Add Atom-optimized strchr and strrchr for x86-64

1 files changed, 282 insertions, 0 deletions

diff --git a/sysdeps/x86_64/multiarch/strchr-sse2-no-bsf.S b/sysdeps/x86_64/multiarch/strchr-sse2-no-bsf.S
new file mode 100644
index 0000000000..e3f080ccb3
--- /dev/null
+++ b/sysdeps/x86_64/multiarch/strchr-sse2-no-bsf.S
@@ -0,0 +1,282 @@
+/* strchr with SSE2 without bsf
+   Copyright (C) 2011 Free Software Foundation, Inc.
+   Contributed by Intel Corporation.
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef  NOT_IN_libc
+
+# include <sysdep.h>
+# include "asm-syntax.h"
+
+	.text
+ENTRY (__strchr_sse2_no_bsf)
+	movd	%esi, %xmm1
+	movq	%rdi, %rcx
+	punpcklbw %xmm1, %xmm1
+	andq	$~15, %rdi
+	pxor	%xmm2, %xmm2
+	punpcklbw %xmm1, %xmm1
+	orl	$0xffffffff, %esi
+	movdqa	(%rdi), %xmm0
+	pshufd	$0, %xmm1, %xmm1
+	subq	%rdi, %rcx
+	movdqa	%xmm0, %xmm3
+	leaq	16(%rdi), %rdi
+	pcmpeqb	%xmm1, %xmm0
+	pcmpeqb	%xmm2, %xmm3
+	shl	%cl, %esi
+	pmovmskb %xmm0, %eax
+	pmovmskb %xmm3, %edx
+	andl	%esi, %eax
+	andl	%esi, %edx
+	test	%eax, %eax
+	jnz	L(matches)
+	test	%edx, %edx
+	jnz	L(return_null)
+
+L(loop):
+	movdqa	(%rdi), %xmm0
+	leaq	16(%rdi), %rdi
+	movdqa	%xmm0, %xmm3
+	pcmpeqb	%xmm1, %xmm0
+	pcmpeqb	%xmm2, %xmm3
+	pmovmskb %xmm0, %eax
+	pmovmskb %xmm3, %edx
+	or	%eax, %edx
+	jz	L(loop)
+
+	pmovmskb %xmm3, %edx
+	test	%eax, %eax
+	jnz	L(matches)
+
+/* Return NULL.  */
+	.p2align 4
+L(return_null):
+	xor	%rax, %rax
+	ret
+
+L(matches):
+	/* There is a match.  First find where NULL is.  */
+	leaq	-16(%rdi), %rdi
+	test	%edx, %edx
+	jz	L(match_case1)
+
+	.p2align 4
+L(match_case2):
+	test	%al, %al
+	jz	L(match_high_case2)
+
+	mov	%al, %cl
+	and	$15, %cl
+	jnz	L(match_case2_4)
+
+	mov	%dl, %ch
+	and	$15, %ch
+	jnz	L(return_null)
+
+	test	$0x10, %al
+	jnz	L(Exit5)
+	test	$0x10, %dl
+	jnz	L(return_null)
+	test	$0x20, %al
+	jnz	L(Exit6)
+	test	$0x20, %dl
+	jnz	L(return_null)
+	test	$0x40, %al
+	jnz	L(Exit7)
+	test	$0x40, %dl
+	jnz	L(return_null)
+	lea	7(%rdi), %rax
+	ret
+
+	.p2align 4
+L(match_case2_4):
+	test	$0x01, %al
+	jnz	L(Exit1)
+	test	$0x01, %dl
+	jnz	L(return_null)
+	test	$0x02, %al
+	jnz	L(Exit2)
+	test	$0x02, %dl
+	jnz	L(return_null)
+	test	$0x04, %al
+	jnz	L(Exit3)
+	test	$0x04, %dl
+	jnz	L(return_null)
+	lea	3(%rdi), %rax
+	ret
+
+	.p2align 4
+L(match_high_case2):
+	test	%dl, %dl
+	jnz	L(return_null)
+
+	mov	%ah, %cl
+	and	$15, %cl
+	jnz	L(match_case2_12)
+
+	mov	%dh, %ch
+	and	$15, %ch
+	jnz	L(return_null)
+
+	test	$0x10, %ah
+	jnz	L(Exit13)
+	test	$0x10, %dh
+	jnz	L(return_null)
+	test	$0x20, %ah
+	jnz	L(Exit14)
+	test	$0x20, %dh
+	jnz	L(return_null)
+	test	$0x40, %ah
+	jnz	L(Exit15)
+	test	$0x40, %dh
+	jnz	L(return_null)
+	lea	15(%rdi), %rax
+	ret
+
+	.p2align 4
+L(match_case2_12):
+	test	$0x01, %ah
+	jnz	L(Exit9)
+	test	$0x01, %dh
+	jnz	L(return_null)
+	test	$0x02, %ah
+	jnz	L(Exit10)
+	test	$0x02, %dh
+	jnz	L(return_null)
+	test	$0x04, %ah
+	jnz	L(Exit11)
+	test	$0x04, %dh
+	jnz	L(return_null)
+	lea	11(%rdi), %rax
+	ret
+
+	.p2align 4
+L(match_case1):
+	test	%al, %al
+	jz	L(match_high_case1)
+
+	test	$0x01, %al
+	jnz	L(Exit1)
+	test	$0x02, %al
+	jnz	L(Exit2)
+	test	$0x04, %al
+	jnz	L(Exit3)
+	test	$0x08, %al
+	jnz	L(Exit4)
+	test	$0x10, %al
+	jnz	L(Exit5)
+	test	$0x20, %al
+	jnz	L(Exit6)
+	test	$0x40, %al
+	jnz	L(Exit7)
+	lea	7(%rdi), %rax
+	ret
+
+	.p2align 4
+L(match_high_case1):
+	test	$0x01, %ah
+	jnz	L(Exit9)
+	test	$0x02, %ah
+	jnz	L(Exit10)
+	test	$0x04, %ah
+	jnz	L(Exit11)
+	test	$0x08, %ah
+	jnz	L(Exit12)
+	test	$0x10, %ah
+	jnz	L(Exit13)
+	test	$0x20, %ah
+	jnz	L(Exit14)
+	test	$0x40, %ah
+	jnz	L(Exit15)
+	lea	15(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit1):
+	lea	(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit2):
+	lea	1(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit3):
+	lea	2(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit4):
+	lea	3(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit5):
+	lea	4(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit6):
+	lea	5(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit7):
+	lea	6(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit9):
+	lea	8(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit10):
+	lea	9(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit11):
+	lea	10(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit12):
+	lea	11(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit13):
+	lea	12(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit14):
+	lea	13(%rdi), %rax
+	ret
+
+	.p2align 4
+L(Exit15):
+	lea	14(%rdi), %rax
+	ret
+
+END (__strchr_sse2_no_bsf)
+#endif
+