Use unsigned comparison in sse memcpy/memset

5 files changed, 90 insertions, 81 deletions

diff --git a/ChangeLog b/ChangeLog
index 7812fa6a48..cc4855fc27 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,14 @@
+2010-02-12  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* sysdeps/i386/i686/multiarch/memcpy-ssse3-rep.S: Use unsigned
+	conditional jumps.
+	(shl_0_gobble_cache_loop_tail): Removed.
+	* sysdeps/i386/i686/multiarch/memcpy-ssse3.S: Likewise.
+
+	* sysdeps/i386/i686/multiarch/memset-sse2-rep.S: Use unsigned
+	conditional jumps.
+	* sysdeps/i386/i686/multiarch/memset-sse2.S: Likewise.
+
 2009-10-27  Aurelien Jarno  <aurelien@aurel32.net>
 
 	[BZ #10855]
diff --git a/sysdeps/i386/i686/multiarch/memcpy-ssse3-rep.S b/sysdeps/i386/i686/multiarch/memcpy-ssse3-rep.S
index b26037d279..f85049185c 100644
--- a/sysdeps/i386/i686/multiarch/memcpy-ssse3-rep.S
+++ b/sysdeps/i386/i686/multiarch/memcpy-ssse3-rep.S
@@ -128,7 +128,7 @@ ENTRY (MEMCPY)
 	jb	L(copy_forward)
 	je	L(fwd_write_0bytes)
 	cmp	$32, %ecx
-	jge	L(memmove_bwd)
+	jae	L(memmove_bwd)
 	jmp	L(bk_write_less32bytes_2)
 L(memmove_bwd):
 	add	%ecx, %eax
@@ -139,12 +139,12 @@ L(memmove_bwd):
 L(copy_forward):
 #endif
 	cmp	$48, %ecx
-	jge	L(48bytesormore)
+	jae	L(48bytesormore)
 
 L(fwd_write_less32bytes):
 #ifndef USE_AS_MEMMOVE
 	cmp	%dl, %al
-	jl	L(bk_write)
+	jb	L(bk_write)
 #endif
 	add	%ecx, %edx
 	add	%ecx, %eax
@@ -181,7 +181,7 @@ L(48bytesormore):
 #endif
 
 	mov	%eax, %edi
-	jge	L(large_page)
+	jae	L(large_page)
 	and	$0xf, %edi
 	jz	L(shl_0)
 
@@ -201,7 +201,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -209,7 +209,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -217,7 +217,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -251,7 +251,7 @@ L(shl_0_gobble):
 	shr	$3, %esi
 	sub	%esi, %edi
 	cmp	%edi, %ecx
-	jge	L(shl_0_gobble_mem_start)
+	jae	L(shl_0_gobble_mem_start)
 	lea	-128(%ecx), %ecx
 	ALIGN (4)
 L(shl_0_gobble_cache_loop):
@@ -275,8 +275,7 @@ L(shl_0_gobble_cache_loop):
 	movaps	%xmm7, 0x70(%edx)
 	lea	0x80(%edx), %edx
 
-	jge	L(shl_0_gobble_cache_loop)
-L(shl_0_gobble_cache_loop_tail):
+	jae	L(shl_0_gobble_cache_loop)
 	cmp	$-0x40, %ecx
 	lea	0x80(%ecx), %ecx
 	jl	L(shl_0_cache_less_64bytes)
@@ -297,7 +296,7 @@ L(shl_0_gobble_cache_loop_tail):
 	add	$0x40, %edx
 L(shl_0_cache_less_64bytes):
 	cmp	$0x20, %ecx
-	jl	L(shl_0_cache_less_32bytes)
+	jb	L(shl_0_cache_less_32bytes)
 	movdqa	(%eax), %xmm0
 	sub	$0x20, %ecx
 	movdqa	0x10(%eax), %xmm1
@@ -307,7 +306,7 @@ L(shl_0_cache_less_64bytes):
 	add	$0x20, %edx
 L(shl_0_cache_less_32bytes):
 	cmp	$0x10, %ecx
-	jl	L(shl_0_cache_less_16bytes)
+	jb	L(shl_0_cache_less_16bytes)
 	sub	$0x10, %ecx
 	movdqa	(%eax), %xmm0
 	add	$0x10, %eax
@@ -352,7 +351,7 @@ L(shl_0_gobble_mem_loop):
 	movaps	%xmm7, 0x70(%edx)
 	lea	0x80(%edx), %edx
 
-	jge	L(shl_0_gobble_mem_loop)
+	jae	L(shl_0_gobble_mem_loop)
 	cmp	$-0x40, %ecx
 	lea	0x80(%ecx), %ecx
 	jl	L(shl_0_mem_less_64bytes)
@@ -373,7 +372,7 @@ L(shl_0_gobble_mem_loop):
 	add	$0x40, %edx
 L(shl_0_mem_less_64bytes):
 	cmp	$0x20, %ecx
-	jl	L(shl_0_mem_less_32bytes)
+	jb	L(shl_0_mem_less_32bytes)
 	movdqa	(%eax), %xmm0
 	sub	$0x20, %ecx
 	movdqa	0x10(%eax), %xmm1
@@ -383,7 +382,7 @@ L(shl_0_mem_less_64bytes):
 	add	$0x20, %edx
 L(shl_0_mem_less_32bytes):
 	cmp	$0x10, %ecx
-	jl	L(shl_0_mem_less_16bytes)
+	jb	L(shl_0_mem_less_16bytes)
 	sub	$0x10, %ecx
 	movdqa	(%eax), %xmm0
 	add	$0x10, %eax
@@ -418,7 +417,7 @@ L(shl_1_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_1_end)
+	jb	L(shl_1_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -461,7 +460,7 @@ L(shl_2_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_2_end)
+	jb	L(shl_2_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -504,7 +503,7 @@ L(shl_3_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_3_end)
+	jb	L(shl_3_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -547,7 +546,7 @@ L(shl_4_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_4_end)
+	jb	L(shl_4_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -590,7 +589,7 @@ L(shl_5_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_5_end)
+	jb	L(shl_5_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -634,7 +633,7 @@ L(shl_6_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_6_end)
+	jb	L(shl_6_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -677,7 +676,7 @@ L(shl_7_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_7_end)
+	jb	L(shl_7_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -720,7 +719,7 @@ L(shl_8_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_8_end)
+	jb	L(shl_8_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -763,7 +762,7 @@ L(shl_9_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_9_end)
+	jb	L(shl_9_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -806,7 +805,7 @@ L(shl_10_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_10_end)
+	jb	L(shl_10_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -849,7 +848,7 @@ L(shl_11_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_11_end)
+	jb	L(shl_11_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -892,7 +891,7 @@ L(shl_12_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_12_end)
+	jb	L(shl_12_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -935,7 +934,7 @@ L(shl_13_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_13_end)
+	jb	L(shl_13_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -978,7 +977,7 @@ L(shl_14_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_14_end)
+	jb	L(shl_14_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -1022,7 +1021,7 @@ L(shl_15_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_15_end)
+	jb	L(shl_15_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -1298,7 +1297,7 @@ L(large_page_loop):
 	sub	$0x40, %ecx
 L(large_page_less_64bytes):
 	cmp	$32, %ecx
-	jl	L(large_page_less_32bytes)
+	jb	L(large_page_less_32bytes)
 	movdqu	(%eax), %xmm0
 	movdqu	0x10(%eax), %xmm1
 	lea	0x20(%eax), %eax
@@ -1665,11 +1664,11 @@ L(copy_backward):
 
 L(bk_aligned_4):
 	cmp	$64, %ecx
-	jge	L(bk_write_more64bytes)
+	jae	L(bk_write_more64bytes)
 
 L(bk_write_64bytesless):
 	cmp	$32, %ecx
-	jl	L(bk_write_less32bytes)
+	jb	L(bk_write_less32bytes)
 
 L(bk_write_more32bytes):
 	/* Copy 32 bytes at a time.  */
@@ -1704,7 +1703,7 @@ L(bk_write_less32bytes_2):
 	ALIGN (4)
 L(bk_align):
 	cmp	$8, %ecx
-	jle	L(bk_write_less32bytes)
+	jbe	L(bk_write_less32bytes)
 	testl	$1, %edx
 	/* We get here only if (EDX & 3 ) != 0 so if (EDX & 1) ==0,
 	   then (EDX & 2) must be != 0.  */
@@ -1760,7 +1759,7 @@ L(bk_ssse3_align):
 
 L(bk_ssse3_cpy_pre):
 	cmp	$64, %ecx
-	jl	L(bk_write_more32bytes)
+	jb	L(bk_write_more32bytes)
 
 L(bk_ssse3_cpy):
 	sub	$64, %esi
@@ -1775,7 +1774,7 @@ L(bk_ssse3_cpy):
 	movdqu	(%esi), %xmm0
 	movdqa	%xmm0, (%edx)
 	cmp	$64, %ecx
-	jge	L(bk_ssse3_cpy)
+	jae	L(bk_ssse3_cpy)
 	jmp	L(bk_write_64bytesless)
 
 #endif
diff --git a/sysdeps/i386/i686/multiarch/memcpy-ssse3.S b/sysdeps/i386/i686/multiarch/memcpy-ssse3.S
index 749c82d379..c512b0e812 100644
--- a/sysdeps/i386/i686/multiarch/memcpy-ssse3.S
+++ b/sysdeps/i386/i686/multiarch/memcpy-ssse3.S
@@ -128,7 +128,7 @@ ENTRY (MEMCPY)
 	jb	L(copy_forward)
 	je	L(fwd_write_0bytes)
 	cmp	$32, %ecx
-	jge	L(memmove_bwd)
+	jae	L(memmove_bwd)
 	jmp	L(bk_write_less32bytes_2)
 L(memmove_bwd):
 	add	%ecx, %eax
@@ -139,12 +139,12 @@ L(memmove_bwd):
 L(copy_forward):
 #endif
 	cmp	$48, %ecx
-	jge	L(48bytesormore)
+	jae	L(48bytesormore)
 
 L(fwd_write_less32bytes):
 #ifndef USE_AS_MEMMOVE
 	cmp	%dl, %al
-	jl	L(bk_write)
+	jb	L(bk_write)
 #endif
 	add	%ecx, %edx
 	add	%ecx, %eax
@@ -181,7 +181,7 @@ L(48bytesormore):
 #endif
 
 	mov	%eax, %edi
-	jge	L(large_page)
+	jae	L(large_page)
 	and	$0xf, %edi
 	jz	L(shl_0)
 
@@ -202,7 +202,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -210,7 +210,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -218,7 +218,7 @@ L(shl_0_loop):
 	movdqa	%xmm0, (%edx, %edi)
 	movdqa	%xmm1, 16(%edx, %edi)
 	lea	32(%edi), %edi
-	jl	L(shl_0_end)
+	jb	L(shl_0_end)
 
 	movdqa	(%eax, %edi), %xmm0
 	movdqa	16(%eax, %edi), %xmm1
@@ -250,7 +250,7 @@ L(shl_0_gobble):
 
 	POP (%edi)
 	lea	-128(%ecx), %ecx
-	jge	L(shl_0_gobble_mem_loop)
+	jae	L(shl_0_gobble_mem_loop)
 L(shl_0_gobble_cache_loop):
 	movdqa	(%eax), %xmm0
 	movdqa	0x10(%eax), %xmm1
@@ -272,8 +272,7 @@ L(shl_0_gobble_cache_loop):
 	movdqa	%xmm7, 0x70(%edx)
 	lea	0x80(%edx), %edx
 
-	jge	L(shl_0_gobble_cache_loop)
-L(shl_0_gobble_cache_loop_tail):
+	jae	L(shl_0_gobble_cache_loop)
 	cmp	$-0x40, %ecx
 	lea	0x80(%ecx), %ecx
 	jl	L(shl_0_cache_less_64bytes)
@@ -294,7 +293,7 @@ L(shl_0_gobble_cache_loop_tail):
 	add	$0x40, %edx
 L(shl_0_cache_less_64bytes):
 	cmp	$0x20, %ecx
-	jl	L(shl_0_cache_less_32bytes)
+	jb	L(shl_0_cache_less_32bytes)
 	movdqa	(%eax), %xmm0
 	sub	$0x20, %ecx
 	movdqa	0x10(%eax), %xmm1
@@ -304,7 +303,7 @@ L(shl_0_cache_less_64bytes):
 	add	$0x20, %edx
 L(shl_0_cache_less_32bytes):
 	cmp	$0x10, %ecx
-	jl	L(shl_0_cache_less_16bytes)
+	jb	L(shl_0_cache_less_16bytes)
 	sub	$0x10, %ecx
 	movdqa	(%eax), %xmm0
 	add	$0x10, %eax
@@ -342,7 +341,7 @@ L(shl_0_gobble_mem_loop):
 	movdqa	%xmm7, 0x70(%edx)
 	lea	0x80(%edx), %edx
 
-	jge	L(shl_0_gobble_mem_loop)
+	jae	L(shl_0_gobble_mem_loop)
 	cmp	$-0x40, %ecx
 	lea	0x80(%ecx), %ecx
 	jl	L(shl_0_mem_less_64bytes)
@@ -363,7 +362,7 @@ L(shl_0_gobble_mem_loop):
 	add	$0x40, %edx
 L(shl_0_mem_less_64bytes):
 	cmp	$0x20, %ecx
-	jl	L(shl_0_mem_less_32bytes)
+	jb	L(shl_0_mem_less_32bytes)
 	movdqa	(%eax), %xmm0
 	sub	$0x20, %ecx
 	movdqa	0x10(%eax), %xmm1
@@ -373,7 +372,7 @@ L(shl_0_mem_less_64bytes):
 	add	$0x20, %edx
 L(shl_0_mem_less_32bytes):
 	cmp	$0x10, %ecx
-	jl	L(shl_0_mem_less_16bytes)
+	jb	L(shl_0_mem_less_16bytes)
 	sub	$0x10, %ecx
 	movdqa	(%eax), %xmm0
 	add	$0x10, %eax
@@ -406,7 +405,7 @@ L(shl_1_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_1_end)
+	jb	L(shl_1_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -449,7 +448,7 @@ L(shl_2_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_2_end)
+	jb	L(shl_2_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -492,7 +491,7 @@ L(shl_3_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_3_end)
+	jb	L(shl_3_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -535,7 +534,7 @@ L(shl_4_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_4_end)
+	jb	L(shl_4_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -578,7 +577,7 @@ L(shl_5_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_5_end)
+	jb	L(shl_5_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -622,7 +621,7 @@ L(shl_6_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_6_end)
+	jb	L(shl_6_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -665,7 +664,7 @@ L(shl_7_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_7_end)
+	jb	L(shl_7_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -708,7 +707,7 @@ L(shl_8_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_8_end)
+	jb	L(shl_8_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -751,7 +750,7 @@ L(shl_9_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_9_end)
+	jb	L(shl_9_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -794,7 +793,7 @@ L(shl_10_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_10_end)
+	jb	L(shl_10_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -837,7 +836,7 @@ L(shl_11_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_11_end)
+	jb	L(shl_11_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -880,7 +879,7 @@ L(shl_12_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_12_end)
+	jb	L(shl_12_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -923,7 +922,7 @@ L(shl_13_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_13_end)
+	jb	L(shl_13_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -966,7 +965,7 @@ L(shl_14_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_14_end)
+	jb	L(shl_14_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -1010,7 +1009,7 @@ L(shl_15_loop):
 	movdqa	%xmm2, -32(%edx, %edi)
 	movdqa	%xmm3, -16(%edx, %edi)
 
-	jl	L(shl_15_end)
+	jb	L(shl_15_end)
 
 	movdqa	16(%eax, %edi), %xmm2
 	sub	$32, %ecx
@@ -1281,7 +1280,7 @@ L(large_page_loop):
 	sub	$0x40, %ecx
 L(large_page_less_64bytes):
 	cmp	$32, %ecx
-	jl	L(large_page_less_32bytes)
+	jb	L(large_page_less_32bytes)
 	movdqu	(%eax), %xmm0
 	movdqu	0x10(%eax), %xmm1
 	lea	0x20(%eax), %eax
@@ -1617,11 +1616,11 @@ L(copy_backward):
 
 L(bk_aligned_4):
 	cmp	$64, %ecx
-	jge	L(bk_write_more64bytes)
+	jae	L(bk_write_more64bytes)
 
 L(bk_write_64bytesless):
 	cmp	$32, %ecx
-	jl	L(bk_write_less32bytes)
+	jb	L(bk_write_less32bytes)
 
 L(bk_write_more32bytes):
 	/* Copy 32 bytes at a time.  */
@@ -1656,7 +1655,7 @@ L(bk_write_less32bytes_2):
 	ALIGN (4)
 L(bk_align):
 	cmp	$8, %ecx
-	jle	L(bk_write_less32bytes)
+	jbe	L(bk_write_less32bytes)
 	testl	$1, %edx
 	/* We get here only if (EDX & 3 ) != 0 so if (EDX & 1) ==0,
 	   then (EDX & 2) must be != 0.  */
@@ -1712,7 +1711,7 @@ L(bk_ssse3_align):
 
 L(bk_ssse3_cpy_pre):
 	cmp	$64, %ecx
-	jl	L(bk_write_more32bytes)
+	jb	L(bk_write_more32bytes)
 
 L(bk_ssse3_cpy):
 	sub	$64, %esi
@@ -1727,7 +1726,7 @@ L(bk_ssse3_cpy):
 	movdqu	(%esi), %xmm0
 	movdqa	%xmm0, (%edx)
 	cmp	$64, %ecx
-	jge	L(bk_ssse3_cpy)
+	jae	L(bk_ssse3_cpy)
 	jmp	L(bk_write_64bytesless)
 
 #endif
diff --git a/sysdeps/i386/i686/multiarch/memset-sse2-rep.S b/sysdeps/i386/i686/multiarch/memset-sse2-rep.S
index 84afffeb66..d4bf9b7d3e 100644
--- a/sysdeps/i386/i686/multiarch/memset-sse2-rep.S
+++ b/sysdeps/i386/i686/multiarch/memset-sse2-rep.S
@@ -261,7 +261,7 @@ L(not_aligned_16):
 	ALIGN (4)
 L(aligned_16):
 	cmp	$128, %ecx
-	jge	L(128bytesormore)
+	jae	L(128bytesormore)
 
 L(aligned_16_less128bytes):
 	BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
@@ -306,7 +306,7 @@ L(128bytesormore_normal):
 	movdqa	%xmm0, 0x60(%edx)
 	movdqa	%xmm0, 0x70(%edx)
 	lea	128(%edx), %edx
-	jl	L(128bytesless_normal)
+	jb	L(128bytesless_normal)
 
 
 	sub	$128, %ecx
@@ -319,7 +319,7 @@ L(128bytesormore_normal):
 	movdqa	%xmm0, 0x60(%edx)
 	movdqa	%xmm0, 0x70(%edx)
 	lea	128(%edx), %edx
-	jge	L(128bytesormore_normal)
+	jae	L(128bytesormore_normal)
 
 L(128bytesless_normal):
 	POP (%edi)
diff --git a/sysdeps/i386/i686/multiarch/memset-sse2.S b/sysdeps/i386/i686/multiarch/memset-sse2.S
index b2b979193e..00e552e44e 100644
--- a/sysdeps/i386/i686/multiarch/memset-sse2.S
+++ b/sysdeps/i386/i686/multiarch/memset-sse2.S
@@ -261,7 +261,7 @@ L(not_aligned_16):
 	ALIGN (4)
 L(aligned_16):
 	cmp	$128, %ecx
-	jge	L(128bytesormore)
+	jae	L(128bytesormore)
 
 L(aligned_16_less128bytes):
 	BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
@@ -312,7 +312,7 @@ L(128bytesormore_normal):
 	movdqa	%xmm0, 0x60(%edx)
 	movdqa	%xmm0, 0x70(%edx)
 	lea	128(%edx), %edx
-	jl	L(128bytesless_normal)
+	jb	L(128bytesless_normal)
 
 
 	sub	$128, %ecx
@@ -325,7 +325,7 @@ L(128bytesormore_normal):
 	movdqa	%xmm0, 0x60(%edx)
 	movdqa	%xmm0, 0x70(%edx)
 	lea	128(%edx), %edx
-	jge	L(128bytesormore_normal)
+	jae	L(128bytesormore_normal)
 
 L(128bytesless_normal):
 	lea	128(%ecx), %ecx
@@ -346,7 +346,7 @@ L(128bytes_L2_normal):
 	movaps	%xmm0, 0x70(%edx)
 	add	$128, %edx
 	cmp	$128, %ecx
-	jge	L(128bytes_L2_normal)
+	jae	L(128bytes_L2_normal)
 
 L(128bytesless_L2_normal):
 	BRANCH_TO_JMPTBL_ENTRY (L(table_16_128bytes))
@@ -368,7 +368,7 @@ L(128bytesormore_shared_cache_loop):
 	movdqa	%xmm0, 0x70(%edx)
 	add	$0x80, %edx
 	cmp	$0x80, %ebx
-	jge	L(128bytesormore_shared_cache_loop)
+	jae	L(128bytesormore_shared_cache_loop)
 	cmp	$0x80, %ecx
 	jb	L(shared_cache_loop_end)
 	ALIGN (4)
@@ -384,7 +384,7 @@ L(128bytesormore_nt):
 	movntdq	%xmm0, 0x70(%edx)
 	add	$0x80, %edx
 	cmp	$0x80, %ecx
-	jge	L(128bytesormore_nt)
+	jae	L(128bytesormore_nt)
 	sfence
 L(shared_cache_loop_end):
 #if defined DATA_CACHE_SIZE || !defined SHARED