meson.build | 2 + 6 files changed, 209 insertions(+) create mode 100644 drivers/ml/cnxk/cn10k_ml_model.c create mode 100644 drivers/ml/cnxk/cn10k_ml_model.h diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 00d23eb3ca..7cf6268115 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -214,6 +214,9 @@ struct cn10k_ml_dev { /* Firmware */ struct cn10k_ml_fw fw; + + /* Number of models loaded */ + uint16_t nb_models_loaded; }; uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c new file mode 100644 index 0000000000..39ed707396 --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -0,0 +1,5 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#include "cn10k_ml_model.h" diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h new file mode 100644 index 0000000000..a9f7b169de --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -0,0 +1,40 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _CN10K_ML_MODEL_H_ +#define _CN10K_ML_MODEL_H_ + +#include + +#include + +#include "cn10k_ml_dev.h" + +/* Model state */ +enum cn10k_ml_model_state { + ML_CN10K_MODEL_STATE_LOADED, + ML_CN10K_MODEL_STATE_JOB_ACTIVE, + ML_CN10K_MODEL_STATE_STARTED, + ML_CN10K_MODEL_STATE_UNKNOWN, +}; + +/* Model Object */ +struct cn10k_ml_model { + /* Device reference */ + struct cn10k_ml_dev *mldev; + + /* Name */ + char name[RTE_ML_STR_MAX]; + + /* ID */ + uint16_t model_id; + + /* Spinlock, used to update model state */ + plt_spinlock_t lock; + + /* State */ + enum cn10k_ml_model_state state; +}; + +#endif /* _CN10K_ML_MODEL_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 82670330d1..0955fa0d76 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -6,8 +6,12 @@ #include #include "cn10k_ml_dev.h" +#include "cn10k_ml_model.h" #include "cn10k_ml_ops.h" +/* ML model macros */ +#define CN10K_ML_MODEL_MEMZONE_NAME "ml_cn10k_model_mz" + static void qp_memzone_name_get(char *name, int size, int dev_id, int qp_id) { @@ -120,9 +124,11 @@ static int cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *conf) { struct rte_ml_dev_info dev_info; + struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; struct cn10k_ml_qp *qp; uint32_t mz_size; + uint16_t model_id; uint16_t qp_id; int ret; @@ -203,6 +209,48 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c } dev->data->nb_queue_pairs = conf->nb_queue_pairs; + /* Allocate ML models */ + if (dev->data->models == NULL) { + mz_size = sizeof(dev->data->models[0]) * conf->nb_models; + dev->data->models = rte_zmalloc("cn10k_mldev_models", mz_size, RTE_CACHE_LINE_SIZE); + if (dev->data->models == NULL) { + dev->data->nb_models = 0; + plt_err("Failed to get memory for ml_models, nb_models %u", + conf->nb_models); + ret = -ENOMEM; + goto error; + } + } else { + /* Re-configure */ + void **models; + + /* Unload all models */ + for (model_id = 0; model_id < dev->data->nb_models; model_id++) { + model = dev->data->models[model_id]; + if (model != NULL) { + if (model->state == ML_CN10K_MODEL_STATE_LOADED) { + if (cn10k_ml_model_unload(dev, model_id) != 0) + plt_err("Could not unload model %u", model_id); + } + dev->data->models[model_id] = NULL; + } + } + + models = dev->data->models; + models = rte_realloc(models, sizeof(models[0]) * conf->nb_models, + RTE_CACHE_LINE_SIZE); + if (models == NULL) { + dev->data->nb_models = 0; + plt_err("Failed to realloc ml_models, nb_models = %u", conf->nb_models); + ret = -ENOMEM; + goto error; + } + memset(models, 0, sizeof(models[0]) * conf->nb_models); + dev->data->models = models; + } + dev->data->nb_models = conf->nb_models; + + mldev->nb_models_loaded = 0; mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; return 0; @@ -211,14 +259,19 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c if (dev->data->queue_pairs != NULL) rte_free(dev->data->queue_pairs); + if (dev->data->models != NULL) + rte_free(dev->data->models); + return ret; } static int cn10k_ml_dev_close(struct rte_ml_dev *dev) { + struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; struct cn10k_ml_qp *qp; + uint16_t model_id; uint16_t qp_id; if (dev == NULL) @@ -226,6 +279,21 @@ cn10k_ml_dev_close(struct rte_ml_dev *dev) mldev = dev->data->dev_private; + /* Unload all models */ + for (model_id = 0; model_id < dev->data->nb_models; model_id++) { + model = dev->data->models[model_id]; + if (model != NULL) { + if (model->state == ML_CN10K_MODEL_STATE_LOADED) { + if (cn10k_ml_model_unload(dev, model_id) != 0) + plt_err("Could not unload model %u", model_id); + } + dev->data->models[model_id] = NULL; + } + } + + if (dev->data->models) + rte_free(dev->data->models); + /* Destroy all queue pairs */ for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { qp = dev->data->queue_pairs[qp_id]; @@ -337,6 +405,88 @@ cn10k_ml_dev_queue_pair_setup(struct rte_ml_dev *dev, uint16_t queue_pair_id, return 0; } +int +cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint16_t *model_id) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + + char str[RTE_MEMZONE_NAMESIZE]; + const struct plt_memzone *mz; + uint64_t mz_size; + uint16_t idx; + bool found; + + PLT_SET_USED(params); + + mldev = dev->data->dev_private; + + /* Find model ID */ + found = false; + for (idx = 0; idx < dev->data->nb_models; idx++) { + if (dev->data->models[idx] == NULL) { + found = true; + break; + } + } + + if (!found) { + plt_err("No slots available to load new model"); + return -ENOMEM; + } + + /* Compute memzone size */ + mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE); + + /* Allocate memzone for model object and model data */ + snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", CN10K_ML_MODEL_MEMZONE_NAME, idx); + mz = plt_memzone_reserve_aligned(str, mz_size, 0, ML_CN10K_ALIGN_SIZE); + if (!mz) { + plt_err("plt_memzone_reserve failed : %s", str); + return -ENOMEM; + } + + model = mz->addr; + model->mldev = mldev; + model->model_id = idx; + + plt_spinlock_init(&model->lock); + model->state = ML_CN10K_MODEL_STATE_LOADED; + dev->data->models[idx] = model; + mldev->nb_models_loaded++; + + *model_id = idx; + + return 0; +} + +int +cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id) +{ + char str[RTE_MEMZONE_NAMESIZE]; + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + + mldev = dev->data->dev_private; + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + if (model->state != ML_CN10K_MODEL_STATE_LOADED) { + plt_err("Cannot unload. Model in use."); + return -EBUSY; + } + + dev->data->models[model_id] = NULL; + mldev->nb_models_loaded--; + + snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", CN10K_ML_MODEL_MEMZONE_NAME, model_id); + return plt_memzone_free(plt_memzone_lookup(str)); +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -348,4 +498,8 @@ struct rte_ml_dev_ops cn10k_ml_ops = { /* Queue-pair handling ops */ .dev_queue_pair_setup = cn10k_ml_dev_queue_pair_setup, .dev_queue_pair_release = cn10k_ml_dev_queue_pair_release, + + /* Model ops */ + .model_load = cn10k_ml_model_load, + .model_unload = cn10k_ml_model_unload, }; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 289c7c5587..d7842ecd73 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -53,4 +53,9 @@ struct cn10k_ml_qp { /* Device ops */ extern struct rte_ml_dev_ops cn10k_ml_ops; +/* Slow-path ops */ +int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, + uint16_t *model_id); +int cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id); + #endif /* _CN10K_ML_OPS_H_ */ diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build index 7dc8a29a80..bf7a9c0225 100644 --- a/drivers/ml/cnxk/meson.build +++ b/drivers/ml/cnxk/meson.build @@ -10,11 +10,13 @@ endif driver_sdk_headers = files( 'cn10k_ml_dev.h', 'cn10k_ml_ops.h', + 'cn10k_ml_model.h', ) sources = files( 'cn10k_ml_dev.c', 'cn10k_ml_ops.c', + 'cn10k_ml_model.c', ) deps += ['mldev', 'common_cnxk', 'kvargs'] From patchwork Fri Mar 10 08:19:48 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124958 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 2B50541D3D; Fri, 10 Mar 2023 09:21:37 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 6975D42D3B; Fri, 10 Mar 2023 09:20:40 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 1E31140685 for ; Fri, 10 Mar 2023 09:20:28 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId7021875 for ; Fri, 10 Mar 2023 00:20:27 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=1mFFs36pR14664ZPGuht5pnfoDu9yiKGFojvDjY6gYw=; b=ZhO5QwJ/e5rXqBPajKWUHZOZ5hDTNw764M2Ou+YegDlp1qDE3/q82y64gvoUAdtftANb 1b13ud5AmFcmPm8g66hHNHaM6ppOOb9TSiU4WZ8RzxVF53WgQB0BdWlcpqoN8IyCydGt Q4NR3XxyXcLG66MwkAvedLNRcpyhxdIg3f4svNNaNrmT3OWlFgNKbxnjC1PVC+iHTHFN N4oqgC4XSGDfvE9mg6FdeclMhTpdVBphr1m3A3mMcEQhXPNi8FLMwY9Ymeyy3ND7/k5U VnELuiJ5CPRg29tarq1IpfhRCL20qe2DmmJ0vv1+ixUOrtzm2GbsxLnkZqZ8szveRsUm 0Q== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-6 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:27 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 396CD5C6907; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 12/39] ml/cnxk: enable validity checks for model metadata Date: Fri, 10 Mar 2023 00:19:48 -0800 Message-ID: <20230310082015.20200-13-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: XPm_IG4i1WbQqiqHHm6LD85Pie4sO4A2 X-Proofpoint-ORIG-GUID: XPm_IG4i1WbQqiqHHm6LD85Pie4sO4A2 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added model metadata structure and enabled metadata check during model load. Remap cnxk IO types with RTE IO types. Store and update model metadata in model structure. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_model.c | 211 +++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_model.h | 312 +++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 14 +- drivers/ml/cnxk/meson.build | 2 +- 4 files changed, 537 insertions(+), 2 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c index 39ed707396..dfa814bbe0 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.c +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -2,4 +2,215 @@ * Copyright (c) 2022 Marvell. */ +#include + +#include + +#include "cn10k_ml_dev.h" #include "cn10k_ml_model.h" + +static enum rte_ml_io_type +cn10k_ml_io_type_map(uint8_t type) +{ + switch (type) { + case 1: + return RTE_ML_IO_TYPE_INT8; + case 2: + return RTE_ML_IO_TYPE_UINT8; + case 3: + return RTE_ML_IO_TYPE_INT16; + case 4: + return RTE_ML_IO_TYPE_UINT16; + case 5: + return RTE_ML_IO_TYPE_INT32; + case 6: + return RTE_ML_IO_TYPE_UINT32; + case 7: + return RTE_ML_IO_TYPE_FP16; + case 8: + return RTE_ML_IO_TYPE_FP32; + } + + return RTE_ML_IO_TYPE_UNKNOWN; +} + +int +cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size) +{ + struct cn10k_ml_model_metadata *metadata; + uint32_t payload_crc32c; + uint32_t header_crc32c; + uint8_t version[4]; + uint8_t i; + + metadata = (struct cn10k_ml_model_metadata *)buffer; + + /* Header CRC check */ + if (metadata->metadata_header.header_crc32c != 0) { + header_crc32c = rte_hash_crc( + buffer, sizeof(metadata->metadata_header) - sizeof(uint32_t), 0); + + if (header_crc32c != metadata->metadata_header.header_crc32c) { + plt_err("Invalid model, Header CRC mismatch"); + return -EINVAL; + } + } + + /* Payload CRC check */ + if (metadata->metadata_header.payload_crc32c != 0) { + payload_crc32c = rte_hash_crc(buffer + sizeof(metadata->metadata_header), + size - sizeof(metadata->metadata_header), 0); + + if (payload_crc32c != metadata->metadata_header.payload_crc32c) { + plt_err("Invalid model, Payload CRC mismatch"); + return -EINVAL; + } + } + + /* Model magic string */ + if (strncmp((char *)metadata->metadata_header.magic, MRVL_ML_MODEL_MAGIC_STRING, 4) != 0) { + plt_err("Invalid model, magic = %s", metadata->metadata_header.magic); + return -EINVAL; + } + + /* Target architecture */ + if (metadata->metadata_header.target_architecture != MRVL_ML_MODEL_TARGET_ARCH) { + plt_err("Model target architecture (%u) not supported", + metadata->metadata_header.target_architecture); + return -ENOTSUP; + } + + /* Header version */ + rte_memcpy(version, metadata->metadata_header.version, 4 * sizeof(uint8_t)); + if (version[0] * 1000 + version[1] * 100 < MRVL_ML_MODEL_VERSION) { + plt_err("Metadata version = %u.%u.%u.%u (< %u.%u.%u.%u) not supported", version[0], + version[1], version[2], version[3], (MRVL_ML_MODEL_VERSION / 1000) % 10, + (MRVL_ML_MODEL_VERSION / 100) % 10, (MRVL_ML_MODEL_VERSION / 10) % 10, + MRVL_ML_MODEL_VERSION % 10); + return -ENOTSUP; + } + + /* Init section */ + if (metadata->init_model.file_size == 0) { + plt_err("Invalid metadata, init_model.file_size = %u", + metadata->init_model.file_size); + return -EINVAL; + } + + /* Main section */ + if (metadata->main_model.file_size == 0) { + plt_err("Invalid metadata, main_model.file_size = %u", + metadata->main_model.file_size); + return -EINVAL; + } + + /* Finish section */ + if (metadata->finish_model.file_size == 0) { + plt_err("Invalid metadata, finish_model.file_size = %u", + metadata->finish_model.file_size); + return -EINVAL; + } + + /* Weights and Bias */ + if (metadata->weights_bias.file_size == 0) { + plt_err("Invalid metadata, weights_bias.file_size = %u", + metadata->weights_bias.file_size); + return -EINVAL; + } + + if (metadata->weights_bias.relocatable != 1) { + plt_err("Model not supported, non-relocatable weights and bias"); + return -ENOTSUP; + } + + /* Check input count */ + if (metadata->model.num_input > MRVL_ML_INPUT_OUTPUT_SIZE) { + plt_err("Invalid metadata, num_input = %u (> %u)", metadata->model.num_input, + MRVL_ML_INPUT_OUTPUT_SIZE); + return -EINVAL; + } + + /* Check output count */ + if (metadata->model.num_output > MRVL_ML_INPUT_OUTPUT_SIZE) { + plt_err("Invalid metadata, num_output = %u (> %u)", metadata->model.num_output, + MRVL_ML_INPUT_OUTPUT_SIZE); + return -EINVAL; + } + + /* Inputs */ + for (i = 0; i < metadata->model.num_input; i++) { + if (rte_ml_io_type_size_get(cn10k_ml_io_type_map(metadata->input[i].input_type)) <= + 0) { + plt_err("Invalid metadata, input[%u] : input_type = %u", i, + metadata->input[i].input_type); + return -EINVAL; + } + + if (rte_ml_io_type_size_get( + cn10k_ml_io_type_map(metadata->input[i].model_input_type)) <= 0) { + plt_err("Invalid metadata, input[%u] : model_input_type = %u", i, + metadata->input[i].model_input_type); + return -EINVAL; + } + + if (metadata->input[i].relocatable != 1) { + plt_err("Model not supported, non-relocatable input: %u", i); + return -ENOTSUP; + } + } + + /* Outputs */ + for (i = 0; i < metadata->model.num_output; i++) { + if (rte_ml_io_type_size_get( + cn10k_ml_io_type_map(metadata->output[i].output_type)) <= 0) { + plt_err("Invalid metadata, output[%u] : output_type = %u", i, + metadata->output[i].output_type); + return -EINVAL; + } + + if (rte_ml_io_type_size_get( + cn10k_ml_io_type_map(metadata->output[i].model_output_type)) <= 0) { + plt_err("Invalid metadata, output[%u] : model_output_type = %u", i, + metadata->output[i].model_output_type); + return -EINVAL; + } + + if (metadata->output[i].relocatable != 1) { + plt_err("Model not supported, non-relocatable output: %u", i); + return -ENOTSUP; + } + } + + return 0; +} + +void +cn10k_ml_model_metadata_update(struct cn10k_ml_model_metadata *metadata) +{ + uint8_t i; + + for (i = 0; i < metadata->model.num_input; i++) { + metadata->input[i].input_type = cn10k_ml_io_type_map(metadata->input[i].input_type); + metadata->input[i].model_input_type = + cn10k_ml_io_type_map(metadata->input[i].model_input_type); + + if (metadata->input[i].shape.w == 0) + metadata->input[i].shape.w = 1; + + if (metadata->input[i].shape.x == 0) + metadata->input[i].shape.x = 1; + + if (metadata->input[i].shape.y == 0) + metadata->input[i].shape.y = 1; + + if (metadata->input[i].shape.z == 0) + metadata->input[i].shape.z = 1; + } + + for (i = 0; i < metadata->model.num_output; i++) { + metadata->output[i].output_type = + cn10k_ml_io_type_map(metadata->output[i].output_type); + metadata->output[i].model_output_type = + cn10k_ml_io_type_map(metadata->output[i].model_output_type); + } +} diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index a9f7b169de..dc30bc2aa7 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -19,6 +19,309 @@ enum cn10k_ml_model_state { ML_CN10K_MODEL_STATE_UNKNOWN, }; +/* Model Metadata : v 2.1.0.2 */ +#define MRVL_ML_MODEL_MAGIC_STRING "MRVL" +#define MRVL_ML_MODEL_TARGET_ARCH 128 +#define MRVL_ML_MODEL_VERSION 2100 +#define MRVL_ML_MODEL_NAME_LEN 64 +#define MRVL_ML_INPUT_NAME_LEN 16 +#define MRVL_ML_OUTPUT_NAME_LEN 16 +#define MRVL_ML_INPUT_OUTPUT_SIZE 8 + +/* Model file metadata structure */ +struct cn10k_ml_model_metadata { + /* Header (256-byte) */ + struct { + /* Magic string ('M', 'R', 'V', 'L') */ + uint8_t magic[4]; + + /* Metadata version */ + uint8_t version[4]; + + /* Metadata size */ + uint32_t metadata_size; + + /* Unique ID */ + uint8_t uuid[128]; + + /* Model target architecture + * 0 = Undefined + * 1 = M1K + * 128 = MLIP + * 256 = Experimental + */ + uint32_t target_architecture; + uint8_t reserved[104]; + + /* CRC of data after metadata_header (i.e. after first 256 bytes) */ + uint32_t payload_crc32c; + + /* CRC of first 252 bytes of metadata_header, after payload_crc calculation */ + uint32_t header_crc32c; + } metadata_header; + + /* Model information (256-byte) */ + struct { + /* Model name string */ + uint8_t name[MRVL_ML_MODEL_NAME_LEN]; + + /* Model version info (xx.xx.xx.xx) */ + uint8_t version[4]; + + /* Model code size (Init + Main + Finish) */ + uint32_t code_size; + + /* Model data size (Weights and Bias) */ + uint32_t data_size; + + /* OCM start offset, set to ocm_wb_range_start */ + uint32_t ocm_start; + + /* OCM start offset, set to max OCM size */ + uint32_t ocm_end; + + /* Relocatable flag (always yes) + * 0 = Not relocatable + * 1 = Relocatable + */ + uint8_t ocm_relocatable; + + /* Tile relocatable flag (always yes) + * 0 = Not relocatable + * 1 = Relocatable + */ + uint8_t tile_relocatable; + + /* Start tile (Always 0) */ + uint8_t tile_start; + + /* End tile (num_tiles - 1) */ + uint8_t tile_end; + + /* Inference batch size */ + uint8_t batch_size; + + /* Number of input tensors (Max 8) */ + uint8_t num_input; + + /* Number of output tensors (Max 8) */ + uint8_t num_output; + uint8_t reserved1; + + /* Total input size in bytes */ + uint32_t input_size; + + /* Total output size in bytes */ + uint32_t output_size; + + /* Table size in bytes */ + uint32_t table_size; + + /* Number of layers in the network */ + uint32_t num_layers; + uint32_t reserved2; + + /* Floor of absolute OCM region */ + uint64_t ocm_tmp_range_floor; + + /* Relative OCM start address of WB data block */ + uint64_t ocm_wb_range_start; + + /* Relative OCM end address of WB data block */ + uint64_t ocm_wb_range_end; + + /* Relative DDR start address of WB data block */ + uint64_t ddr_wb_range_start; + + /* Relative DDR end address of all outputs */ + uint64_t ddr_wb_range_end; + + /* Relative DDR start address of all inputs */ + uint64_t ddr_input_range_start; + + /* Relative DDR end address of all inputs */ + uint64_t ddr_input_range_end; + + /* Relative DDR start address of all outputs */ + uint64_t ddr_output_range_start; + + /* Relative DDR end address of all outputs */ + uint64_t ddr_output_range_end; + + /* Compiler version */ + uint8_t compiler_version[8]; + + /* CDK version */ + uint8_t cdk_version[4]; + + /* Lower batch optimization support + * 0 - No, + * 1 - Yes + */ + uint8_t supports_lower_batch_size_optimization; + uint8_t reserved3[59]; + } model; + + /* Init section (64-byte) */ + struct { + uint32_t file_offset; + uint32_t file_size; + uint8_t reserved[56]; + } init_model; + + /* Main section (64-byte) */ + struct { + uint32_t file_offset; + uint32_t file_size; + uint8_t reserved[56]; + } main_model; + + /* Finish section (64-byte) */ + struct { + uint32_t file_offset; + uint32_t file_size; + uint8_t reserved[56]; + } finish_model; + + uint8_t reserved1[512]; /* End of 2k bytes */ + + /* Weights and Bias (64-byte) */ + struct { + /* Memory offset, set to ddr_wb_range_start */ + uint64_t mem_offset; + uint32_t file_offset; + uint32_t file_size; + + /* Relocatable flag for WB + * 1 = Relocatable + * 2 = Not relocatable + */ + uint8_t relocatable; + uint8_t reserved[47]; + } weights_bias; + + /* Input (512-byte, 64-byte per input) provisioned for 8 inputs */ + struct { + /* DDR offset (in OCM absolute addresses for input) */ + uint64_t mem_offset; + + /* Relocatable flag + * 1 = Relocatable + * 2 = Not relocatable + */ + uint8_t relocatable; + + /* Input quantization + * 1 = Requires quantization + * 2 = Pre-quantized + */ + uint8_t quantize; + + /* Type of incoming input + * 1 = INT8, 2 = UINT8, 3 = INT16, 4 = UINT16, + * 5 = INT32, 6 = UINT32, 7 = FP16, 8 = FP32 + */ + uint8_t input_type; + + /* Type of input required by model + * 1 = INT8, 2 = UINT8, 3 = INT16, 4 = UINT16, + * 5 = INT32, 6 = UINT32, 7 = FP16, 8 = FP32 + */ + uint8_t model_input_type; + + /* float_32 qscale value + * quantized = non-quantized * qscale + */ + float qscale; + + /* Input shape */ + struct { + /* Input format + * 1 = NCHW + * 2 = NHWC + */ + uint8_t format; + uint8_t reserved[3]; + uint32_t w; + uint32_t x; + uint32_t y; + uint32_t z; + } shape; + uint8_t reserved[4]; + + /* Name of input */ + uint8_t input_name[MRVL_ML_INPUT_NAME_LEN]; + + /* DDR range end + * new = mem_offset + size_bytes - 1 + */ + uint64_t ddr_range_end; + } input[MRVL_ML_INPUT_OUTPUT_SIZE]; + + /* Output (512 byte, 64-byte per input) provisioned for 8 outputs */ + struct { + /* DDR offset in OCM absolute addresses for output */ + uint64_t mem_offset; + + /* Relocatable flag + * 1 = Relocatable + * 2 = Not relocatable + */ + uint8_t relocatable; + + /* Output dequantization + * 1 = De-quantization required + * 2 = De-quantization not required + */ + uint8_t dequantize; + + /* Type of outgoing output + * 1 = INT8, 2 = UINT8, 3 = INT16, 4 = UINT16 + * 5 = INT32, 6 = UINT32, 7 = FP16, 8 = FP32 + */ + uint8_t output_type; + + /* Type of output produced by model + * 1 = INT8, 2 = UINT8, 3 = INT16, 4 = UINT16 + * 5 = INT32, 6 = UINT32, 7 = FP16, 8 = FP32 + */ + uint8_t model_output_type; + + /* float_32 dscale value + * dequantized = quantized * dscale + */ + float dscale; + + /* Number of items in the output */ + uint32_t size; + uint8_t reserved[20]; + + /* DDR range end + * new = mem_offset + size_bytes - 1 + */ + uint64_t ddr_range_end; + uint8_t output_name[MRVL_ML_OUTPUT_NAME_LEN]; + } output[MRVL_ML_INPUT_OUTPUT_SIZE]; + + uint8_t reserved2[1792]; + + /* Model data */ + struct { + uint8_t reserved1[4068]; + + /* Beta: xx.xx.xx.xx, + * Later: YYYYMM.xx.xx + */ + uint8_t compiler_version[8]; + + /* M1K CDK version (xx.xx.xx.xx) */ + uint8_t m1k_cdk_version[4]; + } data; + + /* Hidden 16 bytes of magic code */ + uint8_t reserved3[16]; +}; + /* Model Object */ struct cn10k_ml_model { /* Device reference */ @@ -30,6 +333,12 @@ struct cn10k_ml_model { /* ID */ uint16_t model_id; + /* Batch size */ + uint32_t batch_size; + + /* Metadata */ + struct cn10k_ml_model_metadata metadata; + /* Spinlock, used to update model state */ plt_spinlock_t lock; @@ -37,4 +346,7 @@ struct cn10k_ml_model { enum cn10k_ml_model_state state; }; +int cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size); +void cn10k_ml_model_metadata_update(struct cn10k_ml_model_metadata *metadata); + #endif /* _CN10K_ML_MODEL_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 0955fa0d76..2cde795903 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -416,8 +416,11 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint64_t mz_size; uint16_t idx; bool found; + int ret; - PLT_SET_USED(params); + ret = cn10k_ml_model_metadata_check(params->addr, params->size); + if (ret != 0) + return ret; mldev = dev->data->dev_private; @@ -450,6 +453,15 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, model->mldev = mldev; model->model_id = idx; + rte_memcpy(&model->metadata, params->addr, sizeof(struct cn10k_ml_model_metadata)); + cn10k_ml_model_metadata_update(&model->metadata); + + /* Enable support for batch_size of 256 */ + if (model->metadata.model.batch_size == 0) + model->batch_size = 256; + else + model->batch_size = model->metadata.model.batch_size; + plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; dev->data->models[idx] = model; diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build index bf7a9c0225..799e8f2470 100644 --- a/drivers/ml/cnxk/meson.build +++ b/drivers/ml/cnxk/meson.build @@ -19,7 +19,7 @@ sources = files( 'cn10k_ml_model.c', ) -deps += ['mldev', 'common_cnxk', 'kvargs'] +deps += ['mldev', 'common_cnxk', 'kvargs', 'hash'] if get_option('buildtype').contains('debug') cflags += [ '-DCNXK_ML_DEV_DEBUG' ] From patchwork Fri Mar 10 08:19:49 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124957 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 4B74241D3D; Fri, 10 Mar 2023 09:21:31 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id F306C42D39; Fri, 10 Mar 2023 09:20:38 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id CD4344161A for ; Fri, 10 Mar 2023 09:20:27 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1D009730 for ; Fri, 10 Mar 2023 00:20:27 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=bEzzf8hUf98dRIH02qWrNrwfHIlzSsptJ2JA9OpgoUs=; b=SKK4FRRC/kL++c2cBgHLK+vEOwpp6/12kk2bNylBrnNTrLqoEA/pkBkc3Jdx/3WtF5u2 LAlA1GTBcaLsP3j5Q6HgyQo94FZMWd9gYVtY7CU9/QQQ0ChVYxK/43fEDineY2BTeYJx d8Unf0JzKDVg0lWqKWoWE/UVXcLUAISUZKv+KQ4QeuCSXfflE2iMvBRGfgnc4HBXyKHl qpqcSiaRrXJcY0aon6O7S+JVTCxW1pLow5vGjVdFwH4RWO6JxZXrXsLRzJlpIxqKBxlB NpLHO2nC2TMhU9v/wZchcYxPe2N08LVPpGeDx/hlIrIHmGncM0FLdVKVb4bVBGIPk3gl LA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-9 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:26 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 65D563F7095; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 13/39] ml/cnxk: add internal structures for derived info Date: Fri, 10 Mar 2023 00:19:49 -0800 Message-ID: <20230310082015.20200-14-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: n7VhP_mfS40TSVTuDBBb7nSQTjeht83U X-Proofpoint-ORIG-GUID: n7VhP_mfS40TSVTuDBBb7nSQTjeht83U X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added internal structures to handle derived address fields and enabled support to compute DMA addresses for model start. Enabled updating internal model fields. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_model.c | 89 ++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_model.h | 80 ++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 18 ++++++- 3 files changed, 186 insertions(+), 1 deletion(-) diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c index dfa814bbe0..2530beb80e 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.c +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -214,3 +214,92 @@ cn10k_ml_model_metadata_update(struct cn10k_ml_model_metadata *metadata) cn10k_ml_io_type_map(metadata->output[i].model_output_type); } } + +void +cn10k_ml_model_addr_update(struct cn10k_ml_model *model, uint8_t *buffer, uint8_t *base_dma_addr) +{ + struct cn10k_ml_model_metadata *metadata; + struct cn10k_ml_model_addr *addr; + size_t model_data_size; + uint8_t *dma_addr_load; + uint8_t *dma_addr_run; + uint8_t i; + int fpos; + + metadata = &model->metadata; + addr = &model->addr; + model_data_size = metadata->init_model.file_size + metadata->main_model.file_size + + metadata->finish_model.file_size + metadata->weights_bias.file_size; + + /* Base address */ + addr->base_dma_addr_load = base_dma_addr; + addr->base_dma_addr_run = PLT_PTR_ADD(addr->base_dma_addr_load, model_data_size); + + /* Init section */ + dma_addr_load = addr->base_dma_addr_load; + dma_addr_run = addr->base_dma_addr_run; + fpos = sizeof(struct cn10k_ml_model_metadata); + addr->init_load_addr = dma_addr_load; + addr->init_run_addr = dma_addr_run; + rte_memcpy(dma_addr_load, PLT_PTR_ADD(buffer, fpos), metadata->init_model.file_size); + + /* Main section */ + dma_addr_load += metadata->init_model.file_size; + dma_addr_run += metadata->init_model.file_size; + fpos += metadata->init_model.file_size; + addr->main_load_addr = dma_addr_load; + addr->main_run_addr = dma_addr_run; + rte_memcpy(dma_addr_load, PLT_PTR_ADD(buffer, fpos), metadata->main_model.file_size); + + /* Finish section */ + dma_addr_load += metadata->main_model.file_size; + dma_addr_run += metadata->main_model.file_size; + fpos += metadata->main_model.file_size; + addr->finish_load_addr = dma_addr_load; + addr->finish_run_addr = dma_addr_run; + rte_memcpy(dma_addr_load, PLT_PTR_ADD(buffer, fpos), metadata->finish_model.file_size); + + /* Weights and Bias section */ + dma_addr_load += metadata->finish_model.file_size; + fpos += metadata->finish_model.file_size; + addr->wb_base_addr = PLT_PTR_SUB(dma_addr_load, metadata->weights_bias.mem_offset); + addr->wb_load_addr = PLT_PTR_ADD(addr->wb_base_addr, metadata->weights_bias.mem_offset); + rte_memcpy(addr->wb_load_addr, PLT_PTR_ADD(buffer, fpos), metadata->weights_bias.file_size); + + /* Inputs */ + addr->total_input_sz_d = 0; + addr->total_input_sz_q = 0; + for (i = 0; i < metadata->model.num_input; i++) { + addr->input[i].nb_elements = + model->metadata.input[i].shape.w * model->metadata.input[i].shape.x * + model->metadata.input[i].shape.y * model->metadata.input[i].shape.z; + addr->input[i].sz_d = addr->input[i].nb_elements * + rte_ml_io_type_size_get(metadata->input[i].input_type); + addr->input[i].sz_q = addr->input[i].nb_elements * + rte_ml_io_type_size_get(metadata->input[i].model_input_type); + addr->total_input_sz_d += addr->input[i].sz_d; + addr->total_input_sz_q += addr->input[i].sz_q; + + plt_ml_dbg("model_id = %u, input[%u] - w:%u x:%u y:%u z:%u, sz_d = %u sz_q = %u", + model->model_id, i, metadata->input[i].shape.w, + metadata->input[i].shape.x, metadata->input[i].shape.y, + metadata->input[i].shape.z, addr->input[i].sz_d, addr->input[i].sz_q); + } + + /* Outputs */ + addr->total_output_sz_q = 0; + addr->total_output_sz_d = 0; + for (i = 0; i < metadata->model.num_output; i++) { + addr->output[i].nb_elements = metadata->output[i].size; + addr->output[i].sz_d = addr->output[i].nb_elements * + rte_ml_io_type_size_get(metadata->output[i].output_type); + addr->output[i].sz_q = + addr->output[i].nb_elements * + rte_ml_io_type_size_get(metadata->output[i].model_output_type); + addr->total_output_sz_q += addr->output[i].sz_q; + addr->total_output_sz_d += addr->output[i].sz_d; + + plt_ml_dbg("model_id = %u, output[%u] - sz_d = %u, sz_q = %u", model->model_id, i, + addr->output[i].sz_d, addr->output[i].sz_q); + } +} diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index dc30bc2aa7..5345160a74 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -322,6 +322,81 @@ struct cn10k_ml_model_metadata { uint8_t reserved3[16]; }; +/* Model address structure */ +struct cn10k_ml_model_addr { + /* Base DMA address for load */ + void *base_dma_addr_load; + + /* Base DMA address for run */ + void *base_dma_addr_run; + + /* Init section load address */ + void *init_load_addr; + + /* Init section run address */ + void *init_run_addr; + + /* Main section load address */ + void *main_load_addr; + + /* Main section run address */ + void *main_run_addr; + + /* Finish section load address */ + void *finish_load_addr; + + /* Finish section run address */ + void *finish_run_addr; + + /* Weights and Bias base address */ + void *wb_base_addr; + + /* Weights and bias load address */ + void *wb_load_addr; + + /* Start tile */ + uint8_t tile_start; + + /* End tile */ + uint8_t tile_end; + + /* Input address and size */ + struct { + /* Number of elements */ + uint32_t nb_elements; + + /* Dequantized input size */ + uint32_t sz_d; + + /* Quantized input size */ + uint32_t sz_q; + } input[MRVL_ML_INPUT_OUTPUT_SIZE]; + + /* Output address and size */ + struct { + /* Number of elements */ + uint32_t nb_elements; + + /* Dequantize output size */ + uint32_t sz_d; + + /* Quantized output size */ + uint32_t sz_q; + } output[MRVL_ML_INPUT_OUTPUT_SIZE]; + + /* Total size of quantized input */ + uint32_t total_input_sz_q; + + /* Total size of dequantized input */ + uint32_t total_input_sz_d; + + /* Total size of quantized output */ + uint32_t total_output_sz_q; + + /* Total size of dequantized output */ + uint32_t total_output_sz_d; +}; + /* Model Object */ struct cn10k_ml_model { /* Device reference */ @@ -339,6 +414,9 @@ struct cn10k_ml_model { /* Metadata */ struct cn10k_ml_model_metadata metadata; + /* Address structure */ + struct cn10k_ml_model_addr addr; + /* Spinlock, used to update model state */ plt_spinlock_t lock; @@ -348,5 +426,7 @@ struct cn10k_ml_model { int cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size); void cn10k_ml_model_metadata_update(struct cn10k_ml_model_metadata *metadata); +void cn10k_ml_model_addr_update(struct cn10k_ml_model *model, uint8_t *buffer, + uint8_t *base_dma_addr); #endif /* _CN10K_ML_MODEL_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 2cde795903..b11228f2cb 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -408,11 +408,14 @@ cn10k_ml_dev_queue_pair_setup(struct rte_ml_dev *dev, uint16_t queue_pair_id, int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint16_t *model_id) { + struct cn10k_ml_model_metadata *metadata; struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; char str[RTE_MEMZONE_NAMESIZE]; const struct plt_memzone *mz; + size_t model_data_size; + uint8_t *base_dma_addr; uint64_t mz_size; uint16_t idx; bool found; @@ -439,7 +442,12 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, } /* Compute memzone size */ - mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE); + metadata = (struct cn10k_ml_model_metadata *)params->addr; + model_data_size = metadata->init_model.file_size + metadata->main_model.file_size + + metadata->finish_model.file_size + metadata->weights_bias.file_size; + model_data_size = PLT_ALIGN_CEIL(model_data_size, ML_CN10K_ALIGN_SIZE); + mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + + 2 * model_data_size; /* Allocate memzone for model object and model data */ snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", CN10K_ML_MODEL_MEMZONE_NAME, idx); @@ -462,6 +470,14 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, else model->batch_size = model->metadata.model.batch_size; + /* Set DMA base address */ + base_dma_addr = PLT_PTR_ADD( + mz->addr, PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE)); + cn10k_ml_model_addr_update(model, params->addr, base_dma_addr); + + /* Copy data from load to run. run address to be used by MLIP */ + rte_memcpy(model->addr.base_dma_addr_run, model->addr.base_dma_addr_load, model_data_size); + plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; dev->data->models[idx] = model; From patchwork Fri Mar 10 08:19:50 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124959 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 45CBD41D3D; Fri, 10 Mar 2023 09:21:44 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 9C94742D3F; Fri, 10 Mar 2023 09:20:41 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 527DC427F5 for ; Fri, 10 Mar 2023 09:20:28 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1E009730 for ; Fri, 10 Mar 2023 00:20:27 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=ZADOEej5RxfDo5a2Hi6AYBTKk6QrXsyhFkGo1nksyAg=; b=K10XnZIyYfIX84S8zTmnIQjUof5is9sZURACR9FcZ7ux7qCvhoCrRVyuRA/4hg9Znh4Y hq3XdyI3pJQ5P2kzoyldyT2OUC76Q+6nacPprxro+TyqfKQVn5ixhcyQUJ7t7MR0mrnR L4lGopNBjfiNF73g+yiL4GUpHMdbWpTTtMqXS18QFRI6n8BzvUiXtsJhOzNnYXov+p73 PZNoHa9KkaDwrfwd7S5eM2w9N6QJym7fUfvh/xwuJs/0yAUj71Jt49TtLq4wBNIST94n zXaMahhFSpaP8Q6vfjw3XdEE7IBHIK1jKlxdxN5A7nQeDkIsBExnlCBnphfR+Flh6Eas zw== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-10 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:27 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 938F83F7097; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 14/39] ml/cnxk: add internal structures for tiles and OCM Date: Fri, 10 Mar 2023 00:19:50 -0800 Message-ID: <20230310082015.20200-15-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: PgNgP15pAKnlqIDBxv5-uQQBJgkCa3bS X-Proofpoint-ORIG-GUID: PgNgP15pAKnlqIDBxv5-uQQBJgkCa3bS X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added internal structures to handle tile and OCM information and OCM to model memory mapping. Initialize the fields to platform specific defaults and compute the OCM / tile requirements for model. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 5 ++ drivers/ml/cnxk/cn10k_ml_model.c | 53 +++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_model.h | 6 +++ drivers/ml/cnxk/cn10k_ml_ocm.c | 5 ++ drivers/ml/cnxk/cn10k_ml_ocm.h | 79 ++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 31 ++++++++++++- drivers/ml/cnxk/meson.build | 2 + 7 files changed, 180 insertions(+), 1 deletion(-) create mode 100644 drivers/ml/cnxk/cn10k_ml_ocm.c create mode 100644 drivers/ml/cnxk/cn10k_ml_ocm.h diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 7cf6268115..02a4496c97 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -7,6 +7,8 @@ #include +#include "cn10k_ml_ocm.h" + /* Marvell OCTEON CN10K ML PMD device name */ #define MLDEV_NAME_CN10K_PMD ml_cn10k @@ -215,6 +217,9 @@ struct cn10k_ml_dev { /* Firmware */ struct cn10k_ml_fw fw; + /* OCM info */ + struct cn10k_ml_ocm ocm; + /* Number of models loaded */ uint16_t nb_models_loaded; }; diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c index 2530beb80e..69d6306104 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.c +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -8,6 +8,7 @@ #include "cn10k_ml_dev.h" #include "cn10k_ml_model.h" +#include "cn10k_ml_ocm.h" static enum rte_ml_io_type cn10k_ml_io_type_map(uint8_t type) @@ -303,3 +304,55 @@ cn10k_ml_model_addr_update(struct cn10k_ml_model *model, uint8_t *buffer, uint8_ addr->output[i].sz_d, addr->output[i].sz_q); } } + +int +cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, uint8_t *buffer, + uint16_t *wb_pages, uint16_t *scratch_pages) +{ + struct cn10k_ml_model_metadata *metadata; + struct cn10k_ml_ocm *ocm; + uint64_t scratch_size; + uint64_t wb_size; + + metadata = (struct cn10k_ml_model_metadata *)buffer; + ocm = &mldev->ocm; + + /* Assume wb_size is zero for non-relocatable models */ + if (metadata->model.ocm_relocatable) + wb_size = metadata->model.ocm_wb_range_end - metadata->model.ocm_wb_range_start + 1; + else + wb_size = 0; + + if (wb_size % ocm->page_size) + *wb_pages = wb_size / ocm->page_size + 1; + else + *wb_pages = wb_size / ocm->page_size; + plt_ml_dbg("model_id = %u, wb_size = %" PRIu64 ", wb_pages = %u", model_id, wb_size, + *wb_pages); + + scratch_size = ocm->size_per_tile - metadata->model.ocm_tmp_range_floor; + if (metadata->model.ocm_tmp_range_floor % ocm->page_size) + *scratch_pages = scratch_size / ocm->page_size + 1; + else + *scratch_pages = scratch_size / ocm->page_size; + plt_ml_dbg("model_id = %u, scratch_size = %" PRIu64 ", scratch_pages = %u", model_id, + scratch_size, *scratch_pages); + + /* Check if the model can be loaded on OCM */ + if ((*wb_pages + *scratch_pages) > ML_CN10K_OCM_NUMPAGES) { + plt_err("Cannot create the model, OCM relocatable = %u", + metadata->model.ocm_relocatable); + plt_err("wb_pages (%u) + scratch_pages (%u) > %u", *wb_pages, *scratch_pages, + ML_CN10K_OCM_NUMPAGES); + return -ENOMEM; + } + + /* Update scratch_pages to block the full tile for OCM non-relocatable model. This would + * prevent the library from allocating the remaining space on the tile to other models. + */ + if (!metadata->model.ocm_relocatable) + *scratch_pages = + PLT_MAX(PLT_U64_CAST(*scratch_pages), PLT_U64_CAST(ML_CN10K_OCM_NUMPAGES)); + + return 0; +} diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index 5345160a74..7893635787 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -10,6 +10,7 @@ #include #include "cn10k_ml_dev.h" +#include "cn10k_ml_ocm.h" /* Model state */ enum cn10k_ml_model_state { @@ -417,6 +418,9 @@ struct cn10k_ml_model { /* Address structure */ struct cn10k_ml_model_addr addr; + /* Tile and memory information object */ + struct cn10k_ml_ocm_model_map model_mem_map; + /* Spinlock, used to update model state */ plt_spinlock_t lock; @@ -428,5 +432,7 @@ int cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size); void cn10k_ml_model_metadata_update(struct cn10k_ml_model_metadata *metadata); void cn10k_ml_model_addr_update(struct cn10k_ml_model *model, uint8_t *buffer, uint8_t *base_dma_addr); +int cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, uint8_t *buffer, + uint16_t *wb_pages, uint16_t *scratch_pages); #endif /* _CN10K_ML_MODEL_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c new file mode 100644 index 0000000000..b1c62f2963 --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -0,0 +1,5 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#include "cn10k_ml_ocm.h" diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h new file mode 100644 index 0000000000..44390396f9 --- /dev/null +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -0,0 +1,79 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright (c) 2022 Marvell. + */ + +#ifndef _CN10K_ML_OCM_H_ +#define _CN10K_ML_OCM_H_ + +#include + +/* Page size in bytes. */ +#define ML_CN10K_OCM_PAGESIZE 0x4000 + +/* Number of OCM tiles. */ +#define ML_CN10K_OCM_NUMTILES 0x8 + +/* OCM in bytes, per tile. */ +#define ML_CN10K_OCM_TILESIZE 0x100000 + +/* OCM pages, per tile. */ +#define ML_CN10K_OCM_NUMPAGES (ML_CN10K_OCM_TILESIZE / ML_CN10K_OCM_PAGESIZE) + +/* Maximum OCM mask words, per tile, 8 bit words. */ +#define ML_CN10K_OCM_MASKWORDS (ML_CN10K_OCM_NUMPAGES / 8) + +/* OCM and Tile information structure */ +struct cn10k_ml_ocm_tile_info { + /* Mask of used / allotted pages on tile's OCM */ + uint8_t ocm_mask[ML_CN10K_OCM_MASKWORDS]; + + /* Last pages in the tile's OCM used for weights and bias, default = -1 */ + int last_wb_page; + + /* Number pages used for scratch memory on the tile's OCM */ + uint16_t scratch_pages; +}; + +/* Model OCM map structure */ +struct cn10k_ml_ocm_model_map { + /* Status of OCM reservation */ + bool ocm_reserved; + + /* Mask of OCM tiles for the model */ + uint64_t tilemask; + + /* Start page for the model load, default = -1 */ + int wb_page_start; + + /* Number of pages required for weights and bias */ + uint16_t wb_pages; + + /* Number of pages required for scratch memory */ + uint16_t scratch_pages; +}; + +/* OCM state structure */ +struct cn10k_ml_ocm { + /* OCM spinlock, used to update OCM state */ + rte_spinlock_t lock; + + /* Number of OCM tiles */ + uint8_t num_tiles; + + /* OCM size per each tile */ + uint64_t size_per_tile; + + /* Size of OCM page */ + uint64_t page_size; + + /* Number of OCM pages */ + uint16_t num_pages; + + /* Words per OCM mask */ + uint16_t mask_words; + + /* OCM memory info and status*/ + struct cn10k_ml_ocm_tile_info tile_ocm_info[ML_CN10K_OCM_NUMTILES]; +}; + +#endif /* _CN10K_ML_OCM_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index b11228f2cb..302ce8a452 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -126,9 +126,11 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c struct rte_ml_dev_info dev_info; struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; struct cn10k_ml_qp *qp; - uint32_t mz_size; uint16_t model_id; + uint32_t mz_size; + uint16_t tile_id; uint16_t qp_id; int ret; @@ -250,6 +252,18 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c } dev->data->nb_models = conf->nb_models; + ocm = &mldev->ocm; + ocm->num_tiles = ML_CN10K_OCM_NUMTILES; + ocm->size_per_tile = ML_CN10K_OCM_TILESIZE; + ocm->page_size = ML_CN10K_OCM_PAGESIZE; + ocm->num_pages = ocm->size_per_tile / ocm->page_size; + ocm->mask_words = ocm->num_pages / (8 * sizeof(uint8_t)); + + for (tile_id = 0; tile_id < ocm->num_tiles; tile_id++) + ocm->tile_ocm_info[tile_id].last_wb_page = -1; + + rte_spinlock_init(&ocm->lock); + mldev->nb_models_loaded = 0; mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; @@ -416,6 +430,8 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, const struct plt_memzone *mz; size_t model_data_size; uint8_t *base_dma_addr; + uint16_t scratch_pages; + uint16_t wb_pages; uint64_t mz_size; uint16_t idx; bool found; @@ -441,6 +457,11 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, return -ENOMEM; } + /* Get WB and scratch pages, check if model can be loaded. */ + ret = cn10k_ml_model_ocm_pages_count(mldev, idx, params->addr, &wb_pages, &scratch_pages); + if (ret < 0) + return ret; + /* Compute memzone size */ metadata = (struct cn10k_ml_model_metadata *)params->addr; model_data_size = metadata->init_model.file_size + metadata->main_model.file_size + @@ -478,6 +499,14 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, /* Copy data from load to run. run address to be used by MLIP */ rte_memcpy(model->addr.base_dma_addr_run, model->addr.base_dma_addr_load, model_data_size); + /* Initialize model_mem_map */ + memset(&model->model_mem_map, 0, sizeof(struct cn10k_ml_ocm_model_map)); + model->model_mem_map.ocm_reserved = false; + model->model_mem_map.tilemask = 0; + model->model_mem_map.wb_page_start = -1; + model->model_mem_map.wb_pages = wb_pages; + model->model_mem_map.scratch_pages = scratch_pages; + plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; dev->data->models[idx] = model; diff --git a/drivers/ml/cnxk/meson.build b/drivers/ml/cnxk/meson.build index 799e8f2470..393bc629b0 100644 --- a/drivers/ml/cnxk/meson.build +++ b/drivers/ml/cnxk/meson.build @@ -11,12 +11,14 @@ driver_sdk_headers = files( 'cn10k_ml_dev.h', 'cn10k_ml_ops.h', 'cn10k_ml_model.h', + 'cn10k_ml_ocm.h', ) sources = files( 'cn10k_ml_dev.c', 'cn10k_ml_ops.c', 'cn10k_ml_model.c', + 'cn10k_ml_ocm.c', ) deps += ['mldev', 'common_cnxk', 'kvargs', 'hash'] From patchwork Fri Mar 10 08:19:51 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124965 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id BA62241D3D; Fri, 10 Mar 2023 09:22:31 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 0402442D88; Fri, 10 Mar 2023 09:20:51 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id CC1FE42BB1 for ; Fri, 10 Mar 2023 09:20:29 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1I009730 for ; Fri, 10 Mar 2023 00:20:29 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=v33YgRIAk6wnQyoc86Zqjn89o2cq1Sw6cDn4LFZfYrE=; b=i7Yu9VC4xg+435TEkAz7K2QqM5RgIfhCnCtUWo7Dunjmf6B2J0XSfX6rmuiFDCSt8uS+ tssGVdT+Mov6gbLa6578R68qObFkkdHd8eSf/pToRLHPFEpkhXde+xGPAVs8YDJb40ly sbsx0IVRZ2HKQm0hGkbHsfO3cjyaqeVIW9WD1uGu02pdthG/4nReMgCkHLntgjvGiHEl 9VlvAFD3fUNtsastFgog16Og43I0ISyMeEJ39Lf8w7FzfCabt3lXsC7mcrgC7FDtY9TA ibS/fhdm72NG2dWS2T1soAurz0m1z9JZxusF1IypZjZATqtGTFH66J3f7mFJSR1QNyJV ng== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-13 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:28 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id BD5293F709C; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 15/39] ml/cnxk: add structures for slow and fast path JDs Date: Fri, 10 Mar 2023 00:19:51 -0800 Message-ID: <20230310082015.20200-16-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: AdxpFuMxEdBIxrGhYm0QiXz5aAMRvQ3P X-Proofpoint-ORIG-GUID: AdxpFuMxEdBIxrGhYm0QiXz5aAMRvQ3P X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added JD structures for load, unload and run jobs. Initialize job command and allocate memory for request structures for slow path jobs. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 99 ++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_model.h | 4 ++ drivers/ml/cnxk/cn10k_ml_ops.c | 19 +++++- drivers/ml/cnxk/cn10k_ml_ops.h | 4 ++ 4 files changed, 125 insertions(+), 1 deletion(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 02a4496c97..68fcc957fa 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -188,6 +188,105 @@ struct cn10k_ml_jd { uint8_t rsvd[8]; } fw_load; + + struct cn10k_ml_jd_section_model_start { + /* Source model start address in DDR relative to ML_MLR_BASE */ + uint64_t model_src_ddr_addr; + + /* Destination model start address in DDR relative to ML_MLR_BASE */ + uint64_t model_dst_ddr_addr; + + /* Offset to model init section in the model */ + uint64_t model_init_offset : 32; + + /* Size of init section in the model */ + uint64_t model_init_size : 32; + + /* Offset to model main section in the model */ + uint64_t model_main_offset : 32; + + /* Size of main section in the model */ + uint64_t model_main_size : 32; + + /* Offset to model finish section in the model */ + uint64_t model_finish_offset : 32; + + /* Size of finish section in the model */ + uint64_t model_finish_size : 32; + + /* Offset to WB in model bin */ + uint64_t model_wb_offset : 32; + + /* Number of model layers */ + uint64_t num_layers : 8; + + /* Number of gather entries, 0 means linear input mode (= no gather) */ + uint64_t num_gather_entries : 8; + + /* Number of scatter entries 0 means linear input mode (= no scatter) */ + uint64_t num_scatter_entries : 8; + + /* Tile mask to load model */ + uint64_t tilemask : 8; + + /* Batch size of model */ + uint64_t batch_size : 32; + + /* OCM WB base address */ + uint64_t ocm_wb_base_address : 32; + + /* OCM WB range start */ + uint64_t ocm_wb_range_start : 32; + + /* OCM WB range End */ + uint64_t ocm_wb_range_end : 32; + + /* DDR WB address */ + uint64_t ddr_wb_base_address; + + /* DDR WB range start */ + uint64_t ddr_wb_range_start : 32; + + /* DDR WB range end */ + uint64_t ddr_wb_range_end : 32; + + union { + /* Points to gather list if num_gather_entries > 0 */ + void *gather_list; + struct { + /* Linear input mode */ + uint64_t ddr_range_start : 32; + uint64_t ddr_range_end : 32; + } s; + } input; + + union { + /* Points to scatter list if num_scatter_entries > 0 */ + void *scatter_list; + struct { + /* Linear output mode */ + uint64_t ddr_range_start : 32; + uint64_t ddr_range_end : 32; + } s; + } output; + } model_start; + + struct cn10k_ml_jd_section_model_stop { + uint8_t rsvd[96]; + } model_stop; + + struct cn10k_ml_jd_section_model_run { + /* Address of the input for the run relative to ML_MLR_BASE */ + uint64_t input_ddr_addr; + + /* Address of the output for the run relative to ML_MLR_BASE */ + uint64_t output_ddr_addr; + + /* Number of batches to run in variable batch processing */ + uint16_t num_batches; + + uint8_t rsvd[78]; + } model_run; }; }; diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index 7893635787..355915deeb 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -11,6 +11,7 @@ #include "cn10k_ml_dev.h" #include "cn10k_ml_ocm.h" +#include "cn10k_ml_ops.h" /* Model state */ enum cn10k_ml_model_state { @@ -426,6 +427,9 @@ struct cn10k_ml_model { /* State */ enum cn10k_ml_model_state state; + + /* Slow-path operations request pointer */ + struct cn10k_ml_req *req; }; int cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 302ce8a452..56adce12ea 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -12,6 +12,10 @@ /* ML model macros */ #define CN10K_ML_MODEL_MEMZONE_NAME "ml_cn10k_model_mz" +/* ML Job descriptor flags */ +#define ML_FLAGS_POLL_COMPL BIT(0) +#define ML_FLAGS_SSO_COMPL BIT(1) + static void qp_memzone_name_get(char *name, int size, int dev_id, int qp_id) { @@ -65,6 +69,7 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des struct cn10k_ml_qp *qp; uint32_t len; uint8_t *va; + uint64_t i; /* Allocate queue pair */ qp = rte_zmalloc_socket("cn10k_ml_pmd_queue_pair", sizeof(struct cn10k_ml_qp), ROC_ALIGN, @@ -95,6 +100,12 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des qp->queue.wait_cycles = ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); qp->nb_desc = nb_desc; + /* Initialize job command */ + for (i = 0; i < qp->nb_desc; i++) { + memset(&qp->queue.reqs[i].jd, 0, sizeof(struct cn10k_ml_jd)); + qp->queue.reqs[i].jcmd.w1.s.jobptr = PLT_U64_CAST(&qp->queue.reqs[i].jd); + } + return qp; qp_free: @@ -468,7 +479,8 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, metadata->finish_model.file_size + metadata->weights_bias.file_size; model_data_size = PLT_ALIGN_CEIL(model_data_size, ML_CN10K_ALIGN_SIZE); mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + - 2 * model_data_size; + 2 * model_data_size + + PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_req), ML_CN10K_ALIGN_SIZE); /* Allocate memzone for model object and model data */ snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", CN10K_ML_MODEL_MEMZONE_NAME, idx); @@ -507,6 +519,11 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, model->model_mem_map.wb_pages = wb_pages; model->model_mem_map.scratch_pages = scratch_pages; + /* Set slow-path request address and state */ + model->req = PLT_PTR_ADD( + mz->addr, PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + + 2 * model_data_size); + plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; dev->data->models[idx] = model; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index d7842ecd73..c86ce66f19 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -6,6 +6,7 @@ #define _CN10K_ML_OPS_H_ #include +#include #include @@ -21,6 +22,9 @@ struct cn10k_ml_req { /* Status field for poll mode requests */ volatile uint64_t status; + + /* Job command */ + struct ml_job_cmd_s jcmd; } __rte_aligned(ROC_ALIGN); /* Request queue */ From patchwork Fri Mar 10 08:19:52 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124962 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5D0B941D3D; Fri, 10 Mar 2023 09:22:10 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id E692D42D79; Fri, 10 Mar 2023 09:20:46 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 6A36C42B8E for ; Fri, 10 Mar 2023 09:20:29 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uId9021875 for ; Fri, 10 Mar 2023 00:20:28 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=F94bsZ2WeTem3ulcLpWQPjqCN/0b0TFwrKEhwhKc7+o=; b=D+gf1ZwFiHiz/ZgiTwTJxjl8/iTShB0YOdYdRCwlX/wMRIVNdf7HN4dc3H2UDu6miajY ZqjfhkrgnmFgRmnFuOYBWAeUmx3+5By6qINK/IfCJQbbWD3m/l4S26R2zqRPu2XBAs34 p3WG4Pv0BYH0565CMl4iVB7mydYSk6i1FysgSz56hiIr8ZlIg33G2ijU/1IkjHwGlqaC Ie82b++2GlM4/JloKyX/CkEut/C8Jg8Yv/ohNIpagH6LzP/Rk3bmQCqMttxRnuxKGfHQ J8SpfaLT/IfbLSNasJY2DPwvKjTJi0YaL/1lF3HIbTRc52BZCkipG/y+Ef/U1GHz05kq Zg== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-8 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:28 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:22 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id E4DE03F709F; Fri, 10 Mar 2023 00:20:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 16/39] ml/cnxk: find OCM mask and page slots for a model Date: Fri, 10 Mar 2023 00:19:52 -0800 Message-ID: <20230310082015.20200-17-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: wFS3ZNHw8500LJNwS1RUaSqmd5-BsqF1 X-Proofpoint-ORIG-GUID: wFS3ZNHw8500LJNwS1RUaSqmd5-BsqF1 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support to compute OCM tilemask and page start for a model. The computed tilemask and page start are used during model start to copy model weights and bias to OCM. OCM slot for a model is allocated from the tiles with maximum amount of free memory. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ocm.c | 330 +++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ocm.h | 5 + 2 files changed, 335 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c index b1c62f2963..df2fa4c514 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.c +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -2,4 +2,334 @@ * Copyright (c) 2022 Marvell. */ +#include + +#include "cn10k_ml_dev.h" #include "cn10k_ml_ocm.h" + +#include "roc_api.h" + +/* OCM macros */ +#define BYTE_LEN 8 +#define OCM_MAP_WORD_SIZE (sizeof(uint8_t) * BYTE_LEN) +#define SET_BIT(num, n) ((num) | (1 << (n))) + +/* Left shift multi-word mask by 1 bit. + * + * For example, given a mask of two uint8_t words + * Input: [00110101] [00110111] + * Output: [01101010] [01101110] + */ +static void +lshift_mask(uint8_t *mask, int nwords) +{ + int i; + int word_sz; + + word_sz = sizeof(uint8_t) * BYTE_LEN; + for (i = nwords - 1; i >= 0; i--) { + mask[i] = mask[i] << 1; + if (i != 0) + mask[i] = mask[i] | (mask[i - 1] >> (word_sz - 1)); + } +} + +/* Get the index of the first unused slot in a multi-word mask (base_mask). Unused slots only after + * the start_pos are considered. An unused slot is a sequence of slot_sz continuous unset bits in + * the multi-word mask. For example given a multi-word mask, + * + * The program creates a search_mask with slot_sz bits set. Uses a sliding windows approach to scan + * the mask to identify the available first slot. search_mask slides left from start_pos to end. + * + * [10111000] [01001001] + * - WORD 1 --- WORD 0 - + * + * When start = 0, + * Index of the first unused slot of size 4 is 7. + * Index of the first unused slot of size 3 is 7. + * Index of the first unused slot of size 2 is 1. + * Index of the first unused slot of size 1 is 1. + * + * When start = 2, + * Index of the first unused slot of size 4 is 7. + * Index of the first unused slot of size 2 is 4. + * Index of the first unused slot of size 1 is 2. + * + * When unable to find a valid slot, return 0 + * When slot_sz is zero, return max_idx + 1 + */ +static int +slot_index_lowest(uint8_t *base_mask, int nwords, int slot_sz, int start_pos) +{ + uint8_t *search_mask; + int word_sz; + int end_pos; + int min_idx; + int max_idx; + bool match; + int i, j; + int idx; + + word_sz = sizeof(uint8_t) * BYTE_LEN; + min_idx = 0; + max_idx = word_sz * nwords; + idx = min_idx - 1; + + if (slot_sz == 0) + return max_idx; + + /* Create a mask with slot_sz bits set */ + search_mask = plt_zmalloc(nwords * sizeof(uint8_t), 0); + if (search_mask == NULL) + goto error; + + for (i = 0; i < nwords; i++) { + if (i < slot_sz / word_sz) + search_mask[i] = 0xFF; + else if (i > slot_sz / word_sz) + search_mask[i] = 0x00; + else + search_mask[i] = (1 << (slot_sz % word_sz)) - 1; + } + + /* Shift search mask by start_pos bits */ + for (i = 0; i < start_pos; i++) + lshift_mask(search_mask, nwords); + + /* Scan for a slot, left shift search mask after every iteration */ + end_pos = nwords * word_sz - slot_sz + 1; + for (j = start_pos; j < end_pos; j++) { + match = true; + for (i = 0; i < nwords; i++) + match = match && (((~base_mask[i]) & search_mask[i]) == search_mask[i]); + + if (match) { + idx = j; + goto found; + } + + lshift_mask(search_mask, nwords); + } + +found: + plt_free(search_mask); + +error: + return idx; +} + +/* Find the largest possible unused slot, with a minimum size of search_sz in a multi-work mask. The + * function returns the start index of the slot and the size of the identified slot (slot_sz). + * + * For example, in multi-word mask + * + * [10111000] [01001001] + * - WORD 1 --- WORD 0 - + * + * When search_sz > 4, return value = -1, slot_sz = 0 + * When search_sz <=4, return value = 7, slot_sz = 4 + */ +static int +slot_index_largest(uint8_t *base_mask, int nwords, int search_sz, int *slot_sz) +{ + uint8_t *search_mask; + int mask_sz; + int word_sz; + int end_pos; + bool match; + int i, j; + int idx; + + word_sz = sizeof(uint8_t) * BYTE_LEN; + mask_sz = nwords * word_sz; + idx = -1; + + /* Create a mask with mask_sz bits set */ + search_mask = plt_zmalloc(mask_sz, 0); + if (search_mask == NULL) + goto error; + +start: + for (i = 0; i < nwords; i++) { + if (i < mask_sz / word_sz) + search_mask[i] = 0xFF; + else if (i > mask_sz / word_sz) + search_mask[i] = 0x00; + else + search_mask[i] = (1 << (mask_sz % word_sz)) - 1; + } + + /* Scan for a slot, left shift search mask after every iteration */ + end_pos = nwords * word_sz - mask_sz + 1; + for (j = 0; j < end_pos; j++) { + match = true; + for (i = 0; i < nwords; i++) + match = match && (((~base_mask[i]) & search_mask[i]) == search_mask[i]); + + if (match) { + idx = j; + goto found; + } + lshift_mask(search_mask, nwords); + } + + mask_sz--; + if (mask_sz >= search_sz) + goto start; + else + mask_sz = 0; + +found: + plt_free(search_mask); + if (search_sz == 0) + idx = word_sz * nwords; + +error: + if (slot_sz) + *slot_sz = mask_sz; + + return idx; +} + +/* Count number of bits in a tilemask. Assumes that all set bits are contiguous. */ +int +cn10k_ml_ocm_tilecount(uint64_t tilemask, int *start, int *end) +{ + uint8_t count; + + PLT_ASSERT(tilemask != 0); + + *start = __builtin_ctzl(tilemask); + *end = 64 - __builtin_clzl(tilemask) - 1; + count = *end - *start + 1; + + PLT_ASSERT(count == __builtin_popcountl(tilemask)); + return count; +} + +/* Find the tiles and wb_page_start to load the model on given 'num_tiles' tiles with the specified + * scratch & WB pages and OCM allocation mode. + */ +int +cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t wb_pages, + uint16_t scratch_pages, uint64_t *tilemask) +{ + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + + uint8_t local_ocm_mask[ML_CN10K_OCM_MASKWORDS] = {0}; + uint16_t used_scratch_pages_max; + uint16_t scratch_page_start; + int used_last_wb_page_max; + uint16_t scratch_page_end; + uint8_t search_start_tile; + uint8_t search_end_tile; + int wb_page_start_curr; + int max_slot_sz_curr; + uint8_t tile_start; + int ocm_alloc_mode; + int wb_page_start; + uint16_t tile_id; + uint16_t word_id; + uint8_t tile_idx; + int max_slot_sz; + int start_tile; + int page_id; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + + if (num_tiles > ML_CN10K_OCM_NUMTILES) { + plt_err("Invalid num_tiles = %u (> ML_CN10K_OCM_NUMTILES)", num_tiles); + return -1; + } + + memset(tilemask, 0, sizeof(uint64_t)); + wb_page_start = -1; + used_scratch_pages_max = 0; + used_last_wb_page_max = -1; + start_tile = -1; + max_slot_sz_curr = 0; + max_slot_sz = 0; + tile_idx = 0; + ocm_alloc_mode = 2; + + if ((start_tile != -1) && (start_tile % num_tiles != 0)) { + plt_err("Invalid start_tile, %d", start_tile); + return -1; + } + + if (start_tile < 0) { + search_start_tile = 0; + search_end_tile = ocm->num_tiles - num_tiles; + } else { + search_start_tile = start_tile; + search_end_tile = start_tile; + } + + tile_start = search_start_tile; +start_search: + used_scratch_pages_max = 0; + used_last_wb_page_max = -1; + for (tile_id = tile_start; tile_id < tile_start + num_tiles; tile_id++) { + used_scratch_pages_max = + PLT_MAX(ocm->tile_ocm_info[tile_id].scratch_pages, used_scratch_pages_max); + used_last_wb_page_max = + PLT_MAX(ocm->tile_ocm_info[tile_id].last_wb_page, used_last_wb_page_max); + } + + memset(local_ocm_mask, 0, sizeof(local_ocm_mask)); + for (tile_id = tile_start; tile_id < tile_start + num_tiles; tile_id++) { + for (word_id = 0; word_id < ocm->mask_words; word_id++) + local_ocm_mask[word_id] |= ocm->tile_ocm_info[tile_id].ocm_mask[word_id]; + } + + if (used_scratch_pages_max < scratch_pages) { /* Check for extra scratch pages */ + if (ocm->num_pages - used_last_wb_page_max - 1 >= + scratch_pages) { /* Pages available */ + scratch_page_start = ocm->num_pages - scratch_pages; + scratch_page_end = ocm->num_pages - 1; + for (page_id = scratch_page_start; page_id <= scratch_page_end; + page_id++) { /* Mark the extra scratch pages as used */ + local_ocm_mask[page_id / OCM_MAP_WORD_SIZE] = + SET_BIT(local_ocm_mask[page_id / OCM_MAP_WORD_SIZE], + page_id % OCM_MAP_WORD_SIZE); + } + } else { /* Pages not available, check for next set of tiles */ + goto next_search; + } + } + + if (ocm_alloc_mode == 1) { + wb_page_start = slot_index_lowest(local_ocm_mask, ocm->mask_words, wb_pages, 0); + if (wb_page_start != -1) { /* Have a valid slot for WB, else next set of tiles */ + tile_idx = tile_start; + goto found; + } + } else if (ocm_alloc_mode == 2) { + wb_page_start_curr = slot_index_largest(local_ocm_mask, ocm->mask_words, wb_pages, + &max_slot_sz_curr); + if (max_slot_sz_curr > max_slot_sz) { + wb_page_start = wb_page_start_curr; + max_slot_sz = max_slot_sz_curr; + tile_idx = tile_start; + } else if (max_slot_sz_curr == max_slot_sz) { + wb_page_start = wb_page_start_curr; + if (wb_page_start == ocm->num_pages) { + tile_idx = tile_start; + goto found; + } + } + } + +next_search: + tile_start = tile_start + num_tiles; + if (tile_start <= search_end_tile) + goto start_search; + +found: + if (wb_page_start != -1) + *tilemask = GENMASK_ULL(tile_idx + num_tiles - 1, tile_idx); + + return wb_page_start; +} diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h index 44390396f9..2e26271a7a 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.h +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -6,6 +6,7 @@ #define _CN10K_ML_OCM_H_ #include +#include /* Page size in bytes. */ #define ML_CN10K_OCM_PAGESIZE 0x4000 @@ -76,4 +77,8 @@ struct cn10k_ml_ocm { struct cn10k_ml_ocm_tile_info tile_ocm_info[ML_CN10K_OCM_NUMTILES]; }; +int cn10k_ml_ocm_tilecount(uint64_t tilemask, int *start, int *end); +int cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t wb_pages, + uint16_t scratch_pages, uint64_t *tilemask); + #endif /* _CN10K_ML_OCM_H_ */ From patchwork Fri Mar 10 08:19:53 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124961 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 7A20741D3D; Fri, 10 Mar 2023 09:22:04 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 97AEA42D73; Fri, 10 Mar 2023 09:20:45 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id F2DAC427F5 for ; Fri, 10 Mar 2023 09:20:28 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1G009730 for ; Fri, 10 Mar 2023 00:20:28 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=hYRzza+oqcCGQ2bCjcsvuKucRwZ26cSXSrQYEIN1o5A=; b=lRvRexmeCLLINEUJcO8fWDNAiPldNaUeOa4jyYNSZ1xhn+/LrhmewL3aSky3nhXn25H6 1WOhq0VN8jF3VNP0SlCGK8YRe1e3HmybYdek0dXWkE/xEN3TLvVF/DMNK2uV29ISEUcb M4Vsbwe8yRcmoT9mbNZpZ1PjnaddpZTf5Gde/LnxRE0lP4i8JdKAwmgJa2ckLGuW2JWO rA8ziyn37VPj9Y4qECqTUGafrTD+O1GWpK35L9//LwjEfSLMxuJ11/qbHVb/D9LHVmV6 YsJJnF/6KUe9rW+EH/jaqSJ8UUUEti1hCRv6arIsoPgquj2jEd/eNtogmpMgCp1Gmtqg BA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-11 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:27 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 1C9EB3F70A1; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 17/39] ml/cnxk: add support to reserve and free OCM pages Date: Fri, 10 Mar 2023 00:19:53 -0800 Message-ID: <20230310082015.20200-18-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: yh19m_KO1Mt5At90AfVAcnwHB6fBxfrl X-Proofpoint-ORIG-GUID: yh19m_KO1Mt5At90AfVAcnwHB6fBxfrl X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support to reserve and free OCM pages for a model. OCM pages are reserved upon completion of model start and are released after model stop. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ocm.c | 131 ++++++++++++++++++++++++++++++++- drivers/ml/cnxk/cn10k_ml_ocm.h | 3 + 2 files changed, 131 insertions(+), 3 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c index df2fa4c514..c3e4de3e9c 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.c +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -5,14 +5,17 @@ #include #include "cn10k_ml_dev.h" +#include "cn10k_ml_model.h" #include "cn10k_ml_ocm.h" #include "roc_api.h" /* OCM macros */ -#define BYTE_LEN 8 -#define OCM_MAP_WORD_SIZE (sizeof(uint8_t) * BYTE_LEN) -#define SET_BIT(num, n) ((num) | (1 << (n))) +#define BYTE_LEN 8 +#define OCM_MAP_WORD_SIZE (sizeof(uint8_t) * BYTE_LEN) +#define IS_BIT_SET(num, n) ((num) & (1 << (n))) +#define SET_BIT(num, n) ((num) | (1 << (n))) +#define CLEAR_BIT(num, n) ((num) &= ~((1) << (n))) /* Left shift multi-word mask by 1 bit. * @@ -333,3 +336,125 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w return wb_page_start; } + +void +cn10k_ml_ocm_reserve_pages(struct rte_ml_dev *dev, uint16_t model_id, uint64_t tilemask, + int wb_page_start, uint16_t wb_pages, uint16_t scratch_pages) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + + int scratch_page_start; + int scratch_page_end; + int wb_page_end; + int tile_start; + int tile_end; + int tile_id; + int page_id; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + model = dev->data->models[model_id]; + + /* Get first set bit, tile_start */ + tile_start = 0; + tile_end = 0; + cn10k_ml_ocm_tilecount(tilemask, &tile_start, &tile_end); + wb_page_end = wb_page_start + wb_pages - 1; + scratch_page_start = ocm->num_pages - scratch_pages; + scratch_page_end = ocm->num_pages - 1; + + /* Update tile_ocm_info */ + for (tile_id = tile_start; tile_id <= tile_end; tile_id++) { + /* Scratch pages */ + for (page_id = scratch_page_start; page_id <= scratch_page_end; page_id++) + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE] = SET_BIT( + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE], + page_id % OCM_MAP_WORD_SIZE); + ocm->tile_ocm_info[tile_id].scratch_pages = + PLT_MAX(ocm->tile_ocm_info[tile_id].scratch_pages, scratch_pages); + + /* WB pages */ + for (page_id = wb_page_start; page_id <= wb_page_end; page_id++) + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE] = SET_BIT( + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE], + page_id % OCM_MAP_WORD_SIZE); + if (wb_pages != 0) + ocm->tile_ocm_info[tile_id].last_wb_page = + PLT_MAX(ocm->tile_ocm_info[tile_id].last_wb_page, wb_page_end); + } + + model->addr.tile_start = tile_start; + model->addr.tile_end = tile_end; + + plt_ml_dbg("model_id = %u, tilemask = 0x%016lx", model_id, tilemask); + plt_ml_dbg("model_id = %u, wb_page_start = %d, wb_page_end = %d", model_id, wb_page_start, + wb_page_end); + plt_ml_dbg("model_id = %u, scratch_page_start = %d, scratch_page_end = %d", model_id, + scratch_page_start, scratch_page_end); +} + +void +cn10k_ml_ocm_free_pages(struct rte_ml_dev *dev, uint16_t model_id) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + + int scratch_resize_pages; + int wb_page_start; + int wb_page_end; + int prev_start; + int curr_start; + int tile_id; + int page_id; + uint16_t i; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + model = dev->data->models[model_id]; + + /* Update OCM info for WB memory */ + wb_page_start = model->model_mem_map.wb_page_start; + wb_page_end = wb_page_start + model->model_mem_map.wb_pages - 1; + for (tile_id = model->addr.tile_start; tile_id <= model->addr.tile_end; tile_id++) { + for (page_id = wb_page_start; page_id <= wb_page_end; page_id++) { + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE] = + CLEAR_BIT(ocm->tile_ocm_info[tile_id] + .ocm_mask[page_id / OCM_MAP_WORD_SIZE], + page_id % OCM_MAP_WORD_SIZE); + } + + /* Update last_wb_page size */ + if (wb_page_end == ocm->tile_ocm_info[tile_id].last_wb_page) + ocm->tile_ocm_info[tile_id].last_wb_page = wb_page_start - 1; + + /* Update scratch page size and clear extra bits */ + scratch_resize_pages = 0; + /* Get max scratch pages required, excluding the current model */ + for (i = 0; i < dev->data->nb_models; i++) { + struct cn10k_ml_model *model = dev->data->models[i]; + + if ((i != model_id) && (model != NULL)) { + if (IS_BIT_SET(model->model_mem_map.tilemask, tile_id)) + scratch_resize_pages = + PLT_MAX((int)model->model_mem_map.scratch_pages, + scratch_resize_pages); + } + } + + /* Clear extra scratch pages */ + if (scratch_resize_pages < ocm->tile_ocm_info[tile_id].scratch_pages) { + prev_start = ocm->num_pages - ocm->tile_ocm_info[tile_id].scratch_pages; + curr_start = ocm->num_pages - scratch_resize_pages; + for (page_id = prev_start; page_id < curr_start; page_id++) { + ocm->tile_ocm_info[tile_id].ocm_mask[page_id / OCM_MAP_WORD_SIZE] = + CLEAR_BIT(ocm->tile_ocm_info[tile_id] + .ocm_mask[page_id / OCM_MAP_WORD_SIZE], + page_id % OCM_MAP_WORD_SIZE); + } + ocm->tile_ocm_info[tile_id].scratch_pages = scratch_resize_pages; + } + } +} diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h index 2e26271a7a..32c9b17afc 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.h +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -80,5 +80,8 @@ struct cn10k_ml_ocm { int cn10k_ml_ocm_tilecount(uint64_t tilemask, int *start, int *end); int cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t wb_pages, uint16_t scratch_pages, uint64_t *tilemask); +void cn10k_ml_ocm_reserve_pages(struct rte_ml_dev *dev, uint16_t model_id, uint64_t tilemask, + int wb_page_start, uint16_t wb_pages, uint16_t scratch_pages); +void cn10k_ml_ocm_free_pages(struct rte_ml_dev *dev, uint16_t model_id); #endif /* _CN10K_ML_OCM_H_ */ From patchwork Fri Mar 10 08:19:54 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124963 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id B864041D3D; Fri, 10 Mar 2023 09:22:16 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 20A5C42D63; Fri, 10 Mar 2023 09:20:48 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 6C2DF42BAC for ; Fri, 10 Mar 2023 09:20:29 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1H009730 for ; Fri, 10 Mar 2023 00:20:28 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=+hM9u53y8o2rREqkBY+BCCGBH9MDkwZ7ZH8iW5y+5wQ=; b=ghDpjqeGcWgwEDOoTGujehqsCfzDMBBkoN6Rl6Tw6Gt0GgwOBLyDx2LsXVwWOLOBgoUA Ak8TmhT/OFBXwhi/C2fN5jJGmEPU9LWMSCOJnpFr60MANVy3kc/hCqlWjm6/qFcjTz7+ bEECshQhM+LpzFr0ZSYMza0H8WZIXikutlSGKYBxMSorvpX7O5Q76uU1B1vRBdtr59eZ LN7jf1T6HnNyybESp4C4KJuL3mAO99HB0gWEKBMgO4TnqnCXYwZqrdrvqSvffHB1t5FD 0fx23/ooJOn6vIxLlxiTDKolZnhS4NZjK6S8jNB383cTqVgyUc4a3g7D3oWX3S6iRKOk oA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-12 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:28 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:22 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 47ED63F70A2; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 18/39] ml/cnxk: enable support to start an ML model Date: Fri, 10 Mar 2023 00:19:54 -0800 Message-ID: <20230310082015.20200-19-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: Mw-CU_SI6UOL-452yXtzk69XhJtKmp8z X-Proofpoint-ORIG-GUID: Mw-CU_SI6UOL-452yXtzk69XhJtKmp8z X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Implemented model start driver function. A model start job is checked for completion in synchronous mode. Tilemask and OCM slot is calculated before starting the model. Model start is enqueued through scratch registers. OCM pages are reserved after model start completion. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 3 + drivers/ml/cnxk/cn10k_ml_ops.c | 207 +++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.h | 4 + 3 files changed, 214 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 68fcc957fa..8f6bc24370 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -33,6 +33,9 @@ /* ML command timeout in seconds */ #define ML_CN10K_CMD_TIMEOUT 5 +/* ML slow-path job flags */ +#define ML_CN10K_SP_FLAGS_OCM_NONRELOCATABLE BIT(0) + /* Poll mode job state */ #define ML_CN10K_POLL_JOB_START 0 #define ML_CN10K_POLL_JOB_FINISH 1 diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 56adce12ea..e8ce65b182 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -114,6 +114,64 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des return NULL; } +static void +cn10k_ml_prep_sp_job_descriptor(struct cn10k_ml_dev *mldev, struct cn10k_ml_model *model, + struct cn10k_ml_req *req, enum cn10k_ml_job_type job_type) +{ + struct cn10k_ml_model_metadata *metadata; + struct cn10k_ml_model_addr *addr; + + metadata = &model->metadata; + addr = &model->addr; + + memset(&req->jd, 0, sizeof(struct cn10k_ml_jd)); + req->jd.hdr.jce.w0.u64 = 0; + req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&req->status); + req->jd.hdr.model_id = model->model_id; + req->jd.hdr.job_type = job_type; + req->jd.hdr.fp_flags = 0x0; + req->jd.hdr.result = roc_ml_addr_ap2mlip(&mldev->roc, &req->result); + + if (job_type == ML_CN10K_JOB_TYPE_MODEL_START) { + if (!model->metadata.model.ocm_relocatable) + req->jd.hdr.sp_flags = ML_CN10K_SP_FLAGS_OCM_NONRELOCATABLE; + else + req->jd.hdr.sp_flags = 0x0; + req->jd.model_start.model_src_ddr_addr = + PLT_U64_CAST(roc_ml_addr_ap2mlip(&mldev->roc, addr->init_load_addr)); + req->jd.model_start.model_dst_ddr_addr = + PLT_U64_CAST(roc_ml_addr_ap2mlip(&mldev->roc, addr->init_run_addr)); + req->jd.model_start.model_init_offset = 0x0; + req->jd.model_start.model_main_offset = metadata->init_model.file_size; + req->jd.model_start.model_finish_offset = + metadata->init_model.file_size + metadata->main_model.file_size; + req->jd.model_start.model_init_size = metadata->init_model.file_size; + req->jd.model_start.model_main_size = metadata->main_model.file_size; + req->jd.model_start.model_finish_size = metadata->finish_model.file_size; + req->jd.model_start.model_wb_offset = metadata->init_model.file_size + + metadata->main_model.file_size + + metadata->finish_model.file_size; + req->jd.model_start.num_layers = metadata->model.num_layers; + req->jd.model_start.num_gather_entries = 0; + req->jd.model_start.num_scatter_entries = 0; + req->jd.model_start.tilemask = 0; /* Updated after reserving pages */ + req->jd.model_start.batch_size = model->batch_size; + req->jd.model_start.ocm_wb_base_address = 0; /* Updated after reserving pages */ + req->jd.model_start.ocm_wb_range_start = metadata->model.ocm_wb_range_start; + req->jd.model_start.ocm_wb_range_end = metadata->model.ocm_wb_range_end; + req->jd.model_start.ddr_wb_base_address = PLT_U64_CAST(roc_ml_addr_ap2mlip( + &mldev->roc, + PLT_PTR_ADD(addr->finish_load_addr, metadata->finish_model.file_size))); + req->jd.model_start.ddr_wb_range_start = metadata->model.ddr_wb_range_start; + req->jd.model_start.ddr_wb_range_end = metadata->model.ddr_wb_range_end; + req->jd.model_start.input.s.ddr_range_start = metadata->model.ddr_input_range_start; + req->jd.model_start.input.s.ddr_range_end = metadata->model.ddr_input_range_end; + req->jd.model_start.output.s.ddr_range_start = + metadata->model.ddr_output_range_start; + req->jd.model_start.output.s.ddr_range_end = metadata->model.ddr_output_range_end; + } +} + static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { @@ -561,6 +619,154 @@ cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id) return plt_memzone_free(plt_memzone_lookup(str)); } +int +cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + struct cn10k_ml_req *req; + + bool job_enqueued; + bool job_dequeued; + uint8_t num_tiles; + uint64_t tilemask; + int wb_page_start; + int tile_start; + int tile_end; + bool locked; + int ret = 0; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + /* Prepare JD */ + req = model->req; + cn10k_ml_prep_sp_job_descriptor(mldev, model, req, ML_CN10K_JOB_TYPE_MODEL_START); + req->result.error_code = 0x0; + req->result.user_ptr = NULL; + + plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + plt_wmb(); + + num_tiles = model->metadata.model.tile_end - model->metadata.model.tile_start + 1; + + locked = false; + while (!locked) { + if (plt_spinlock_trylock(&model->lock) != 0) { + if (model->state == ML_CN10K_MODEL_STATE_STARTED) { + plt_ml_dbg("Model already started, model = 0x%016lx", + PLT_U64_CAST(model)); + plt_spinlock_unlock(&model->lock); + return 1; + } + + if (model->state == ML_CN10K_MODEL_STATE_JOB_ACTIVE) { + plt_err("A slow-path job is active for the model = 0x%016lx", + PLT_U64_CAST(model)); + plt_spinlock_unlock(&model->lock); + return -EBUSY; + } + + model->state = ML_CN10K_MODEL_STATE_JOB_ACTIVE; + plt_spinlock_unlock(&model->lock); + locked = true; + } + } + + while (!model->model_mem_map.ocm_reserved) { + if (plt_spinlock_trylock(&ocm->lock) != 0) { + wb_page_start = cn10k_ml_ocm_tilemask_find( + dev, num_tiles, model->model_mem_map.wb_pages, + model->model_mem_map.scratch_pages, &tilemask); + + if (wb_page_start == -1) { + plt_err("Free pages not available on OCM tiles"); + plt_err("Failed to start model = 0x%016lx, name = %s", + PLT_U64_CAST(model), model->metadata.model.name); + + plt_spinlock_unlock(&ocm->lock); + return -ENOMEM; + } + + model->model_mem_map.tilemask = tilemask; + model->model_mem_map.wb_page_start = wb_page_start; + + cn10k_ml_ocm_reserve_pages( + dev, model->model_id, model->model_mem_map.tilemask, + model->model_mem_map.wb_page_start, model->model_mem_map.wb_pages, + model->model_mem_map.scratch_pages); + model->model_mem_map.ocm_reserved = true; + plt_spinlock_unlock(&ocm->lock); + } + } + + /* Update JD */ + cn10k_ml_ocm_tilecount(model->model_mem_map.tilemask, &tile_start, &tile_end); + req->jd.model_start.tilemask = GENMASK_ULL(tile_end, tile_start); + req->jd.model_start.ocm_wb_base_address = + model->model_mem_map.wb_page_start * ocm->page_size; + + job_enqueued = false; + job_dequeued = false; + do { + if (!job_enqueued) { + req->timeout = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + job_enqueued = roc_ml_scratch_enqueue(&mldev->roc, &req->jd); + } + + if (job_enqueued && !job_dequeued) + job_dequeued = roc_ml_scratch_dequeue(&mldev->roc, &req->jd); + + if (job_dequeued) + break; + } while (plt_tsc_cycles() < req->timeout); + + if (job_dequeued) { + if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { + if (req->result.error_code == 0) + ret = 0; + else + ret = -1; + } + } else { /* Reset scratch registers */ + roc_ml_scratch_queue_reset(&mldev->roc); + ret = -ETIME; + } + + locked = false; + while (!locked) { + if (plt_spinlock_trylock(&model->lock) != 0) { + if (ret == 0) + model->state = ML_CN10K_MODEL_STATE_STARTED; + else + model->state = ML_CN10K_MODEL_STATE_UNKNOWN; + + plt_spinlock_unlock(&model->lock); + locked = true; + } + } + + if (model->state == ML_CN10K_MODEL_STATE_UNKNOWN) { + while (model->model_mem_map.ocm_reserved) { + if (plt_spinlock_trylock(&ocm->lock) != 0) { + cn10k_ml_ocm_free_pages(dev, model->model_id); + model->model_mem_map.ocm_reserved = false; + model->model_mem_map.tilemask = 0x0; + plt_spinlock_unlock(&ocm->lock); + } + } + } + + return ret; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -576,4 +782,5 @@ struct rte_ml_dev_ops cn10k_ml_ops = { /* Model ops */ .model_load = cn10k_ml_model_load, .model_unload = cn10k_ml_model_unload, + .model_start = cn10k_ml_model_start, }; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index c86ce66f19..989af978c4 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -25,6 +25,9 @@ struct cn10k_ml_req { /* Job command */ struct ml_job_cmd_s jcmd; + + /* Timeout cycle */ + uint64_t timeout; } __rte_aligned(ROC_ALIGN); /* Request queue */ @@ -61,5 +64,6 @@ extern struct rte_ml_dev_ops cn10k_ml_ops; int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint16_t *model_id); int cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id); +int cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id); #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:19:55 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124964 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5A0D841D3D; Fri, 10 Mar 2023 09:22:25 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id BBD3A42D84; Fri, 10 Mar 2023 09:20:49 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id E5AA142B8E for ; Fri, 10 Mar 2023 09:20:29 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdA021875 for ; Fri, 10 Mar 2023 00:20:29 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=yIJAlsqHtFlVIXo/vDDUx61Y98G+kO8oSKtFSS2lnSY=; b=PCslzJHizYa9p1cjvNOCYmLCG6q/eE2JZrgyfQUNQ4FtcarLVPxseDk69/QmSyLup1hq pmxF1tncZ9KeUMeQEFDoeKFUs7mEXI6JEMI51uQzTB2T0Uz/Aqet0i3mU/HvS/uzIQoQ DH44ntPI045R+Q/A3f3YgelGnYBE9kfVBMsJ2/qPGwmNgZNPKKgyLohAB8GzYHrau/XQ l1P8/QDZ1O0nasBnifebmQxFU4A9TCgUivKR5HezdT1PX6VlABgWfy42YYZ83Z7v62W1 Bu/GuEea3hL5Osv5J3gFeZxLvznCxMXm8X80u1xMCdhP0i5OIT+JUIFAGbw0j8UwJlNB sQ== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-9 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:29 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 73A5D3F7088; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 19/39] ml/cnxk: enable support to stop an ML models Date: Fri, 10 Mar 2023 00:19:55 -0800 Message-ID: <20230310082015.20200-20-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: 9ozsgbFBnUufWvkHn42vrxcVuBE4cnjl X-Proofpoint-ORIG-GUID: 9ozsgbFBnUufWvkHn42vrxcVuBE4cnjl X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Implemented model stop driver function. A model stop job is enqueued through scratch registers and is checked for completion through polling in a synchronous mode. OCM pages are released after model stop completion. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 115 ++++++++++++++++++++++++++++++++- drivers/ml/cnxk/cn10k_ml_ops.h | 1 + 2 files changed, 114 insertions(+), 2 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index e8ce65b182..77d3728d8d 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -295,10 +295,14 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c /* Re-configure */ void **models; - /* Unload all models */ + /* Stop and unload all models */ for (model_id = 0; model_id < dev->data->nb_models; model_id++) { model = dev->data->models[model_id]; if (model != NULL) { + if (model->state == ML_CN10K_MODEL_STATE_STARTED) { + if (cn10k_ml_model_stop(dev, model_id) != 0) + plt_err("Could not stop model %u", model_id); + } if (model->state == ML_CN10K_MODEL_STATE_LOADED) { if (cn10k_ml_model_unload(dev, model_id) != 0) plt_err("Could not unload model %u", model_id); @@ -362,10 +366,14 @@ cn10k_ml_dev_close(struct rte_ml_dev *dev) mldev = dev->data->dev_private; - /* Unload all models */ + /* Stop and unload all models */ for (model_id = 0; model_id < dev->data->nb_models; model_id++) { model = dev->data->models[model_id]; if (model != NULL) { + if (model->state == ML_CN10K_MODEL_STATE_STARTED) { + if (cn10k_ml_model_stop(dev, model_id) != 0) + plt_err("Could not stop model %u", model_id); + } if (model->state == ML_CN10K_MODEL_STATE_LOADED) { if (cn10k_ml_model_unload(dev, model_id) != 0) plt_err("Could not unload model %u", model_id); @@ -767,6 +775,108 @@ cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id) return ret; } +int +cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + struct cn10k_ml_req *req; + + bool job_enqueued; + bool job_dequeued; + bool locked; + int ret = 0; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + /* Prepare JD */ + req = model->req; + cn10k_ml_prep_sp_job_descriptor(mldev, model, req, ML_CN10K_JOB_TYPE_MODEL_STOP); + req->result.error_code = 0x0; + req->result.user_ptr = NULL; + + plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + plt_wmb(); + + locked = false; + while (!locked) { + if (plt_spinlock_trylock(&model->lock) != 0) { + if (model->state == ML_CN10K_MODEL_STATE_LOADED) { + plt_ml_dbg("Model not started, model = 0x%016lx", + PLT_U64_CAST(model)); + plt_spinlock_unlock(&model->lock); + return 1; + } + + if (model->state == ML_CN10K_MODEL_STATE_JOB_ACTIVE) { + plt_err("A slow-path job is active for the model = 0x%016lx", + PLT_U64_CAST(model)); + plt_spinlock_unlock(&model->lock); + return -EBUSY; + } + + model->state = ML_CN10K_MODEL_STATE_JOB_ACTIVE; + plt_spinlock_unlock(&model->lock); + locked = true; + } + } + + while (model->model_mem_map.ocm_reserved) { + if (plt_spinlock_trylock(&ocm->lock) != 0) { + cn10k_ml_ocm_free_pages(dev, model->model_id); + model->model_mem_map.ocm_reserved = false; + model->model_mem_map.tilemask = 0x0; + plt_spinlock_unlock(&ocm->lock); + } + } + + job_enqueued = false; + job_dequeued = false; + do { + if (!job_enqueued) { + req->timeout = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + job_enqueued = roc_ml_scratch_enqueue(&mldev->roc, &req->jd); + } + + if (job_enqueued && !job_dequeued) + job_dequeued = roc_ml_scratch_dequeue(&mldev->roc, &req->jd); + + if (job_dequeued) + break; + } while (plt_tsc_cycles() < req->timeout); + + if (job_dequeued) { + if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { + if (req->result.error_code == 0x0) + ret = 0; + else + ret = -1; + } + } else { + roc_ml_scratch_queue_reset(&mldev->roc); + ret = -ETIME; + } + + locked = false; + while (!locked) { + if (plt_spinlock_trylock(&model->lock) != 0) { + model->state = ML_CN10K_MODEL_STATE_LOADED; + plt_spinlock_unlock(&model->lock); + locked = true; + } + } + + return ret; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -783,4 +893,5 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .model_load = cn10k_ml_model_load, .model_unload = cn10k_ml_model_unload, .model_start = cn10k_ml_model_start, + .model_stop = cn10k_ml_model_stop, }; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 989af978c4..22576b93c0 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -65,5 +65,6 @@ int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *para uint16_t *model_id); int cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id); int cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id); +int cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id); #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:19:56 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124967 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5CE8841D3D; Fri, 10 Mar 2023 09:22:46 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id BF1FD42D5A; Fri, 10 Mar 2023 09:20:53 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 7053342BB1 for ; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdB021875 for ; Fri, 10 Mar 2023 00:20:29 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=jfFFtagwoLhGzlImTyPjiK2mcsfv/BsxhY/LDUD7hqI=; b=N2ceXZryShyg8bQwY85Yx7zE33uDRJlmYLT/s1fuLvmpuI5izgB2p6VkVnS6vYzcf8U+ 5yhuq+0oN1XZF0j5BuFTO1Ei5yOuLULSl37HFWnxaF+4Wobe9JeElqAUHNsVssoO8HLy D7THFrdMa4WtloaDucFf5XQWJrdeKVOTljDyyPgu5uu1R4N6uGxpGTbfmBMsGV90wz47 eIH0YkPh7y2o10UiC4elMqc6BXDciItaPfwI6HZmWY088UUYoO42h/JllfNmZ4fGrGd5 EW3H7SMwJj27vaKJTFyNTl42sByA/pPGJ8I6QAUP3b9rbXzTeWJ1Gi64KiXgneuRdN3f Sw== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-10 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:29 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 9D5245B6939; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 20/39] ml/cnxk: enable support to get model information Date: Fri, 10 Mar 2023 00:19:56 -0800 Message-ID: <20230310082015.20200-21-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: a05kWtbs0tVseurt0NrBDk9LIIK2PUmA X-Proofpoint-ORIG-GUID: a05kWtbs0tVseurt0NrBDk9LIIK2PUmA X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added driver functions to get model information. Added internal functions to set and get model info. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_model.c | 55 ++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_model.h | 9 ++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 37 ++++++++++++++++++--- 3 files changed, 97 insertions(+), 4 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c index 69d6306104..0ded355d81 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.c +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -356,3 +356,58 @@ cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, ui return 0; } + +void +cn10k_ml_model_info_set(struct rte_ml_dev *dev, struct cn10k_ml_model *model) +{ + struct rte_ml_model_info *info; + struct rte_ml_io_info *output; + struct rte_ml_io_info *input; + uint8_t i; + + info = PLT_PTR_CAST(model->info); + input = PLT_PTR_ADD(info, sizeof(struct rte_ml_model_info)); + output = + PLT_PTR_ADD(input, model->metadata.model.num_input * sizeof(struct rte_ml_io_info)); + + /* Set model info */ + memset(info, 0, sizeof(struct rte_ml_model_info)); + rte_memcpy(info->name, model->metadata.model.name, MRVL_ML_MODEL_NAME_LEN); + snprintf(info->version, RTE_ML_STR_MAX, "%u.%u.%u.%u", model->metadata.model.version[0], + model->metadata.model.version[1], model->metadata.model.version[2], + model->metadata.model.version[3]); + info->model_id = model->model_id; + info->device_id = dev->data->dev_id; + info->batch_size = model->batch_size; + info->nb_inputs = model->metadata.model.num_input; + info->input_info = input; + info->nb_outputs = model->metadata.model.num_output; + info->output_info = output; + info->wb_size = model->metadata.weights_bias.file_size; + + /* Set input info */ + for (i = 0; i < info->nb_inputs; i++) { + rte_memcpy(input[i].name, model->metadata.input[i].input_name, + MRVL_ML_INPUT_NAME_LEN); + input[i].dtype = model->metadata.input[i].input_type; + input[i].qtype = model->metadata.input[i].model_input_type; + input[i].shape.format = model->metadata.input[i].shape.format; + input[i].shape.w = model->metadata.input[i].shape.w; + input[i].shape.x = model->metadata.input[i].shape.x; + input[i].shape.y = model->metadata.input[i].shape.y; + input[i].shape.z = model->metadata.input[i].shape.z; + } + + /* Set output info */ + for (i = 0; i < info->nb_outputs; i++) { + rte_memcpy(output[i].name, model->metadata.output[i].output_name, + MRVL_ML_OUTPUT_NAME_LEN); + output[i].dtype = model->metadata.output[i].output_type; + output[i].qtype = model->metadata.output[i].model_output_type; + output[i].shape.format = RTE_ML_IO_FORMAT_1D; + output[i].shape.w = model->metadata.output[i].size; + output[i].shape.x = 1; + output[i].shape.y = 1; + output[i].shape.z = 1; + } +} diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index 355915deeb..75990fe1e4 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -422,6 +422,14 @@ struct cn10k_ml_model { /* Tile and memory information object */ struct cn10k_ml_ocm_model_map model_mem_map; + /* Internal model information structure + * Size of the buffer = sizeof(struct rte_ml_model_info) + * + num_inputs * sizeof(struct rte_ml_io_info) + * + num_outputs * sizeof(struct rte_ml_io_info). + * Structures would be arranged in the same order in the buffer. + */ + uint8_t *info; + /* Spinlock, used to update model state */ plt_spinlock_t lock; @@ -438,5 +446,6 @@ void cn10k_ml_model_addr_update(struct cn10k_ml_model *model, uint8_t *buffer, uint8_t *base_dma_addr); int cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, uint8_t *buffer, uint16_t *wb_pages, uint16_t *scratch_pages); +void cn10k_ml_model_info_set(struct rte_ml_dev *dev, struct cn10k_ml_model *model); #endif /* _CN10K_ML_MODEL_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 77d3728d8d..ad9b3dfd21 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -506,6 +506,7 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, char str[RTE_MEMZONE_NAMESIZE]; const struct plt_memzone *mz; size_t model_data_size; + size_t model_info_size; uint8_t *base_dma_addr; uint16_t scratch_pages; uint16_t wb_pages; @@ -544,8 +545,13 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, model_data_size = metadata->init_model.file_size + metadata->main_model.file_size + metadata->finish_model.file_size + metadata->weights_bias.file_size; model_data_size = PLT_ALIGN_CEIL(model_data_size, ML_CN10K_ALIGN_SIZE); + model_info_size = sizeof(struct rte_ml_model_info) + + metadata->model.num_input * sizeof(struct rte_ml_io_info) + + metadata->model.num_output * sizeof(struct rte_ml_io_info); + model_info_size = PLT_ALIGN_CEIL(model_info_size, ML_CN10K_ALIGN_SIZE); + mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + - 2 * model_data_size + + 2 * model_data_size + model_info_size + PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_req), ML_CN10K_ALIGN_SIZE); /* Allocate memzone for model object and model data */ @@ -585,10 +591,12 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, model->model_mem_map.wb_pages = wb_pages; model->model_mem_map.scratch_pages = scratch_pages; + /* Set model info */ + model->info = PLT_PTR_ADD(base_dma_addr, 2 * model_data_size); + cn10k_ml_model_info_set(dev, model); + /* Set slow-path request address and state */ - model->req = PLT_PTR_ADD( - mz->addr, PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + - 2 * model_data_size); + model->req = PLT_PTR_ADD(model->info, model_info_size); plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; @@ -877,6 +885,26 @@ cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id) return ret; } +static int +cn10k_ml_model_info_get(struct rte_ml_dev *dev, uint16_t model_id, + struct rte_ml_model_info *model_info) +{ + struct cn10k_ml_model *model; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + rte_memcpy(model_info, model->info, sizeof(struct rte_ml_model_info)); + model_info->input_info = ((struct rte_ml_model_info *)model->info)->input_info; + model_info->output_info = ((struct rte_ml_model_info *)model->info)->output_info; + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -894,4 +922,5 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .model_unload = cn10k_ml_model_unload, .model_start = cn10k_ml_model_start, .model_stop = cn10k_ml_model_stop, + .model_info_get = cn10k_ml_model_info_get, }; From patchwork Fri Mar 10 08:19:57 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124969 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id EC57341D3D; Fri, 10 Mar 2023 09:22:57 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 34CF742DA5; Fri, 10 Mar 2023 09:20:56 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id D86E94113C for ; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdC021875 for ; Fri, 10 Mar 2023 00:20:30 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=b6ePjV7UtJRI8gbp//qRK9WNLaBw/7vtwZHmrx4Ukc8=; b=YVDSFtiNamoOfhE7AgnvRncLxrwrL4HNlrqSbpPX80tNJkR9Kfu/FnyuSdetraod78Ps shKWS0pI5q7xWz3tWWkWtF4T8sPotFoVC2UxmtDGF+z1w/zrnDn22wH0kR8a5rhiBpU0 SLf9iaBNh/LwIcdsGofkQpUq0Xh+Ql8VF5/XcnPWsLwoiMmD6AmevsbSan4x2p3rJIFp 5qdwbN2KKB2fFfL8/AfLgm6U+RPau/tvrOAl2SkUpUT6B8tLKR1b61jBobVe9nDmauOk +JpevlKIZxR3cfn9gchyXUyHd9UAB5N7xwuZqIi/oo15OwrO9mGqbARZRA0Mj/UwdEkL UQ== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-11 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:30 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id C3BCD5B693D; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 21/39] ml/cnxk: enable support to update model params Date: Fri, 10 Mar 2023 00:19:57 -0800 Message-ID: <20230310082015.20200-22-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: _vMT-_H1EVU9BePeOmgmxVMmprOLyZCH X-Proofpoint-ORIG-GUID: _vMT-_H1EVU9BePeOmgmxVMmprOLyZCH X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added cnxk driver functions to update model params or weights and bias after a models is loaded. Updating model params would not require reloading the model. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index ad9b3dfd21..92bf1a0854 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -905,6 +905,36 @@ cn10k_ml_model_info_get(struct rte_ml_dev *dev, uint16_t model_id, return 0; } +static int +cn10k_ml_model_params_update(struct rte_ml_dev *dev, uint16_t model_id, void *buffer) +{ + struct cn10k_ml_model *model; + size_t size; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + if (model->state == ML_CN10K_MODEL_STATE_UNKNOWN) + return -1; + else if (model->state != ML_CN10K_MODEL_STATE_LOADED) + return -EBUSY; + + size = model->metadata.init_model.file_size + model->metadata.main_model.file_size + + model->metadata.finish_model.file_size + model->metadata.weights_bias.file_size; + + /* Update model weights & bias */ + rte_memcpy(model->addr.wb_load_addr, buffer, model->metadata.weights_bias.file_size); + + /* Copy data from load to run. run address to be used by MLIP */ + rte_memcpy(model->addr.base_dma_addr_run, model->addr.base_dma_addr_load, size); + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -923,4 +953,5 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .model_start = cn10k_ml_model_start, .model_stop = cn10k_ml_model_stop, .model_info_get = cn10k_ml_model_info_get, + .model_params_update = cn10k_ml_model_params_update, }; From patchwork Fri Mar 10 08:19:58 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124971 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 9199341D3D; Fri, 10 Mar 2023 09:23:09 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 9188E42DB5; Fri, 10 Mar 2023 09:20:58 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 555724113C for ; Fri, 10 Mar 2023 09:20:31 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdD021875 for ; Fri, 10 Mar 2023 00:20:30 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=SBcnCfK0N/BUFCcrMg405YNoUKA9OOXcGRF4osH3KOE=; b=iyjpkZQXTZ+9+JZTeRHFTXLekMn00KYhNIemSBGvuzP8JcQqVn9JqaPQH6q/sxyeCz7a q1SYyrmMUZtaRecchRda2VWH517kDu71PaFamOpXYA4cv74DBgMugG19ijKLyzRBWtho aYsSAmuDdWMdxmnUip8VJPZp+5DYoQt7w85gTGjUx/cXKbgW93o/GLEp0w+kAcUQCHUh 9MbMK2jDxfjY+4B1AJHU/5YEi+qKx0Jc2JF8Zf8Y2e0sBG4fy65g/b4Wc5M22QgdtTnH D9e5lLlQmvdbGmY5kgrmu6psvBa4jM1V+Wf/NlDN7FTYWA7mqdrSx+66PxRHcfl+x3DB Uw== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-12 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:30 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id EB1E65C68E2; Fri, 10 Mar 2023 00:20:21 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 22/39] ml/cnxk: add support to get IO buffer sizes Date: Fri, 10 Mar 2023 00:19:58 -0800 Message-ID: <20230310082015.20200-23-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: 9wOysjKcnpn4H8xoKFkWFeM-mkr4EHP9 X-Proofpoint-ORIG-GUID: 9wOysjKcnpn4H8xoKFkWFeM-mkr4EHP9 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added driver functions to get input and output buffer sizes for a given batch size. This function would compute the buffer size based on specific requirements of the device. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 52 ++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 92bf1a0854..b5c89bee40 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -935,6 +935,54 @@ cn10k_ml_model_params_update(struct rte_ml_dev *dev, uint16_t model_id, void *bu return 0; } +static int +cn10k_ml_io_input_size_get(struct rte_ml_dev *dev, uint16_t model_id, uint32_t nb_batches, + uint64_t *input_qsize, uint64_t *input_dsize) +{ + struct cn10k_ml_model *model; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + if (input_qsize != NULL) + *input_qsize = PLT_U64_CAST(model->addr.total_input_sz_q * + PLT_DIV_CEIL(nb_batches, model->batch_size)); + + if (input_dsize != NULL) + *input_dsize = PLT_U64_CAST(model->addr.total_input_sz_d * + PLT_DIV_CEIL(nb_batches, model->batch_size)); + + return 0; +} + +static int +cn10k_ml_io_output_size_get(struct rte_ml_dev *dev, uint16_t model_id, uint32_t nb_batches, + uint64_t *output_qsize, uint64_t *output_dsize) +{ + struct cn10k_ml_model *model; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + if (output_qsize != NULL) + *output_qsize = PLT_U64_CAST(model->addr.total_output_sz_q * + PLT_DIV_CEIL(nb_batches, model->batch_size)); + + if (output_dsize != NULL) + *output_dsize = PLT_U64_CAST(model->addr.total_output_sz_d * + PLT_DIV_CEIL(nb_batches, model->batch_size)); + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -954,4 +1002,8 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .model_stop = cn10k_ml_model_stop, .model_info_get = cn10k_ml_model_info_get, .model_params_update = cn10k_ml_model_params_update, + + /* I/O ops */ + .io_input_size_get = cn10k_ml_io_input_size_get, + .io_output_size_get = cn10k_ml_io_output_size_get, }; From patchwork Fri Mar 10 08:19:59 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124973 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id E5DA741D3D; Fri, 10 Mar 2023 09:23:20 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id A165742DC6; Fri, 10 Mar 2023 09:21:00 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id D14A342BDA for ; Fri, 10 Mar 2023 09:20:31 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdE021875 for ; Fri, 10 Mar 2023 00:20:31 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=C4bPUhYr92S/nF2vE8AGx3IoRio0mSEO7kPfyqW87ew=; b=EcysZc0A828u58iHWhtxBPkqsJIrJPuFhY1vDqoFthvYedwjP5Rfxf9fvJsU35+3BEPi pXzjAJXFqZmYUMoLraZt7pKm1KP3TsZ7XxvTQYWb0oaPlrcnPAXC05FcYX+4nNf8EAsQ 04Gim2C5icRBRZa8bVycj92fbCYw5aQD7MLkPXPeFVgHTz2DZz6KwdN2Xs1w/UCavTF7 mMWcG5NLbSOuSg168+Q3syCwzOFTcICzzwoBlfEPQSGGS5NoWdXtZ8VIZjji5m8q+S6x 4BWaqzP1z977iRBZotMSNdPHFigPBtNhOyLE8TckLH7C/A30f/8Wvb7NBCQTpStw1w63 BQ== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-13 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:31 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 1EFB55B6943; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 23/39] ml/cnxk: enable quantization and dequantization Date: Fri, 10 Mar 2023 00:19:59 -0800 Message-ID: <20230310082015.20200-24-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: JLwPRjVdV9zGyZWjYH1VznT6aQloR-Va X-Proofpoint-ORIG-GUID: JLwPRjVdV9zGyZWjYH1VznT6aQloR-Va X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Implemented driver functions to quantize / dequantize input and output data. Support is enabled for multiple batches. Quantization / dequantization use the type conversion functions defined in ML common code. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 151 +++++++++++++++++++++++++++++++++ 1 file changed, 151 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index b5c89bee40..231c9b340b 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -5,6 +5,8 @@ #include #include +#include + #include "cn10k_ml_dev.h" #include "cn10k_ml_model.h" #include "cn10k_ml_ops.h" @@ -983,6 +985,153 @@ cn10k_ml_io_output_size_get(struct rte_ml_dev *dev, uint16_t model_id, uint32_t return 0; } +static int +cn10k_ml_io_quantize(struct rte_ml_dev *dev, uint16_t model_id, uint16_t nb_batches, void *dbuffer, + void *qbuffer) +{ + struct cn10k_ml_model *model; + uint8_t *lcl_dbuffer; + uint8_t *lcl_qbuffer; + uint32_t batch_id; + uint32_t i; + int ret; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + lcl_dbuffer = dbuffer; + lcl_qbuffer = qbuffer; + batch_id = 0; + +next_batch: + for (i = 0; i < model->metadata.model.num_input; i++) { + if (model->metadata.input[i].input_type == + model->metadata.input[i].model_input_type) { + rte_memcpy(lcl_qbuffer, lcl_dbuffer, model->addr.input[i].sz_d); + } else { + switch (model->metadata.input[i].model_input_type) { + case RTE_ML_IO_TYPE_INT8: + ret = rte_ml_io_float32_to_int8(model->metadata.input[i].qscale, + model->addr.input[i].nb_elements, + lcl_dbuffer, lcl_qbuffer); + break; + case RTE_ML_IO_TYPE_UINT8: + ret = rte_ml_io_float32_to_uint8(model->metadata.input[i].qscale, + model->addr.input[i].nb_elements, + lcl_dbuffer, lcl_qbuffer); + break; + case RTE_ML_IO_TYPE_INT16: + ret = rte_ml_io_float32_to_int16(model->metadata.input[i].qscale, + model->addr.input[i].nb_elements, + lcl_dbuffer, lcl_qbuffer); + break; + case RTE_ML_IO_TYPE_UINT16: + ret = rte_ml_io_float32_to_uint16(model->metadata.input[i].qscale, + model->addr.input[i].nb_elements, + lcl_dbuffer, lcl_qbuffer); + break; + case RTE_ML_IO_TYPE_FP16: + ret = rte_ml_io_float32_to_float16(model->addr.input[i].nb_elements, + lcl_dbuffer, lcl_qbuffer); + break; + default: + plt_err("Unsupported model_input_type[%u] : %u", i, + model->metadata.input[i].model_input_type); + ret = -ENOTSUP; + } + if (ret < 0) + return ret; + } + + lcl_dbuffer += model->addr.input[i].sz_d; + lcl_qbuffer += model->addr.input[i].sz_q; + } + + batch_id++; + if (batch_id < PLT_DIV_CEIL(nb_batches, model->batch_size)) + goto next_batch; + + return 0; +} + +static int +cn10k_ml_io_dequantize(struct rte_ml_dev *dev, uint16_t model_id, uint16_t nb_batches, + void *qbuffer, void *dbuffer) +{ + struct cn10k_ml_model *model; + uint8_t *lcl_qbuffer; + uint8_t *lcl_dbuffer; + uint32_t batch_id; + uint32_t i; + int ret; + + model = dev->data->models[model_id]; + + if (model == NULL) { + plt_err("Invalid model_id = %u", model_id); + return -EINVAL; + } + + lcl_dbuffer = dbuffer; + lcl_qbuffer = qbuffer; + batch_id = 0; + +next_batch: + for (i = 0; i < model->metadata.model.num_output; i++) { + if (model->metadata.output[i].output_type == + model->metadata.output[i].model_output_type) { + rte_memcpy(lcl_dbuffer, lcl_qbuffer, model->addr.output[i].sz_q); + } else { + switch (model->metadata.output[i].model_output_type) { + case RTE_ML_IO_TYPE_INT8: + ret = rte_ml_io_int8_to_float32(model->metadata.output[i].dscale, + model->addr.output[i].nb_elements, + lcl_qbuffer, lcl_dbuffer); + break; + case RTE_ML_IO_TYPE_UINT8: + ret = rte_ml_io_uint8_to_float32(model->metadata.output[i].dscale, + model->addr.output[i].nb_elements, + lcl_qbuffer, lcl_dbuffer); + break; + case RTE_ML_IO_TYPE_INT16: + ret = rte_ml_io_int16_to_float32(model->metadata.output[i].dscale, + model->addr.output[i].nb_elements, + lcl_qbuffer, lcl_dbuffer); + break; + case RTE_ML_IO_TYPE_UINT16: + ret = rte_ml_io_uint16_to_float32(model->metadata.output[i].dscale, + model->addr.output[i].nb_elements, + lcl_qbuffer, lcl_dbuffer); + break; + case RTE_ML_IO_TYPE_FP16: + ret = rte_ml_io_float16_to_float32( + model->addr.output[i].nb_elements, lcl_qbuffer, + lcl_dbuffer); + break; + default: + plt_err("Unsupported model_output_type[%u] : %u", i, + model->metadata.output[i].model_output_type); + ret = -ENOTSUP; + } + if (ret < 0) + return ret; + } + + lcl_qbuffer += model->addr.output[i].sz_q; + lcl_dbuffer += model->addr.output[i].sz_d; + } + + batch_id++; + if (batch_id < PLT_DIV_CEIL(nb_batches, model->batch_size)) + goto next_batch; + + return 0; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, @@ -1006,4 +1155,6 @@ struct rte_ml_dev_ops cn10k_ml_ops = { /* I/O ops */ .io_input_size_get = cn10k_ml_io_input_size_get, .io_output_size_get = cn10k_ml_io_output_size_get, + .io_quantize = cn10k_ml_io_quantize, + .io_dequantize = cn10k_ml_io_dequantize, }; From patchwork Fri Mar 10 08:20:00 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124976 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 93EC541DD0; Fri, 10 Mar 2023 09:23:36 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 0F36A42F92; Fri, 10 Mar 2023 09:21:04 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id F11E742C24 for ; Fri, 10 Mar 2023 09:20:32 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdG021875 for ; Fri, 10 Mar 2023 00:20:32 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=GCAJXzyKByo3IyFwqK2tSSbRxaSwe6D9QYTHFXHzsXE=; b=XLuwbZMXJ3gIeIXc/EU3IB4L1/PEFAFAj6khT3Qi2QJpYNezxkTQS0/RiXYdeajCUYC3 +RLXlyF3cY30/x1GHpdOU0CESmCoSCTWhyNEsszZJArILsY1G/QlcBg1eRPcJncbypsW PN6llQG0eJLq5uiN2Hgy7c5HQJin6sKVy6tRy9uLdTEbaT+VcdvtQxctapFMBnT7n0ml UO4xnJLM9IJ18foGbZe1qzMHqM2arI9/pHW79q4+5pq1at44JJTmesrf7gDQk2vO+Fjg 8fRqoGJ2s6OP1SEWJF8NObAFvKOpEQbtTRTP1razjZ/66YU/u8cMuKV8LTPF28weozOT dA== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-15 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:32 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 4A1FF5C68E4; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 24/39] ml/cnxk: enable support to dump device debug info Date: Fri, 10 Mar 2023 00:20:00 -0800 Message-ID: <20230310082015.20200-25-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: DISPlb7n5OKAIGmfVh4mI2nCYRgUJSrz X-Proofpoint-ORIG-GUID: DISPlb7n5OKAIGmfVh4mI2nCYRgUJSrz X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support to dump device debug information. Debug info on cn10k device includes model state info, OCM usage info, firmware debug and exception buffer. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ocm.c | 51 +++++++++ drivers/ml/cnxk/cn10k_ml_ocm.h | 1 + drivers/ml/cnxk/cn10k_ml_ops.c | 189 +++++++++++++++++++++++++++++++++ 3 files changed, 241 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c index c3e4de3e9c..0b04fcc2da 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.c +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -458,3 +458,54 @@ cn10k_ml_ocm_free_pages(struct rte_ml_dev *dev, uint16_t model_id) } } } + +static void +cn10k_ml_ocm_pagemask_to_str(struct cn10k_ml_ocm_tile_info *tile_info, uint16_t nwords, char *str) +{ + char *p = str; + int word; + + /* add prefix 0x */ + *p++ = '0'; + *p++ = 'x'; + + /* build one word at a time */ + for (word = nwords - 1; word >= 0; word--) { + sprintf(p, "%02X", tile_info->ocm_mask[word]); + p += 2; + } + + /* terminate */ + *p++ = 0; +} + +void +cn10k_ml_ocm_print(struct rte_ml_dev *dev, FILE *fp) +{ + char str[ML_CN10K_OCM_NUMPAGES / 4 + 2]; /* nibbles + prefix '0x' */ + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + uint8_t tile_id; + uint8_t word_id; + int wb_pages; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + + fprintf(fp, "OCM State:\n"); + for (tile_id = 0; tile_id < ocm->num_tiles; tile_id++) { + cn10k_ml_ocm_pagemask_to_str(&ocm->tile_ocm_info[tile_id], ocm->mask_words, str); + + wb_pages = 0 - ocm->tile_ocm_info[tile_id].scratch_pages; + for (word_id = 0; word_id < ML_CN10K_OCM_MASKWORDS; word_id++) + wb_pages += + __builtin_popcount(ocm->tile_ocm_info[tile_id].ocm_mask[word_id]); + + fprintf(fp, + "tile = %2u, scratch_pages = %4u," + " wb_pages = %4d, last_wb_page = %4d," + " pagemask = %s\n", + tile_id, ocm->tile_ocm_info[tile_id].scratch_pages, wb_pages, + ocm->tile_ocm_info[tile_id].last_wb_page, str); + } +} diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h index 32c9b17afc..0c7172a671 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.h +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -83,5 +83,6 @@ int cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16 void cn10k_ml_ocm_reserve_pages(struct rte_ml_dev *dev, uint16_t model_id, uint64_t tilemask, int wb_page_start, uint16_t wb_pages, uint16_t scratch_pages); void cn10k_ml_ocm_free_pages(struct rte_ml_dev *dev, uint16_t model_id); +void cn10k_ml_ocm_print(struct rte_ml_dev *dev, FILE *fp); #endif /* _CN10K_ML_OCM_H_ */ diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 231c9b340b..2d7d760536 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -14,10 +14,25 @@ /* ML model macros */ #define CN10K_ML_MODEL_MEMZONE_NAME "ml_cn10k_model_mz" +/* Debug print width */ +#define STR_LEN 12 +#define FIELD_LEN 16 +#define LINE_LEN 90 + /* ML Job descriptor flags */ #define ML_FLAGS_POLL_COMPL BIT(0) #define ML_FLAGS_SSO_COMPL BIT(1) +static void +print_line(FILE *fp, int len) +{ + int i; + + for (i = 0; i < len; i++) + fprintf(fp, "-"); + fprintf(fp, "\n"); +} + static void qp_memzone_name_get(char *name, int size, int dev_id, int qp_id) { @@ -116,6 +131,102 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des return NULL; } +static void +cn10k_ml_model_print(struct rte_ml_dev *dev, uint16_t model_id, FILE *fp) +{ + + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_ocm *ocm; + char str[STR_LEN]; + uint8_t i; + + mldev = dev->data->dev_private; + ocm = &mldev->ocm; + model = dev->data->models[model_id]; + + /* Print debug info */ + print_line(fp, LINE_LEN); + fprintf(fp, " Model Information (%s)\n", model->metadata.model.name); + print_line(fp, LINE_LEN); + fprintf(fp, "%*s : %s\n", FIELD_LEN, "name", model->metadata.model.name); + fprintf(fp, "%*s : %u.%u.%u.%u\n", FIELD_LEN, "version", model->metadata.model.version[0], + model->metadata.model.version[1], model->metadata.model.version[2], + model->metadata.model.version[3]); + if (strlen(model->name) != 0) + fprintf(fp, "%*s : %s\n", FIELD_LEN, "debug_name", model->name); + fprintf(fp, "%*s : 0x%016lx\n", FIELD_LEN, "model", PLT_U64_CAST(model)); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "model_id", model->model_id); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "batch_size", model->metadata.model.batch_size); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "num_layers", model->metadata.model.num_layers); + + /* Print model state */ + if (model->state == ML_CN10K_MODEL_STATE_LOADED) + fprintf(fp, "%*s : %s\n", FIELD_LEN, "state", "loaded"); + if (model->state == ML_CN10K_MODEL_STATE_JOB_ACTIVE) + fprintf(fp, "%*s : %s\n", FIELD_LEN, "state", "job_active"); + if (model->state == ML_CN10K_MODEL_STATE_STARTED) + fprintf(fp, "%*s : %s\n", FIELD_LEN, "state", "started"); + + /* Print OCM status */ + fprintf(fp, "%*s : %" PRIu64 " bytes\n", FIELD_LEN, "wb_size", + model->metadata.model.ocm_wb_range_end - model->metadata.model.ocm_wb_range_start + + 1); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "wb_pages", model->model_mem_map.wb_pages); + fprintf(fp, "%*s : %" PRIu64 " bytes\n", FIELD_LEN, "scratch_size", + ocm->size_per_tile - model->metadata.model.ocm_tmp_range_floor); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "scratch_pages", model->model_mem_map.scratch_pages); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "num_tiles", + model->metadata.model.tile_end - model->metadata.model.tile_start + 1); + + if (model->state == ML_CN10K_MODEL_STATE_STARTED) { + fprintf(fp, "%*s : 0x%0*" PRIx64 "\n", FIELD_LEN, "tilemask", + ML_CN10K_OCM_NUMTILES / 4, model->model_mem_map.tilemask); + fprintf(fp, "%*s : 0x%x\n", FIELD_LEN, "ocm_wb_start", + model->model_mem_map.wb_page_start * ML_CN10K_OCM_PAGESIZE); + } + + fprintf(fp, "%*s : %u\n", FIELD_LEN, "num_inputs", model->metadata.model.num_input); + fprintf(fp, "%*s : %u\n", FIELD_LEN, "num_outputs", model->metadata.model.num_output); + fprintf(fp, "\n"); + + print_line(fp, LINE_LEN); + fprintf(fp, "%8s %16s %12s %18s %12s %14s\n", "input", "input_name", "input_type", + "model_input_type", "quantize", "format"); + print_line(fp, LINE_LEN); + for (i = 0; i < model->metadata.model.num_input; i++) { + fprintf(fp, "%8u ", i); + fprintf(fp, "%*s ", 16, model->metadata.input[i].input_name); + rte_ml_io_type_to_str(model->metadata.input[i].input_type, str, STR_LEN); + fprintf(fp, "%*s ", 12, str); + rte_ml_io_type_to_str(model->metadata.input[i].model_input_type, str, STR_LEN); + fprintf(fp, "%*s ", 18, str); + fprintf(fp, "%*s", 12, (model->metadata.input[i].quantize == 1 ? "Yes" : "No")); + rte_ml_io_format_to_str(model->metadata.input[i].shape.format, str, STR_LEN); + fprintf(fp, "%*s", 16, str); + fprintf(fp, "\n"); + } + fprintf(fp, "\n"); + + print_line(fp, LINE_LEN); + fprintf(fp, "%8s %16s %12s %18s %12s\n", "output", "output_name", "output_type", + "model_output_type", "dequantize"); + print_line(fp, LINE_LEN); + for (i = 0; i < model->metadata.model.num_output; i++) { + fprintf(fp, "%8u ", i); + fprintf(fp, "%*s ", 16, model->metadata.output[i].output_name); + rte_ml_io_type_to_str(model->metadata.output[i].output_type, str, STR_LEN); + fprintf(fp, "%*s ", 12, str); + rte_ml_io_type_to_str(model->metadata.output[i].model_output_type, str, STR_LEN); + fprintf(fp, "%*s ", 18, str); + fprintf(fp, "%*s", 12, (model->metadata.output[i].dequantize == 1 ? "Yes" : "No")); + fprintf(fp, "\n"); + } + fprintf(fp, "\n"); + print_line(fp, LINE_LEN); + fprintf(fp, "\n"); +} + static void cn10k_ml_prep_sp_job_descriptor(struct cn10k_ml_dev *mldev, struct cn10k_ml_model *model, struct cn10k_ml_req *req, enum cn10k_ml_job_type job_type) @@ -498,6 +609,83 @@ cn10k_ml_dev_queue_pair_setup(struct rte_ml_dev *dev, uint16_t queue_pair_id, return 0; } +static int +cn10k_ml_dev_dump(struct rte_ml_dev *dev, FILE *fp) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_fw *fw; + + uint32_t head_loc; + uint32_t tail_loc; + uint16_t model_id; + uint32_t bufsize; + char *head_ptr; + int core_id; + + if (roc_env_is_asim()) + return 0; + + mldev = dev->data->dev_private; + fw = &mldev->fw; + + /* Dump model info */ + for (model_id = 0; model_id < dev->data->nb_models; model_id++) { + model = dev->data->models[model_id]; + if (model != NULL) { + cn10k_ml_model_print(dev, model_id, fp); + fprintf(fp, "\n"); + } + } + + /* Dump OCM state */ + cn10k_ml_ocm_print(dev, fp); + + /* Dump debug buffer */ + for (core_id = 0; core_id <= 1; core_id++) { + bufsize = fw->req->jd.fw_load.debug.debug_buffer_size; + if (core_id == 0) { + head_loc = roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_DBG_BUFFER_HEAD_C0); + tail_loc = roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_DBG_BUFFER_TAIL_C0); + head_ptr = PLT_PTR_CAST(fw->req->jd.fw_load.debug.core0_debug_ptr); + head_ptr = roc_ml_addr_mlip2ap(&mldev->roc, head_ptr); + } else { + head_loc = roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_DBG_BUFFER_HEAD_C1); + tail_loc = roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_DBG_BUFFER_TAIL_C1); + head_ptr = PLT_PTR_CAST(fw->req->jd.fw_load.debug.core1_debug_ptr); + head_ptr = roc_ml_addr_mlip2ap(&mldev->roc, head_ptr); + } + if (head_loc < tail_loc) { + fprintf(fp, "%.*s\n", tail_loc - head_loc, &head_ptr[head_loc]); + } else if (head_loc >= tail_loc + 1) { + fprintf(fp, "%.*s\n", bufsize - tail_loc, &head_ptr[head_loc]); + fprintf(fp, "%.*s\n", tail_loc, &head_ptr[0]); + } + } + + /* Dump exception info */ + for (core_id = 0; core_id <= 1; core_id++) { + bufsize = fw->req->jd.fw_load.debug.exception_state_size; + if ((core_id == 0) && + (roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C0) != 0)) { + head_ptr = PLT_PTR_CAST(fw->req->jd.fw_load.debug.core0_exception_buffer); + fprintf(fp, "ML_SCRATCH_EXCEPTION_SP_C0 = 0x%016lx", + roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C0)); + head_ptr = roc_ml_addr_mlip2ap(&mldev->roc, head_ptr); + fprintf(fp, "%.*s", bufsize, head_ptr); + } else if ((core_id == 1) && + (roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C1) != 0)) { + head_ptr = PLT_PTR_CAST(fw->req->jd.fw_load.debug.core1_exception_buffer); + fprintf(fp, "ML_SCRATCH_EXCEPTION_SP_C1 = 0x%016lx", + roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C1)); + head_ptr = roc_ml_addr_mlip2ap(&mldev->roc, head_ptr); + fprintf(fp, "%.*s", bufsize, head_ptr); + } + } + + return 0; +} + int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint16_t *model_id) { @@ -1139,6 +1327,7 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .dev_close = cn10k_ml_dev_close, .dev_start = cn10k_ml_dev_start, .dev_stop = cn10k_ml_dev_stop, + .dev_dump = cn10k_ml_dev_dump, /* Queue-pair handling ops */ .dev_queue_pair_setup = cn10k_ml_dev_queue_pair_setup, From patchwork Fri Mar 10 08:20:01 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124975 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id DB94B41D3D; Fri, 10 Mar 2023 09:23:31 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id EA5E242F84; Fri, 10 Mar 2023 09:21:02 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 5458942C24 for ; Fri, 10 Mar 2023 09:20:32 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdF021875 for ; Fri, 10 Mar 2023 00:20:31 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=Ar2vR9cjXlqN28pPjqE706UZSMuZZm0oXKz9QCaUprI=; b=Yw01rRz4jjpleehf5yKLz94VAcEvj1Yd/fK/F/PvbbrTXmFpBui3lk9AvIptOB4rb3o7 3xm5dJeyA9X3oCjgEHZijFE88zkM2ryg+uzwspoCxDEOh0Ptp5eHB46KRgV5IHeX0gan DFESpeLRaPSCGLit1Jf5dJuvd7uelykuW31lB2ZS4rFh17WSQ8TXxjoJLCaUQ8gvx0H8 cn1n3Qtw6SPFOL6mt2cilM8U9keKwEdkCvi/H2Uqzc4wUce7yyHX7mgXvylFsm+FzWVA JyJXsXxQ7Q3LKbkgAEPkBcYf6fuH1XaZsKsTlPVNFqlkfxTbys9IFYS6vQQVaQNvl16M 8g== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-14 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:31 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 73AD25C68E6; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 25/39] ml/cnxk: add driver support for device selftest Date: Fri, 10 Mar 2023 00:20:01 -0800 Message-ID: <20230310082015.20200-26-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: lB_-f3rtiu8TG0ax95sk2oFoHiyoZ-A1 X-Proofpoint-ORIG-GUID: lB_-f3rtiu8TG0ax95sk2oFoHiyoZ-A1 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support for device selftest. Device selftest includes checking the status of firmware. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 57 ++++++++++++++++++++++++++++++++++ 1 file changed, 57 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 2d7d760536..2fa0522faf 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -686,6 +686,62 @@ cn10k_ml_dev_dump(struct rte_ml_dev *dev, FILE *fp) return 0; } +static int +cn10k_ml_dev_selftest(struct rte_ml_dev *dev) +{ + const struct plt_memzone *mz; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_req *req; + uint64_t timeout_cycle; + bool timeout; + int ret; + + mldev = dev->data->dev_private; + mz = plt_memzone_reserve_aligned("dev_selftest", sizeof(struct cn10k_ml_req), 0, + ML_CN10K_ALIGN_SIZE); + if (mz == NULL) { + plt_err("Could not allocate reserved memzone"); + return -ENOMEM; + } + req = mz->addr; + + /* Prepare load completion structure */ + memset(&req->jd, 0, sizeof(struct cn10k_ml_jd)); + req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&req->status); + req->jd.hdr.job_type = ML_CN10K_JOB_TYPE_FIRMWARE_SELFTEST; + req->jd.hdr.result = roc_ml_addr_ap2mlip(&mldev->roc, &req->result); + req->jd.fw_load.flags = cn10k_ml_fw_flags_get(&mldev->fw); + plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + plt_wmb(); + + /* Enqueue firmware selftest request through scratch registers */ + timeout = true; + timeout_cycle = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + roc_ml_scratch_enqueue(&mldev->roc, &req->jd); + + plt_rmb(); + do { + if (roc_ml_scratch_is_done_bit_set(&mldev->roc) && + (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH)) { + timeout = false; + break; + } + } while (plt_tsc_cycles() < timeout_cycle); + + /* Check firmware selftest status, clean-up and exit */ + ret = 0; + if (timeout) { + ret = -ETIME; + } else { + if (req->result.error_code != 0) + ret = -1; + } + + plt_memzone_free(mz); + + return ret; +} + int cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint16_t *model_id) { @@ -1328,6 +1384,7 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .dev_start = cn10k_ml_dev_start, .dev_stop = cn10k_ml_dev_stop, .dev_dump = cn10k_ml_dev_dump, + .dev_selftest = cn10k_ml_dev_selftest, /* Queue-pair handling ops */ .dev_queue_pair_setup = cn10k_ml_dev_queue_pair_setup, From patchwork Fri Mar 10 08:20:02 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124966 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id B27B141D3D; Fri, 10 Mar 2023 09:22:39 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 94F1442D98; Fri, 10 Mar 2023 09:20:52 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 3A92B4113C for ; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1J009730 for ; Fri, 10 Mar 2023 00:20:29 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=VKvAbEOA9CFaQgKQjBloR4xLNlsc3sp7vp098xbs7lo=; b=QjZRz6GJp2QPn/Vy1F+KpTk3P3MNqI4SuI+w/iJrDMOXCmD/NP8ZDq0U6yDyIo5frv8F WjHA1fe62Pto9cgyRz/3JVJEwLY+1XGEUcUPLQw7a8R42isXysZ64UQgHc98h8sG1eht gm9fJgnoq+AZvox/IzpAtkWW8eBD3VZWoLDGxDOO02ZcCtW7NzKGqTHNV7EBcA7A17WJ DHijfgpEfdbTvguARNLiTtPbGPkVcoOCet5z0L/kvlNpd2F7aJUd/VW/NX5CIe7TJsVv pQipIvWm+BZok6AWIjj8W78GBTro1JOo+RlTK5Gps8P3EvC/HDTna7Gm6xV11C478Qdy 9Q== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-14 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:29 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 9E0B25C68E8; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 26/39] ml/cnxk: enqueue a burst of inference requests Date: Fri, 10 Mar 2023 00:20:02 -0800 Message-ID: <20230310082015.20200-27-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: TH05vSNzxABpTDeanA6Xm9LUm59UbckX X-Proofpoint-ORIG-GUID: TH05vSNzxABpTDeanA6Xm9LUm59UbckX X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled driver support to enqueue a burst of inference requests to ML device. Enqueue uses internal ML request structure to queue the inferences and job completion through polling. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 96 ++++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.h | 7 +++ 2 files changed, 103 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 2fa0522faf..f024487fc1 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -285,6 +285,28 @@ cn10k_ml_prep_sp_job_descriptor(struct cn10k_ml_dev *mldev, struct cn10k_ml_mode } } +static __rte_always_inline void +cn10k_ml_prep_fp_job_descriptor(struct rte_ml_dev *dev, struct cn10k_ml_req *req, + struct rte_ml_op *op) +{ + struct cn10k_ml_dev *mldev; + + mldev = dev->data->dev_private; + + req->jd.hdr.jce.w0.u64 = 0; + req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&req->status); + req->jd.hdr.model_id = op->model_id; + req->jd.hdr.job_type = ML_CN10K_JOB_TYPE_MODEL_RUN; + req->jd.hdr.fp_flags = ML_FLAGS_POLL_COMPL; + req->jd.hdr.sp_flags = 0x0; + req->jd.hdr.result = roc_ml_addr_ap2mlip(&mldev->roc, &req->result); + req->jd.model_run.input_ddr_addr = + PLT_U64_CAST(roc_ml_addr_ap2mlip(&mldev->roc, op->input.addr)); + req->jd.model_run.output_ddr_addr = + PLT_U64_CAST(roc_ml_addr_ap2mlip(&mldev->roc, op->output.addr)); + req->jd.model_run.num_batches = op->nb_batches; +} + static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { @@ -450,6 +472,8 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c rte_spinlock_init(&ocm->lock); + dev->enqueue_burst = cn10k_ml_enqueue_burst; + mldev->nb_models_loaded = 0; mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; @@ -1376,6 +1400,78 @@ cn10k_ml_io_dequantize(struct rte_ml_dev *dev, uint16_t model_id, uint16_t nb_ba return 0; } +static __rte_always_inline void +queue_index_advance(uint64_t *index, uint64_t nb_desc) +{ + *index = (*index + 1) % nb_desc; +} + +static __rte_always_inline uint64_t +queue_pending_count(uint64_t head, uint64_t tail, uint64_t nb_desc) +{ + return (nb_desc + head - tail) % nb_desc; +} + +static __rte_always_inline uint64_t +queue_free_count(uint64_t head, uint64_t tail, uint64_t nb_desc) +{ + return nb_desc - queue_pending_count(head, tail, nb_desc) - 1; +} + +__rte_hot uint16_t +cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, + uint16_t nb_ops) +{ + struct cn10k_ml_queue *queue; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_req *req; + struct cn10k_ml_qp *qp; + struct rte_ml_op *op; + + uint16_t count; + uint64_t head; + bool enqueued; + + mldev = dev->data->dev_private; + qp = dev->data->queue_pairs[qp_id]; + queue = &qp->queue; + + head = queue->head; + nb_ops = PLT_MIN(nb_ops, queue_free_count(head, queue->tail, qp->nb_desc)); + count = 0; + + if (unlikely(nb_ops == 0)) + return 0; + +enqueue_req: + op = ops[count]; + req = &queue->reqs[head]; + + cn10k_ml_prep_fp_job_descriptor(dev, req, op); + + memset(&req->result, 0, sizeof(struct cn10k_ml_result)); + req->result.user_ptr = op->user_ptr; + + plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + enqueued = roc_ml_jcmdq_enqueue_lf(&mldev->roc, &req->jcmd); + if (unlikely(!enqueued)) + goto jcmdq_full; + + req->timeout = plt_tsc_cycles() + queue->wait_cycles; + req->op = op; + + queue_index_advance(&head, qp->nb_desc); + count++; + + if (count < nb_ops) + goto enqueue_req; + +jcmdq_full: + queue->head = head; + + return count; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 22576b93c0..a1724f6156 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -28,6 +28,9 @@ struct cn10k_ml_req { /* Timeout cycle */ uint64_t timeout; + + /* Op */ + struct rte_ml_op *op; } __rte_aligned(ROC_ALIGN); /* Request queue */ @@ -67,4 +70,8 @@ int cn10k_ml_model_unload(struct rte_ml_dev *dev, uint16_t model_id); int cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id); int cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id); +/* Fast-path ops */ +__rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, + struct rte_ml_op **ops, uint16_t nb_ops); + #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:20:03 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124977 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id A424441DD0; Fri, 10 Mar 2023 09:23:42 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 2E81A42C4D; Fri, 10 Mar 2023 09:21:05 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 6D04A42C24 for ; Fri, 10 Mar 2023 09:20:33 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdH021875 for ; Fri, 10 Mar 2023 00:20:32 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=0OfsuUQcnMiZiUnFbfMx7ZJEFGYb6o65Lwqc3WUTYpg=; b=BZO0eZXMn4k0egTJnKdPvv9qm5Ilw1S1IkswgOBwcoUu7+XtUu7pD7TJfnkEyAHFJol1 jfet7dkTStY1GfD+4TjxiXmoDgNUNjBkqSt8PvEmEl3ffPZ906+MzV+apFuSSYzR+Q8w rKG7gxCHXgeDReRfudugWvas1I+9FmqZ8h858Hlar7lBo0XdQnWfjQEgtV8LeTBJTyiT WUpfGOQb3ikWkgyJ/bG3JCIeZuvw8QlwFvimzkTzMkK44VFHS4i2hQZIoQFmI0kz9Ug9 leWzfEaW7q43O4yVVGZ7oKBCeeMBcOxJC3aup13HbzYyOi9DY2Il/v09k4gfM0WpEiyC 8Q== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-16 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:32 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:23 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id C8FA05C68EB; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 27/39] ml/cnxk: dequeue a burst of inference requests Date: Fri, 10 Mar 2023 00:20:03 -0800 Message-ID: <20230310082015.20200-28-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: NvVnuX4GijTsqeG_FLRPK19UERLaGYZb X-Proofpoint-ORIG-GUID: NvVnuX4GijTsqeG_FLRPK19UERLaGYZb X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled driver support to dequeue inference requests from internal queue. Dequeue checks for request completion by polling the status field of the job request. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 61 ++++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.h | 2 ++ 2 files changed, 63 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index f024487fc1..51f1c92a8d 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -473,6 +473,7 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c rte_spinlock_init(&ocm->lock); dev->enqueue_burst = cn10k_ml_enqueue_burst; + dev->dequeue_burst = cn10k_ml_dequeue_burst; mldev->nb_models_loaded = 0; mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; @@ -1418,6 +1419,23 @@ queue_free_count(uint64_t head, uint64_t tail, uint64_t nb_desc) return nb_desc - queue_pending_count(head, tail, nb_desc) - 1; } +static __rte_always_inline void +cn10k_ml_result_update(struct rte_ml_dev *dev, int qp_id, struct cn10k_ml_result *result, + struct rte_ml_op *op) +{ + PLT_SET_USED(dev); + PLT_SET_USED(qp_id); + + op->impl_opaque = result->error_code; + + if (likely(result->error_code == 0)) + op->status = RTE_ML_OP_STATUS_SUCCESS; + else + op->status = RTE_ML_OP_STATUS_ERROR; + + op->user_ptr = result->user_ptr; +} + __rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, uint16_t nb_ops) @@ -1472,6 +1490,49 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op return count; } +__rte_hot uint16_t +cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, + uint16_t nb_ops) +{ + struct cn10k_ml_queue *queue; + struct cn10k_ml_req *req; + struct cn10k_ml_qp *qp; + + uint64_t status; + uint16_t count; + uint64_t tail; + + qp = dev->data->queue_pairs[qp_id]; + queue = &qp->queue; + + tail = queue->tail; + nb_ops = PLT_MIN(nb_ops, queue_pending_count(queue->head, tail, qp->nb_desc)); + count = 0; + + if (unlikely(nb_ops == 0)) + goto empty_or_active; + +dequeue_req: + req = &queue->reqs[tail]; + status = plt_read64(&req->status); + if (unlikely(status != ML_CN10K_POLL_JOB_FINISH)) + goto empty_or_active; + + cn10k_ml_result_update(dev, qp_id, &req->result, req->op); + ops[count] = req->op; + + queue_index_advance(&tail, qp->nb_desc); + count++; + + if (count < nb_ops) + goto dequeue_req; + +empty_or_active: + queue->tail = tail; + + return count; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index a1724f6156..f6aab4a609 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -73,5 +73,7 @@ int cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id); /* Fast-path ops */ __rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, uint16_t nb_ops); +__rte_hot uint16_t cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, + struct rte_ml_op **ops, uint16_t nb_ops); #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:20:04 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124978 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 7EB8B41DD0; Fri, 10 Mar 2023 09:23:47 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 4366242FA2; Fri, 10 Mar 2023 09:21:06 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id E2C0A42C24 for ; Fri, 10 Mar 2023 09:20:33 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdI021875 for ; Fri, 10 Mar 2023 00:20:33 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=chKjMBg3yM5P1HINXXujah3hK3XVMn61vvJ47XuTeNM=; b=F1bgJk0JNlxxI5mfjlBE2EerOAZhMRNcF7N960shTcIjpvZL5mLTSWAILwFoWOTFQXmM 4PJfRJ6PxH+Cm3/zybpAqeldGiZeUPAK1q4N865ywOnXsEG1u0jrOwNWNrnvltXfpVXM aVR7e/UvT2e2FUOxktMmsU0vne9EeysVotwCuZKzOL3Ik0msA+iZPPfp40GWktzEVw8a m1JIFFeV7+Lb2UXbhRlJUVq1Z59ZvOROTbSyxN8HVfWak62c6W69oTLKoE9dqXbHwUSS wCOcPxvVCpNi7QBqfG9L+Q5f+68dDTNAUt9VQd3XE2oWec8+C5egF5U85u6UWpL3hcBU 8A== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-17 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:33 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:23 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id F0F855C68EE; Fri, 10 Mar 2023 00:20:22 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 28/39] ml/cnxk: add internal function for sync mode run Date: Fri, 10 Mar 2023 00:20:04 -0800 Message-ID: <20230310082015.20200-29-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: Oi7eNntbzeHQToiK-nO8t5JK2phN4Azj X-Proofpoint-ORIG-GUID: Oi7eNntbzeHQToiK-nO8t5JK2phN4Azj X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added internal function to execute ML inference requests in synchronous mode. Sync mode inference execution is used to launch inference requests without using a queue-pair. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 53 ++++++++++++++++++++++++++++++++++ drivers/ml/cnxk/cn10k_ml_ops.h | 1 + 2 files changed, 54 insertions(+) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 51f1c92a8d..87778c37bb 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -1533,6 +1533,59 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op return count; } +__rte_hot int +cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + struct cn10k_ml_req *req; + bool timeout; + int ret = 0; + + mldev = dev->data->dev_private; + model = dev->data->models[op->model_id]; + req = model->req; + + cn10k_ml_prep_fp_job_descriptor(dev, req, op); + + memset(&req->result, 0, sizeof(struct cn10k_ml_result)); + req->result.user_ptr = op->user_ptr; + + plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + req->jcmd.w1.s.jobptr = PLT_U64_CAST(&req->jd); + + timeout = true; + req->timeout = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); + do { + if (roc_ml_jcmdq_enqueue_lf(&mldev->roc, &req->jcmd)) { + req->op = op; + timeout = false; + break; + } + } while (plt_tsc_cycles() < req->timeout); + + if (timeout) { + ret = -EBUSY; + goto error_enqueue; + } + + timeout = true; + do { + if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { + timeout = false; + break; + } + } while (plt_tsc_cycles() < req->timeout); + + if (timeout) + ret = -ETIME; + else + cn10k_ml_result_update(dev, -1, &req->result, req->op); + +error_enqueue: + return ret; +} + struct rte_ml_dev_ops cn10k_ml_ops = { /* Device control ops */ .dev_info_get = cn10k_ml_dev_info_get, diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index f6aab4a609..7c35bf7539 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -75,5 +75,6 @@ __rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id struct rte_ml_op **ops, uint16_t nb_ops); __rte_hot uint16_t cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, uint16_t nb_ops); +__rte_hot int cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op); #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:20:05 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124980 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 26AD441DD0; Fri, 10 Mar 2023 09:23:58 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 4C6E642FA8; Fri, 10 Mar 2023 09:21:08 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 0E6DB42D1D for ; Fri, 10 Mar 2023 09:20:34 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdK021875 for ; Fri, 10 Mar 2023 00:20:34 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=f2xAjJ4TgaG2zcQDqpA3ovW7IPwgrxy2g7o6oBsWNOM=; b=flRD8irR7JbCccHQkgIM7KLrgCxZoB3lN7X5AJR+ZNIpSZrVaRuabHCrGhke+rY/Uffd PCt9kR7jsJlnOfrRLheFqbQ50QCNZtR1MDHCPfG+CpNV2WPtUq+lkZqNs8ccE0yRXYOv JAXsPogFHzPignpnsSLxBhKFaC1GBWuotemMEZqFOZamid6pnl4/IZZg0pM7HzivaGOp d9fI1UqpGC32zNCiSjqvdb9kvHSjFj5KWvtv8eefxNFa5y/8PVPMZmuVw5Pi2fXBY1zq 4/DcyWXqrr1kNucHw+y79uzq0Fy070es6hwr0XxKZq7ptDxfEHXmtofknJb7jCLV1UrH PA== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-19 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:34 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 282735C68EF; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 29/39] ml/cnxk: enable support for firmware error codes Date: Fri, 10 Mar 2023 00:20:05 -0800 Message-ID: <20230310082015.20200-30-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: A7BqKijoVYDKtMy1uB0JYPk4GXTsl5Vx X-Proofpoint-ORIG-GUID: A7BqKijoVYDKtMy1uB0JYPk4GXTsl5Vx X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support for error handling. Added error types and subtypes supported by ML firmware. Enabled support to get device specific error code and message for a completed ML request. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 4 +- drivers/ml/cnxk/cn10k_ml_dev.h | 50 +++++++++++++- drivers/ml/cnxk/cn10k_ml_ops.c | 117 ++++++++++++++++++++++++++++++--- drivers/ml/cnxk/cn10k_ml_ops.h | 2 + 4 files changed, 160 insertions(+), 13 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 837f006bf0..76ed853a3c 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -261,7 +261,7 @@ cn10k_ml_fw_load_asim(struct cn10k_ml_fw *fw) } while (plt_tsc_cycles() < timeout_cycle); /* Check firmware load status, clean-up and exit on failure. */ - if ((!timeout) && (fw->req->result.error_code == 0)) { + if ((!timeout) && (fw->req->result.error_code.u64 == 0)) { cn10k_ml_fw_print_info(fw); } else { /* Set ML to disable new jobs */ @@ -452,7 +452,7 @@ cn10k_ml_fw_load_cn10ka(struct cn10k_ml_fw *fw, void *buffer, uint64_t size) } while (plt_tsc_cycles() < timeout_cycle); /* Check firmware load status, clean-up and exit on failure. */ - if ((!timeout) && (fw->req->result.error_code == 0)) { + if ((!timeout) && (fw->req->result.error_code.u64 == 0)) { cn10k_ml_fw_print_info(fw); } else { /* Set ML to disable new jobs */ diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 8f6bc24370..604a200e26 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -64,6 +64,54 @@ enum cn10k_ml_dev_state { ML_CN10K_DEV_STATE_CLOSED }; +/* Error types enumeration */ +enum cn10k_ml_error_etype { + /* 0x0 */ ML_ETYPE_NO_ERROR = 0, /* No error */ + /* 0x1 */ ML_ETYPE_FW_NONFATAL, /* Firmware non-fatal error */ + /* 0x2 */ ML_ETYPE_HW_NONFATAL, /* Hardware non-fatal error */ + /* 0x3 */ ML_ETYPE_HW_FATAL, /* Hardware fatal error */ + /* 0x4 */ ML_ETYPE_HW_WARNING, /* Hardware warning */ + /* 0x5 */ ML_ETYPE_DRIVER, /* Driver specific error */ + /* 0x6 */ ML_ETYPE_UNKNOWN, /* Unknown error */ +}; + +/* Firmware non-fatal error sub-type */ +enum cn10k_ml_error_stype_fw_nf { + /* 0x0 */ ML_FW_ERR_NOERR = 0, /* No error */ + /* 0x1 */ ML_FW_ERR_UNLOAD_ID_NOT_FOUND, /* Model ID not found during load */ + /* 0x2 */ ML_FW_ERR_LOAD_LUT_OVERFLOW, /* Lookup table overflow at load */ + /* 0x3 */ ML_FW_ERR_ID_IN_USE, /* Model ID already in use */ + /* 0x4 */ ML_FW_ERR_INVALID_TILEMASK, /* Invalid OCM tilemask */ + /* 0x5 */ ML_FW_ERR_RUN_LUT_OVERFLOW, /* Lookup table overflow at run */ + /* 0x6 */ ML_FW_ERR_RUN_ID_NOT_FOUND, /* Model ID not found during run */ + /* 0x7 */ ML_FW_ERR_COMMAND_NOTSUP, /* Unsupported command */ + /* 0x8 */ ML_FW_ERR_DDR_ADDR_RANGE, /* DDR address out of range */ + /* 0x9 */ ML_FW_ERR_NUM_BATCHES_INVALID, /* Invalid number of batches */ + /* 0xA */ ML_FW_ERR_INSSYNC_TIMEOUT, /* INS sync timeout */ +}; + +/* Driver error sub-type */ +enum cn10k_ml_error_stype_driver { + /* 0x0 */ ML_DRIVER_ERR_NOERR = 0, /* No error */ + /* 0x1 */ ML_DRIVER_ERR_UNKNOWN, /* Unable to determine error sub-type */ + /* 0x2 */ ML_DRIVER_ERR_EXCEPTION, /* Firmware exception */ + /* 0x3 */ ML_DRIVER_ERR_FW_ERROR, /* Unknown firmware error */ +}; + +/* Error structure */ +union cn10k_ml_error_code { + struct { + /* Error type */ + uint64_t etype : 4; + + /* Error sub-type */ + uint64_t stype : 60; + } s; + + /* WORD 0 */ + uint64_t u64; +}; + /* Firmware stats */ struct cn10k_ml_fw_stats { /* Firmware start cycle */ @@ -82,7 +130,7 @@ struct cn10k_ml_fw_stats { /* Result structure */ struct cn10k_ml_result { /* Job error code */ - uint64_t error_code; + union cn10k_ml_error_code error_code; /* Firmware stats */ struct cn10k_ml_fw_stats stats; diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 87778c37bb..23a9ca4ff2 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -23,6 +23,49 @@ #define ML_FLAGS_POLL_COMPL BIT(0) #define ML_FLAGS_SSO_COMPL BIT(1) +/* Error message length */ +#define ERRMSG_LEN 32 + +/* Error type database */ +static const struct cn10k_ml_etype_db { + enum cn10k_ml_error_etype etype; + char name[ERRMSG_LEN]; +} ml_etype_db[] = { + {ML_ETYPE_NO_ERROR, "NO_ERROR"}, {ML_ETYPE_FW_NONFATAL, "FW_NON_FATAL"}, + {ML_ETYPE_HW_NONFATAL, "HW_NON_FATAL"}, {ML_ETYPE_HW_FATAL, "HW_FATAL"}, + {ML_ETYPE_HW_WARNING, "HW_WARNING"}, {ML_ETYPE_DRIVER, "DRIVER_ERROR"}, + {ML_ETYPE_UNKNOWN, "UNKNOWN_ERROR"}, +}; + +/* Hardware non-fatal error subtype database */ +static const struct cn10k_ml_stype_db_hw_nf { + enum cn10k_ml_error_stype_fw_nf stype; + char msg[ERRMSG_LEN]; +} ml_stype_db_hw_nf[] = { + {ML_FW_ERR_NOERR, "NO ERROR"}, + {ML_FW_ERR_UNLOAD_ID_NOT_FOUND, "UNLOAD MODEL ID NOT FOUND"}, + {ML_FW_ERR_LOAD_LUT_OVERFLOW, "LOAD LUT OVERFLOW"}, + {ML_FW_ERR_ID_IN_USE, "MODEL ID IN USE"}, + {ML_FW_ERR_INVALID_TILEMASK, "INVALID TILEMASK"}, + {ML_FW_ERR_RUN_LUT_OVERFLOW, "RUN LUT OVERFLOW"}, + {ML_FW_ERR_RUN_ID_NOT_FOUND, "RUN MODEL ID NOT FOUND"}, + {ML_FW_ERR_COMMAND_NOTSUP, "COMMAND NOT SUPPORTED"}, + {ML_FW_ERR_DDR_ADDR_RANGE, "DDR ADDRESS OUT OF RANGE"}, + {ML_FW_ERR_NUM_BATCHES_INVALID, "INVALID BATCHES"}, + {ML_FW_ERR_INSSYNC_TIMEOUT, "INSSYNC TIMEOUT"}, +}; + +/* Driver error subtype database */ +static const struct cn10k_ml_stype_db_driver { + enum cn10k_ml_error_stype_driver stype; + char msg[ERRMSG_LEN]; +} ml_stype_db_driver[] = { + {ML_DRIVER_ERR_NOERR, "NO ERROR"}, + {ML_DRIVER_ERR_UNKNOWN, "UNKNOWN ERROR"}, + {ML_DRIVER_ERR_EXCEPTION, "FW EXCEPTION"}, + {ML_DRIVER_ERR_FW_ERROR, "UNKNOWN FIRMWARE ERROR"}, +}; + static void print_line(FILE *fp, int len) { @@ -474,6 +517,7 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c dev->enqueue_burst = cn10k_ml_enqueue_burst; dev->dequeue_burst = cn10k_ml_dequeue_burst; + dev->op_error_get = cn10k_ml_op_error_get; mldev->nb_models_loaded = 0; mldev->state = ML_CN10K_DEV_STATE_CONFIGURED; @@ -758,7 +802,7 @@ cn10k_ml_dev_selftest(struct rte_ml_dev *dev) if (timeout) { ret = -ETIME; } else { - if (req->result.error_code != 0) + if (req->result.error_code.u64 != 0) ret = -1; } @@ -936,7 +980,7 @@ cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id) /* Prepare JD */ req = model->req; cn10k_ml_prep_sp_job_descriptor(mldev, model, req, ML_CN10K_JOB_TYPE_MODEL_START); - req->result.error_code = 0x0; + req->result.error_code.u64 = 0x0; req->result.user_ptr = NULL; plt_write64(ML_CN10K_POLL_JOB_START, &req->status); @@ -1017,7 +1061,7 @@ cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id) if (job_dequeued) { if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { - if (req->result.error_code == 0) + if (req->result.error_code.u64 == 0) ret = 0; else ret = -1; @@ -1079,7 +1123,7 @@ cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id) /* Prepare JD */ req = model->req; cn10k_ml_prep_sp_job_descriptor(mldev, model, req, ML_CN10K_JOB_TYPE_MODEL_STOP); - req->result.error_code = 0x0; + req->result.error_code.u64 = 0x0; req->result.user_ptr = NULL; plt_write64(ML_CN10K_POLL_JOB_START, &req->status); @@ -1134,7 +1178,7 @@ cn10k_ml_model_stop(struct rte_ml_dev *dev, uint16_t model_id) if (job_dequeued) { if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { - if (req->result.error_code == 0x0) + if (req->result.error_code.u64 == 0x0) ret = 0; else ret = -1; @@ -1426,12 +1470,30 @@ cn10k_ml_result_update(struct rte_ml_dev *dev, int qp_id, struct cn10k_ml_result PLT_SET_USED(dev); PLT_SET_USED(qp_id); - op->impl_opaque = result->error_code; + struct cn10k_ml_dev *mldev; - if (likely(result->error_code == 0)) + if (likely(result->error_code.u64 == 0)) { + op->impl_opaque = result->error_code.u64; op->status = RTE_ML_OP_STATUS_SUCCESS; - else + } else { + /* Handle driver error */ + if (result->error_code.s.etype == ML_ETYPE_DRIVER) { + mldev = dev->data->dev_private; + + /* Check for exception */ + if ((roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C0) != 0) || + (roc_ml_reg_read64(&mldev->roc, ML_SCRATCH_EXCEPTION_SP_C1) != 0)) + result->error_code.s.stype = ML_DRIVER_ERR_EXCEPTION; + else if ((roc_ml_reg_read64(&mldev->roc, ML_CORE_INT_LO) != 0) || + (roc_ml_reg_read64(&mldev->roc, ML_CORE_INT_HI) != 0)) + result->error_code.s.stype = ML_DRIVER_ERR_FW_ERROR; + else + result->error_code.s.stype = ML_DRIVER_ERR_UNKNOWN; + } + + op->impl_opaque = result->error_code.u64; op->status = RTE_ML_OP_STATUS_ERROR; + } op->user_ptr = result->user_ptr; } @@ -1468,6 +1530,7 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op cn10k_ml_prep_fp_job_descriptor(dev, req, op); memset(&req->result, 0, sizeof(struct cn10k_ml_result)); + req->result.error_code.s.etype = ML_ETYPE_UNKNOWN; req->result.user_ptr = op->user_ptr; plt_write64(ML_CN10K_POLL_JOB_START, &req->status); @@ -1515,8 +1578,12 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op dequeue_req: req = &queue->reqs[tail]; status = plt_read64(&req->status); - if (unlikely(status != ML_CN10K_POLL_JOB_FINISH)) - goto empty_or_active; + if (unlikely(status != ML_CN10K_POLL_JOB_FINISH)) { + if (plt_tsc_cycles() < req->timeout) + goto empty_or_active; + else /* Timeout, set indication of driver error */ + req->result.error_code.s.etype = ML_ETYPE_DRIVER; + } cn10k_ml_result_update(dev, qp_id, &req->result, req->op); ops[count] = req->op; @@ -1533,6 +1600,35 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op return count; } +__rte_hot int +cn10k_ml_op_error_get(struct rte_ml_dev *dev, struct rte_ml_op *op, struct rte_ml_op_error *error) +{ + union cn10k_ml_error_code *error_code; + char msg[RTE_ML_STR_MAX]; + + PLT_SET_USED(dev); + + error_code = (union cn10k_ml_error_code *)&op->impl_opaque; + + /* Copy error message */ + plt_strlcpy(msg, ml_etype_db[error_code->s.etype].name, sizeof(msg)); + + /* Copy sub error message */ + if (error_code->s.etype == ML_ETYPE_HW_NONFATAL) { + strcat(msg, " : "); + strcat(msg, ml_stype_db_hw_nf[error_code->s.stype].msg); + } + + if (error_code->s.etype == ML_ETYPE_DRIVER) { + strcat(msg, " : "); + strcat(msg, ml_stype_db_driver[error_code->s.stype].msg); + } + + plt_strlcpy(error->message, msg, sizeof(error->message)); + + return 0; +} + __rte_hot int cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) { @@ -1549,6 +1645,7 @@ cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) cn10k_ml_prep_fp_job_descriptor(dev, req, op); memset(&req->result, 0, sizeof(struct cn10k_ml_result)); + req->result.error_code.s.etype = ML_ETYPE_UNKNOWN; req->result.user_ptr = op->user_ptr; plt_write64(ML_CN10K_POLL_JOB_START, &req->status); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 7c35bf7539..1784900cff 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -75,6 +75,8 @@ __rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id struct rte_ml_op **ops, uint16_t nb_ops); __rte_hot uint16_t cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops, uint16_t nb_ops); +__rte_hot int cn10k_ml_op_error_get(struct rte_ml_dev *dev, struct rte_ml_op *op, + struct rte_ml_op_error *error); __rte_hot int cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op); #endif /* _CN10K_ML_OPS_H_ */ From patchwork Fri Mar 10 08:20:06 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124981 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 7417A41DD0; Fri, 10 Mar 2023 09:24:03 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 5FA9F42FB4; Fri, 10 Mar 2023 09:21:09 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 881A142D1D for ; Fri, 10 Mar 2023 09:20:35 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdL021875 for ; Fri, 10 Mar 2023 00:20:35 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=D4y33g2oviPLrfjSR9zcfXXMI8zlhLDX2QjNNJ+dCxg=; b=Ar+B2yfygvu/XNGItuJALf6DgpYnSQBIlo2Qr3Q+Ojm4l5qCLkdd6t/FddV8AHq6p6at XkyIEKrnEi63WVSygN2VEAjA/BspWMH14ug6JfUc2AnHhlPqJKxzZBYlZCK+CXBwjhTp KV9ZUDk97/lGxeQvnwjuV7L9IF1iya704LcfsYhwuf2gqJ0q6FqwZVXIsZl/C+gU/HRu YWgf5h2qYe3iG1b51oK9e7Tdxv0OU+hMLKtGIpasNKatL3cj7QrJYToFzLwOOctlZkDP f0/wUOsK1Uzibl7BKFQPUHYGrN4AzeivcZYl1sSS/Xd/KPwgxmo45DnHqEYAEttaPl8O 4Q== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-20 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:34 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 50A095C68F2; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 30/39] ml/cnxk: add support to get and reset device stats Date: Fri, 10 Mar 2023 00:20:06 -0800 Message-ID: <20230310082015.20200-31-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: pvAnvOyCSQIPjol2ijz1OHSbqZE3BEMP X-Proofpoint-ORIG-GUID: pvAnvOyCSQIPjol2ijz1OHSbqZE3BEMP X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support to get and reset ML device stats. Device stats include number of requests enqueued/dequeued and error count. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 55 ++++++++++++++++++++++++++++++++-- drivers/ml/cnxk/cn10k_ml_ops.h | 3 ++ 2 files changed, 55 insertions(+), 3 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 23a9ca4ff2..c38f018a50 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -159,6 +159,10 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des qp->queue.tail = 0; qp->queue.wait_cycles = ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); qp->nb_desc = nb_desc; + qp->stats.enqueued_count = 0; + qp->stats.dequeued_count = 0; + qp->stats.enqueue_err_count = 0; + qp->stats.dequeue_err_count = 0; /* Initialize job command */ for (i = 0; i < qp->nb_desc; i++) { @@ -678,6 +682,38 @@ cn10k_ml_dev_queue_pair_setup(struct rte_ml_dev *dev, uint16_t queue_pair_id, return 0; } +static int +cn10k_ml_dev_stats_get(struct rte_ml_dev *dev, struct rte_ml_dev_stats *stats) +{ + struct cn10k_ml_qp *qp; + int qp_id; + + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { + qp = dev->data->queue_pairs[qp_id]; + stats->enqueued_count += qp->stats.enqueued_count; + stats->dequeued_count += qp->stats.dequeued_count; + stats->enqueue_err_count += qp->stats.enqueue_err_count; + stats->dequeue_err_count += qp->stats.dequeue_err_count; + } + + return 0; +} + +static void +cn10k_ml_dev_stats_reset(struct rte_ml_dev *dev) +{ + struct cn10k_ml_qp *qp; + int qp_id; + + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { + qp = dev->data->queue_pairs[qp_id]; + qp->stats.enqueued_count = 0; + qp->stats.dequeued_count = 0; + qp->stats.enqueue_err_count = 0; + qp->stats.dequeue_err_count = 0; + } +} + static int cn10k_ml_dev_dump(struct rte_ml_dev *dev, FILE *fp) { @@ -1467,15 +1503,23 @@ static __rte_always_inline void cn10k_ml_result_update(struct rte_ml_dev *dev, int qp_id, struct cn10k_ml_result *result, struct rte_ml_op *op) { - PLT_SET_USED(dev); - PLT_SET_USED(qp_id); - struct cn10k_ml_dev *mldev; + struct cn10k_ml_qp *qp; if (likely(result->error_code.u64 == 0)) { + if (likely(qp_id >= 0)) { + qp = dev->data->queue_pairs[qp_id]; + qp->stats.dequeued_count++; + } + op->impl_opaque = result->error_code.u64; op->status = RTE_ML_OP_STATUS_SUCCESS; } else { + if (likely(qp_id >= 0)) { + qp = dev->data->queue_pairs[qp_id]; + qp->stats.dequeue_err_count++; + } + /* Handle driver error */ if (result->error_code.s.etype == ML_ETYPE_DRIVER) { mldev = dev->data->dev_private; @@ -1549,6 +1593,7 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op jcmdq_full: queue->head = head; + qp->stats.enqueued_count += count; return count; } @@ -1697,6 +1742,10 @@ struct rte_ml_dev_ops cn10k_ml_ops = { .dev_queue_pair_setup = cn10k_ml_dev_queue_pair_setup, .dev_queue_pair_release = cn10k_ml_dev_queue_pair_release, + /* Stats ops */ + .dev_stats_get = cn10k_ml_dev_stats_get, + .dev_stats_reset = cn10k_ml_dev_stats_reset, + /* Model ops */ .model_load = cn10k_ml_model_load, .model_unload = cn10k_ml_model_unload, diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 1784900cff..65ae8b44f3 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -58,6 +58,9 @@ struct cn10k_ml_qp { /* Request queue */ struct cn10k_ml_queue queue; + + /* Statistics per queue-pair */ + struct rte_ml_dev_stats stats; }; /* Device ops */ From patchwork Fri Mar 10 08:20:07 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124982 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id CD12941DD0; Fri, 10 Mar 2023 09:24:14 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id F3B3742FF8; Fri, 10 Mar 2023 09:21:11 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 1C72042D33 for ; Fri, 10 Mar 2023 09:20:36 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7WbXG003211 for ; Fri, 10 Mar 2023 00:20:36 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=rfkPirM4inrsSCBzI3PorUzI0IpkIYsoyjYiNNgkp/c=; b=PMj/ZBCaB72LTcdqj5TGn+06LP1LUUh/GkI6Sa4QVWqeLdjTcLF5jKt7St7koboBI4K5 C4Xfq2Dh4GDPlMlbWc240/z8BdqDyDTIhDjEBoIpO/3RHAxVYkCkkPgz9PrDm+3KORFX 3roHjyZZAnqfoXDTFXKdLZruQg5savlXjeDLMG4o4Qr+6Yw/hZLu9YEyRsJTgfPQ9KFt ppemcZTT2b3hSOTvAl3MTnQifYIT/EeThd7twq61bwUyibpKR+mSw+IabXsjDBUbELZc qt5kDnRrNgY63FBR2pf3bBkG02Dh7fkdOS71WaYuClsoGrz+yHA97xkJeMNusDzYQjpx 4Q== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0bf-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:35 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 79AAE3F708D; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 31/39] ml/cnxk: add support to handle extended dev stats Date: Fri, 10 Mar 2023 00:20:07 -0800 Message-ID: <20230310082015.20200-32-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: gtv2cc3zqqQeyOHmiulYrZtz08jqktCs X-Proofpoint-ORIG-GUID: gtv2cc3zqqQeyOHmiulYrZtz08jqktCs X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support to handle ML device extended stats. Support is enabled to get xstats names and stats values and reset xstats. Supported xstats include avg, min and max hardware and firmware latency. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.h | 3 + drivers/ml/cnxk/cn10k_ml_model.h | 57 +++++ drivers/ml/cnxk/cn10k_ml_ops.c | 356 ++++++++++++++++++++++++++++++- 3 files changed, 415 insertions(+), 1 deletion(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 604a200e26..b7ff369ba8 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -372,6 +372,9 @@ struct cn10k_ml_dev { /* Number of models loaded */ uint16_t nb_models_loaded; + + /* xstats status */ + bool xstats_enabled; }; uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); diff --git a/drivers/ml/cnxk/cn10k_ml_model.h b/drivers/ml/cnxk/cn10k_ml_model.h index 75990fe1e4..1bc748265d 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.h +++ b/drivers/ml/cnxk/cn10k_ml_model.h @@ -399,6 +399,57 @@ struct cn10k_ml_model_addr { uint32_t total_output_sz_d; }; +/* Extended stats types enum */ +enum cn10k_ml_model_xstats_type { + /* Average hardware latency */ + avg_hw_latency = 0, + + /* Minimum hardware latency */ + min_hw_latency, + + /* Maximum hardware latency */ + max_hw_latency, + + /* Average firmware latency */ + avg_fw_latency, + + /* Minimum firmware latency */ + min_fw_latency, + + /* Maximum firmware latency */ + max_fw_latency, +}; + +/* Model fast-path stats */ +struct cn10k_ml_model_stats { + /* Total hardware latency, sum of all inferences */ + uint64_t hw_latency_tot; + + /* Minimum hardware latency */ + uint64_t hw_latency_min; + + /* Maximum hardware latency */ + uint64_t hw_latency_max; + + /* Total firmware latency, sum of all inferences */ + uint64_t fw_latency_tot; + + /* Minimum firmware latency */ + uint64_t fw_latency_min; + + /* Maximum firmware latency */ + uint64_t fw_latency_max; + + /* Total jobs dequeued */ + uint64_t dequeued_count; + + /* Hardware stats reset index */ + uint64_t hw_reset_count; + + /* Firmware stats reset index */ + uint64_t fw_reset_count; +}; + /* Model Object */ struct cn10k_ml_model { /* Device reference */ @@ -438,6 +489,12 @@ struct cn10k_ml_model { /* Slow-path operations request pointer */ struct cn10k_ml_req *req; + + /* Stats for burst ops */ + struct cn10k_ml_model_stats *burst_stats; + + /* Stats for sync ops */ + struct cn10k_ml_model_stats *sync_stats; }; int cn10k_ml_model_metadata_check(uint8_t *buffer, uint64_t size); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index c38f018a50..880bb6a5a9 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -354,6 +354,134 @@ cn10k_ml_prep_fp_job_descriptor(struct rte_ml_dev *dev, struct cn10k_ml_req *req req->jd.model_run.num_batches = op->nb_batches; } +#define ML_AVG_FOREACH_QP(dev, model, qp_id, str, value, count) \ + do { \ + value = 0; \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { \ + value += model->burst_stats[qp_id].str##_latency_tot; \ + count += model->burst_stats[qp_id].dequeued_count - \ + model->burst_stats[qp_id].str##_reset_count; \ + } \ + value = value / count; \ + } while (0) + +#define ML_MIN_FOREACH_QP(dev, model, qp_id, str, value, count) \ + do { \ + value = UINT64_MAX; \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { \ + value = PLT_MIN(value, model->burst_stats[qp_id].str##_latency_min); \ + count += model->burst_stats[qp_id].dequeued_count - \ + model->burst_stats[qp_id].str##_reset_count; \ + } \ + if (count == 0) \ + value = 0; \ + } while (0) + +#define ML_MAX_FOREACH_QP(dev, model, qp_id, str, value, count) \ + do { \ + value = 0; \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { \ + value = PLT_MAX(value, model->burst_stats[qp_id].str##_latency_max); \ + count += model->burst_stats[qp_id].dequeued_count - \ + model->burst_stats[qp_id].str##_reset_count; \ + } \ + if (count == 0) \ + value = 0; \ + } while (0) + +static uint64_t +cn10k_ml_model_xstat_get(struct rte_ml_dev *dev, uint16_t model_id, + enum cn10k_ml_model_xstats_type type) +{ + struct cn10k_ml_model *model; + uint64_t count = 0; + uint64_t value; + uint32_t qp_id; + + model = dev->data->models[model_id]; + if (model == NULL) + return 0; + + switch (type) { + case avg_hw_latency: + ML_AVG_FOREACH_QP(dev, model, qp_id, hw, value, count); + break; + case min_hw_latency: + ML_MIN_FOREACH_QP(dev, model, qp_id, hw, value, count); + break; + case max_hw_latency: + ML_MAX_FOREACH_QP(dev, model, qp_id, hw, value, count); + break; + case avg_fw_latency: + ML_AVG_FOREACH_QP(dev, model, qp_id, fw, value, count); + break; + case min_fw_latency: + ML_MIN_FOREACH_QP(dev, model, qp_id, fw, value, count); + break; + case max_fw_latency: + ML_MAX_FOREACH_QP(dev, model, qp_id, fw, value, count); + break; + default: + value = 0; + } + + return value; +} + +#define ML_AVG_RESET_FOREACH_QP(dev, model, qp_id, str) \ + do { \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) { \ + model->burst_stats[qp_id].str##_latency_tot = 0; \ + model->burst_stats[qp_id].str##_reset_count = \ + model->burst_stats[qp_id].dequeued_count; \ + } \ + } while (0) + +#define ML_MIN_RESET_FOREACH_QP(dev, model, qp_id, str) \ + do { \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) \ + model->burst_stats[qp_id].str##_latency_min = UINT64_MAX; \ + } while (0) + +#define ML_MAX_RESET_FOREACH_QP(dev, model, qp_id, str) \ + do { \ + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs; qp_id++) \ + model->burst_stats[qp_id].str##_latency_max = 0; \ + } while (0) + +static void +cn10k_ml_model_xstat_reset(struct rte_ml_dev *dev, uint16_t model_id, + enum cn10k_ml_model_xstats_type type) +{ + struct cn10k_ml_model *model; + uint32_t qp_id; + + model = dev->data->models[model_id]; + + switch (type) { + case avg_hw_latency: + ML_AVG_RESET_FOREACH_QP(dev, model, qp_id, hw); + break; + case min_hw_latency: + ML_MIN_RESET_FOREACH_QP(dev, model, qp_id, hw); + break; + case max_hw_latency: + ML_MAX_RESET_FOREACH_QP(dev, model, qp_id, hw); + break; + case avg_fw_latency: + ML_AVG_RESET_FOREACH_QP(dev, model, qp_id, fw); + break; + case min_fw_latency: + ML_MIN_RESET_FOREACH_QP(dev, model, qp_id, fw); + break; + case max_fw_latency: + ML_MAX_RESET_FOREACH_QP(dev, model, qp_id, fw); + break; + default: + return; + } +} + static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { @@ -519,6 +647,13 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c rte_spinlock_init(&ocm->lock); + /* Check firmware stats */ + if ((mldev->fw.req->jd.fw_load.cap.s.hw_stats) && + (mldev->fw.req->jd.fw_load.cap.s.fw_stats)) + mldev->xstats_enabled = true; + else + mldev->xstats_enabled = false; + dev->enqueue_burst = cn10k_ml_enqueue_burst; dev->dequeue_burst = cn10k_ml_dequeue_burst; dev->op_error_get = cn10k_ml_op_error_get; @@ -714,6 +849,170 @@ cn10k_ml_dev_stats_reset(struct rte_ml_dev *dev) } } +/* Model xstats names */ +struct rte_ml_dev_xstats_map cn10k_ml_model_xstats_table[] = { + {avg_hw_latency, "Avg-HW-Latency"}, {min_hw_latency, "Min-HW-Latency"}, + {max_hw_latency, "Max-HW-Latency"}, {avg_fw_latency, "Avg-FW-Latency"}, + {min_fw_latency, "Min-FW-Latency"}, {max_fw_latency, "Max-FW-Latency"}, +}; + +static int +cn10k_ml_dev_xstats_names_get(struct rte_ml_dev *dev, struct rte_ml_dev_xstats_map *xstats_map, + uint32_t size) +{ + struct rte_ml_dev_info dev_info; + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + uint32_t model_id; + uint32_t count; + uint32_t type; + uint32_t id; + + mldev = dev->data->dev_private; + if (!mldev->xstats_enabled) + return 0; + + if (xstats_map == NULL) + return PLT_DIM(cn10k_ml_model_xstats_table) * mldev->nb_models_loaded; + + /* Model xstats names */ + count = 0; + cn10k_ml_dev_info_get(dev, &dev_info); + + for (id = 0; id < PLT_DIM(cn10k_ml_model_xstats_table) * dev_info.max_models; id++) { + model_id = id / PLT_DIM(cn10k_ml_model_xstats_table); + model = dev->data->models[model_id]; + + if (model == NULL) + continue; + + xstats_map[count].id = id; + type = id % PLT_DIM(cn10k_ml_model_xstats_table); + + snprintf(xstats_map[count].name, RTE_ML_STR_MAX, "%s-%s-cycles", + model->metadata.model.name, cn10k_ml_model_xstats_table[type].name); + + count++; + if (count == size) + break; + } + + return count; +} + +static int +cn10k_ml_dev_xstats_by_name_get(struct rte_ml_dev *dev, const char *name, uint16_t *stat_id, + uint64_t *value) +{ + struct rte_ml_dev_xstats_map *xstats_map; + struct rte_ml_dev_info dev_info; + struct cn10k_ml_dev *mldev; + uint32_t num_xstats; + uint32_t model_id; + uint32_t type; + uint32_t id; + + mldev = dev->data->dev_private; + if (!mldev->xstats_enabled) + return 0; + + num_xstats = PLT_DIM(cn10k_ml_model_xstats_table) * mldev->nb_models_loaded; + xstats_map = rte_zmalloc("cn10k_ml_xstats_map", + sizeof(struct rte_ml_dev_xstats_map) * num_xstats, 0); + cn10k_ml_dev_xstats_names_get(dev, xstats_map, num_xstats); + + cn10k_ml_dev_info_get(dev, &dev_info); + for (id = 0; id < PLT_DIM(cn10k_ml_model_xstats_table) * dev_info.max_models; id++) { + if (strncmp(name, xstats_map[id].name, strlen(name)) == 0) { + *stat_id = id; + rte_free(xstats_map); + break; + } + } + + if (id == PLT_DIM(cn10k_ml_model_xstats_table) * dev_info.max_models) + return -EINVAL; + + model_id = id / PLT_DIM(cn10k_ml_model_xstats_table); + type = id % PLT_DIM(cn10k_ml_model_xstats_table); + *value = cn10k_ml_model_xstat_get(dev, model_id, type); + + return 0; +} + +static int +cn10k_ml_dev_xstats_get(struct rte_ml_dev *dev, const uint16_t *stat_ids, uint64_t *values, + uint16_t nb_ids) +{ + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + uint32_t model_id; + uint32_t count; + uint32_t type; + uint32_t i; + + mldev = dev->data->dev_private; + if (!mldev->xstats_enabled) + return 0; + + count = 0; + for (i = 0; i < nb_ids; i++) { + model_id = stat_ids[i] / PLT_DIM(cn10k_ml_model_xstats_table); + model = dev->data->models[model_id]; + + if (model == NULL) + continue; + + type = stat_ids[i] % PLT_DIM(cn10k_ml_model_xstats_table); + values[i] = cn10k_ml_model_xstat_get(dev, model_id, type); + count++; + } + + return count; +} + +static int +cn10k_ml_dev_xstats_reset(struct rte_ml_dev *dev, const uint16_t *stat_ids, uint16_t nb_ids) +{ + struct rte_ml_dev_info dev_info; + struct cn10k_ml_model *model; + struct cn10k_ml_dev *mldev; + uint32_t model_id; + uint32_t type; + uint32_t i; + + mldev = dev->data->dev_private; + if (!mldev->xstats_enabled) + return 0; + + cn10k_ml_dev_info_get(dev, &dev_info); + if (stat_ids == NULL) { + for (i = 0; i < PLT_DIM(cn10k_ml_model_xstats_table) * dev_info.max_models; i++) { + model_id = i / PLT_DIM(cn10k_ml_model_xstats_table); + model = dev->data->models[model_id]; + + if (model == NULL) + continue; + + type = i % PLT_DIM(cn10k_ml_model_xstats_table); + cn10k_ml_model_xstat_reset(dev, model_id, type); + } + } else { + for (i = 0; i < nb_ids; i++) { + model_id = stat_ids[i] / PLT_DIM(cn10k_ml_model_xstats_table); + model = dev->data->models[model_id]; + + if (model == NULL) + continue; + + type = stat_ids[i] % PLT_DIM(cn10k_ml_model_xstats_table); + cn10k_ml_model_xstat_reset(dev, model_id, type); + } + } + + return 0; +} + static int cn10k_ml_dev_dump(struct rte_ml_dev *dev, FILE *fp) { @@ -856,6 +1155,7 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, char str[RTE_MEMZONE_NAMESIZE]; const struct plt_memzone *mz; + size_t model_stats_size; size_t model_data_size; size_t model_info_size; uint8_t *base_dma_addr; @@ -864,6 +1164,7 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, uint64_t mz_size; uint16_t idx; bool found; + int qp_id; int ret; ret = cn10k_ml_model_metadata_check(params->addr, params->size); @@ -900,10 +1201,12 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, metadata->model.num_input * sizeof(struct rte_ml_io_info) + metadata->model.num_output * sizeof(struct rte_ml_io_info); model_info_size = PLT_ALIGN_CEIL(model_info_size, ML_CN10K_ALIGN_SIZE); + model_stats_size = (dev->data->nb_queue_pairs + 1) * sizeof(struct cn10k_ml_model_stats); mz_size = PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_model), ML_CN10K_ALIGN_SIZE) + 2 * model_data_size + model_info_size + - PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_req), ML_CN10K_ALIGN_SIZE); + PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_req), ML_CN10K_ALIGN_SIZE) + + model_stats_size; /* Allocate memzone for model object and model data */ snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", CN10K_ML_MODEL_MEMZONE_NAME, idx); @@ -949,6 +1252,24 @@ cn10k_ml_model_load(struct rte_ml_dev *dev, struct rte_ml_model_params *params, /* Set slow-path request address and state */ model->req = PLT_PTR_ADD(model->info, model_info_size); + /* Reset burst and sync stats */ + model->burst_stats = PLT_PTR_ADD( + model->req, PLT_ALIGN_CEIL(sizeof(struct cn10k_ml_req), ML_CN10K_ALIGN_SIZE)); + for (qp_id = 0; qp_id < dev->data->nb_queue_pairs + 1; qp_id++) { + model->burst_stats[qp_id].hw_latency_tot = 0; + model->burst_stats[qp_id].hw_latency_min = UINT64_MAX; + model->burst_stats[qp_id].hw_latency_max = 0; + model->burst_stats[qp_id].fw_latency_tot = 0; + model->burst_stats[qp_id].fw_latency_min = UINT64_MAX; + model->burst_stats[qp_id].fw_latency_max = 0; + model->burst_stats[qp_id].hw_reset_count = 0; + model->burst_stats[qp_id].fw_reset_count = 0; + model->burst_stats[qp_id].dequeued_count = 0; + } + model->sync_stats = + PLT_PTR_ADD(model->burst_stats, + dev->data->nb_queue_pairs * sizeof(struct cn10k_ml_model_stats)); + plt_spinlock_init(&model->lock); model->state = ML_CN10K_MODEL_STATE_LOADED; dev->data->models[idx] = model; @@ -1503,15 +1824,44 @@ static __rte_always_inline void cn10k_ml_result_update(struct rte_ml_dev *dev, int qp_id, struct cn10k_ml_result *result, struct rte_ml_op *op) { + struct cn10k_ml_model_stats *stats; + struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; struct cn10k_ml_qp *qp; + uint64_t hw_latency; + uint64_t fw_latency; if (likely(result->error_code.u64 == 0)) { + model = dev->data->models[op->model_id]; if (likely(qp_id >= 0)) { qp = dev->data->queue_pairs[qp_id]; qp->stats.dequeued_count++; + stats = &model->burst_stats[qp_id]; + } else { + stats = model->sync_stats; + } + + if (unlikely(stats->dequeued_count == stats->hw_reset_count)) { + stats->hw_latency_min = UINT64_MAX; + stats->hw_latency_max = 0; } + if (unlikely(stats->dequeued_count == stats->fw_reset_count)) { + stats->fw_latency_min = UINT64_MAX; + stats->fw_latency_max = 0; + } + + hw_latency = result->stats.hw_end - result->stats.hw_start; + fw_latency = result->stats.fw_end - result->stats.fw_start - hw_latency; + + stats->hw_latency_tot += hw_latency; + stats->hw_latency_min = PLT_MIN(stats->hw_latency_min, hw_latency); + stats->hw_latency_max = PLT_MAX(stats->hw_latency_max, hw_latency); + stats->fw_latency_tot += fw_latency; + stats->fw_latency_min = PLT_MIN(stats->fw_latency_min, fw_latency); + stats->fw_latency_max = PLT_MAX(stats->fw_latency_max, fw_latency); + stats->dequeued_count++; + op->impl_opaque = result->error_code.u64; op->status = RTE_ML_OP_STATUS_SUCCESS; } else { @@ -1745,6 +2095,10 @@ struct rte_ml_dev_ops cn10k_ml_ops = { /* Stats ops */ .dev_stats_get = cn10k_ml_dev_stats_get, .dev_stats_reset = cn10k_ml_dev_stats_reset, + .dev_xstats_names_get = cn10k_ml_dev_xstats_names_get, + .dev_xstats_by_name_get = cn10k_ml_dev_xstats_by_name_get, + .dev_xstats_get = cn10k_ml_dev_xstats_get, + .dev_xstats_reset = cn10k_ml_dev_xstats_reset, /* Model ops */ .model_load = cn10k_ml_model_load, From patchwork Fri Mar 10 08:20:08 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124979 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id E95C041DD0; Fri, 10 Mar 2023 09:23:52 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 3F58042F98; Fri, 10 Mar 2023 09:21:07 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 7432A42D1A for ; Fri, 10 Mar 2023 09:20:34 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7uIdJ021875 for ; Fri, 10 Mar 2023 00:20:33 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=xeAijMZVsgE559mrWl2bAGVKpgMsX1wl/FCZ2hK+wEE=; b=V4vgMmXMLkVRr76YCjHW9OtFMwUND0sHoGE7Om0Z3qNYpmeJ2deR1FBc2iU782h4SZBE 14uyXIt5fQct8ZS4vU2b8UYTdjIa5X0UBnK8v7+Ui49IXp6ghypj26xgvlrc3P56WXrH lBQEi5ZpimUaBbYtdpnjutyX9KYSQiUTXvCYOScutRYfaKsssmPoppFN8D/hJs/kbCUZ 69oLSYc3J2FTqeY9QT/aarQAkHOseuvrjNoi0l9hCUOgtkE8ONOZKI1YhSuakm1k+xsr TId+R2QfwJY6nnnzexXqt0WtPUvBajjPria1DKwtYjVM64rEm7ZYK1wdmteENmTf7Y42 Ig== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3p80k782tf-18 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:33 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id A35E13F7093; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 32/39] ml/cnxk: enable support to get xstats in cycles Date: Fri, 10 Mar 2023 00:20:08 -0800 Message-ID: <20230310082015.20200-33-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: bDWOBLQuV-LXdEnZiq4VA5zUr5BG5VnF X-Proofpoint-ORIG-GUID: bDWOBLQuV-LXdEnZiq4VA5zUr5BG5VnF X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support to retrieve xstats in either cycles or ns. Access to sclk is enabled only if an RVU device is probed during initialization. Driver would return the xstats in nanoseconds only when an RVU device is probed, else would fallback to cycles. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_ops.c | 19 +++++++++++++++++-- 1 file changed, 17 insertions(+), 2 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 880bb6a5a9..5689fbfcb2 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -394,6 +394,8 @@ cn10k_ml_model_xstat_get(struct rte_ml_dev *dev, uint16_t model_id, enum cn10k_ml_model_xstats_type type) { struct cn10k_ml_model *model; + uint16_t rclk_freq; /* MHz */ + uint16_t sclk_freq; /* MHz */ uint64_t count = 0; uint64_t value; uint32_t qp_id; @@ -425,6 +427,10 @@ cn10k_ml_model_xstat_get(struct rte_ml_dev *dev, uint16_t model_id, value = 0; } + roc_clk_freq_get(&rclk_freq, &sclk_freq); + if (sclk_freq != 0) /* return in ns */ + value = (value * 1000ULL) / sclk_freq; + return value; } @@ -863,6 +869,8 @@ cn10k_ml_dev_xstats_names_get(struct rte_ml_dev *dev, struct rte_ml_dev_xstats_m struct rte_ml_dev_info dev_info; struct cn10k_ml_model *model; struct cn10k_ml_dev *mldev; + uint16_t rclk_freq; + uint16_t sclk_freq; uint32_t model_id; uint32_t count; uint32_t type; @@ -878,6 +886,7 @@ cn10k_ml_dev_xstats_names_get(struct rte_ml_dev *dev, struct rte_ml_dev_xstats_m /* Model xstats names */ count = 0; cn10k_ml_dev_info_get(dev, &dev_info); + roc_clk_freq_get(&rclk_freq, &sclk_freq); for (id = 0; id < PLT_DIM(cn10k_ml_model_xstats_table) * dev_info.max_models; id++) { model_id = id / PLT_DIM(cn10k_ml_model_xstats_table); @@ -889,8 +898,14 @@ cn10k_ml_dev_xstats_names_get(struct rte_ml_dev *dev, struct rte_ml_dev_xstats_m xstats_map[count].id = id; type = id % PLT_DIM(cn10k_ml_model_xstats_table); - snprintf(xstats_map[count].name, RTE_ML_STR_MAX, "%s-%s-cycles", - model->metadata.model.name, cn10k_ml_model_xstats_table[type].name); + if (sclk_freq == 0) + snprintf(xstats_map[count].name, RTE_ML_STR_MAX, "%s-%s-cycles", + model->metadata.model.name, + cn10k_ml_model_xstats_table[type].name); + else + snprintf(xstats_map[count].name, RTE_ML_STR_MAX, "%s-%s-ns", + model->metadata.model.name, + cn10k_ml_model_xstats_table[type].name); count++; if (count == size) From patchwork Fri Mar 10 08:20:09 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124985 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 3836041DD0; Fri, 10 Mar 2023 09:24:30 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 4A68842FD6; Fri, 10 Mar 2023 09:21:15 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 57C2542D2C for ; Fri, 10 Mar 2023 09:20:38 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7WbXJ003211 for ; Fri, 10 Mar 2023 00:20:37 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=bQid8g7/TgRhAVCXeUZsdULRHG2l4aU4Wt668oxEs5I=; b=E+nGD/RRaxIU+HeQ/AryVLbTKmpwUC8siOxyRQBbz8aEDmQn+5FciQe5plbZXOYDWVIk HcONSOzBju6MmDuqt2aW+VeQZdZYwP5tphuDU3g3kVbXq5O3YAllehguXkfBr4a+vzV0 yNVpqbpISNS7RaFjBaHjpK8iLPzYKoShhL4qY1hS+9UoGWZMai2I3CBnxXCIXtxfdMXO 02D3mBQnaG2fOJFVxM0GiXaldV0oFDuO+fOgrupNMBNzjEBFPbToPad6/6nTidA+oL0o EdnnIM/CyE4VQ6usLtCgyRWWkdNKi+VqMJFE+jS2MxtkSbA7r5h5Tx5hegOj/c/+llOL Aw== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0bf-4 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:37 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:25 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:25 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id CC4785C68ED; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 33/39] ml/cnxk: add support to report DPE FW warnings Date: Fri, 10 Mar 2023 00:20:09 -0800 Message-ID: <20230310082015.20200-34-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: Xx6ojhQLZO_2TeFD0G9FMtSHDAmys9mq X-Proofpoint-ORIG-GUID: Xx6ojhQLZO_2TeFD0G9FMtSHDAmys9mq X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Enabled support to enable and report DPE warnings from ML firmware. Configure firmware load flags based on the device arguments. Default values: enable_dpe_errors = 1 report_dpe_errors = 0 Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 94 +++++++++++++++++++++++++++++++--- drivers/ml/cnxk/cn10k_ml_dev.h | 6 +++ 2 files changed, 93 insertions(+), 7 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 76ed853a3c..ac6592891b 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -17,9 +17,13 @@ #include "cn10k_ml_dev.h" #include "cn10k_ml_ops.h" -#define CN10K_ML_FW_PATH "fw_path" +#define CN10K_ML_FW_PATH "fw_path" +#define CN10K_ML_FW_ENABLE_DPE_WARNINGS "enable_dpe_warnings" +#define CN10K_ML_FW_REPORT_DPE_WARNINGS "report_dpe_warnings" -#define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" +#define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" +#define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 +#define CN10K_ML_FW_REPORT_DPE_WARNINGS_DEFAULT 0 /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -28,9 +32,13 @@ #define FW_EXCEPTION_BUFFER_SIZE 0x400 #define FW_LINKER_OFFSET 0x80000 #define FW_WAIT_CYCLES 100 -#define FW_LOAD_FLAGS 0x1 -static const char *const valid_args[] = {CN10K_ML_FW_PATH, NULL}; +/* Firmware flags */ +#define FW_ENABLE_DPE_WARNING_BITMASK BIT(0) +#define FW_REPORT_DPE_WARNING_BITMASK BIT(1) + +static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_FW_ENABLE_DPE_WARNINGS, + CN10K_ML_FW_REPORT_DPE_WARNINGS, NULL}; /* Dummy operations for ML device */ struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; @@ -49,9 +57,25 @@ parse_string_arg(const char *key __rte_unused, const char *value, void *extra_ar return 0; } +static int +parse_integer_arg(const char *key __rte_unused, const char *value, void *extra_args) +{ + int *i = (int *)extra_args; + + *i = atoi(value); + if (*i < 0) { + plt_err("Argument has to be positive."); + return -EINVAL; + } + + return 0; +} + static int cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mldev) { + bool enable_dpe_warnings_set = false; + bool report_dpe_warnings_set = false; struct rte_kvargs *kvlist = NULL; bool fw_path_set = false; char *fw_path = NULL; @@ -76,6 +100,30 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde fw_path_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_FW_ENABLE_DPE_WARNINGS) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_FW_ENABLE_DPE_WARNINGS, + &parse_integer_arg, &mldev->fw.enable_dpe_warnings); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", + CN10K_ML_FW_ENABLE_DPE_WARNINGS); + ret = -EINVAL; + goto exit; + } + enable_dpe_warnings_set = true; + } + + if (rte_kvargs_count(kvlist, CN10K_ML_FW_REPORT_DPE_WARNINGS) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_FW_REPORT_DPE_WARNINGS, + &parse_integer_arg, &mldev->fw.report_dpe_warnings); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", + CN10K_ML_FW_REPORT_DPE_WARNINGS); + ret = -EINVAL; + goto exit; + } + report_dpe_warnings_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -83,6 +131,30 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde mldev->fw.path = fw_path; plt_info("ML: %s = %s", CN10K_ML_FW_PATH, mldev->fw.path); + if (!enable_dpe_warnings_set) { + mldev->fw.enable_dpe_warnings = CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT; + } else { + if ((mldev->fw.enable_dpe_warnings < 0) || (mldev->fw.enable_dpe_warnings > 1)) { + plt_err("Invalid argument, %s = %d\n", CN10K_ML_FW_ENABLE_DPE_WARNINGS, + mldev->fw.enable_dpe_warnings); + ret = -EINVAL; + goto exit; + } + } + plt_info("ML: %s = %d", CN10K_ML_FW_ENABLE_DPE_WARNINGS, mldev->fw.enable_dpe_warnings); + + if (!report_dpe_warnings_set) { + mldev->fw.report_dpe_warnings = CN10K_ML_FW_REPORT_DPE_WARNINGS_DEFAULT; + } else { + if ((mldev->fw.report_dpe_warnings < 0) || (mldev->fw.report_dpe_warnings > 1)) { + plt_err("Invalid argument, %s = %d\n", CN10K_ML_FW_REPORT_DPE_WARNINGS, + mldev->fw.report_dpe_warnings); + ret = -EINVAL; + goto exit; + } + } + plt_info("ML: %s = %d", CN10K_ML_FW_REPORT_DPE_WARNINGS, mldev->fw.report_dpe_warnings); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -208,9 +280,15 @@ cn10k_ml_fw_print_info(struct cn10k_ml_fw *fw) uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw) { - PLT_SET_USED(fw); + uint64_t flags = 0x0; + + if (fw->enable_dpe_warnings) + flags = flags | FW_ENABLE_DPE_WARNING_BITMASK; + + if (fw->report_dpe_warnings) + flags = flags | FW_REPORT_DPE_WARNING_BITMASK; - return FW_LOAD_FLAGS; + return flags; } static int @@ -614,4 +692,6 @@ RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); -RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH "="); +RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, + CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS + "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS "=<0|1>"); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index b7ff369ba8..9ba56ffba6 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -349,6 +349,12 @@ struct cn10k_ml_fw { /* Firmware file path */ const char *path; + /* Enable DPE warnings */ + int enable_dpe_warnings; + + /* Report DPE warnings */ + int report_dpe_warnings; + /* Data buffer */ uint8_t *data; From patchwork Fri Mar 10 08:20:10 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124968 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5884041D3D; Fri, 10 Mar 2023 09:22:52 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id F0D7342DA2; Fri, 10 Mar 2023 09:20:54 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id A028E42BDA for ; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1K009730 for ; Fri, 10 Mar 2023 00:20:29 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=w0qFTY2B2fqCZwLj3I5X6QsR/sqZ49/AZa0cy5DjNok=; b=IGg3QHSA4bivtKTV0UbbPs4B257HGfNjNd5tjkPLkdckLQYZ9Eym/0/2IXsBAnsxe+yO hJ4WJafiaUwxUbrf+FlUx3EvUyzLf7g0GRDIFLhgya5aiZAedJE1FcJs7c0un0g1xD2p thleY84q3l2O6FLfFqNxVNHMKZhcNcaabCZm9Pwuz+HmRM6zJMfQf1wu0lDq9QcmarqZ vaLBL9sN63qCJJpLTSxUogwhawSxQUgUILuEEJ8VcJpO8m+5gCFWLwwGKeBJ6rz4K8f1 NYZ391AxPQ1Wn+l0ts7qXkC3qWvBdiOFsOlbolOuM9N4ci1VIH+Sd/iERtMZgDTm0xkq qA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-15 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:29 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 00DD95C68F3; Fri, 10 Mar 2023 00:20:23 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 34/39] ml/cnxk: add support to enable model data caching Date: Fri, 10 Mar 2023 00:20:10 -0800 Message-ID: <20230310082015.20200-35-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: 8bnmfH6-STk8L8qLlf6owcmqw46TWs5e X-Proofpoint-ORIG-GUID: 8bnmfH6-STk8L8qLlf6owcmqw46TWs5e X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added device argument 'cache_model_data' to enable model data caching. An inference request would be executed with dummy data in synchronous mode during model start stage. This run would cache the model weights and bias in the memory and result in improved inference throughput. cache_model_data = 1, enable (default) cache_model_data = 0, disable Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 33 ++++++++++++++++++++-- drivers/ml/cnxk/cn10k_ml_dev.h | 3 ++ drivers/ml/cnxk/cn10k_ml_ops.c | 50 ++++++++++++++++++++++++++++++++++ 3 files changed, 84 insertions(+), 2 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index ac6592891b..948708a420 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -20,10 +20,12 @@ #define CN10K_ML_FW_PATH "fw_path" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS "enable_dpe_warnings" #define CN10K_ML_FW_REPORT_DPE_WARNINGS "report_dpe_warnings" +#define CN10K_ML_DEV_CACHE_MODEL_DATA "cache_model_data" #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 #define CN10K_ML_FW_REPORT_DPE_WARNINGS_DEFAULT 0 +#define CN10K_ML_DEV_CACHE_MODEL_DATA_DEFAULT 1 /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -38,7 +40,8 @@ #define FW_REPORT_DPE_WARNING_BITMASK BIT(1) static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_FW_ENABLE_DPE_WARNINGS, - CN10K_ML_FW_REPORT_DPE_WARNINGS, NULL}; + CN10K_ML_FW_REPORT_DPE_WARNINGS, + CN10K_ML_DEV_CACHE_MODEL_DATA, NULL}; /* Dummy operations for ML device */ struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; @@ -76,6 +79,7 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde { bool enable_dpe_warnings_set = false; bool report_dpe_warnings_set = false; + bool cache_model_data_set = false; struct rte_kvargs *kvlist = NULL; bool fw_path_set = false; char *fw_path = NULL; @@ -124,6 +128,18 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde report_dpe_warnings_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_DEV_CACHE_MODEL_DATA) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_DEV_CACHE_MODEL_DATA, &parse_integer_arg, + &mldev->cache_model_data); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", + CN10K_ML_DEV_CACHE_MODEL_DATA); + ret = -EINVAL; + goto exit; + } + cache_model_data_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -155,6 +171,18 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde } plt_info("ML: %s = %d", CN10K_ML_FW_REPORT_DPE_WARNINGS, mldev->fw.report_dpe_warnings); + if (!cache_model_data_set) { + mldev->cache_model_data = CN10K_ML_DEV_CACHE_MODEL_DATA_DEFAULT; + } else { + if ((mldev->cache_model_data < 0) || (mldev->cache_model_data > 1)) { + plt_err("Invalid argument, %s = %d\n", CN10K_ML_DEV_CACHE_MODEL_DATA, + mldev->cache_model_data); + ret = -EINVAL; + goto exit; + } + } + plt_info("ML: %s = %d", CN10K_ML_DEV_CACHE_MODEL_DATA, mldev->cache_model_data); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -694,4 +722,5 @@ RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS - "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS "=<0|1>"); + "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS + "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA "=<0|1>"); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 9ba56ffba6..718edadde7 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -381,6 +381,9 @@ struct cn10k_ml_dev { /* xstats status */ bool xstats_enabled; + + /* Enable / disable model data caching */ + int cache_model_data; }; uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 5689fbfcb2..d69df42b27 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -488,6 +488,49 @@ cn10k_ml_model_xstat_reset(struct rte_ml_dev *dev, uint16_t model_id, } } +static int +cn10k_ml_cache_model_data(struct rte_ml_dev *dev, uint16_t model_id) +{ + struct cn10k_ml_model *model; + struct rte_ml_op op; + + char str[RTE_MEMZONE_NAMESIZE]; + const struct plt_memzone *mz; + uint64_t isize = 0; + uint64_t osize = 0; + int ret = 0; + + model = dev->data->models[model_id]; + + /* Create input and output buffers. */ + rte_ml_io_input_size_get(dev->data->dev_id, model_id, model->batch_size, &isize, NULL); + rte_ml_io_output_size_get(dev->data->dev_id, model_id, model->batch_size, &osize, NULL); + + snprintf(str, RTE_MEMZONE_NAMESIZE, "%s_%u", "ml_dummy_io", model_id); + mz = plt_memzone_reserve_aligned(str, isize + osize, 0, ML_CN10K_ALIGN_SIZE); + if (mz == NULL) + return -ENOMEM; + memset(mz->addr, 0, isize + osize); + + op.model_id = model_id; + op.nb_batches = model->batch_size; + op.mempool = NULL; + + op.input.addr = mz->addr; + op.input.length = isize; + op.input.next = NULL; + + op.output.addr = PLT_PTR_ADD(op.input.addr, isize); + op.output.length = osize; + op.output.next = NULL; + + memset(model->req, 0, sizeof(struct cn10k_ml_req)); + ret = cn10k_ml_inference_sync(dev, &op); + plt_memzone_free(mz); + + return ret; +} + static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { @@ -1467,6 +1510,13 @@ cn10k_ml_model_start(struct rte_ml_dev *dev, uint16_t model_id) } } + if (ret < 0) { /* Call unload to update model and FW state, ignore error */ + rte_ml_model_stop(dev->data->dev_id, model_id); + } else { + if (mldev->cache_model_data && roc_model_is_cn10ka()) + ret = cn10k_ml_cache_model_data(dev, model_id); + } + return ret; } From patchwork Fri Mar 10 08:20:11 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124970 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id D81B241D3D; Fri, 10 Mar 2023 09:23:03 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 623F742DAD; Fri, 10 Mar 2023 09:20:57 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 0FE9C42BFE for ; Fri, 10 Mar 2023 09:20:30 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1L009730 for ; Fri, 10 Mar 2023 00:20:30 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=lP+8w4lJ4bygBYRZC8TUAnKIerEvl+BRypJgebjxcsU=; b=LZPDBrI605HTkmO/YiEMCaZKdA+3O6awItiucFnDUzrKDqNG0MC+99nKjAKLPNakE1fw vIfuEm0KjJ6SF0oUa0EJ2H2FPWz/garF4Ea3rM8XDINg+VCqb/rngFZpi7/r+bh1m/2w 375PSrQ+zTtsZRHHalCKRbycAZtNh3kxZHCDH9bb4PdjU4sQZCEH+0HQUou9bGNuAQCZ mn0+cMOptzb3F855PJVYW7xPriFvxl+nw7IehzwU/dTXsNsmPVlmBXbHiVrmwparwAcu ya7CvhHk9v2Kb4dC7eu4Vq91CiqZwN1HFzLogd/+VwpTXy5B9gkCbaq60hShxpGrZPNc KA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-16 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:30 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 2A0FC5C68F5; Fri, 10 Mar 2023 00:20:24 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 35/39] ml/cnxk: add support to select OCM allocation mode Date: Fri, 10 Mar 2023 00:20:11 -0800 Message-ID: <20230310082015.20200-36-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: gkX6FjGdiNAGGgp9Tb4rzaLFXKmnUy2f X-Proofpoint-ORIG-GUID: gkX6FjGdiNAGGgp9Tb4rzaLFXKmnUy2f X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added device argument "ocm_alloc_mode" to select OCM allocation method during model start. Two modes are supported by the driver. Added implementation for ocm_alloc_mode lowest as default. ocm_alloc_mode: lowest: Allocate from first available free slot / lowest tile ID in OCM (default) largest: Allocate from a slot with maximum free memory Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 45 +++++++++++++++++++++++++++++----- drivers/ml/cnxk/cn10k_ml_ocm.c | 6 ++--- drivers/ml/cnxk/cn10k_ml_ocm.h | 3 +++ 3 files changed, 44 insertions(+), 10 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 948708a420..5c02d67c8e 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -21,11 +21,13 @@ #define CN10K_ML_FW_ENABLE_DPE_WARNINGS "enable_dpe_warnings" #define CN10K_ML_FW_REPORT_DPE_WARNINGS "report_dpe_warnings" #define CN10K_ML_DEV_CACHE_MODEL_DATA "cache_model_data" +#define CN10K_ML_OCM_ALLOC_MODE "ocm_alloc_mode" #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 #define CN10K_ML_FW_REPORT_DPE_WARNINGS_DEFAULT 0 #define CN10K_ML_DEV_CACHE_MODEL_DATA_DEFAULT 1 +#define CN10K_ML_OCM_ALLOC_MODE_DEFAULT "lowest" /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -39,9 +41,12 @@ #define FW_ENABLE_DPE_WARNING_BITMASK BIT(0) #define FW_REPORT_DPE_WARNING_BITMASK BIT(1) -static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_FW_ENABLE_DPE_WARNINGS, +static const char *const valid_args[] = {CN10K_ML_FW_PATH, + CN10K_ML_FW_ENABLE_DPE_WARNINGS, CN10K_ML_FW_REPORT_DPE_WARNINGS, - CN10K_ML_DEV_CACHE_MODEL_DATA, NULL}; + CN10K_ML_DEV_CACHE_MODEL_DATA, + CN10K_ML_OCM_ALLOC_MODE, + NULL}; /* Dummy operations for ML device */ struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; @@ -81,6 +86,8 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde bool report_dpe_warnings_set = false; bool cache_model_data_set = false; struct rte_kvargs *kvlist = NULL; + bool ocm_alloc_mode_set = false; + char *ocm_alloc_mode = NULL; bool fw_path_set = false; char *fw_path = NULL; int ret = 0; @@ -140,6 +147,17 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde cache_model_data_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_OCM_ALLOC_MODE) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_OCM_ALLOC_MODE, &parse_string_arg, + &ocm_alloc_mode); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", CN10K_ML_OCM_ALLOC_MODE); + ret = -EINVAL; + goto exit; + } + ocm_alloc_mode_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -183,6 +201,20 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde } plt_info("ML: %s = %d", CN10K_ML_DEV_CACHE_MODEL_DATA, mldev->cache_model_data); + if (!ocm_alloc_mode_set) { + mldev->ocm.alloc_mode = CN10K_ML_OCM_ALLOC_MODE_DEFAULT; + } else { + if (!((strcmp(ocm_alloc_mode, "lowest") == 0) || + (strcmp(ocm_alloc_mode, "largest") == 0))) { + plt_err("Invalid argument, %s = %s\n", CN10K_ML_OCM_ALLOC_MODE, + ocm_alloc_mode); + ret = -EINVAL; + goto exit; + } + mldev->ocm.alloc_mode = ocm_alloc_mode; + } + plt_info("ML: %s = %s", CN10K_ML_OCM_ALLOC_MODE, mldev->ocm.alloc_mode); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -720,7 +752,8 @@ RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); -RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, - CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS - "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS - "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA "=<0|1>"); +RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH + "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS + "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS + "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA + "=<0|1>" CN10K_ML_OCM_ALLOC_MODE "="); diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c index 0b04fcc2da..551faef7eb 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.c +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -230,7 +230,6 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w int wb_page_start_curr; int max_slot_sz_curr; uint8_t tile_start; - int ocm_alloc_mode; int wb_page_start; uint16_t tile_id; uint16_t word_id; @@ -255,7 +254,6 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w max_slot_sz_curr = 0; max_slot_sz = 0; tile_idx = 0; - ocm_alloc_mode = 2; if ((start_tile != -1) && (start_tile % num_tiles != 0)) { plt_err("Invalid start_tile, %d", start_tile); @@ -303,13 +301,13 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w } } - if (ocm_alloc_mode == 1) { + if (strcmp(ocm->alloc_mode, "lowest") == 0) { wb_page_start = slot_index_lowest(local_ocm_mask, ocm->mask_words, wb_pages, 0); if (wb_page_start != -1) { /* Have a valid slot for WB, else next set of tiles */ tile_idx = tile_start; goto found; } - } else if (ocm_alloc_mode == 2) { + } else if (strcmp(ocm->alloc_mode, "largest") == 0) { wb_page_start_curr = slot_index_largest(local_ocm_mask, ocm->mask_words, wb_pages, &max_slot_sz_curr); if (max_slot_sz_curr > max_slot_sz) { diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h index 0c7172a671..5f018b410a 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.h +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -58,6 +58,9 @@ struct cn10k_ml_ocm { /* OCM spinlock, used to update OCM state */ rte_spinlock_t lock; + /* OCM allocation mode */ + const char *alloc_mode; + /* Number of OCM tiles */ uint8_t num_tiles; From patchwork Fri Mar 10 08:20:12 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124983 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 0E07F41DD0; Fri, 10 Mar 2023 09:24:19 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 1480242FFE; Fri, 10 Mar 2023 09:21:13 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 76EBE42D2C for ; Fri, 10 Mar 2023 09:20:37 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7WbXH003211 for ; Fri, 10 Mar 2023 00:20:36 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=vx20V2D80PGWxIsspRqUd+ZyUBBpX/ks2ExSzlGNrQA=; b=czE2vcvNxfE+yH5Kg8Hj11EFaA9+1BzfBwjfQbay8FC1ItrLgVSX0nz9Lp65u6LXI2mS UN4dIHVjYQPBLSQS57YngQWrAAmPGhs+oFAR7W4Nh5grVTIa6V2BQa6+7WRiWRU/sj2o b9EshuUcfwofTrX86725FxGv+Fq5qAF/Hb5XChxpxeiqy1s2MuQw57eYK9HDNfd3Jfvx UV1G/mg+AmtfREdWkWrRcAtTl0pzS+7heFcSmBAbPTt5MluYmITJ90nNAo+LN0+Y9cOK rh/TAEjYcd5aSosQKMVqIsme3ls2ToBIC2f9u5Zp4lR3snsAdAR9q4yHJsSm8RToVJgR ag== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0bf-2 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:36 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:25 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 549D55B6939; Fri, 10 Mar 2023 00:20:24 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 36/39] ml/cnxk: add support to use lock during jcmd enq Date: Fri, 10 Mar 2023 00:20:12 -0800 Message-ID: <20230310082015.20200-37-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: tbvvWlVuUo-1jbVYLp5DZJ9CzWpweBiF X-Proofpoint-ORIG-GUID: tbvvWlVuUo-1jbVYLp5DZJ9CzWpweBiF X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added device argument "hw_queue_lock" to select the JCMDQ enqueue ROC function to be used in fast path. hw_queue_lock: 0: Disable, use lock free version of JCMDQ enqueue ROC function for job queuing. To avoid race condition in request queuing to hardware, disabling hw_queue_lock restricts the number of queue-pairs supported by cnxk driver to 1. 1: Enable, (default) use spin-lock version of JCMDQ enqueue ROC function for job queuing. Enabling spinlock version would disable restrictions on the number of queue-pairs that can be created. Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 31 ++++++++++++++++++++++++++++++- drivers/ml/cnxk/cn10k_ml_dev.h | 13 +++++++++++-- drivers/ml/cnxk/cn10k_ml_ops.c | 20 +++++++++++++++++--- 3 files changed, 58 insertions(+), 6 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index 5c02d67c8e..aa503b2691 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -22,12 +22,14 @@ #define CN10K_ML_FW_REPORT_DPE_WARNINGS "report_dpe_warnings" #define CN10K_ML_DEV_CACHE_MODEL_DATA "cache_model_data" #define CN10K_ML_OCM_ALLOC_MODE "ocm_alloc_mode" +#define CN10K_ML_DEV_HW_QUEUE_LOCK "hw_queue_lock" #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 #define CN10K_ML_FW_REPORT_DPE_WARNINGS_DEFAULT 0 #define CN10K_ML_DEV_CACHE_MODEL_DATA_DEFAULT 1 #define CN10K_ML_OCM_ALLOC_MODE_DEFAULT "lowest" +#define CN10K_ML_DEV_HW_QUEUE_LOCK_DEFAULT 1 /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -46,6 +48,7 @@ static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_FW_REPORT_DPE_WARNINGS, CN10K_ML_DEV_CACHE_MODEL_DATA, CN10K_ML_OCM_ALLOC_MODE, + CN10K_ML_DEV_HW_QUEUE_LOCK, NULL}; /* Dummy operations for ML device */ @@ -87,6 +90,7 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde bool cache_model_data_set = false; struct rte_kvargs *kvlist = NULL; bool ocm_alloc_mode_set = false; + bool hw_queue_lock_set = false; char *ocm_alloc_mode = NULL; bool fw_path_set = false; char *fw_path = NULL; @@ -158,6 +162,18 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde ocm_alloc_mode_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_DEV_HW_QUEUE_LOCK) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_DEV_HW_QUEUE_LOCK, &parse_integer_arg, + &mldev->hw_queue_lock); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", + CN10K_ML_DEV_HW_QUEUE_LOCK); + ret = -EINVAL; + goto exit; + } + hw_queue_lock_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -215,6 +231,18 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde } plt_info("ML: %s = %s", CN10K_ML_OCM_ALLOC_MODE, mldev->ocm.alloc_mode); + if (!hw_queue_lock_set) { + mldev->hw_queue_lock = CN10K_ML_DEV_HW_QUEUE_LOCK_DEFAULT; + } else { + if ((mldev->hw_queue_lock < 0) || (mldev->hw_queue_lock > 1)) { + plt_err("Invalid argument, %s = %d\n", CN10K_ML_DEV_HW_QUEUE_LOCK, + mldev->hw_queue_lock); + ret = -EINVAL; + goto exit; + } + } + plt_info("ML: %s = %d", CN10K_ML_DEV_HW_QUEUE_LOCK, mldev->hw_queue_lock); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -756,4 +784,5 @@ RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA - "=<0|1>" CN10K_ML_OCM_ALLOC_MODE "="); + "=<0|1>" CN10K_ML_OCM_ALLOC_MODE + "=" CN10K_ML_DEV_HW_QUEUE_LOCK "=<0|1>"); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 718edadde7..49676ac9e7 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -21,8 +21,11 @@ /* Maximum number of models per device */ #define ML_CN10K_MAX_MODELS 16 -/* Maximum number of queue-pairs per device */ -#define ML_CN10K_MAX_QP_PER_DEVICE 1 +/* Maximum number of queue-pairs per device, spinlock version */ +#define ML_CN10K_MAX_QP_PER_DEVICE_SL 16 + +/* Maximum number of queue-pairs per device, lock-free version */ +#define ML_CN10K_MAX_QP_PER_DEVICE_LF 1 /* Maximum number of descriptors per queue-pair */ #define ML_CN10K_MAX_DESC_PER_QP 1024 @@ -384,6 +387,12 @@ struct cn10k_ml_dev { /* Enable / disable model data caching */ int cache_model_data; + + /* Use spinlock version of ROC enqueue */ + int hw_queue_lock; + + /* JCMD enqueue function handler */ + bool (*ml_jcmdq_enqueue)(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd); }; uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index d69df42b27..f92f778e23 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -534,13 +534,21 @@ cn10k_ml_cache_model_data(struct rte_ml_dev *dev, uint16_t model_id) static int cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) { + struct cn10k_ml_dev *mldev; + if (dev_info == NULL) return -EINVAL; + mldev = dev->data->dev_private; + memset(dev_info, 0, sizeof(struct rte_ml_dev_info)); dev_info->driver_name = dev->device->driver->name; dev_info->max_models = ML_CN10K_MAX_MODELS; - dev_info->max_queue_pairs = ML_CN10K_MAX_QP_PER_DEVICE; + if (mldev->hw_queue_lock) + dev_info->max_queue_pairs = ML_CN10K_MAX_QP_PER_DEVICE_SL; + else + dev_info->max_queue_pairs = ML_CN10K_MAX_QP_PER_DEVICE_LF; + dev_info->max_desc = ML_CN10K_MAX_DESC_PER_QP; dev_info->max_segments = ML_CN10K_MAX_SEGMENTS; dev_info->min_align_size = ML_CN10K_ALIGN_SIZE; @@ -703,6 +711,12 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c else mldev->xstats_enabled = false; + /* Set JCMDQ enqueue function */ + if (mldev->hw_queue_lock == 1) + mldev->ml_jcmdq_enqueue = roc_ml_jcmdq_enqueue_sl; + else + mldev->ml_jcmdq_enqueue = roc_ml_jcmdq_enqueue_lf; + dev->enqueue_burst = cn10k_ml_enqueue_burst; dev->dequeue_burst = cn10k_ml_dequeue_burst; dev->op_error_get = cn10k_ml_op_error_get; @@ -1993,7 +2007,7 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op req->result.user_ptr = op->user_ptr; plt_write64(ML_CN10K_POLL_JOB_START, &req->status); - enqueued = roc_ml_jcmdq_enqueue_lf(&mldev->roc, &req->jcmd); + enqueued = mldev->ml_jcmdq_enqueue(&mldev->roc, &req->jcmd); if (unlikely(!enqueued)) goto jcmdq_full; @@ -2114,7 +2128,7 @@ cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) timeout = true; req->timeout = plt_tsc_cycles() + ML_CN10K_CMD_TIMEOUT * plt_tsc_hz(); do { - if (roc_ml_jcmdq_enqueue_lf(&mldev->roc, &req->jcmd)) { + if (mldev->ml_jcmdq_enqueue(&mldev->roc, &req->jcmd)) { req->op = op; timeout = false; break; From patchwork Fri Mar 10 08:20:13 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124984 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 5243E41DD0; Fri, 10 Mar 2023 09:24:24 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 1D0E642FF2; Fri, 10 Mar 2023 09:21:14 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id ED1E242BAC for ; Fri, 10 Mar 2023 09:20:37 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7WbXI003211 for ; Fri, 10 Mar 2023 00:20:37 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=AijxDeQYX7HBXLogvBCXzfab1c0wdr1Fmd9o9IB1LFQ=; b=hKZ/LB6w+rKLdkWl+07C+onZd4iMD/DDlWLyrSJNm0tz2T0mNEAdI8aZYWhoXSlDgRrJ ARbA8Lsf9HdYPJtgcOxhLh9A35HYkRYQSP2PlqlmDgnem8C9CfuNT4LEHq1bgNxccSVi vQTeiHEnUn4GAEpEXfv4wu2mO7TyjaERZmKJ4AubhLsF+Cnp3d/x1HPpezCuZYbwhggj hymYQCAYtoyPfPb3McZHXWQszgHWDlOsGw1o0yDEBsCoQSt1BBQXp2/46rGOXadUPgo5 BuzWfUeMJL7ezNTP883VeCURMwXkAbQGeAJx5OoInw4NkKYM3fxDE6HQc2cSWFewMaDy AA== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0bf-3 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:36 -0800 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:24 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:24 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 7F5703F7097; Fri, 10 Mar 2023 00:20:24 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 37/39] ml/cnxk: add support to select poll memory region Date: Fri, 10 Mar 2023 00:20:13 -0800 Message-ID: <20230310082015.20200-38-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: nbxEwqv5C52t_zNdLX37KjxWy-iCkoTi X-Proofpoint-ORIG-GUID: nbxEwqv5C52t_zNdLX37KjxWy-iCkoTi X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added device argument "poll_mem" to select the memory region to be used for polling in fast-path requests. Implemented support to use scratch registers for polling. Available pool of scratch registers one-to-one mapped with the internal request queue. poll_mem: ddr: Use DDR memory location for polling (default) register: Use scratch registers polling Signed-off-by: Srikanth Yalavarthi --- drivers/ml/cnxk/cn10k_ml_dev.c | 47 +++++++++++-- drivers/ml/cnxk/cn10k_ml_dev.h | 24 +++++++ drivers/ml/cnxk/cn10k_ml_ops.c | 124 +++++++++++++++++++++++++++++++-- drivers/ml/cnxk/cn10k_ml_ops.h | 9 +++ 4 files changed, 192 insertions(+), 12 deletions(-) diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index aa503b2691..a746a66849 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -23,6 +23,7 @@ #define CN10K_ML_DEV_CACHE_MODEL_DATA "cache_model_data" #define CN10K_ML_OCM_ALLOC_MODE "ocm_alloc_mode" #define CN10K_ML_DEV_HW_QUEUE_LOCK "hw_queue_lock" +#define CN10K_ML_FW_POLL_MEM "poll_mem" #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 @@ -30,6 +31,7 @@ #define CN10K_ML_DEV_CACHE_MODEL_DATA_DEFAULT 1 #define CN10K_ML_OCM_ALLOC_MODE_DEFAULT "lowest" #define CN10K_ML_DEV_HW_QUEUE_LOCK_DEFAULT 1 +#define CN10K_ML_FW_POLL_MEM_DEFAULT "ddr" /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -42,6 +44,7 @@ /* Firmware flags */ #define FW_ENABLE_DPE_WARNING_BITMASK BIT(0) #define FW_REPORT_DPE_WARNING_BITMASK BIT(1) +#define FW_USE_DDR_POLL_ADDR_FP BIT(2) static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_FW_ENABLE_DPE_WARNINGS, @@ -49,6 +52,7 @@ static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_DEV_CACHE_MODEL_DATA, CN10K_ML_OCM_ALLOC_MODE, CN10K_ML_DEV_HW_QUEUE_LOCK, + CN10K_ML_FW_POLL_MEM, NULL}; /* Dummy operations for ML device */ @@ -92,7 +96,9 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde bool ocm_alloc_mode_set = false; bool hw_queue_lock_set = false; char *ocm_alloc_mode = NULL; + bool poll_mem_set = false; bool fw_path_set = false; + char *poll_mem = NULL; char *fw_path = NULL; int ret = 0; @@ -174,6 +180,17 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde hw_queue_lock_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_FW_POLL_MEM) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_FW_POLL_MEM, &parse_string_arg, + &poll_mem); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", CN10K_ML_FW_POLL_MEM); + ret = -EINVAL; + goto exit; + } + poll_mem_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -243,6 +260,18 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde } plt_info("ML: %s = %d", CN10K_ML_DEV_HW_QUEUE_LOCK, mldev->hw_queue_lock); + if (!poll_mem_set) { + mldev->fw.poll_mem = CN10K_ML_FW_POLL_MEM_DEFAULT; + } else { + if (!((strcmp(poll_mem, "ddr") == 0) || (strcmp(poll_mem, "register") == 0))) { + plt_err("Invalid argument, %s = %s\n", CN10K_ML_FW_POLL_MEM, poll_mem); + ret = -EINVAL; + goto exit; + } + mldev->fw.poll_mem = poll_mem; + } + plt_info("ML: %s = %s", CN10K_ML_FW_POLL_MEM, mldev->fw.poll_mem); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -376,6 +405,11 @@ cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw) if (fw->report_dpe_warnings) flags = flags | FW_REPORT_DPE_WARNING_BITMASK; + if (strcmp(fw->poll_mem, "ddr") == 0) + flags = flags | FW_USE_DDR_POLL_ADDR_FP; + else if (strcmp(fw->poll_mem, "register") == 0) + flags = flags & ~FW_USE_DDR_POLL_ADDR_FP; + return flags; } @@ -780,9 +814,10 @@ RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); -RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH - "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS - "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS - "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA - "=<0|1>" CN10K_ML_OCM_ALLOC_MODE - "=" CN10K_ML_DEV_HW_QUEUE_LOCK "=<0|1>"); +RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, + CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS + "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS + "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA + "=<0|1>" CN10K_ML_OCM_ALLOC_MODE + "=" CN10K_ML_DEV_HW_QUEUE_LOCK + "=<0|1>" CN10K_ML_FW_POLL_MEM "="); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 49676ac9e7..966d92e027 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -43,6 +43,18 @@ #define ML_CN10K_POLL_JOB_START 0 #define ML_CN10K_POLL_JOB_FINISH 1 +/* Memory barrier macros */ +#if defined(RTE_ARCH_ARM) +#define dmb_st ({ asm volatile("dmb st" : : : "memory"); }) +#define dsb_st ({ asm volatile("dsb st" : : : "memory"); }) +#else +#define dmb_st +#define dsb_st +#endif + +struct cn10k_ml_req; +struct cn10k_ml_qp; + /* Job types */ enum cn10k_ml_job_type { ML_CN10K_JOB_TYPE_MODEL_RUN = 0, @@ -358,6 +370,9 @@ struct cn10k_ml_fw { /* Report DPE warnings */ int report_dpe_warnings; + /* Memory to be used for polling in fast-path requests */ + const char *poll_mem; + /* Data buffer */ uint8_t *data; @@ -393,6 +408,15 @@ struct cn10k_ml_dev { /* JCMD enqueue function handler */ bool (*ml_jcmdq_enqueue)(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd); + + /* Poll handling function pointers */ + void (*set_poll_addr)(struct cn10k_ml_qp *qp, struct cn10k_ml_req *req, uint64_t idx); + void (*set_poll_ptr)(struct roc_ml *roc_ml, struct cn10k_ml_req *req); + uint64_t (*get_poll_ptr)(struct roc_ml *roc_ml, struct cn10k_ml_req *req); + + /* Memory barrier function pointers to handle synchronization */ + void (*set_enq_barrier)(void); + void (*set_deq_barrier)(void); }; uint64_t cn10k_ml_fw_flags_get(struct cn10k_ml_fw *fw); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index f92f778e23..61e6d023c5 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -23,6 +23,11 @@ #define ML_FLAGS_POLL_COMPL BIT(0) #define ML_FLAGS_SSO_COMPL BIT(1) +/* Scratch register range for poll mode requests */ +#define ML_POLL_REGISTER_SYNC 1023 +#define ML_POLL_REGISTER_START 1024 +#define ML_POLL_REGISTER_END 2047 + /* Error message length */ #define ERRMSG_LEN 32 @@ -76,6 +81,80 @@ print_line(FILE *fp, int len) fprintf(fp, "\n"); } +static inline void +cn10k_ml_set_poll_addr_ddr(struct cn10k_ml_qp *qp, struct cn10k_ml_req *req, uint64_t idx) +{ + PLT_SET_USED(qp); + PLT_SET_USED(idx); + + req->compl_W1 = PLT_U64_CAST(&req->status); +} + +static inline void +cn10k_ml_set_poll_addr_reg(struct cn10k_ml_qp *qp, struct cn10k_ml_req *req, uint64_t idx) +{ + req->compl_W1 = ML_SCRATCH(qp->block_start + idx % qp->block_size); +} + +static inline void +cn10k_ml_set_poll_ptr_ddr(struct roc_ml *roc_ml, struct cn10k_ml_req *req) +{ + PLT_SET_USED(roc_ml); + + plt_write64(ML_CN10K_POLL_JOB_START, req->compl_W1); +} + +static inline void +cn10k_ml_set_poll_ptr_reg(struct roc_ml *roc_ml, struct cn10k_ml_req *req) +{ + roc_ml_reg_write64(roc_ml, ML_CN10K_POLL_JOB_START, req->compl_W1); +} + +static inline uint64_t +cn10k_ml_get_poll_ptr_ddr(struct roc_ml *roc_ml, struct cn10k_ml_req *req) +{ + PLT_SET_USED(roc_ml); + + return plt_read64(req->compl_W1); +} + +static inline uint64_t +cn10k_ml_get_poll_ptr_reg(struct roc_ml *roc_ml, struct cn10k_ml_req *req) +{ + return roc_ml_reg_read64(roc_ml, req->compl_W1); +} + +static inline void +cn10k_ml_set_sync_addr(struct cn10k_ml_dev *mldev, struct cn10k_ml_req *req) +{ + if (strcmp(mldev->fw.poll_mem, "ddr") == 0) + req->compl_W1 = PLT_U64_CAST(&req->status); + else if (strcmp(mldev->fw.poll_mem, "register") == 0) + req->compl_W1 = ML_SCRATCH(ML_POLL_REGISTER_SYNC); +} + +static inline void +cn10k_ml_enq_barrier_ddr(void) +{ +} + +static inline void +cn10k_ml_deq_barrier_ddr(void) +{ +} + +static inline void +cn10k_ml_enq_barrier_register(void) +{ + dmb_st; +} + +static inline void +cn10k_ml_deq_barrier_register(void) +{ + dsb_st; +} + static void qp_memzone_name_get(char *name, int size, int dev_id, int qp_id) { @@ -163,6 +242,9 @@ cn10k_ml_qp_create(const struct rte_ml_dev *dev, uint16_t qp_id, uint32_t nb_des qp->stats.dequeued_count = 0; qp->stats.enqueue_err_count = 0; qp->stats.dequeue_err_count = 0; + qp->block_size = + (ML_POLL_REGISTER_END - ML_POLL_REGISTER_START + 1) / dev->data->nb_queue_pairs; + qp->block_start = ML_POLL_REGISTER_START + qp_id * qp->block_size; /* Initialize job command */ for (i = 0; i < qp->nb_desc; i++) { @@ -341,7 +423,7 @@ cn10k_ml_prep_fp_job_descriptor(struct rte_ml_dev *dev, struct cn10k_ml_req *req mldev = dev->data->dev_private; req->jd.hdr.jce.w0.u64 = 0; - req->jd.hdr.jce.w1.u64 = PLT_U64_CAST(&req->status); + req->jd.hdr.jce.w1.u64 = req->compl_W1; req->jd.hdr.model_id = op->model_id; req->jd.hdr.job_type = ML_CN10K_JOB_TYPE_MODEL_RUN; req->jd.hdr.fp_flags = ML_FLAGS_POLL_COMPL; @@ -549,7 +631,11 @@ cn10k_ml_dev_info_get(struct rte_ml_dev *dev, struct rte_ml_dev_info *dev_info) else dev_info->max_queue_pairs = ML_CN10K_MAX_QP_PER_DEVICE_LF; - dev_info->max_desc = ML_CN10K_MAX_DESC_PER_QP; + if (strcmp(mldev->fw.poll_mem, "register") == 0) + dev_info->max_desc = ML_CN10K_MAX_DESC_PER_QP / dev_info->max_queue_pairs; + else if (strcmp(mldev->fw.poll_mem, "ddr") == 0) + dev_info->max_desc = ML_CN10K_MAX_DESC_PER_QP; + dev_info->max_segments = ML_CN10K_MAX_SEGMENTS; dev_info->min_align_size = ML_CN10K_ALIGN_SIZE; @@ -717,6 +803,26 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c else mldev->ml_jcmdq_enqueue = roc_ml_jcmdq_enqueue_lf; + /* Set polling function pointers */ + if (strcmp(mldev->fw.poll_mem, "ddr") == 0) { + mldev->set_poll_addr = cn10k_ml_set_poll_addr_ddr; + mldev->set_poll_ptr = cn10k_ml_set_poll_ptr_ddr; + mldev->get_poll_ptr = cn10k_ml_get_poll_ptr_ddr; + } else if (strcmp(mldev->fw.poll_mem, "register") == 0) { + mldev->set_poll_addr = cn10k_ml_set_poll_addr_reg; + mldev->set_poll_ptr = cn10k_ml_set_poll_ptr_reg; + mldev->get_poll_ptr = cn10k_ml_get_poll_ptr_reg; + } + + /* Set barrier function pointers */ + if (strcmp(mldev->fw.poll_mem, "ddr") == 0) { + mldev->set_enq_barrier = cn10k_ml_enq_barrier_ddr; + mldev->set_deq_barrier = cn10k_ml_deq_barrier_ddr; + } else if (strcmp(mldev->fw.poll_mem, "register") == 0) { + mldev->set_enq_barrier = cn10k_ml_enq_barrier_register; + mldev->set_deq_barrier = cn10k_ml_deq_barrier_register; + } + dev->enqueue_burst = cn10k_ml_enqueue_burst; dev->dequeue_burst = cn10k_ml_dequeue_burst; dev->op_error_get = cn10k_ml_op_error_get; @@ -2000,13 +2106,15 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op op = ops[count]; req = &queue->reqs[head]; + mldev->set_poll_addr(qp, req, head); cn10k_ml_prep_fp_job_descriptor(dev, req, op); memset(&req->result, 0, sizeof(struct cn10k_ml_result)); req->result.error_code.s.etype = ML_ETYPE_UNKNOWN; req->result.user_ptr = op->user_ptr; + mldev->set_enq_barrier(); - plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + mldev->set_poll_ptr(&mldev->roc, req); enqueued = mldev->ml_jcmdq_enqueue(&mldev->roc, &req->jcmd); if (unlikely(!enqueued)) goto jcmdq_full; @@ -2032,6 +2140,7 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op uint16_t nb_ops) { struct cn10k_ml_queue *queue; + struct cn10k_ml_dev *mldev; struct cn10k_ml_req *req; struct cn10k_ml_qp *qp; @@ -2039,6 +2148,7 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op uint16_t count; uint64_t tail; + mldev = dev->data->dev_private; qp = dev->data->queue_pairs[qp_id]; queue = &qp->queue; @@ -2051,7 +2161,7 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op dequeue_req: req = &queue->reqs[tail]; - status = plt_read64(&req->status); + status = mldev->get_poll_ptr(&mldev->roc, req); if (unlikely(status != ML_CN10K_POLL_JOB_FINISH)) { if (plt_tsc_cycles() < req->timeout) goto empty_or_active; @@ -2059,6 +2169,7 @@ cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op req->result.error_code.s.etype = ML_ETYPE_DRIVER; } + mldev->set_deq_barrier(); cn10k_ml_result_update(dev, qp_id, &req->result, req->op); ops[count] = req->op; @@ -2116,13 +2227,14 @@ cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) model = dev->data->models[op->model_id]; req = model->req; + cn10k_ml_set_sync_addr(mldev, req); cn10k_ml_prep_fp_job_descriptor(dev, req, op); memset(&req->result, 0, sizeof(struct cn10k_ml_result)); req->result.error_code.s.etype = ML_ETYPE_UNKNOWN; req->result.user_ptr = op->user_ptr; - plt_write64(ML_CN10K_POLL_JOB_START, &req->status); + mldev->set_poll_ptr(&mldev->roc, req); req->jcmd.w1.s.jobptr = PLT_U64_CAST(&req->jd); timeout = true; @@ -2142,7 +2254,7 @@ cn10k_ml_inference_sync(struct rte_ml_dev *dev, struct rte_ml_op *op) timeout = true; do { - if (plt_read64(&req->status) == ML_CN10K_POLL_JOB_FINISH) { + if (mldev->get_poll_ptr(&mldev->roc, req) == ML_CN10K_POLL_JOB_FINISH) { timeout = false; break; } diff --git a/drivers/ml/cnxk/cn10k_ml_ops.h b/drivers/ml/cnxk/cn10k_ml_ops.h index 65ae8b44f3..58c992720a 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.h +++ b/drivers/ml/cnxk/cn10k_ml_ops.h @@ -26,6 +26,9 @@ struct cn10k_ml_req { /* Job command */ struct ml_job_cmd_s jcmd; + /* Job completion W1 */ + uint64_t compl_W1; + /* Timeout cycle */ uint64_t timeout; @@ -61,6 +64,12 @@ struct cn10k_ml_qp { /* Statistics per queue-pair */ struct rte_ml_dev_stats stats; + + /* Register block start for polling */ + uint32_t block_start; + + /* Register block end for polling */ + uint32_t block_size; }; /* Device ops */ From patchwork Fri Mar 10 08:20:14 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124972 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 0E2D341D3D; Fri, 10 Mar 2023 09:23:15 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id A401442DBA; Fri, 10 Mar 2023 09:20:59 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id 81C1142C24 for ; Fri, 10 Mar 2023 09:20:31 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1M009730; Fri, 10 Mar 2023 00:20:30 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=QXP5s9lMw0HalDEJt6ny2e60FqPucBaYSMAZa11xPuE=; b=NKgDgVBbJqSkbh3Yd3hOCg9GH4AWy3Pdtqgi9aDYH7fyw2nnxr+9q/pNmkVwRd7+gDXV cqggBZzcYuilZuECu1JEC0y50LXxO5yxyPPWlab+23mV6L3k97ZbB9Q55Mwhqc7rMU6H 2C39YkiBd5KkglMg4e7pLQ5F9PmOBvz/DiRDOGRkC9UqEluXk+FBePTFmwuek93ZImuz +fR4MHF2g/pd2ljISVA9l2MA1A3q+90lmYecKCm1w5K7cevv/JMkLP3X7UoB5hQskrkP WRM2m6weKc9SFL+NnG1zX+zfFuLPbhCjnDqqewEcNEHnehGMFzDs0w+wnd4OKX4V5+EN 1w== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-17 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Fri, 10 Mar 2023 00:20:30 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:25 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:25 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id AD8523F7098; Fri, 10 Mar 2023 00:20:24 -0800 (PST) From: Srikanth Yalavarthi To: Thomas Monjalon , Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 38/39] ml/cnxk: add user guide for marvell cnxk ml driver Date: Fri, 10 Mar 2023 00:20:14 -0800 Message-ID: <20230310082015.20200-39-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: JwHtI9B2uDTWPou51w0WcSJ5LXlMGvhC X-Proofpoint-ORIG-GUID: JwHtI9B2uDTWPou51w0WcSJ5LXlMGvhC X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added user guide for Marvell cnxk ML driver for Marvell Octeon cnxk Soc family. Added details about device initialization, debug options and runtime device args supported by the driver. Signed-off-by: Srikanth Yalavarthi --- MAINTAINERS | 1 + doc/guides/index.rst | 1 + doc/guides/mldevs/cnxk.rst | 238 ++++++++++++++++++++++++++++++++++++ doc/guides/mldevs/index.rst | 14 +++ 4 files changed, 254 insertions(+) create mode 100644 doc/guides/mldevs/cnxk.rst create mode 100644 doc/guides/mldevs/index.rst diff --git a/MAINTAINERS b/MAINTAINERS index 8f695516c7..ae6c4decbe 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1450,6 +1450,7 @@ M: Srikanth Yalavarthi F: drivers/common/cnxk/hw/ml.h F: drivers/common/cnxk/roc_ml* F: drivers/ml/cnxk/ +F: doc/guides/mldevs/cnxk.rst Packet processing diff --git a/doc/guides/index.rst b/doc/guides/index.rst index 5eb5bd9c9a..0bd729530a 100644 --- a/doc/guides/index.rst +++ b/doc/guides/index.rst @@ -26,6 +26,7 @@ DPDK documentation eventdevs/index rawdevs/index mempool/index + mldevs/index platform/index contributing/index rel_notes/index diff --git a/doc/guides/mldevs/cnxk.rst b/doc/guides/mldevs/cnxk.rst new file mode 100644 index 0000000000..da40336299 --- /dev/null +++ b/doc/guides/mldevs/cnxk.rst @@ -0,0 +1,238 @@ +.. SPDX-License-Identifier: BSD-3-Clause + Copyright (c) 2022 Marvell. + +Marvell cnxk Machine Learning Poll Mode Driver +============================================== + +The cnxk ML poll mode driver provides support for offloading Machine +Learning inference operations to Machine Learning accelerator units +on the **Marvell OCTEON cnxk** SoC family. + +The cnxk ML PMD code is organized into multiple files with all file names +starting with cn10k, providing support for CN106XX and CN106XXS. + +More information about OCTEON cnxk SoCs may be obtained from ``_ + +Supported OCTEON cnxk SoCs +-------------------------- + +- CN106XX +- CN106XXS + +Features +-------- + +The OCTEON cnxk ML PMD provides support for the following set of operations: + +Slow-path device and ML model handling: + +* ``Device probing, configuration and close`` +* ``Device start / stop`` +* ``Model loading and unloading`` +* ``Model start / stop`` +* ``Data quantization and dequantization`` + +Fast-path Inference: + +* ``Inference execution`` +* ``Error handling`` + + +Installation +------------ + +The OCTEON cnxk ML PMD may be compiled natively on an OCTEON cnxk platform +or cross-compiled on an x86 platform. + +Refer to :doc:`../platform/cnxk` for instructions to build your DPDK +application. + + +Initialization +-------------- + +``CN10K Initialization`` + +List the ML PF devices available on cn10k platform: + +.. code-block:: console + + lspci -d:a092 + +``a092`` is the ML device PF id. You should see output similar to: + +.. code-block:: console + + 0000:00:10.0 System peripheral: Cavium, Inc. Device a092 + +Bind the ML PF device to the vfio_pci driver: + +.. code-block:: console + + cd + ./usertools/dpdk-devbind.py -u 0000:00:10.0 + ./usertools/dpdk-devbind.py -b vfio-pci 0000:00:10.0 + +Runtime Config Options +---------------------- + +- ``Firmware file path`` (default ``/lib/firmware/mlip-fw.bin``) + + Path to the firmware binary to be loaded during device configuration. + The ``fw_path`` ``devargs`` parameter can be used by the user to load + ML firmware from a custom path. + + For example:: + + -a 0000:00:10.0,fw_path="/home/user/ml_fw.bin" + + With the above configuration, driver loads the firmware from the path + "/home/user/ml_fw.bin". + +- ``Enable DPE warnings`` (default ``1``) + + ML firmware can be configured during load to handle the DPE errors reported + by ML inference engine. When enabled, firmware would mask the DPE non-fatal + hardware errors as warnings. The parameter ``enable_dpe_warnings`` ``devargs`` + is used fo this configuration. + + For example:: + + -a 0000:00:10.0,enable_dpe_warnings=0 + + With the above configuration, DPE non-fatal errors reported by HW are + considered as errors. + + +- ``Model data caching`` (default ``1``) + + Enable caching model data on ML ACC cores. Enabling this option executes a + dummy inference request in synchronous mode during model start stage. Caching + of model data improves the inferencing throughput / latency for the model. + The parameter ``cache_model_data`` ``devargs`` is used to enable data caching. + + For example:: + + -a 0000:00:10.0,cache_model_data=0 + + With the above configuration, model data caching is disabled. + + +- ``OCM allocation mode`` (default ``lowest``) + + Option to specify the method to be used while allocating OCM memory for a + model during model start. Two modes are supported by the driver. The + parameter ``ocm_alloc_mode`` ``devargs`` is used to select the OCM + allocation mode. + + ``lowest`` - Allocate OCM for the model from first available free slot. Search + for the free slot is done starting from the lowest tile ID and lowest page ID. + ``largest`` - Allocate OCM for the model from the slot with largest amount of + free space. + + For example:: + + -a 0000:00:10.0,ocm_alloc_mode=lowest + + With the above configuration, OCM allocation fo the model would be done from + the first available free slot / from the lowest possible tile ID. + + +- ``Enable hardware queue lock`` (default ``0``) + + Option to select the job request enqueue function to used to queue the requests + to hardware queue. The parameter ``hw_queue_lock`` ``devargs`` is used to select + the enqueue function. + + ``0`` - Disable (default), use lock free version of hardware enqueue function + for job queuing in enqueue burst operation. To avoid race condition in request + queuing to hardware, disabling hw_queue_lock restricts the number of queue-pairs + supported by cnxk driver to 1. + ``1`` - Enable, use spin-lock version of hardware enqueue function for job queuing. + Enabling spinlock version would disable restrictions on the number of queue-pairs + that can be supported by the driver. + + For example:: + + -a 0000:00:10.0,hw_queue_lock=1 + + With the above configuration, spinlock version of hardware enqueue function is used + in the fast path enqueue burst operation. + + +- ``Polling memory location`` (default ``ddr``) + + ML cnxk driver provides the option to select the memory location to be used + for polling to check the inference request completion. Driver supports using + the either DDR address space (``ddr``) or ML registers (``register``) as + polling locations. The parameter ``poll_mem`` ``devargs`` is used to specify + the poll location. + + For example:: + + -a 0000:00:10.0,poll_mem="register" + + With the above configuration, ML cnxk driver is configured to use ML registers + for polling in fastpath requests. + + +Debugging Options +----------------- + +.. _table_octeon_cnxk_ml_debug_options: + +.. table:: OCTEON cnxk ML PMD debug options + + +---+------------+-------------------------------------------------------+ + | # | Component | EAL log command | + +===+============+=======================================================+ + | 1 | ML | --log-level='pmd\.ml\.cnxk,8' | + +---+------------+-------------------------------------------------------+ + + +Extended stats +-------------- + +Marvell cnxk ML PMD supports reporting the inference latencies through extended +stats. The PMD supports the below list of 6 extended stats types per each model. +Total number of extended stats would be equal to 6 x number of models loaded. + +.. _table_octeon_cnxk_ml_xstats_names: + +.. table:: OCTEON cnxk ML PMD xstats names + + +---+---------------------+----------------------------------------------+ + | # | Type | Description | + +===+=====================+==============================================+ + | 1 | Avg-HW-Latency | Average hardware latency | + +---+---------------------+----------------------------------------------+ + | 2 | Min-HW-Latency | Minimum hardware latency | + +---+---------------------+----------------------------------------------+ + | 3 | Max-HW-Latency | Maximum hardware latency | + +---+---------------------+----------------------------------------------+ + | 4 | Avg-HW-Latency | Average firmware latency | + +---+---------------------+----------------------------------------------+ + | 5 | Avg-HW-Latency | Minimum firmware latency | + +---+---------------------+----------------------------------------------+ + | 6 | Avg-HW-Latency | Maximum firmware latency | + +---+---------------------+----------------------------------------------+ + +Latency values reported by the PMD through xstats can have units, either in +cycles or nano seconds. The units of the latency is determined during DPDK +initialization and would depend on the availability of SCLK. Latencies are +reported in nao seconds when the SCLK is available and in cycles otherwise. +Application needs to initialize at least one RVU for the clock to be available. + +xstats names are dynamically generated by the PMD and would have the format +"Model--Type-". + +For example:: + Model-1-Avg-FW-Latency-ns + +The above xstat name would report average firmware latency in nano seconds for +model with model ID 1. + +Number of xstats made available by the PMD change dynamically. The number would +increase with loading a model and would decrease with unloading a model. +Application needs to update the xstats map after a model is either loaded or +unloaded. diff --git a/doc/guides/mldevs/index.rst b/doc/guides/mldevs/index.rst new file mode 100644 index 0000000000..f201e54175 --- /dev/null +++ b/doc/guides/mldevs/index.rst @@ -0,0 +1,14 @@ +.. SPDX-License-Identifier: BSD-3-Clause + Copyright (c) 2022 Marvell. + +Machine Learning Device Driver +============================== + +The following are a list of ML device PMDs, which can be used from an +application through the ML device API. + +.. toctree:: + :maxdepth: 2 + :numbered: + + cnxk From patchwork Fri Mar 10 08:20:15 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 124974 X-Patchwork-Delegate: jerinj@marvell.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 2F2B441D3D; Fri, 10 Mar 2023 09:23:26 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id C30A842F86; Fri, 10 Mar 2023 09:21:01 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0a-0016f401.pphosted.com [67.231.148.174]) by mails.dpdk.org (Postfix) with ESMTP id E82B042BFE for ; Fri, 10 Mar 2023 09:20:31 +0100 (CET) Received: from pps.filterd (m0045849.ppops.net [127.0.0.1]) by mx0a-0016f401.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 32A7aV1N009730 for ; Fri, 10 Mar 2023 00:20:31 -0800 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=pfpt0220; bh=mLw/boaeUKCS/S/THxRyDeUFO1CIEDV0UD/Oaedr4jw=; b=Yn9oHLjgwmNfGzAV8xr9rofpYfJCl+kDh57lxS3B5YBU1SKqCVWMKXInwKpJlZojEQHP 1MhF6EtJYfqknhCq5enRJItfQXpynN6iR70jldIF+FHJHv5dLlsMpO9XDmoVD2V/WhkT MnKrngvD4QXoH8kVDs7aVuVJF0EOWoaulHHBvr1mG1+iK4i90j9ZLD3J4t0WFs9Pqdjh pwNCjX7npFbyZ8qlc73w4OBcYb9XXcYu+7hfUNYd3VYwxMMMdoyXacKiPv7SFD22xTJz GrV+wYSE3P5cnM8IdFCm6vQicS1ZAJJXN+1NKJmm53Q2UiWLua29xh7sPr4k2tWpOg5A vA== Received: from dc5-exch02.marvell.com ([199.233.59.182]) by mx0a-0016f401.pphosted.com (PPS) with ESMTPS id 3p7n7dj0a4-18 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT) for ; Fri, 10 Mar 2023 00:20:30 -0800 Received: from DC5-EXCH02.marvell.com (10.69.176.39) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server (TLS) id 15.0.1497.42; Fri, 10 Mar 2023 00:20:25 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.42 via Frontend Transport; Fri, 10 Mar 2023 00:20:25 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id DBF773F709F; Fri, 10 Mar 2023 00:20:24 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi CC: , , , , , Subject: [PATCH v6 39/39] ml/cnxk: add support for configurable ocm page Date: Fri, 10 Mar 2023 00:20:15 -0800 Message-ID: <20230310082015.20200-40-syalavarthi@marvell.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20230310082015.20200-1-syalavarthi@marvell.com> References: <20221208200220.20267-1-syalavarthi@marvell.com> <20230310082015.20200-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-GUID: af59YqRy3GNR9dHYwB4VpgrL6WqVTZoz X-Proofpoint-ORIG-GUID: af59YqRy3GNR9dHYwB4VpgrL6WqVTZoz X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-03-10_02,2023-03-09_01,2023-02-09_01 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Added support for configurable OCM page size. A new device argument "ocm_page_size" is added to specify the page size for OCM management. Supported page sizes are 1KB, 2KB, 4KB, 8KB and 16KB. Default page size is 16KB. Signed-off-by: Srikanth Yalavarthi --- doc/guides/mldevs/cnxk.rst | 16 +++++++++ drivers/ml/cnxk/cn10k_ml_dev.c | 61 ++++++++++++++++++++++++++++---- drivers/ml/cnxk/cn10k_ml_dev.h | 3 ++ drivers/ml/cnxk/cn10k_ml_model.c | 6 ++-- drivers/ml/cnxk/cn10k_ml_ocm.c | 18 +++++++--- drivers/ml/cnxk/cn10k_ml_ocm.h | 14 +++----- drivers/ml/cnxk/cn10k_ml_ops.c | 17 ++++++--- 7 files changed, 107 insertions(+), 28 deletions(-) diff --git a/doc/guides/mldevs/cnxk.rst b/doc/guides/mldevs/cnxk.rst index da40336299..f7f61e8bfa 100644 --- a/doc/guides/mldevs/cnxk.rst +++ b/doc/guides/mldevs/cnxk.rst @@ -175,6 +175,22 @@ Runtime Config Options With the above configuration, ML cnxk driver is configured to use ML registers for polling in fastpath requests. +- ``OCM page size`` (default ``16384``) + + Option to specify the page size in bytes to be used for OCM management. Available + OCM is split into multiple pages of specified sizes and the pages are allocated to + the models. The parameter ``ocm_page_size`` ``devargs`` is used to specify the page + size to be used. + + Supported page sizes by the driver are 1 KB, 2 KB, 4 KB, 8 KB and 16 KB. Default + page size is 16 KB. + + For example:: + + -a 0000:00:10.0,ocm_page_size=8192 + + With the above configuration, page size of OCM is set to 8192 bytes / 8 KB. + Debugging Options ----------------- diff --git a/drivers/ml/cnxk/cn10k_ml_dev.c b/drivers/ml/cnxk/cn10k_ml_dev.c index a746a66849..6f9a1015a6 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.c +++ b/drivers/ml/cnxk/cn10k_ml_dev.c @@ -24,6 +24,7 @@ #define CN10K_ML_OCM_ALLOC_MODE "ocm_alloc_mode" #define CN10K_ML_DEV_HW_QUEUE_LOCK "hw_queue_lock" #define CN10K_ML_FW_POLL_MEM "poll_mem" +#define CN10K_ML_OCM_PAGE_SIZE "ocm_page_size" #define CN10K_ML_FW_PATH_DEFAULT "/lib/firmware/mlip-fw.bin" #define CN10K_ML_FW_ENABLE_DPE_WARNINGS_DEFAULT 1 @@ -32,6 +33,7 @@ #define CN10K_ML_OCM_ALLOC_MODE_DEFAULT "lowest" #define CN10K_ML_DEV_HW_QUEUE_LOCK_DEFAULT 1 #define CN10K_ML_FW_POLL_MEM_DEFAULT "ddr" +#define CN10K_ML_OCM_PAGE_SIZE_DEFAULT 16384 /* ML firmware macros */ #define FW_MEMZONE_NAME "ml_cn10k_fw_mz" @@ -53,8 +55,12 @@ static const char *const valid_args[] = {CN10K_ML_FW_PATH, CN10K_ML_OCM_ALLOC_MODE, CN10K_ML_DEV_HW_QUEUE_LOCK, CN10K_ML_FW_POLL_MEM, + CN10K_ML_OCM_PAGE_SIZE, NULL}; +/* Supported OCM page sizes: 1KB, 2KB, 4KB, 8KB and 16KB */ +static const int valid_ocm_page_size[] = {1024, 2048, 4096, 8192, 16384}; + /* Dummy operations for ML device */ struct rte_ml_dev_ops ml_dev_dummy_ops = {0}; @@ -95,12 +101,15 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde struct rte_kvargs *kvlist = NULL; bool ocm_alloc_mode_set = false; bool hw_queue_lock_set = false; + bool ocm_page_size_set = false; char *ocm_alloc_mode = NULL; bool poll_mem_set = false; bool fw_path_set = false; char *poll_mem = NULL; char *fw_path = NULL; int ret = 0; + bool found; + uint8_t i; if (devargs == NULL) goto check_args; @@ -191,6 +200,17 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde poll_mem_set = true; } + if (rte_kvargs_count(kvlist, CN10K_ML_OCM_PAGE_SIZE) == 1) { + ret = rte_kvargs_process(kvlist, CN10K_ML_OCM_PAGE_SIZE, &parse_integer_arg, + &mldev->ocm_page_size); + if (ret < 0) { + plt_err("Error processing arguments, key = %s\n", CN10K_ML_OCM_PAGE_SIZE); + ret = -EINVAL; + goto exit; + } + ocm_page_size_set = true; + } + check_args: if (!fw_path_set) mldev->fw.path = CN10K_ML_FW_PATH_DEFAULT; @@ -272,6 +292,32 @@ cn10k_mldev_parse_devargs(struct rte_devargs *devargs, struct cn10k_ml_dev *mlde } plt_info("ML: %s = %s", CN10K_ML_FW_POLL_MEM, mldev->fw.poll_mem); + if (!ocm_page_size_set) { + mldev->ocm_page_size = CN10K_ML_OCM_PAGE_SIZE_DEFAULT; + } else { + if (mldev->ocm_page_size < 0) { + plt_err("Invalid argument, %s = %d\n", CN10K_ML_OCM_PAGE_SIZE, + mldev->ocm_page_size); + ret = -EINVAL; + goto exit; + } + + found = false; + for (i = 0; i < PLT_DIM(valid_ocm_page_size); i++) { + if (mldev->ocm_page_size == valid_ocm_page_size[i]) { + found = true; + break; + } + } + + if (!found) { + plt_err("Unsupported ocm_page_size = %d\n", mldev->ocm_page_size); + ret = -EINVAL; + goto exit; + } + } + plt_info("ML: %s = %d", CN10K_ML_OCM_PAGE_SIZE, mldev->ocm_page_size); + exit: if (kvlist) rte_kvargs_free(kvlist); @@ -814,10 +860,11 @@ RTE_PMD_REGISTER_PCI(MLDEV_NAME_CN10K_PMD, cn10k_mldev_pmd); RTE_PMD_REGISTER_PCI_TABLE(MLDEV_NAME_CN10K_PMD, pci_id_ml_table); RTE_PMD_REGISTER_KMOD_DEP(MLDEV_NAME_CN10K_PMD, "vfio-pci"); -RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, - CN10K_ML_FW_PATH "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS - "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS - "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA - "=<0|1>" CN10K_ML_OCM_ALLOC_MODE - "=" CN10K_ML_DEV_HW_QUEUE_LOCK - "=<0|1>" CN10K_ML_FW_POLL_MEM "="); +RTE_PMD_REGISTER_PARAM_STRING(MLDEV_NAME_CN10K_PMD, CN10K_ML_FW_PATH + "=" CN10K_ML_FW_ENABLE_DPE_WARNINGS + "=<0|1>" CN10K_ML_FW_REPORT_DPE_WARNINGS + "=<0|1>" CN10K_ML_DEV_CACHE_MODEL_DATA + "=<0|1>" CN10K_ML_OCM_ALLOC_MODE + "=" CN10K_ML_DEV_HW_QUEUE_LOCK + "=<0|1>" CN10K_ML_FW_POLL_MEM "=" CN10K_ML_OCM_PAGE_SIZE + "=<1024|2048|4096|8192|16384>"); diff --git a/drivers/ml/cnxk/cn10k_ml_dev.h b/drivers/ml/cnxk/cn10k_ml_dev.h index 966d92e027..b4e46899c0 100644 --- a/drivers/ml/cnxk/cn10k_ml_dev.h +++ b/drivers/ml/cnxk/cn10k_ml_dev.h @@ -406,6 +406,9 @@ struct cn10k_ml_dev { /* Use spinlock version of ROC enqueue */ int hw_queue_lock; + /* OCM page size */ + int ocm_page_size; + /* JCMD enqueue function handler */ bool (*ml_jcmdq_enqueue)(struct roc_ml *roc_ml, struct ml_job_cmd_s *job_cmd); diff --git a/drivers/ml/cnxk/cn10k_ml_model.c b/drivers/ml/cnxk/cn10k_ml_model.c index 0ded355d81..ceffde8459 100644 --- a/drivers/ml/cnxk/cn10k_ml_model.c +++ b/drivers/ml/cnxk/cn10k_ml_model.c @@ -339,11 +339,11 @@ cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, ui scratch_size, *scratch_pages); /* Check if the model can be loaded on OCM */ - if ((*wb_pages + *scratch_pages) > ML_CN10K_OCM_NUMPAGES) { + if ((*wb_pages + *scratch_pages) > mldev->ocm.num_pages) { plt_err("Cannot create the model, OCM relocatable = %u", metadata->model.ocm_relocatable); plt_err("wb_pages (%u) + scratch_pages (%u) > %u", *wb_pages, *scratch_pages, - ML_CN10K_OCM_NUMPAGES); + mldev->ocm.num_pages); return -ENOMEM; } @@ -352,7 +352,7 @@ cn10k_ml_model_ocm_pages_count(struct cn10k_ml_dev *mldev, uint16_t model_id, ui */ if (!metadata->model.ocm_relocatable) *scratch_pages = - PLT_MAX(PLT_U64_CAST(*scratch_pages), PLT_U64_CAST(ML_CN10K_OCM_NUMPAGES)); + PLT_MAX(PLT_U64_CAST(*scratch_pages), PLT_U64_CAST(mldev->ocm.num_pages)); return 0; } diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.c b/drivers/ml/cnxk/cn10k_ml_ocm.c index 551faef7eb..d8d2c71a3c 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.c +++ b/drivers/ml/cnxk/cn10k_ml_ocm.c @@ -220,13 +220,13 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w struct cn10k_ml_dev *mldev; struct cn10k_ml_ocm *ocm; - uint8_t local_ocm_mask[ML_CN10K_OCM_MASKWORDS] = {0}; uint16_t used_scratch_pages_max; uint16_t scratch_page_start; int used_last_wb_page_max; uint16_t scratch_page_end; uint8_t search_start_tile; uint8_t search_end_tile; + uint8_t *local_ocm_mask; int wb_page_start_curr; int max_slot_sz_curr; uint8_t tile_start; @@ -268,6 +268,9 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w search_end_tile = start_tile; } + /* nibbles + prefix '0x' */ + local_ocm_mask = rte_zmalloc("local_ocm_mask", mldev->ocm.mask_words, RTE_CACHE_LINE_SIZE); + tile_start = search_start_tile; start_search: used_scratch_pages_max = 0; @@ -279,7 +282,7 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w PLT_MAX(ocm->tile_ocm_info[tile_id].last_wb_page, used_last_wb_page_max); } - memset(local_ocm_mask, 0, sizeof(local_ocm_mask)); + memset(local_ocm_mask, 0, mldev->ocm.mask_words); for (tile_id = tile_start; tile_id < tile_start + num_tiles; tile_id++) { for (word_id = 0; word_id < ocm->mask_words; word_id++) local_ocm_mask[word_id] |= ocm->tile_ocm_info[tile_id].ocm_mask[word_id]; @@ -332,6 +335,8 @@ cn10k_ml_ocm_tilemask_find(struct rte_ml_dev *dev, uint8_t num_tiles, uint16_t w if (wb_page_start != -1) *tilemask = GENMASK_ULL(tile_idx + num_tiles - 1, tile_idx); + rte_free(local_ocm_mask); + return wb_page_start; } @@ -480,7 +485,7 @@ cn10k_ml_ocm_pagemask_to_str(struct cn10k_ml_ocm_tile_info *tile_info, uint16_t void cn10k_ml_ocm_print(struct rte_ml_dev *dev, FILE *fp) { - char str[ML_CN10K_OCM_NUMPAGES / 4 + 2]; /* nibbles + prefix '0x' */ + char *str; struct cn10k_ml_dev *mldev; struct cn10k_ml_ocm *ocm; uint8_t tile_id; @@ -490,12 +495,15 @@ cn10k_ml_ocm_print(struct rte_ml_dev *dev, FILE *fp) mldev = dev->data->dev_private; ocm = &mldev->ocm; + /* nibbles + prefix '0x' */ + str = rte_zmalloc("ocm_mask_str", mldev->ocm.num_pages / 4 + 2, RTE_CACHE_LINE_SIZE); + fprintf(fp, "OCM State:\n"); for (tile_id = 0; tile_id < ocm->num_tiles; tile_id++) { cn10k_ml_ocm_pagemask_to_str(&ocm->tile_ocm_info[tile_id], ocm->mask_words, str); wb_pages = 0 - ocm->tile_ocm_info[tile_id].scratch_pages; - for (word_id = 0; word_id < ML_CN10K_OCM_MASKWORDS; word_id++) + for (word_id = 0; word_id < mldev->ocm.mask_words; word_id++) wb_pages += __builtin_popcount(ocm->tile_ocm_info[tile_id].ocm_mask[word_id]); @@ -506,4 +514,6 @@ cn10k_ml_ocm_print(struct rte_ml_dev *dev, FILE *fp) tile_id, ocm->tile_ocm_info[tile_id].scratch_pages, wb_pages, ocm->tile_ocm_info[tile_id].last_wb_page, str); } + + rte_free(str); } diff --git a/drivers/ml/cnxk/cn10k_ml_ocm.h b/drivers/ml/cnxk/cn10k_ml_ocm.h index 5f018b410a..3404e7fd65 100644 --- a/drivers/ml/cnxk/cn10k_ml_ocm.h +++ b/drivers/ml/cnxk/cn10k_ml_ocm.h @@ -8,25 +8,16 @@ #include #include -/* Page size in bytes. */ -#define ML_CN10K_OCM_PAGESIZE 0x4000 - /* Number of OCM tiles. */ #define ML_CN10K_OCM_NUMTILES 0x8 /* OCM in bytes, per tile. */ #define ML_CN10K_OCM_TILESIZE 0x100000 -/* OCM pages, per tile. */ -#define ML_CN10K_OCM_NUMPAGES (ML_CN10K_OCM_TILESIZE / ML_CN10K_OCM_PAGESIZE) - -/* Maximum OCM mask words, per tile, 8 bit words. */ -#define ML_CN10K_OCM_MASKWORDS (ML_CN10K_OCM_NUMPAGES / 8) - /* OCM and Tile information structure */ struct cn10k_ml_ocm_tile_info { /* Mask of used / allotted pages on tile's OCM */ - uint8_t ocm_mask[ML_CN10K_OCM_MASKWORDS]; + uint8_t *ocm_mask; /* Last pages in the tile's OCM used for weights and bias, default = -1 */ int last_wb_page; @@ -78,6 +69,9 @@ struct cn10k_ml_ocm { /* OCM memory info and status*/ struct cn10k_ml_ocm_tile_info tile_ocm_info[ML_CN10K_OCM_NUMTILES]; + + /* Memory for ocm_mask */ + uint8_t *ocm_mask; }; int cn10k_ml_ocm_tilecount(uint64_t tilemask, int *start, int *end); diff --git a/drivers/ml/cnxk/cn10k_ml_ops.c b/drivers/ml/cnxk/cn10k_ml_ops.c index 61e6d023c5..5b77e47322 100644 --- a/drivers/ml/cnxk/cn10k_ml_ops.c +++ b/drivers/ml/cnxk/cn10k_ml_ops.c @@ -311,8 +311,8 @@ cn10k_ml_model_print(struct rte_ml_dev *dev, uint16_t model_id, FILE *fp) if (model->state == ML_CN10K_MODEL_STATE_STARTED) { fprintf(fp, "%*s : 0x%0*" PRIx64 "\n", FIELD_LEN, "tilemask", ML_CN10K_OCM_NUMTILES / 4, model->model_mem_map.tilemask); - fprintf(fp, "%*s : 0x%x\n", FIELD_LEN, "ocm_wb_start", - model->model_mem_map.wb_page_start * ML_CN10K_OCM_PAGESIZE); + fprintf(fp, "%*s : 0x%" PRIx64 "\n", FIELD_LEN, "ocm_wb_start", + model->model_mem_map.wb_page_start * mldev->ocm.page_size); } fprintf(fp, "%*s : %u\n", FIELD_LEN, "num_inputs", model->metadata.model.num_input); @@ -781,12 +781,18 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c ocm = &mldev->ocm; ocm->num_tiles = ML_CN10K_OCM_NUMTILES; ocm->size_per_tile = ML_CN10K_OCM_TILESIZE; - ocm->page_size = ML_CN10K_OCM_PAGESIZE; + ocm->page_size = mldev->ocm_page_size; ocm->num_pages = ocm->size_per_tile / ocm->page_size; ocm->mask_words = ocm->num_pages / (8 * sizeof(uint8_t)); - for (tile_id = 0; tile_id < ocm->num_tiles; tile_id++) + /* Allocate memory for ocm_mask */ + ocm->ocm_mask = + rte_zmalloc("ocm_mask", ocm->mask_words * ocm->num_tiles, RTE_CACHE_LINE_SIZE); + + for (tile_id = 0; tile_id < ocm->num_tiles; tile_id++) { + ocm->tile_ocm_info[tile_id].ocm_mask = ocm->ocm_mask + tile_id * ocm->mask_words; ocm->tile_ocm_info[tile_id].last_wb_page = -1; + } rte_spinlock_init(&ocm->lock); @@ -856,6 +862,9 @@ cn10k_ml_dev_close(struct rte_ml_dev *dev) mldev = dev->data->dev_private; + /* Release ocm_mask memory */ + rte_free(mldev->ocm.ocm_mask); + /* Stop and unload all models */ for (model_id = 0; model_id < dev->data->nb_models; model_id++) { model = dev->data->models[model_id];