diff options
| author | H.J. Lu <hjl.tools@gmail.com> | 2021-03-05 07:26:42 -0800 |
|---|---|---|
| committer | H.J. Lu <hjl.tools@gmail.com> | 2021-03-29 07:40:17 -0700 |
| commit | 7ebba91361badf7531d4e75050627a88d424872f (patch) | |
| tree | d99781a37b47b95441ad358d119ec3741960d405 /sysdeps/x86_64/multiarch/strlen-avx2.S | |
| parent | 91264fe3577fe887b4860923fa6142b5274c8965 (diff) | |
| download | glibc-7ebba91361badf7531d4e75050627a88d424872f.tar.gz glibc-7ebba91361badf7531d4e75050627a88d424872f.tar.xz glibc-7ebba91361badf7531d4e75050627a88d424872f.zip | |
x86-64: Add AVX optimized string/memory functions for RTM
Since VZEROUPPER triggers RTM abort while VZEROALL won't, select AVX optimized string/memory functions with xtest jz 1f vzeroall ret 1: vzeroupper ret at function exit on processors with usable RTM, but without 256-bit EVEX instructions to avoid VZEROUPPER inside a transactionally executing RTM region.
Diffstat (limited to 'sysdeps/x86_64/multiarch/strlen-avx2.S')
| -rw-r--r-- | sysdeps/x86_64/multiarch/strlen-avx2.S | 43 |
1 files changed, 18 insertions, 25 deletions
diff --git a/sysdeps/x86_64/multiarch/strlen-avx2.S b/sysdeps/x86_64/multiarch/strlen-avx2.S index caa615970c..1caae9e6bc 100644 --- a/sysdeps/x86_64/multiarch/strlen-avx2.S +++ b/sysdeps/x86_64/multiarch/strlen-avx2.S @@ -36,9 +36,13 @@ # define VZEROUPPER vzeroupper # endif +# ifndef SECTION +# define SECTION(p) p##.avx +# endif + # define VEC_SIZE 32 - .section .text.avx,"ax",@progbits + .section SECTION(.text),"ax",@progbits ENTRY (STRLEN) # ifdef USE_AS_STRNLEN /* Check for zero length. */ @@ -111,8 +115,8 @@ L(cros_page_boundary): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret +L(return_vzeroupper): + ZERO_UPPER_VEC_REGISTERS_RETURN .p2align 4 L(aligned_more): @@ -231,8 +235,7 @@ L(last_4x_vec_or_less): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(last_2x_vec): @@ -253,8 +256,7 @@ L(last_2x_vec): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x0_check): @@ -267,8 +269,7 @@ L(first_vec_x0_check): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x1_check): @@ -282,8 +283,7 @@ L(first_vec_x1_check): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x2_check): @@ -297,8 +297,7 @@ L(first_vec_x2_check): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x3_check): @@ -312,8 +311,7 @@ L(first_vec_x3_check): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(max): @@ -321,8 +319,7 @@ L(max): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(zero): @@ -338,8 +335,7 @@ L(first_vec_x0): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x1): @@ -350,8 +346,7 @@ L(first_vec_x1): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(first_vec_x2): @@ -362,8 +357,7 @@ L(first_vec_x2): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN .p2align 4 L(4x_vec_end): @@ -389,8 +383,7 @@ L(first_vec_x3): # ifdef USE_AS_WCSLEN shrq $2, %rax # endif - VZEROUPPER - ret + VZEROUPPER_RETURN END (STRLEN) #endif |