bionic/x86: Optimization for string routines

Optimized strcpy, strcat,
strncpy, strncat, strlcpy, strlcat,
memchr, memrchr, strchr, strrchr, index,
strnlen, strlen, wcslen, wmemcmp, wcscmp,
wcschr, wcsrchr, wcscpy, wcscat

Change-Id: I82b29132edf9a2e144e0bb3ee4ff5217df8d2a6d
Signed-off-by: Liubov Dmitrieva <liubov.dmitrieva@intel.com>
diff --git a/libc/arch-x86/string/sse2-memchr-atom.S b/libc/arch-x86/string/sse2-memchr-atom.S
new file mode 100644
index 0000000..013af9b
--- /dev/null
+++ b/libc/arch-x86/string/sse2-memchr-atom.S
@@ -0,0 +1,556 @@
+/*
+Copyright (c) 2011, Intel Corporation
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice,
+    * this list of conditions and the following disclaimer.
+
+    * Redistributions in binary form must reproduce the above copyright notice,
+    * this list of conditions and the following disclaimer in the documentation
+    * and/or other materials provided with the distribution.
+
+    * Neither the name of Intel Corporation nor the names of its contributors
+    * may be used to endorse or promote products derived from this software
+    * without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
+ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
+ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*/
+
+#ifndef L
+# define L(label)	.L##label
+#endif
+
+#ifndef cfi_startproc
+# define cfi_startproc	.cfi_startproc
+#endif
+
+#ifndef cfi_endproc
+# define cfi_endproc	.cfi_endproc
+#endif
+
+#ifndef cfi_rel_offset
+# define cfi_rel_offset(reg, off)	.cfi_rel_offset reg, off
+#endif
+
+#ifndef cfi_restore
+# define cfi_restore(reg)	.cfi_restore reg
+#endif
+
+#ifndef cfi_adjust_cfa_offset
+# define cfi_adjust_cfa_offset(off)	.cfi_adjust_cfa_offset off
+#endif
+
+#ifndef ENTRY
+# define ENTRY(name)	\
+	.type name,  @function;	\
+	.globl name;	\
+	.p2align 4;	\
+name:	\
+	cfi_startproc
+#endif
+
+#ifndef END
+# define END(name)	\
+	cfi_endproc;	\
+	.size name,	.-name
+#endif
+
+#define CFI_PUSH(REG)	\
+	cfi_adjust_cfa_offset (4);	\
+	cfi_rel_offset (REG, 0)
+
+#define CFI_POP(REG)	\
+	cfi_adjust_cfa_offset (-4);	\
+	cfi_restore (REG)
+
+#define PUSH(REG) pushl REG; CFI_PUSH (REG)
+#define POP(REG) popl REG; CFI_POP (REG)
+
+#define ENTRANCE PUSH (%edi);
+#define PARMS  8
+#define RETURN  POP (%edi); ret; CFI_PUSH (%edi);
+
+#define STR1  PARMS
+#define STR2  STR1+4
+#define LEN   STR2+4
+
+	.text
+ENTRY (memchr)
+	ENTRANCE
+	mov	STR1(%esp), %ecx
+	movd	STR2(%esp), %xmm1
+	mov	LEN(%esp), %edx
+	test	%edx, %edx
+	jz	L(return_null)
+
+	punpcklbw %xmm1, %xmm1
+	mov	%ecx, %edi
+	punpcklbw %xmm1, %xmm1
+
+	and	$63, %ecx
+	pshufd	$0, %xmm1, %xmm1
+	cmp	$48, %ecx
+	ja	L(crosscache)
+
+	movdqu	(%edi), %xmm0
+	pcmpeqb	%xmm1, %xmm0
+	pmovmskb %xmm0, %eax
+	test	%eax, %eax
+	jnz	L(match_case2_prolog)
+
+	sub	$16, %edx
+	jbe	L(return_null)
+	lea	16(%edi), %edi
+	and	$15, %ecx
+	and	$-16, %edi
+	add	%ecx, %edx
+	sub	$64, %edx
+	jbe	L(exit_loop)
+	jmp	L(loop_prolog)
+
+	.p2align 4
+L(crosscache):
+	and	$15, %ecx
+	and	$-16, %edi
+	movdqa	(%edi), %xmm0
+	pcmpeqb	%xmm1, %xmm0
+	pmovmskb %xmm0, %eax
+	sar	%cl, %eax
+	test	%eax, %eax
+
+	jnz	L(match_case2_prolog1)
+	lea	-16(%edx), %edx
+	add	%ecx, %edx
+	jle	L(return_null)
+	lea	16(%edi), %edi
+	sub	$64, %edx
+	jbe	L(exit_loop)
+
+	.p2align 4
+L(loop_prolog):
+	movdqa	(%edi), %xmm0
+	pcmpeqb	%xmm1, %xmm0
+	xor	%ecx, %ecx
+	pmovmskb %xmm0, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	16(%edi), %xmm2
+	pcmpeqb	%xmm1, %xmm2
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm2, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	32(%edi), %xmm3
+	pcmpeqb	%xmm1, %xmm3
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm3, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	48(%edi), %xmm4
+	pcmpeqb	%xmm1, %xmm4
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm4, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	lea	64(%edi), %edi
+	sub	$64, %edx
+	jbe	L(exit_loop)
+
+	movdqa	(%edi), %xmm0
+	pcmpeqb	%xmm1, %xmm0
+	xor	%ecx, %ecx
+	pmovmskb %xmm0, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	16(%edi), %xmm2
+	pcmpeqb	%xmm1, %xmm2
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm2, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	32(%edi), %xmm3
+	pcmpeqb	%xmm1, %xmm3
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm3, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	48(%edi), %xmm4
+	pcmpeqb	%xmm1, %xmm4
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm4, %eax
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	lea	64(%edi), %edi
+	mov	%edi, %ecx
+	and	$-64, %edi
+	and	$63, %ecx
+	add	%ecx, %edx
+
+	.p2align 4
+L(align64_loop):
+	sub	$64, %edx
+	jbe	L(exit_loop)
+	movdqa	(%edi), %xmm0
+	movdqa	16(%edi), %xmm2
+	movdqa	32(%edi), %xmm3
+	movdqa	48(%edi), %xmm4
+	pcmpeqb	%xmm1, %xmm0
+	pcmpeqb	%xmm1, %xmm2
+	pcmpeqb	%xmm1, %xmm3
+	pcmpeqb	%xmm1, %xmm4
+
+	pmaxub	%xmm0, %xmm3
+	pmaxub	%xmm2, %xmm4
+	pmaxub	%xmm3, %xmm4
+	add	$64, %edi
+	pmovmskb %xmm4, %eax
+
+	test	%eax, %eax
+	jz	L(align64_loop)
+
+	sub	$64, %edi
+
+	pmovmskb %xmm0, %eax
+	xor	%ecx, %ecx
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	pmovmskb %xmm2, %eax
+	lea	16(%ecx), %ecx
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	movdqa	32(%edi), %xmm3
+	pcmpeqb	%xmm1, %xmm3
+	pmovmskb %xmm3, %eax
+	lea	16(%ecx), %ecx
+	test	%eax, %eax
+	jnz	L(match_case1)
+
+	pcmpeqb	48(%edi), %xmm1
+	pmovmskb %xmm1, %eax
+	lea	16(%ecx), %ecx
+
+	.p2align 4
+L(match_case1):
+	add	%ecx, %edi
+	test	%al, %al
+	jz	L(match_case1_high)
+	mov	%al, %cl
+	and	$15, %cl
+	jz	L(match_case1_8)
+	test	$0x01, %al
+	jnz	L(exit_case1_1)
+	test	$0x02, %al
+	jnz	L(exit_case1_2)
+	test	$0x04, %al
+	jnz	L(exit_case1_3)
+	lea	3(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case1_8):
+	test	$0x10, %al
+	jnz	L(exit_case1_5)
+	test	$0x20, %al
+	jnz	L(exit_case1_6)
+	test	$0x40, %al
+	jnz	L(exit_case1_7)
+	lea	7(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case1_high):
+	mov	%ah, %ch
+	and	$15, %ch
+	jz	L(match_case1_high_8)
+	test	$0x01, %ah
+	jnz	L(exit_case1_9)
+	test	$0x02, %ah
+	jnz	L(exit_case1_10)
+	test	$0x04, %ah
+	jnz	L(exit_case1_11)
+	lea	11(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case1_high_8):
+	test	$0x10, %ah
+	jnz	L(exit_case1_13)
+	test	$0x20, %ah
+	jnz	L(exit_case1_14)
+	test	$0x40, %ah
+	jnz	L(exit_case1_15)
+	lea	15(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_loop):
+	add	$64, %edx
+
+	movdqa	(%edi), %xmm0
+	pcmpeqb	%xmm1, %xmm0
+	xor	%ecx, %ecx
+	pmovmskb %xmm0, %eax
+	test	%eax, %eax
+	jnz	L(match_case2)
+	cmp	$16, %edx
+	jbe	L(return_null)
+
+	movdqa	16(%edi), %xmm2
+	pcmpeqb	%xmm1, %xmm2
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm2, %eax
+	test	%eax, %eax
+	jnz	L(match_case2)
+	cmp	$32, %edx
+	jbe	L(return_null)
+
+	movdqa	32(%edi), %xmm3
+	pcmpeqb	%xmm1, %xmm3
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm3, %eax
+	test	%eax, %eax
+	jnz	L(match_case2)
+	cmp	$48, %edx
+	jbe	L(return_null)
+
+	pcmpeqb	48(%edi), %xmm1
+	lea	16(%ecx), %ecx
+	pmovmskb %xmm1, %eax
+	test	%eax, %eax
+	jnz	L(match_case2)
+
+	xor	%eax, %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_1):
+	mov	%edi, %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_2):
+	lea	1(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_3):
+	lea	2(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_5):
+	lea	4(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_6):
+	lea	5(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_7):
+	lea	6(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_9):
+	lea	8(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_10):
+	lea	9(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_11):
+	lea	10(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_13):
+	lea	12(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_14):
+	lea	13(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case1_15):
+	lea	14(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case2):
+	sub	%ecx, %edx
+L(match_case2_prolog1):
+	add	%ecx, %edi
+L(match_case2_prolog):
+	test	%al, %al
+	jz	L(match_case2_high)
+	mov	%al, %cl
+	and	$15, %cl
+	jz	L(match_case2_8)
+	test	$0x01, %al
+	jnz	L(exit_case2_1)
+	test	$0x02, %al
+	jnz	L(exit_case2_2)
+	test	$0x04, %al
+	jnz	L(exit_case2_3)
+	sub	$4, %edx
+	jb	L(return_null)
+	lea	3(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case2_8):
+	test	$0x10, %al
+	jnz	L(exit_case2_5)
+	test	$0x20, %al
+	jnz	L(exit_case2_6)
+	test	$0x40, %al
+	jnz	L(exit_case2_7)
+	sub	$8, %edx
+	jb	L(return_null)
+	lea	7(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case2_high):
+	mov	%ah, %ch
+	and	$15, %ch
+	jz	L(match_case2_high_8)
+	test	$0x01, %ah
+	jnz	L(exit_case2_9)
+	test	$0x02, %ah
+	jnz	L(exit_case2_10)
+	test	$0x04, %ah
+	jnz	L(exit_case2_11)
+	sub	$12, %edx
+	jb	L(return_null)
+	lea	11(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(match_case2_high_8):
+	test	$0x10, %ah
+	jnz	L(exit_case2_13)
+	test	$0x20, %ah
+	jnz	L(exit_case2_14)
+	test	$0x40, %ah
+	jnz	L(exit_case2_15)
+	sub	$16, %edx
+	jb	L(return_null)
+	lea	15(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_1):
+	mov	%edi, %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_2):
+	sub	$2, %edx
+	jb	L(return_null)
+	lea	1(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_3):
+	sub	$3, %edx
+	jb	L(return_null)
+	lea	2(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_5):
+	sub	$5, %edx
+	jb	L(return_null)
+	lea	4(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_6):
+	sub	$6, %edx
+	jb	L(return_null)
+	lea	5(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_7):
+	sub	$7, %edx
+	jb	L(return_null)
+	lea	6(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_9):
+	sub	$9, %edx
+	jb	L(return_null)
+	lea	8(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_10):
+	sub	$10, %edx
+	jb	L(return_null)
+	lea	9(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_11):
+	sub	$11, %edx
+	jb	L(return_null)
+	lea	10(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_13):
+	sub	$13, %edx
+	jb	L(return_null)
+	lea	12(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_14):
+	sub	$14, %edx
+	jb	L(return_null)
+	lea	13(%edi), %eax
+	RETURN
+
+	.p2align 4
+L(exit_case2_15):
+	sub	$15, %edx
+	jb	L(return_null)
+	lea	14(%edi), %eax
+	RETURN
+	.p2align 4
+L(return_null):
+	xor	%eax, %eax
+	RETURN
+END (memchr)