diff options
| author | Andrew Senkevich <andrew.senkevich@intel.com> | 2014-12-29 14:39:46 +0300 |
|---|---|---|
| committer | H.J. Lu <hjl.tools@gmail.com> | 2014-12-30 07:19:38 -0800 |
| commit | 8b4416d83c79ba77b0669203741c712880a09ae4 (patch) | |
| tree | c0701090d02b9e3c9ddcb840e2ad62084c498b4a /sysdeps/i386/i686/multiarch/ifunc-impl-list.c | |
| parent | 6d6d7fde04c8ef830205a9900bf101597a2f4b18 (diff) | |
| download | glibc-8b4416d83c79ba77b0669203741c712880a09ae4.tar.gz | |
i386: memcpy functions with SSE2 unaligned load/store
These new memcpy functions are the 32-bit version of x86_64 SSE2 unaligned
memcpy. Memcpy average performace benefit is 18% on Silvermont, other
platforms also improved about 35%, benchmarked on Silvermont, Haswell, Ivy
Bridge, Sandy Bridge and Westmere, performance results attached in
https://sourceware.org/ml/libc-alpha/2014-07/msg00157.html
* sysdeps/i386/i686/multiarch/bcopy-sse2-unaligned.S: New file.
* sysdeps/i386/i686/multiarch/memcpy-sse2-unaligned.S: Likewise.
* sysdeps/i386/i686/multiarch/memmove-sse2-unaligned.S: Likewise.
* sysdeps/i386/i686/multiarch/mempcpy-sse2-unaligned.S: Likewise.
* sysdeps/i386/i686/multiarch/bcopy.S: Select the sse2_unaligned
version if bit_Fast_Unaligned_Load is set.
* sysdeps/i386/i686/multiarch/memcpy.S: Likewise.
* sysdeps/i386/i686/multiarch/memcpy_chk.S: Likewise.
* sysdeps/i386/i686/multiarch/memmove.S: Likewise.
* sysdeps/i386/i686/multiarch/memmove_chk.S: Likewise.
* sysdeps/i386/i686/multiarch/mempcpy.S: Likewise.
* sysdeps/i386/i686/multiarch/mempcpy_chk.S: Likewise.
* sysdeps/i386/i686/multiarch/Makefile (sysdep_routines): Add
bcopy-sse2-unaligned, memcpy-sse2-unaligned,
memmove-sse2-unaligned and mempcpy-sse2-unaligned.
* sysdeps/i386/i686/multiarch/ifunc-impl-list.c (MAX_IFUNC): Set
to 4.
(__libc_ifunc_impl_list): Test __bcopy_sse2_unaligned,
__memmove_chk_sse2_unaligned, __memmove_sse2_unaligned,
__memcpy_chk_sse2_unaligned, __memcpy_sse2_unaligned,
__mempcpy_chk_sse2_unaligned, and __mempcpy_sse2_unaligned.
Diffstat (limited to 'sysdeps/i386/i686/multiarch/ifunc-impl-list.c')
| -rw-r--r-- | sysdeps/i386/i686/multiarch/ifunc-impl-list.c | 16 |
1 files changed, 15 insertions, 1 deletions
diff --git a/sysdeps/i386/i686/multiarch/ifunc-impl-list.c b/sysdeps/i386/i686/multiarch/ifunc-impl-list.c index e47577690d..4efa9c5ef9 100644 --- a/sysdeps/i386/i686/multiarch/ifunc-impl-list.c +++ b/sysdeps/i386/i686/multiarch/ifunc-impl-list.c @@ -23,7 +23,7 @@ #include "init-arch.h" /* Maximum number of IFUNC implementations. */ -#define MAX_IFUNC 3 +#define MAX_IFUNC 4 /* Fill ARRAY of MAX elements with IFUNC implementations for function NAME and return the number of valid entries. */ @@ -41,6 +41,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL_ADD (array, i, bcopy, HAS_SSSE3, __bcopy_ssse3_rep) IFUNC_IMPL_ADD (array, i, bcopy, HAS_SSSE3, __bcopy_ssse3) + IFUNC_IMPL_ADD (array, i, bcopy, HAS_SSE2, + __bcopy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, bcopy, 1, __bcopy_ia32)) /* Support sysdeps/i386/i686/multiarch/bzero.S. */ @@ -69,6 +71,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __memmove_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __memmove_chk, HAS_SSSE3, __memmove_chk_ssse3) + IFUNC_IMPL_ADD (array, i, __memmove_chk, HAS_SSE2, + __memmove_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __memmove_chk, 1, __memmove_chk_ia32)) @@ -78,6 +82,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __memmove_ssse3_rep) IFUNC_IMPL_ADD (array, i, memmove, HAS_SSSE3, __memmove_ssse3) + IFUNC_IMPL_ADD (array, i, memmove, HAS_SSE2, + __memmove_sse2_unaligned) IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_ia32)) /* Support sysdeps/i386/i686/multiarch/memrchr.S. */ @@ -268,6 +274,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __memcpy_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __memcpy_chk, HAS_SSSE3, __memcpy_chk_ssse3) + IFUNC_IMPL_ADD (array, i, __memcpy_chk, HAS_SSE2, + __memcpy_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __memcpy_chk, 1, __memcpy_chk_ia32)) @@ -276,6 +284,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL_ADD (array, i, memcpy, HAS_SSSE3, __memcpy_ssse3_rep) IFUNC_IMPL_ADD (array, i, memcpy, HAS_SSSE3, __memcpy_ssse3) + IFUNC_IMPL_ADD (array, i, memcpy, HAS_SSE2, + __memcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/mempcpy_chk.S. */ @@ -284,6 +294,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __mempcpy_chk_ssse3_rep) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, HAS_SSSE3, __mempcpy_chk_ssse3) + IFUNC_IMPL_ADD (array, i, __mempcpy_chk, HAS_SSE2, + __mempcpy_chk_sse2_unaligned) IFUNC_IMPL_ADD (array, i, __mempcpy_chk, 1, __mempcpy_chk_ia32)) @@ -293,6 +305,8 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, __mempcpy_ssse3_rep) IFUNC_IMPL_ADD (array, i, mempcpy, HAS_SSSE3, __mempcpy_ssse3) + IFUNC_IMPL_ADD (array, i, mempcpy, HAS_SSE2, + __mempcpy_sse2_unaligned) IFUNC_IMPL_ADD (array, i, mempcpy, 1, __mempcpy_ia32)) /* Support sysdeps/i386/i686/multiarch/strlen.S. */ |