glibc/sysdeps/x86_64/multiarch/strchr-avx2.S
Noah Goldstein a6fbf4d51e x86: Code cleanup in strchr-avx2 and comment justifying branch
Small code cleanup for size: -53 bytes.

Add comment justifying using a branch to do NULL/non-null return.

All string/memory tests pass and no regressions in benchtests.

geometric_mean(N=20) of all benchmarks Original / New: 1.00
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
2022-03-25 11:46:13 -05:00

314 lines
7.2 KiB
ArmAsm

/* strchr/strchrnul optimized with AVX2.
Copyright (C) 2017-2022 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
#if IS_IN (libc)
# include <sysdep.h>
# ifndef STRCHR
# define STRCHR __strchr_avx2
# endif
# ifdef USE_AS_WCSCHR
# define VPBROADCAST vpbroadcastd
# define VPCMPEQ vpcmpeqd
# define VPMINU vpminud
# define CHAR_REG esi
# else
# define VPBROADCAST vpbroadcastb
# define VPCMPEQ vpcmpeqb
# define VPMINU vpminub
# define CHAR_REG sil
# endif
# ifndef VZEROUPPER
# define VZEROUPPER vzeroupper
# endif
# ifndef SECTION
# define SECTION(p) p##.avx
# endif
# define VEC_SIZE 32
# define PAGE_SIZE 4096
.section SECTION(.text),"ax",@progbits
ENTRY_P2ALIGN (STRCHR, 5)
/* Broadcast CHAR to YMM0. */
vmovd %esi, %xmm0
movl %edi, %eax
andl $(PAGE_SIZE - 1), %eax
VPBROADCAST %xmm0, %ymm0
vpxor %xmm1, %xmm1, %xmm1
/* Check if we cross page boundary with one vector load. */
cmpl $(PAGE_SIZE - VEC_SIZE), %eax
ja L(cross_page_boundary)
/* Check the first VEC_SIZE bytes. Search for both CHAR and the
null byte. */
vmovdqu (%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jz L(aligned_more)
tzcntl %eax, %eax
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
/* NB: Use a branch instead of cmovcc here. The expectation is
that with strchr the user will branch based on input being
null. Since this branch will be 100% predictive of the user
branch a branch miss here should save what otherwise would
be branch miss in the user code. Otherwise using a branch 1)
saves code size and 2) is faster in highly predictable
environments. */
jne L(zero)
# endif
addq %rdi, %rax
L(return_vzeroupper):
ZERO_UPPER_VEC_REGISTERS_RETURN
# ifndef USE_AS_STRCHRNUL
L(zero):
xorl %eax, %eax
VZEROUPPER_RETURN
# endif
.p2align 4
L(first_vec_x1):
/* Use bsf to save code size. */
bsfl %eax, %eax
incq %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4,, 10
L(first_vec_x2):
/* Use bsf to save code size. */
bsfl %eax, %eax
addq $(VEC_SIZE + 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4,, 8
L(first_vec_x3):
/* Use bsf to save code size. */
bsfl %eax, %eax
addq $(VEC_SIZE * 2 + 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4,, 10
L(first_vec_x4):
/* Use bsf to save code size. */
bsfl %eax, %eax
addq $(VEC_SIZE * 3 + 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4
L(aligned_more):
/* Align data to VEC_SIZE - 1. This is the same number of
instructions as using andq -VEC_SIZE but saves 4 bytes of code
on x4 check. */
orq $(VEC_SIZE - 1), %rdi
L(cross_page_continue):
/* Check the next 4 * VEC_SIZE. Only one VEC_SIZE at a time
since data is only aligned to VEC_SIZE. */
vmovdqa 1(%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jnz L(first_vec_x1)
vmovdqa (VEC_SIZE + 1)(%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jnz L(first_vec_x2)
vmovdqa (VEC_SIZE * 2 + 1)(%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jnz L(first_vec_x3)
vmovdqa (VEC_SIZE * 3 + 1)(%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jnz L(first_vec_x4)
/* Align data to VEC_SIZE * 4 - 1. */
incq %rdi
orq $(VEC_SIZE * 4 - 1), %rdi
.p2align 4
L(loop_4x_vec):
/* Compare 4 * VEC at a time forward. */
vmovdqa 1(%rdi), %ymm6
vmovdqa (VEC_SIZE + 1)(%rdi), %ymm7
/* Leaves only CHARS matching esi as 0. */
vpxor %ymm6, %ymm0, %ymm2
vpxor %ymm7, %ymm0, %ymm3
VPMINU %ymm2, %ymm6, %ymm2
VPMINU %ymm3, %ymm7, %ymm3
vmovdqa (VEC_SIZE * 2 + 1)(%rdi), %ymm6
vmovdqa (VEC_SIZE * 3 + 1)(%rdi), %ymm7
vpxor %ymm6, %ymm0, %ymm4
vpxor %ymm7, %ymm0, %ymm5
VPMINU %ymm4, %ymm6, %ymm4
VPMINU %ymm5, %ymm7, %ymm5
VPMINU %ymm2, %ymm3, %ymm6
VPMINU %ymm4, %ymm5, %ymm7
VPMINU %ymm6, %ymm7, %ymm7
VPCMPEQ %ymm7, %ymm1, %ymm7
vpmovmskb %ymm7, %ecx
subq $-(VEC_SIZE * 4), %rdi
testl %ecx, %ecx
jz L(loop_4x_vec)
VPCMPEQ %ymm2, %ymm1, %ymm2
vpmovmskb %ymm2, %eax
testl %eax, %eax
jnz L(last_vec_x0)
VPCMPEQ %ymm3, %ymm1, %ymm3
vpmovmskb %ymm3, %eax
testl %eax, %eax
jnz L(last_vec_x1)
VPCMPEQ %ymm4, %ymm1, %ymm4
vpmovmskb %ymm4, %eax
/* rcx has combined result from all 4 VEC. It will only be used
if the first 3 other VEC all did not contain a match. */
salq $32, %rcx
orq %rcx, %rax
tzcntq %rax, %rax
subq $(VEC_SIZE * 2 - 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero_end)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4,, 10
L(last_vec_x0):
/* Use bsf to save code size. */
bsfl %eax, %eax
addq $-(VEC_SIZE * 4 - 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero_end)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
.p2align 4,, 10
L(last_vec_x1):
tzcntl %eax, %eax
subq $(VEC_SIZE * 3 - 1), %rdi
# ifndef USE_AS_STRCHRNUL
/* Found CHAR or the null byte. */
cmp (%rdi, %rax), %CHAR_REG
jne L(zero_end)
# endif
addq %rdi, %rax
VZEROUPPER_RETURN
# ifndef USE_AS_STRCHRNUL
L(zero_end):
xorl %eax, %eax
VZEROUPPER_RETURN
# endif
/* Cold case for crossing page with first load. */
.p2align 4,, 8
L(cross_page_boundary):
movq %rdi, %rdx
/* Align rdi to VEC_SIZE - 1. */
orq $(VEC_SIZE - 1), %rdi
vmovdqa -(VEC_SIZE - 1)(%rdi), %ymm2
VPCMPEQ %ymm2, %ymm0, %ymm3
VPCMPEQ %ymm2, %ymm1, %ymm2
vpor %ymm3, %ymm2, %ymm3
vpmovmskb %ymm3, %eax
/* Remove the leading bytes. sarxl only uses bits [5:0] of COUNT
so no need to manually mod edx. */
sarxl %edx, %eax, %eax
testl %eax, %eax
jz L(cross_page_continue)
tzcntl %eax, %eax
# ifndef USE_AS_STRCHRNUL
xorl %ecx, %ecx
/* Found CHAR or the null byte. */
cmp (%rdx, %rax), %CHAR_REG
jne L(zero_end)
# endif
addq %rdx, %rax
VZEROUPPER_RETURN
END (STRCHR)
#endif