diff mbox series

[v3,5/5] eal/x86: force inlining of all memcpy and mov helpers

Message ID 20190529130420.6428-6-maxime.coquelin@redhat.com (mailing list archive)
State Accepted, archived
Delegated to: Maxime Coquelin
Headers
Series vhost: I-cache pressure optimizations |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation fail Compilation issues

Commit Message

Maxime Coquelin May 29, 2019, 1:04 p.m. UTC 
  Some helpers in the header file are forced inlined other are
only inlined, this patch forces inline for all.

It will avoid it to be embedded as functions when called multiple
times in the same object file. For example, when we added packed
ring support in vhost-user library, rte_memcpy_generic got no
more inlined.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 .../common/include/arch/x86/rte_memcpy.h       | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

Comments

Bruce Richardson June 5, 2019, 12:53 p.m. UTC | #1 
On Wed, May 29, 2019 at 03:04:20PM +0200, Maxime Coquelin wrote:
> Some helpers in the header file are forced inlined other are
> only inlined, this patch forces inline for all.
> 
> It will avoid it to be embedded as functions when called multiple
> times in the same object file. For example, when we added packed
> ring support in vhost-user library, rte_memcpy_generic got no
> more inlined.
> 
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Maxime Coquelin June 6, 2019, 9:33 a.m. UTC | #2 
On 6/5/19 2:53 PM, Bruce Richardson wrote:
> On Wed, May 29, 2019 at 03:04:20PM +0200, Maxime Coquelin wrote:
>> Some helpers in the header file are forced inlined other are
>> only inlined, this patch forces inline for all.
>>
>> It will avoid it to be embedded as functions when called multiple
>> times in the same object file. For example, when we added packed
>> ring support in vhost-user library, rte_memcpy_generic got no
>> more inlined.
>>
>> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> 

Thanks, finally applied it on dpdk-next-virtio tree as per Thomas'
request.
diff mbox series

Patch

diff --git a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
index 7b758094df..ba44c4a328 100644
--- a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
+++ b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
@@ -115,7 +115,7 @@  rte_mov256(uint8_t *dst, const uint8_t *src)
  * Copy 128-byte blocks from one location to another,
  * locations should not overlap.
  */
-static inline void
+static __rte_always_inline void
 rte_mov128blocks(uint8_t *dst, const uint8_t *src, size_t n)
 {
 	__m512i zmm0, zmm1;
@@ -163,7 +163,7 @@  rte_mov512blocks(uint8_t *dst, const uint8_t *src, size_t n)
 	}
 }
 
-static inline void *
+static __rte_always_inline void *
 rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	uintptr_t dstu = (uintptr_t)dst;
@@ -330,7 +330,7 @@  rte_mov64(uint8_t *dst, const uint8_t *src)
  * Copy 128 bytes from one location to another,
  * locations should not overlap.
  */
-static inline void
+static __rte_always_inline void
 rte_mov128(uint8_t *dst, const uint8_t *src)
 {
 	rte_mov32((uint8_t *)dst + 0 * 32, (const uint8_t *)src + 0 * 32);
@@ -343,7 +343,7 @@  rte_mov128(uint8_t *dst, const uint8_t *src)
  * Copy 128-byte blocks from one location to another,
  * locations should not overlap.
  */
-static inline void
+static __rte_always_inline void
 rte_mov128blocks(uint8_t *dst, const uint8_t *src, size_t n)
 {
 	__m256i ymm0, ymm1, ymm2, ymm3;
@@ -363,7 +363,7 @@  rte_mov128blocks(uint8_t *dst, const uint8_t *src, size_t n)
 	}
 }
 
-static inline void *
+static __rte_always_inline void *
 rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	uintptr_t dstu = (uintptr_t)dst;
@@ -523,7 +523,7 @@  rte_mov64(uint8_t *dst, const uint8_t *src)
  * Copy 128 bytes from one location to another,
  * locations should not overlap.
  */
-static inline void
+static __rte_always_inline void
 rte_mov128(uint8_t *dst, const uint8_t *src)
 {
 	rte_mov16((uint8_t *)dst + 0 * 16, (const uint8_t *)src + 0 * 16);
@@ -655,7 +655,7 @@  __extension__ ({                                                      \
     }                                                                 \
 })
 
-static inline void *
+static __rte_always_inline void *
 rte_memcpy_generic(void *dst, const void *src, size_t n)
 {
 	__m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7, xmm8;
@@ -800,7 +800,7 @@  rte_memcpy_generic(void *dst, const void *src, size_t n)
 
 #endif /* RTE_MACHINE_CPUFLAG */
 
-static inline void *
+static __rte_always_inline void *
 rte_memcpy_aligned(void *dst, const void *src, size_t n)
 {
 	void *ret = dst;
@@ -860,7 +860,7 @@  rte_memcpy_aligned(void *dst, const void *src, size_t n)
 	return ret;
 }
 
-static inline void *
+static __rte_always_inline void *
 rte_memcpy(void *dst, const void *src, size_t n)
 {
 	if (!(((uintptr_t)dst | (uintptr_t)src) & ALIGNMENT_MASK))