From 16d12015c57701b08d7bbed6ec536641bcafb428 Mon Sep 17 00:00:00 2001 From: Noah Goldstein Date: Mon, 17 May 2021 13:56:52 -0400 Subject: x86: Optimize memcmp-avx2-movbe.S No bug. This commit optimizes memcmp-avx2.S. The optimizations include adding a new vec compare path for small sizes, reorganizing the entry control flow, and removing some unnecissary ALU instructions from the main loop. test-memcmp and test-wmemcmp are both passing. Signed-off-by: Noah Goldstein Reviewed-by: H.J. Lu --- sysdeps/x86_64/multiarch/ifunc-impl-list.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'sysdeps/x86_64/multiarch/ifunc-impl-list.c') diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c index 2d811a550b..15eda47667 100644 --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c @@ -63,16 +63,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL (i, name, memcmp, IFUNC_IMPL_ADD (array, i, memcmp, (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE)), __memcmp_avx2_movbe) IFUNC_IMPL_ADD (array, i, memcmp, (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE) && CPU_FEATURE_USABLE (RTM)), __memcmp_avx2_movbe_rtm) IFUNC_IMPL_ADD (array, i, memcmp, (CPU_FEATURE_USABLE (AVX512VL) && CPU_FEATURE_USABLE (AVX512BW) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE)), __memcmp_evex_movbe) IFUNC_IMPL_ADD (array, i, memcmp, CPU_FEATURE_USABLE (SSE4_1), @@ -729,16 +732,19 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, IFUNC_IMPL (i, name, wmemcmp, IFUNC_IMPL_ADD (array, i, wmemcmp, (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE)), __wmemcmp_avx2_movbe) IFUNC_IMPL_ADD (array, i, wmemcmp, (CPU_FEATURE_USABLE (AVX2) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE) && CPU_FEATURE_USABLE (RTM)), __wmemcmp_avx2_movbe_rtm) IFUNC_IMPL_ADD (array, i, wmemcmp, (CPU_FEATURE_USABLE (AVX512VL) && CPU_FEATURE_USABLE (AVX512BW) + && CPU_FEATURE_USABLE (BMI2) && CPU_FEATURE_USABLE (MOVBE)), __wmemcmp_evex_movbe) IFUNC_IMPL_ADD (array, i, wmemcmp, CPU_FEATURE_USABLE (SSE4_1), -- cgit 1.4.1