diff options
Diffstat (limited to 'sysdeps/x86_64/multiarch')
| -rw-r--r-- | sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S | 6 | ||||
| -rw-r--r-- | sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S | 34 |
2 files changed, 26 insertions, 14 deletions
diff --git a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S index 6506ce6252..cd7a32c7c2 100644 --- a/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S +++ b/sysdeps/x86_64/multiarch/memset-avx512-no-vzeroupper.S @@ -29,12 +29,16 @@ .section .text.avx512,"ax",@progbits #if defined PIC ENTRY (MEMSET_CHK) - cmpq %rdx, %rcx + cmp %RDX_LP, %RCX_LP jb HIDDEN_JUMPTARGET (__chk_fail) END (MEMSET_CHK) #endif ENTRY (MEMSET) +# ifdef __ILP32__ + /* Clear the upper 32 bits. */ + mov %edx, %edx +# endif vpxor %xmm0, %xmm0, %xmm0 vmovd %esi, %xmm1 lea (%rdi, %rdx), %rsi diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S index 71a0887c68..5e0d307d85 100644 --- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S +++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S @@ -75,8 +75,8 @@ .section SECTION(.text),"ax",@progbits #if VEC_SIZE == 16 && IS_IN (libc) ENTRY (__bzero) - movq %rdi, %rax /* Set return value. */ - movq %rsi, %rdx /* Set n. */ + mov %RDI_LP, %RAX_LP /* Set return value. */ + mov %RSI_LP, %RDX_LP /* Set n. */ pxor %xmm0, %xmm0 jmp L(entry_from_bzero) END (__bzero) @@ -86,13 +86,13 @@ weak_alias (__bzero, bzero) #if IS_IN (libc) # if defined SHARED ENTRY_CHK (WMEMSET_CHK_SYMBOL (__wmemset_chk, unaligned)) - cmpq %rdx, %rcx + cmp %RDX_LP, %RCX_LP jb HIDDEN_JUMPTARGET (__chk_fail) END_CHK (WMEMSET_CHK_SYMBOL (__wmemset_chk, unaligned)) # endif ENTRY (WMEMSET_SYMBOL (__wmemset, unaligned)) - shlq $2, %rdx + shl $2, %RDX_LP WMEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi) jmp L(entry_from_bzero) END (WMEMSET_SYMBOL (__wmemset, unaligned)) @@ -100,13 +100,17 @@ END (WMEMSET_SYMBOL (__wmemset, unaligned)) #if defined SHARED && IS_IN (libc) ENTRY_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned)) - cmpq %rdx, %rcx + cmp %RDX_LP, %RCX_LP jb HIDDEN_JUMPTARGET (__chk_fail) END_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned)) #endif ENTRY (MEMSET_SYMBOL (__memset, unaligned)) MEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi) +# ifdef __ILP32__ + /* Clear the upper 32 bits. */ + mov %edx, %edx +# endif L(entry_from_bzero): cmpq $VEC_SIZE, %rdx jb L(less_vec) @@ -122,14 +126,14 @@ END (MEMSET_SYMBOL (__memset, unaligned)) # if VEC_SIZE == 16 ENTRY (__memset_chk_erms) - cmpq %rdx, %rcx + cmp %RDX_LP, %RCX_LP jb HIDDEN_JUMPTARGET (__chk_fail) END (__memset_chk_erms) /* Only used to measure performance of REP STOSB. */ ENTRY (__memset_erms) /* Skip zero length. */ - testq %rdx, %rdx + test %RDX_LP, %RDX_LP jnz L(stosb) movq %rdi, %rax ret @@ -141,11 +145,11 @@ ENTRY (MEMSET_SYMBOL (__memset, erms)) L(stosb): /* Issue vzeroupper before rep stosb. */ VZEROUPPER - movq %rdx, %rcx + mov %RDX_LP, %RCX_LP movzbl %sil, %eax - movq %rdi, %rdx + mov %RDI_LP, %RDX_LP rep stosb - movq %rdx, %rax + mov %RDX_LP, %RAX_LP ret # if VEC_SIZE == 16 END (__memset_erms) @@ -155,16 +159,20 @@ END (MEMSET_SYMBOL (__memset, erms)) # if defined SHARED && IS_IN (libc) ENTRY_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned_erms)) - cmpq %rdx, %rcx + cmp %RDX_LP, %RCX_LP jb HIDDEN_JUMPTARGET (__chk_fail) END_CHK (MEMSET_CHK_SYMBOL (__memset_chk, unaligned_erms)) # endif ENTRY (MEMSET_SYMBOL (__memset, unaligned_erms)) MEMSET_VDUP_TO_VEC0_AND_SET_RETURN (%esi, %rdi) - cmpq $VEC_SIZE, %rdx +# ifdef __ILP32__ + /* Clear the upper 32 bits. */ + mov %edx, %edx +# endif + cmp $VEC_SIZE, %RDX_LP jb L(less_vec) - cmpq $(VEC_SIZE * 2), %rdx + cmp $(VEC_SIZE * 2), %RDX_LP ja L(stosb_more_2x_vec) /* From VEC and to 2 * VEC. No branch when size == VEC_SIZE. */ VMOVU %VEC(0), -VEC_SIZE(%rdi,%rdx) |