/* strcat with 256-bit EVEX instructions.
Copyright (C) 2021-2022 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
. */
#if IS_IN (libc)
# include
# ifndef STRCAT
# define STRCAT __strcat_evex
# endif
# define VMOVU vmovdqu64
# define VMOVA vmovdqa64
/* zero register */
# define XMMZERO xmm16
# define YMMZERO ymm16
# define YMM0 ymm17
# define YMM1 ymm18
# define USE_AS_STRCAT
/* Number of bytes in a vector register */
# define VEC_SIZE 32
.section .text.evex,"ax",@progbits
ENTRY (STRCAT)
mov %rdi, %r9
# ifdef USE_AS_STRNCAT
mov %rdx, %r8
# endif
xor %eax, %eax
mov %edi, %ecx
and $((VEC_SIZE * 4) - 1), %ecx
vpxorq %XMMZERO, %XMMZERO, %XMMZERO
cmp $(VEC_SIZE * 3), %ecx
ja L(fourth_vector_boundary)
vpcmpb $0, (%rdi), %YMMZERO, %k0
kmovd %k0, %edx
test %edx, %edx
jnz L(exit_null_on_first_vector)
mov %rdi, %rax
and $-VEC_SIZE, %rax
jmp L(align_vec_size_start)
L(fourth_vector_boundary):
mov %rdi, %rax
and $-VEC_SIZE, %rax
vpcmpb $0, (%rax), %YMMZERO, %k0
mov $-1, %r10d
sub %rax, %rcx
shl %cl, %r10d
kmovd %k0, %edx
and %r10d, %edx
jnz L(exit)
L(align_vec_size_start):
vpcmpb $0, VEC_SIZE(%rax), %YMMZERO, %k0
kmovd %k0, %edx
test %edx, %edx
jnz L(exit_null_on_second_vector)
vpcmpb $0, (VEC_SIZE * 2)(%rax), %YMMZERO, %k1
kmovd %k1, %edx
test %edx, %edx
jnz L(exit_null_on_third_vector)
vpcmpb $0, (VEC_SIZE * 3)(%rax), %YMMZERO, %k2
kmovd %k2, %edx
test %edx, %edx
jnz L(exit_null_on_fourth_vector)
vpcmpb $0, (VEC_SIZE * 4)(%rax), %YMMZERO, %k3
kmovd %k3, %edx
test %edx, %edx
jnz L(exit_null_on_fifth_vector)
vpcmpb $0, (VEC_SIZE * 5)(%rax), %YMMZERO, %k4
add $(VEC_SIZE * 4), %rax
kmovd %k4, %edx
test %edx, %edx
jnz L(exit_null_on_second_vector)
vpcmpb $0, (VEC_SIZE * 2)(%rax), %YMMZERO, %k1
kmovd %k1, %edx
test %edx, %edx
jnz L(exit_null_on_third_vector)
vpcmpb $0, (VEC_SIZE * 3)(%rax), %YMMZERO, %k2
kmovd %k2, %edx
test %edx, %edx
jnz L(exit_null_on_fourth_vector)
vpcmpb $0, (VEC_SIZE * 4)(%rax), %YMMZERO, %k3
kmovd %k3, %edx
test %edx, %edx
jnz L(exit_null_on_fifth_vector)
vpcmpb $0, (VEC_SIZE * 5)(%rax), %YMMZERO, %k4
kmovd %k4, %edx
add $(VEC_SIZE * 4), %rax
test %edx, %edx
jnz L(exit_null_on_second_vector)
vpcmpb $0, (VEC_SIZE * 2)(%rax), %YMMZERO, %k1
kmovd %k1, %edx
test %edx, %edx
jnz L(exit_null_on_third_vector)
vpcmpb $0, (VEC_SIZE * 3)(%rax), %YMMZERO, %k2
kmovd %k2, %edx
test %edx, %edx
jnz L(exit_null_on_fourth_vector)
vpcmpb $0, (VEC_SIZE * 4)(%rax), %YMMZERO, %k3
kmovd %k3, %edx
test %edx, %edx
jnz L(exit_null_on_fifth_vector)
vpcmpb $0, (VEC_SIZE * 5)(%rax), %YMMZERO, %k4
add $(VEC_SIZE * 4), %rax
kmovd %k4, %edx
test %edx, %edx
jnz L(exit_null_on_second_vector)
vpcmpb $0, (VEC_SIZE * 2)(%rax), %YMMZERO, %k1
kmovd %k1, %edx
test %edx, %edx
jnz L(exit_null_on_third_vector)
vpcmpb $0, (VEC_SIZE * 3)(%rax), %YMMZERO, %k2
kmovd %k2, %edx
test %edx, %edx
jnz L(exit_null_on_fourth_vector)
vpcmpb $0, (VEC_SIZE * 4)(%rax), %YMMZERO, %k3
kmovd %k3, %edx
test %edx, %edx
jnz L(exit_null_on_fifth_vector)
test $((VEC_SIZE * 4) - 1), %rax
jz L(align_four_vec_loop)
vpcmpb $0, (VEC_SIZE * 5)(%rax), %YMMZERO, %k4
add $(VEC_SIZE * 5), %rax
kmovd %k4, %edx
test %edx, %edx
jnz L(exit)
test $((VEC_SIZE * 4) - 1), %rax
jz L(align_four_vec_loop)
vpcmpb $0, VEC_SIZE(%rax), %YMMZERO, %k0
add $VEC_SIZE, %rax
kmovd %k0, %edx
test %edx, %edx
jnz L(exit)
test $((VEC_SIZE * 4) - 1), %rax
jz L(align_four_vec_loop)
vpcmpb $0, VEC_SIZE(%rax), %YMMZERO, %k0
add $VEC_SIZE, %rax
kmovd %k0, %edx
test %edx, %edx
jnz L(exit)
test $((VEC_SIZE * 4) - 1), %rax
jz L(align_four_vec_loop)
vpcmpb $0, VEC_SIZE(%rax), %YMMZERO, %k1
add $VEC_SIZE, %rax
kmovd %k1, %edx
test %edx, %edx
jnz L(exit)
add $VEC_SIZE, %rax
.p2align 4
L(align_four_vec_loop):
VMOVA (%rax), %YMM0
VMOVA (VEC_SIZE * 2)(%rax), %YMM1
vpminub VEC_SIZE(%rax), %YMM0, %YMM0
vpminub (VEC_SIZE * 3)(%rax), %YMM1, %YMM1
vpminub %YMM0, %YMM1, %YMM0
/* If K0 != 0, there is a null byte. */
vpcmpb $0, %YMM0, %YMMZERO, %k0
add $(VEC_SIZE * 4), %rax
ktestd %k0, %k0
jz L(align_four_vec_loop)
vpcmpb $0, -(VEC_SIZE * 4)(%rax), %YMMZERO, %k0
sub $(VEC_SIZE * 5), %rax
kmovd %k0, %edx
test %edx, %edx
jnz L(exit_null_on_second_vector)
vpcmpb $0, (VEC_SIZE * 2)(%rax), %YMMZERO, %k1
kmovd %k1, %edx
test %edx, %edx
jnz L(exit_null_on_third_vector)
vpcmpb $0, (VEC_SIZE * 3)(%rax), %YMMZERO, %k2
kmovd %k2, %edx
test %edx, %edx
jnz L(exit_null_on_fourth_vector)
vpcmpb $0, (VEC_SIZE * 4)(%rax), %YMMZERO, %k3
kmovd %k3, %edx
sub %rdi, %rax
bsf %rdx, %rdx
add %rdx, %rax
add $(VEC_SIZE * 4), %rax
jmp L(StartStrcpyPart)
.p2align 4
L(exit):
sub %rdi, %rax
L(exit_null_on_first_vector):
bsf %rdx, %rdx
add %rdx, %rax
jmp L(StartStrcpyPart)
.p2align 4
L(exit_null_on_second_vector):
sub %rdi, %rax
bsf %rdx, %rdx
add %rdx, %rax
add $VEC_SIZE, %rax
jmp L(StartStrcpyPart)
.p2align 4
L(exit_null_on_third_vector):
sub %rdi, %rax
bsf %rdx, %rdx
add %rdx, %rax
add $(VEC_SIZE * 2), %rax
jmp L(StartStrcpyPart)
.p2align 4
L(exit_null_on_fourth_vector):
sub %rdi, %rax
bsf %rdx, %rdx
add %rdx, %rax
add $(VEC_SIZE * 3), %rax
jmp L(StartStrcpyPart)
.p2align 4
L(exit_null_on_fifth_vector):
sub %rdi, %rax
bsf %rdx, %rdx
add %rdx, %rax
add $(VEC_SIZE * 4), %rax
.p2align 4
L(StartStrcpyPart):
lea (%r9, %rax), %rdi
mov %rsi, %rcx
mov %r9, %rax /* save result */
# ifdef USE_AS_STRNCAT
test %r8, %r8
jz L(ExitZero)
# define USE_AS_STRNCPY
# endif
# include "strcpy-evex.S"
#endif