From patchwork Sun Jan 7 15:28:11 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Srikanth Yalavarthi X-Patchwork-Id: 135777 X-Patchwork-Delegate: thomas@monjalon.net 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 BB01043857; Sun, 7 Jan 2024 16:28:34 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 6258E402F1; Sun, 7 Jan 2024 16:28:27 +0100 (CET) Received: from mx0b-0016f401.pphosted.com (mx0b-0016f401.pphosted.com [67.231.156.173]) by mails.dpdk.org (Postfix) with ESMTP id 9015240608 for ; Sun, 7 Jan 2024 16:28:25 +0100 (CET) Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.17.1.24/8.17.1.24) with ESMTP id 407Et9Wf023658; Sun, 7 Jan 2024 07:28:22 -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-transfer-encoding:content-type; s= pfpt0220; bh=zOEwlv6b+T8O6bWdzno1nkOK2Bt4Zms8Xc1QYDR0TVs=; b=b4q KS0MnSiCTwAUtYmPLeYZGsy6OfgMPzzV08C/DnfoZbxCAQN+5e0S5ZvBnN+Urm0W efDQw1az9Cz6CIlPEND7Gx4U53xJF3cjOE2yLrvSy8y1ovwVMdYX+r0ZPMTt6KAL 8xDuR6Yft+VzE2YRJa9G6yjLq77StfSR1QYJ+QKSn/fe3hCPIO5ANAuivxoI50nD kRSsHU5xiSOh/HfNf529Tpjkycr08XFhO3uf80wiSfcl6XlI0wrpwtnCoEZSzyql Fbvx5FC3+aSksGoObvM4kDDJB0EbYPvIIcnYK1wZ1BoMwvNxuMbhvZHGQiiEfZtT XDrollI7C5Mah98xlaA== Received: from dc5-exch01.marvell.com ([199.233.59.181]) by mx0b-0016f401.pphosted.com (PPS) with ESMTPS id 3vf78n29nt-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Sun, 07 Jan 2024 07:28:22 -0800 (PST) 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 15.0.1497.48; Sun, 7 Jan 2024 07:28:20 -0800 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH02.marvell.com (10.69.176.39) with Microsoft SMTP Server id 15.0.1497.48 via Frontend Transport; Sun, 7 Jan 2024 07:28:20 -0800 Received: from ml-host-33.caveonetworks.com (unknown [10.110.143.233]) by maili.marvell.com (Postfix) with ESMTP id 54D8D3F709E; Sun, 7 Jan 2024 07:28:20 -0800 (PST) From: Srikanth Yalavarthi To: Srikanth Yalavarthi , Ruifeng Wang CC: , , , Subject: [PATCH 2/3] mldev: add support for 64-integer data type Date: Sun, 7 Jan 2024 07:28:11 -0800 Message-ID: <20240107152813.2668-3-syalavarthi@marvell.com> X-Mailer: git-send-email 2.42.0 In-Reply-To: <20240107152813.2668-1-syalavarthi@marvell.com> References: <20240107152813.2668-1-syalavarthi@marvell.com> MIME-Version: 1.0 X-Proofpoint-ORIG-GUID: PJz01FH56BpV9nNx_rbpR3vFulNzzail X-Proofpoint-GUID: PJz01FH56BpV9nNx_rbpR3vFulNzzail X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.272,Aquarius:18.0.997,Hydra:6.0.619,FMLib:17.11.176.26 definitions=2023-12-09_02,2023-12-07_01,2023-05-22_02 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 in mldev spec for 64-bit integer types. Added routines to convert data from 64-bit integer type to float32_t and vice-versa. Signed-off-by: Srikanth Yalavarthi --- lib/mldev/mldev_utils.c | 4 + lib/mldev/mldev_utils.h | 92 ++++++++++ lib/mldev/mldev_utils_neon.c | 324 +++++++++++++++++++++++++++++++++ lib/mldev/mldev_utils_scalar.c | 98 ++++++++++ lib/mldev/rte_mldev.h | 4 + lib/mldev/version.map | 4 + 6 files changed, 526 insertions(+) diff --git a/lib/mldev/mldev_utils.c b/lib/mldev/mldev_utils.c index ccd2c39ca89..13ac615e9fc 100644 --- a/lib/mldev/mldev_utils.c +++ b/lib/mldev/mldev_utils.c @@ -32,6 +32,10 @@ rte_ml_io_type_size_get(enum rte_ml_io_type type) return sizeof(int32_t); case RTE_ML_IO_TYPE_UINT32: return sizeof(uint32_t); + case RTE_ML_IO_TYPE_INT64: + return sizeof(int64_t); + case RTE_ML_IO_TYPE_UINT64: + return sizeof(uint64_t); case RTE_ML_IO_TYPE_FP8: return sizeof(uint8_t); case RTE_ML_IO_TYPE_FP16: diff --git a/lib/mldev/mldev_utils.h b/lib/mldev/mldev_utils.h index 1d041531b43..6daae6d0a1c 100644 --- a/lib/mldev/mldev_utils.h +++ b/lib/mldev/mldev_utils.h @@ -328,6 +328,98 @@ __rte_internal int rte_ml_io_uint32_to_float32(float scale, uint64_t nb_elements, void *input, void *output); +/** + * @internal + * + * Convert a buffer containing numbers in single precision floating format (float32) to signed + * 64-bit integer format (INT64). + * + * @param[in] scale + * Scale factor for conversion. + * @param[in] nb_elements + * Number of elements in the buffer. + * @param[in] input + * Input buffer containing float32 numbers. Size of buffer is equal to (nb_elements * 4) bytes. + * @param[out] output + * Output buffer to store INT64 numbers. Size of buffer is equal to (nb_elements * 8) bytes. + * + * @return + * - 0, Success. + * - < 0, Error code on failure. + */ +__rte_internal +int +rte_ml_io_float32_to_int64(float scale, uint64_t nb_elements, void *input, void *output); + +/** + * @internal + * + * Convert a buffer containing numbers in signed 64-bit integer format (INT64) to single precision + * floating format (float32). + * + * @param[in] scale + * Scale factor for conversion. + * @param[in] nb_elements + * Number of elements in the buffer. + * @param[in] input + * Input buffer containing INT64 numbers. Size of buffer is equal to (nb_elements * 8) bytes. + * @param[out] output + * Output buffer to store float32 numbers. Size of buffer is equal to (nb_elements * 4) bytes. + * + * @return + * - 0, Success. + * - < 0, Error code on failure. + */ +__rte_internal +int +rte_ml_io_int64_to_float32(float scale, uint64_t nb_elements, void *input, void *output); + +/** + * @internal + * + * Convert a buffer containing numbers in single precision floating format (float32) to unsigned + * 64-bit integer format (UINT64). + * + * @param[in] scale + * Scale factor for conversion. + * @param[in] nb_elements + * Number of elements in the buffer. + * @param[in] input + * Input buffer containing float32 numbers. Size of buffer is equal to (nb_elements * 4) bytes. + * @param[out] output + * Output buffer to store UINT64 numbers. Size of buffer is equal to (nb_elements * 8) bytes. + * + * @return + * - 0, Success. + * - < 0, Error code on failure. + */ +__rte_internal +int +rte_ml_io_float32_to_uint64(float scale, uint64_t nb_elements, void *input, void *output); + +/** + * @internal + * + * Convert a buffer containing numbers in unsigned 64-bit integer format (UINT64) to single + * precision floating format (float32). + * + * @param[in] scale + * Scale factor for conversion. + * @param[in] nb_elements + * Number of elements in the buffer. + * @param[in] input + * Input buffer containing UINT64 numbers. Size of buffer is equal to (nb_elements * 8) bytes. + * @param[out] output + * Output buffer to store float32 numbers. Size of buffer is equal to (nb_elements * 4) bytes. + * + * @return + * - 0, Success. + * - < 0, Error code on failure. + */ +__rte_internal +int +rte_ml_io_uint64_to_float32(float scale, uint64_t nb_elements, void *input, void *output); + /** * @internal * diff --git a/lib/mldev/mldev_utils_neon.c b/lib/mldev/mldev_utils_neon.c index 250fa43fa73..4cde2ebabd3 100644 --- a/lib/mldev/mldev_utils_neon.c +++ b/lib/mldev/mldev_utils_neon.c @@ -842,6 +842,330 @@ rte_ml_io_uint32_to_float32(float scale, uint64_t nb_elements, void *input, void return 0; } +static inline void +__float32_to_int64_neon_s64x2(float scale, float *input, int64_t *output) +{ + float32x2_t f32x2; + float64x2_t f64x2; + int64x2_t s64x2; + + /* load 2 x float elements */ + f32x2 = vld1_f32(input); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* convert to float64x2_t */ + f64x2 = vcvt_f64_f32(f32x2); + + /* convert to int64x2_t */ + s64x2 = vcvtaq_s64_f64(f64x2); + + /* store 2 elements */ + vst1q_s64(output, s64x2); +} + +static inline void +__float32_to_int64_neon_s64x1(float scale, float *input, int64_t *output) +{ + float32x2_t f32x2; + float64x2_t f64x2; + int64x2_t s64x2; + + /* load 1 x float element */ + f32x2 = vdup_n_f32(*input); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* convert to float64x2_t */ + f64x2 = vcvt_f64_f32(f32x2); + + /* convert to int64x2_t */ + s64x2 = vcvtaq_s64_f64(f64x2); + + /* store lane 0 of int64x2_t */ + vst1q_lane_s64(output, s64x2, 0); +} + +int +rte_ml_io_float32_to_int64(float scale, uint64_t nb_elements, void *input, void *output) +{ + float *input_buffer; + int64_t *output_buffer; + uint64_t nb_iterations; + uint32_t vlen; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (float *)input; + output_buffer = (int64_t *)output; + vlen = 4 * sizeof(float) / sizeof(int64_t); + nb_iterations = nb_elements / vlen; + + /* convert vlen elements in each iteration */ + for (i = 0; i < nb_iterations; i++) { + __float32_to_int64_neon_s64x2(scale, input_buffer, output_buffer); + input_buffer += vlen; + output_buffer += vlen; + } + + /* convert leftover elements */ + i = i * vlen; + for (; i < nb_elements; i++) { + __float32_to_int64_neon_s64x1(scale, input_buffer, output_buffer); + input_buffer++; + output_buffer++; + } + + return 0; +} + +static inline void +__int64_to_float32_neon_f32x2(float scale, int64_t *input, float *output) +{ + int64x2_t s64x2; + float64x2_t f64x2; + float32x2_t f32x2; + + /* load 2 x int64_t elements */ + s64x2 = vld1q_s64(input); + + /* convert int64x2_t to float64x2_t */ + f64x2 = vcvtq_f64_s64(s64x2); + + /* convert float64x2_t to float32x2_t */ + f32x2 = vcvt_f32_f64(f64x2); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* store float32x2_t */ + vst1_f32(output, f32x2); +} + +static inline void +__int64_to_float32_neon_f32x1(float scale, int64_t *input, float *output) +{ + int64x2_t s64x2; + float64x2_t f64x2; + float32x2_t f32x2; + + /* load 2 x int64_t elements */ + s64x2 = vld1q_lane_s64(input, vdupq_n_s64(0), 0); + + /* convert int64x2_t to float64x2_t */ + f64x2 = vcvtq_f64_s64(s64x2); + + /* convert float64x2_t to float32x2_t */ + f32x2 = vcvt_f32_f64(f64x2); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* store float32x2_t */ + vst1_lane_f32(output, f32x2, 0); +} + +int +rte_ml_io_int64_to_float32(float scale, uint64_t nb_elements, void *input, void *output) +{ + int64_t *input_buffer; + float *output_buffer; + uint64_t nb_iterations; + uint32_t vlen; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (int64_t *)input; + output_buffer = (float *)output; + vlen = 4 * sizeof(float) / sizeof(int64_t); + nb_iterations = nb_elements / vlen; + + /* convert vlen elements in each iteration */ + for (i = 0; i < nb_iterations; i++) { + __int64_to_float32_neon_f32x2(scale, input_buffer, output_buffer); + input_buffer += vlen; + output_buffer += vlen; + } + + /* convert leftover elements */ + i = i * vlen; + for (; i < nb_elements; i++) { + __int64_to_float32_neon_f32x1(scale, input_buffer, output_buffer); + input_buffer++; + output_buffer++; + } + + return 0; +} + +static inline void +__float32_to_uint64_neon_u64x2(float scale, float *input, uint64_t *output) +{ + float32x2_t f32x2; + float64x2_t f64x2; + uint64x2_t u64x2; + + /* load 2 x float elements */ + f32x2 = vld1_f32(input); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* convert to float64x2_t */ + f64x2 = vcvt_f64_f32(f32x2); + + /* convert to int64x2_t */ + u64x2 = vcvtaq_u64_f64(f64x2); + + /* store 2 elements */ + vst1q_u64(output, u64x2); +} + +static inline void +__float32_to_uint64_neon_u64x1(float scale, float *input, uint64_t *output) +{ + float32x2_t f32x2; + float64x2_t f64x2; + uint64x2_t u64x2; + + /* load 1 x float element */ + f32x2 = vld1_lane_f32(input, vdup_n_f32(0), 0); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* convert to float64x2_t */ + f64x2 = vcvt_f64_f32(f32x2); + + /* convert to int64x2_t */ + u64x2 = vcvtaq_u64_f64(f64x2); + + /* store 2 elements */ + vst1q_lane_u64(output, u64x2, 0); +} + +int +rte_ml_io_float32_to_uint64(float scale, uint64_t nb_elements, void *input, void *output) +{ + float *input_buffer; + uint64_t *output_buffer; + uint64_t nb_iterations; + uint32_t vlen; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (float *)input; + output_buffer = (uint64_t *)output; + vlen = 4 * sizeof(float) / sizeof(uint64_t); + nb_iterations = nb_elements / vlen; + + /* convert vlen elements in each iteration */ + for (i = 0; i < nb_iterations; i++) { + __float32_to_uint64_neon_u64x2(scale, input_buffer, output_buffer); + input_buffer += vlen; + output_buffer += vlen; + } + + /* convert leftover elements */ + i = i * vlen; + for (; i < nb_elements; i++) { + __float32_to_uint64_neon_u64x1(scale, input_buffer, output_buffer); + input_buffer++; + output_buffer++; + } + + return 0; +} + +static inline void +__uint64_to_float32_neon_f32x2(float scale, uint64_t *input, float *output) +{ + uint64x2_t u64x2; + float64x2_t f64x2; + float32x2_t f32x2; + + /* load 2 x int64_t elements */ + u64x2 = vld1q_u64(input); + + /* convert int64x2_t to float64x2_t */ + f64x2 = vcvtq_f64_u64(u64x2); + + /* convert float64x2_t to float32x2_t */ + f32x2 = vcvt_f32_f64(f64x2); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* store float32x2_t */ + vst1_f32(output, f32x2); +} + +static inline void +__uint64_to_float32_neon_f32x1(float scale, uint64_t *input, float *output) +{ + uint64x2_t u64x2; + float64x2_t f64x2; + float32x2_t f32x2; + + /* load 2 x int64_t elements */ + u64x2 = vld1q_lane_u64(input, vdupq_n_u64(0), 0); + + /* convert int64x2_t to float64x2_t */ + f64x2 = vcvtq_f64_u64(u64x2); + + /* convert float64x2_t to float32x2_t */ + f32x2 = vcvt_f32_f64(f64x2); + + /* scale */ + f32x2 = vmul_n_f32(f32x2, scale); + + /* store float32x2_t */ + vst1_lane_f32(output, f32x2, 0); +} + +int +rte_ml_io_uint64_to_float32(float scale, uint64_t nb_elements, void *input, void *output) +{ + uint64_t *input_buffer; + float *output_buffer; + uint64_t nb_iterations; + uint32_t vlen; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (uint64_t *)input; + output_buffer = (float *)output; + vlen = 4 * sizeof(float) / sizeof(uint64_t); + nb_iterations = nb_elements / vlen; + + /* convert vlen elements in each iteration */ + for (i = 0; i < nb_iterations; i++) { + __uint64_to_float32_neon_f32x2(scale, input_buffer, output_buffer); + input_buffer += vlen; + output_buffer += vlen; + } + + /* convert leftover elements */ + i = i * vlen; + for (; i < nb_elements; i++) { + __uint64_to_float32_neon_f32x1(scale, input_buffer, output_buffer); + input_buffer++; + output_buffer++; + } + + return 0; +} + static inline void __float32_to_float16_neon_f16x4(float32_t *input, float16_t *output) { diff --git a/lib/mldev/mldev_utils_scalar.c b/lib/mldev/mldev_utils_scalar.c index af1a3a103b2..63a9900cc8c 100644 --- a/lib/mldev/mldev_utils_scalar.c +++ b/lib/mldev/mldev_utils_scalar.c @@ -327,6 +327,104 @@ rte_ml_io_uint32_to_float32(float scale, uint64_t nb_elements, void *input, void return 0; } +int +rte_ml_io_float32_to_int64(float scale, uint64_t nb_elements, void *input, void *output) +{ + float *input_buffer; + int64_t *output_buffer; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (float *)input; + output_buffer = (int64_t *)output; + + for (i = 0; i < nb_elements; i++) { + *output_buffer = (int64_t)round((*input_buffer) * scale); + + input_buffer++; + output_buffer++; + } + + return 0; +} + +int +rte_ml_io_int64_to_float32(float scale, uint64_t nb_elements, void *input, void *output) +{ + int64_t *input_buffer; + float *output_buffer; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (int64_t *)input; + output_buffer = (float *)output; + + for (i = 0; i < nb_elements; i++) { + *output_buffer = scale * (float)(*input_buffer); + + input_buffer++; + output_buffer++; + } + + return 0; +} + +int +rte_ml_io_float32_to_uint64(float scale, uint64_t nb_elements, void *input, void *output) +{ + float *input_buffer; + uint64_t *output_buffer; + int64_t i64; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (float *)input; + output_buffer = (uint64_t *)output; + + for (i = 0; i < nb_elements; i++) { + i64 = (int64_t)round((*input_buffer) * scale); + + if (i64 < 0) + i64 = 0; + + *output_buffer = (uint64_t)i64; + + input_buffer++; + output_buffer++; + } + + return 0; +} + +int +rte_ml_io_uint64_to_float32(float scale, uint64_t nb_elements, void *input, void *output) +{ + uint64_t *input_buffer; + float *output_buffer; + uint64_t i; + + if ((scale == 0) || (nb_elements == 0) || (input == NULL) || (output == NULL)) + return -EINVAL; + + input_buffer = (uint64_t *)input; + output_buffer = (float *)output; + + for (i = 0; i < nb_elements; i++) { + *output_buffer = scale * (float)(*input_buffer); + + input_buffer++; + output_buffer++; + } + + return 0; +} + /* Convert a single precision floating point number (float32) into a half precision * floating point number (float16) using round to nearest rounding mode. */ diff --git a/lib/mldev/rte_mldev.h b/lib/mldev/rte_mldev.h index 5cf6f0566f1..27e372fbcf1 100644 --- a/lib/mldev/rte_mldev.h +++ b/lib/mldev/rte_mldev.h @@ -874,6 +874,10 @@ enum rte_ml_io_type { /**< 32-bit integer */ RTE_ML_IO_TYPE_UINT32, /**< 32-bit unsigned integer */ + RTE_ML_IO_TYPE_INT64, + /**< 32-bit integer */ + RTE_ML_IO_TYPE_UINT64, + /**< 32-bit unsigned integer */ RTE_ML_IO_TYPE_FP8, /**< 8-bit floating point number */ RTE_ML_IO_TYPE_FP16, diff --git a/lib/mldev/version.map b/lib/mldev/version.map index 2e8f1555225..1978695314e 100644 --- a/lib/mldev/version.map +++ b/lib/mldev/version.map @@ -61,6 +61,10 @@ INTERNAL { rte_ml_io_int32_to_float32; rte_ml_io_float32_to_uint32; rte_ml_io_uint32_to_float32; + rte_ml_io_float32_to_int64; + rte_ml_io_int64_to_float32; + rte_ml_io_float32_to_uint64; + rte_ml_io_uint64_to_float32; rte_ml_io_float32_to_float16; rte_ml_io_float16_to_float32; rte_ml_io_float32_to_bfloat16;