[PATCH v1] x86: Move wcslen SSE2 implementation to multiarch/wcslen-sse2.S

[Noah Goldstein]

This commit doesn't affect libc.so.6, its just housekeeping to prepare
for adding explicit ISA level support.

Tested build on x86_64 and x86_32 with/without multiarch.
---
 sysdeps/x86_64/multiarch/wcslen-sse2.S | 221 ++++++++++++++++++++++++-
 sysdeps/x86_64/wcslen.S                | 216 +-----------------------
 2 files changed, 218 insertions(+), 219 deletions(-)

diff --git a/sysdeps/x86_64/multiarch/wcslen-sse2.S b/sysdeps/x86_64/multiarch/wcslen-sse2.S
index 2b3a9efd64..944c3bd9c6 100644
--- a/sysdeps/x86_64/multiarch/wcslen-sse2.S
+++ b/sysdeps/x86_64/multiarch/wcslen-sse2.S
@@ -17,10 +17,221 @@
    <https://www.gnu.org/licenses/>.  */
 
 #if IS_IN (libc)
-# define __wcslen __wcslen_sse2
-
-# undef weak_alias
-# define weak_alias(__wcslen, wcslen)
+# ifndef WCSLEN
+#  define WCSLEN __wcslen_sse2
+# endif
 #endif
 
-#include "../wcslen.S"
+#include <sysdep.h>
+
+	.text
+ENTRY (WCSLEN)
+	cmpl	$0, (%rdi)
+	jz	L(exit_tail0)
+	cmpl	$0, 4(%rdi)
+	jz	L(exit_tail1)
+	cmpl	$0, 8(%rdi)
+	jz	L(exit_tail2)
+	cmpl	$0, 12(%rdi)
+	jz	L(exit_tail3)
+	cmpl	$0, 16(%rdi)
+	jz	L(exit_tail4)
+	cmpl	$0, 20(%rdi)
+	jz	L(exit_tail5)
+	cmpl	$0, 24(%rdi)
+	jz	L(exit_tail6)
+	cmpl	$0, 28(%rdi)
+	jz	L(exit_tail7)
+
+	pxor	%xmm0, %xmm0
+
+	lea	32(%rdi), %rax
+	addq	$16, %rdi
+	and	$-16, %rax
+
+	pcmpeqd	(%rax), %xmm0
+	pmovmskb %xmm0, %edx
+	pxor	%xmm1, %xmm1
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm1
+	pmovmskb %xmm1, %edx
+	pxor	%xmm2, %xmm2
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm2
+	pmovmskb %xmm2, %edx
+	pxor	%xmm3, %xmm3
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm3
+	pmovmskb %xmm3, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm0
+	pmovmskb %xmm0, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm1
+	pmovmskb %xmm1, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm2
+	pmovmskb %xmm2, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm3
+	pmovmskb %xmm3, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm0
+	pmovmskb %xmm0, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm1
+	pmovmskb %xmm1, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm2
+	pmovmskb %xmm2, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	(%rax), %xmm3
+	pmovmskb %xmm3, %edx
+	addq	$16, %rax
+	test	%edx, %edx
+	jnz	L(exit)
+
+	and	$-0x40, %rax
+
+	.p2align 4
+L(aligned_64_loop):
+	movaps	(%rax), %xmm0
+	movaps	16(%rax), %xmm1
+	movaps	32(%rax), %xmm2
+	movaps	48(%rax), %xmm6
+
+	pminub	%xmm1, %xmm0
+	pminub	%xmm6, %xmm2
+	pminub	%xmm0, %xmm2
+	pcmpeqd	%xmm3, %xmm2
+	pmovmskb %xmm2, %edx
+	addq	$64, %rax
+	test	%edx, %edx
+	jz	L(aligned_64_loop)
+
+	pcmpeqd	-64(%rax), %xmm3
+	pmovmskb %xmm3, %edx
+    addq	$48, %rdi
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	%xmm1, %xmm3
+	pmovmskb %xmm3, %edx
+    addq	$-16, %rdi
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	-32(%rax), %xmm3
+	pmovmskb %xmm3, %edx
+    addq	$-16, %rdi
+	test	%edx, %edx
+	jnz	L(exit)
+
+	pcmpeqd	%xmm6, %xmm3
+	pmovmskb %xmm3, %edx
+    addq	$-16, %rdi
+	test	%edx, %edx
+	jz	L(aligned_64_loop)
+
+	.p2align 4
+L(exit):
+	sub	%rdi, %rax
+	shr	$2, %rax
+	test	%dl, %dl
+	jz	L(exit_high)
+
+	andl	$15, %edx
+	jz	L(exit_1)
+	ret
+
+	/* No align here. Naturally aligned % 16 == 1.  */
+L(exit_high):
+	andl	$(15 << 8), %edx
+	jz	L(exit_3)
+	add	$2, %rax
+	ret
+
+	.p2align 3
+L(exit_1):
+	add	$1, %rax
+	ret
+
+	.p2align 3
+L(exit_3):
+	add	$3, %rax
+	ret
+
+	.p2align 3
+L(exit_tail0):
+	xorl	%eax, %eax
+	ret
+
+	.p2align 3
+L(exit_tail1):
+	movl	$1, %eax
+	ret
+
+	.p2align 3
+L(exit_tail2):
+	movl	$2, %eax
+	ret
+
+	.p2align 3
+L(exit_tail3):
+	movl	$3, %eax
+	ret
+
+	.p2align 3
+L(exit_tail4):
+	movl	$4, %eax
+	ret
+
+	.p2align 3
+L(exit_tail5):
+	movl	$5, %eax
+	ret
+
+	.p2align 3
+L(exit_tail6):
+	movl	$6, %eax
+	ret
+
+	.p2align 3
+L(exit_tail7):
+	movl	$7, %eax
+	ret
+
+END (WCSLEN)
diff --git a/sysdeps/x86_64/wcslen.S b/sysdeps/x86_64/wcslen.S
index d641141d75..588a0fbe01 100644
--- a/sysdeps/x86_64/wcslen.S
+++ b/sysdeps/x86_64/wcslen.S
@@ -16,218 +16,6 @@
    License along with the GNU C Library; if not, see
    <https://www.gnu.org/licenses/>.  */
 
-#include <sysdep.h>
-
-	.text
-ENTRY (__wcslen)
-	cmpl	$0, (%rdi)
-	jz	L(exit_tail0)
-	cmpl	$0, 4(%rdi)
-	jz	L(exit_tail1)
-	cmpl	$0, 8(%rdi)
-	jz	L(exit_tail2)
-	cmpl	$0, 12(%rdi)
-	jz	L(exit_tail3)
-	cmpl	$0, 16(%rdi)
-	jz	L(exit_tail4)
-	cmpl	$0, 20(%rdi)
-	jz	L(exit_tail5)
-	cmpl	$0, 24(%rdi)
-	jz	L(exit_tail6)
-	cmpl	$0, 28(%rdi)
-	jz	L(exit_tail7)
-
-	pxor	%xmm0, %xmm0
-
-	lea	32(%rdi), %rax
-	addq	$16, %rdi
-	and	$-16, %rax
-
-	pcmpeqd	(%rax), %xmm0
-	pmovmskb %xmm0, %edx
-	pxor	%xmm1, %xmm1
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm1
-	pmovmskb %xmm1, %edx
-	pxor	%xmm2, %xmm2
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm2
-	pmovmskb %xmm2, %edx
-	pxor	%xmm3, %xmm3
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm3
-	pmovmskb %xmm3, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm0
-	pmovmskb %xmm0, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm1
-	pmovmskb %xmm1, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm2
-	pmovmskb %xmm2, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm3
-	pmovmskb %xmm3, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm0
-	pmovmskb %xmm0, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm1
-	pmovmskb %xmm1, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm2
-	pmovmskb %xmm2, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	(%rax), %xmm3
-	pmovmskb %xmm3, %edx
-	addq	$16, %rax
-	test	%edx, %edx
-	jnz	L(exit)
-
-	and	$-0x40, %rax
-
-	.p2align 4
-L(aligned_64_loop):
-	movaps	(%rax), %xmm0
-	movaps	16(%rax), %xmm1
-	movaps	32(%rax), %xmm2
-	movaps	48(%rax), %xmm6
-
-	pminub	%xmm1, %xmm0
-	pminub	%xmm6, %xmm2
-	pminub	%xmm0, %xmm2
-	pcmpeqd	%xmm3, %xmm2
-	pmovmskb %xmm2, %edx
-	addq	$64, %rax
-	test	%edx, %edx
-	jz	L(aligned_64_loop)
-
-	pcmpeqd	-64(%rax), %xmm3
-	pmovmskb %xmm3, %edx
-    addq	$48, %rdi
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	%xmm1, %xmm3
-	pmovmskb %xmm3, %edx
-    addq	$-16, %rdi
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	-32(%rax), %xmm3
-	pmovmskb %xmm3, %edx
-    addq	$-16, %rdi
-	test	%edx, %edx
-	jnz	L(exit)
-
-	pcmpeqd	%xmm6, %xmm3
-	pmovmskb %xmm3, %edx
-    addq	$-16, %rdi
-	test	%edx, %edx
-	jz	L(aligned_64_loop)
-
-	.p2align 4
-L(exit):
-	sub	%rdi, %rax
-	shr	$2, %rax
-	test	%dl, %dl
-	jz	L(exit_high)
-
-	andl	$15, %edx
-	jz	L(exit_1)
-	ret
-
-	/* No align here. Naturally aligned % 16 == 1.  */
-L(exit_high):
-	andl	$(15 << 8), %edx
-	jz	L(exit_3)
-	add	$2, %rax
-	ret
-
-	.p2align 3
-L(exit_1):
-	add	$1, %rax
-	ret
-
-	.p2align 3
-L(exit_3):
-	add	$3, %rax
-	ret
-
-	.p2align 3
-L(exit_tail0):
-	xorl	%eax, %eax
-	ret
-
-	.p2align 3
-L(exit_tail1):
-	movl	$1, %eax
-	ret
-
-	.p2align 3
-L(exit_tail2):
-	movl	$2, %eax
-	ret
-
-	.p2align 3
-L(exit_tail3):
-	movl	$3, %eax
-	ret
-
-	.p2align 3
-L(exit_tail4):
-	movl	$4, %eax
-	ret
-
-	.p2align 3
-L(exit_tail5):
-	movl	$5, %eax
-	ret
-
-	.p2align 3
-L(exit_tail6):
-	movl	$6, %eax
-	ret
-
-	.p2align 3
-L(exit_tail7):
-	movl	$7, %eax
-	ret
-
-END (__wcslen)
-
+#define WCSLEN	__wcslen
+#include "multiarch/wcslen-sse2.S"
 weak_alias(__wcslen, wcslen)
-- 
2.34.1