BasicOpsX86: regenerated with final 4.4.0 release of GCC

Regenerated BasicOps assemblies with final 4.4.0 release of GCC,
providing some more minor optimizations.

CMakeLists.txt

@@ -441,7 +441,7 @@ FOREACH(opt_target ${opt_targets})
 	ELSE(EXISTS "$ENV{SVN_C_COMPILER}")
 		SET(C_COMPILER ${CMAKE_C_COMPILER})
 	ENDIF(EXISTS "$ENV{SVN_C_COMPILER}")
-	ADD_CUSTOM_TARGET(regen-basic-ops-${opt_target} COMMAND ${C_COMPILER} -O2 -ftree-vectorize -ftree-vectorizer-verbose=2 -fomit-frame-pointer -c -S -I${CMAKE_SOURCE_DIR}/include -I${CMAKE_BINARY_DIR} -g0 -DBUILD_${OPT_TARGET} -m${opt_target} ${FPMATH_FLAGS} -o ${BASIC_OPS_X86_TARGET_S} ${BASIC_OPS_X86_C} DEPENDS ${BASIC_OPS_X86_C})
+	ADD_CUSTOM_TARGET(regen-basic-ops-${opt_target} COMMAND ${C_COMPILER} -O2 -fno-stack-protector -ftree-vectorize -ftree-vectorizer-verbose=2 -fomit-frame-pointer -c -S -I${CMAKE_SOURCE_DIR}/include -I${CMAKE_BINARY_DIR} -g0 -DBUILD_${OPT_TARGET} -m${opt_target} ${FPMATH_FLAGS} -o ${BASIC_OPS_X86_TARGET_S} ${BASIC_OPS_X86_C} DEPENDS ${BASIC_OPS_X86_C})
 	ADD_CUSTOM_COMMAND(OUTPUT ${BASIC_OPS_X86_TARGET_O} COMMAND ${CMAKE_C_COMPILER} ARGS ${BASIC_OPS_X86_TARGET_S} -c -o ${BASIC_OPS_X86_TARGET_O} DEPENDS ${BASIC_OPS_X86_TARGET_S})
 	ADD_DEPENDENCIES(regen-basic-ops regen-basic-ops-${opt_target})
 	SET(opt_target_objects ${opt_target_objects} ${BASIC_OPS_X86_TARGET_O})

src/core/basic_ops_x86_mmx.s

@@ -103,5 +103,5 @@ movq    %mm0, 56(%edx)
 	emms
 	ret
 	.size	alignedMemClearMMX, .-alignedMemClearMMX
-	.ident	"GCC: (GNU) 4.4.0 20090304 (experimental)"
+	.ident	"GCC: (Ubuntu 4.4.0-0ubuntu2) 4.4.0"
 	.section	.note.GNU-stack,"",@progbits

src/core/basic_ops_x86_sse.s

@@ -230,48 +230,42 @@ alignedBufWetDryMixSplittedSSE:
 	pushl	%edi
 	pushl	%esi
 	pushl	%ebx
-	subl	$140, %esp
-	movl	180(%esp), %eax
-	flds	172(%esp)
-	movl	160(%esp), %edx
-	movl	164(%esp), %esi
+	subl	$124, %esp
+	movl	164(%esp), %eax
+	movl	144(%esp), %edx
+	movl	148(%esp), %esi
+	movl	152(%esp), %ecx
 	testl	%eax, %eax
-	movl	168(%esp), %ecx
-	flds	176(%esp)
-	jle	.L43
-	movl	180(%esp), %eax
+	jle	.L39
+	movl	164(%esp), %eax
 	subl	$1, %eax
 	shrl	%eax
 	addl	$1, %eax
 	movl	%eax, %ebp
-	movl	%eax, 120(%esp)
+	movl	%eax, 104(%esp)
 	shrl	$2, %ebp
-	cmpl	$3, 120(%esp)
+	cmpl	$3, 104(%esp)
 	leal	0(,%ebp,4), %eax
-	movl	%eax, 124(%esp)
+	movl	%eax, 108(%esp)
 	jbe	.L40
 	testl	%eax, %eax
 	jne	.L34
 .L40:
-	fxch	%st(1)
 	xorl	%edi, %edi
 	jmp	.L36
 	.p2align 4,,7
 	.p2align 3
 .L34:
-	fsts	12(%esp)
-	fxch	%st(1)
+	movss	160(%esp), %xmm0
 	xorps	%xmm7, %xmm7
-	movss	12(%esp), %xmm0
 	movl	%esi, %ebx
-	fsts	12(%esp)
 	xorl	%eax, %eax
 	xorl	%edi, %edi
 	shufps	$0, %xmm0, %xmm0
-	movaps	%xmm0, 32(%esp)
-	movss	12(%esp), %xmm0
-	shufps	$0, %xmm0, %xmm0
 	movaps	%xmm0, 16(%esp)
+	movss	156(%esp), %xmm0
+	shufps	$0, %xmm0, %xmm0
+	movaps	%xmm0, (%esp)
 	.p2align 4,,7
 	.p2align 3
 .L37:
@@ -282,28 +276,28 @@ alignedBufWetDryMixSplittedSSE:
 	shufps	$136, %xmm6, %xmm0
 	movaps	32(%edx,%eax,2), %xmm4
 	shufps	$221, %xmm6, %xmm5
-	movaps	%xmm0, 96(%esp)
+	movaps	%xmm0, 80(%esp)
 	movaps	48(%edx,%eax,2), %xmm3
 	movaps	%xmm4, %xmm0
 	shufps	$136, %xmm3, %xmm0
-	movaps	96(%esp), %xmm2
+	movaps	80(%esp), %xmm2
 	shufps	$221, %xmm3, %xmm4
 	movaps	%xmm7, %xmm6
 	movlps	(%ebx), %xmm6
-	movaps	%xmm5, 80(%esp)
+	movaps	%xmm5, 64(%esp)
 	movhps	8(%ebx), %xmm6
 	shufps	$136, %xmm0, %xmm2
-	movaps	%xmm0, 64(%esp)
+	movaps	%xmm0, 48(%esp)
 	movaps	%xmm7, %xmm5
 	movaps	%xmm6, %xmm0
 	movlps	16(%ebx), %xmm5
 	movhps	24(%ebx), %xmm5
 	shufps	$136, %xmm5, %xmm0
-	mulps	32(%esp), %xmm2
+	mulps	16(%esp), %xmm2
 	shufps	$221, %xmm5, %xmm6
-	movaps	%xmm4, 48(%esp)
+	movaps	%xmm4, 32(%esp)
 	addl	$32, %ebx
-	mulps	16(%esp), %xmm0
+	mulps	(%esp), %xmm0
 	movaps	%xmm7, %xmm4
 	movlps	(%eax,%ecx), %xmm4
 	movaps	%xmm7, %xmm3
@@ -313,21 +307,21 @@ alignedBufWetDryMixSplittedSSE:
 	movhps	24(%ecx,%eax), %xmm3
 	shufps	$136, %xmm3, %xmm1
 	addps	%xmm0, %xmm2
-	movaps	80(%esp), %xmm0
+	movaps	64(%esp), %xmm0
 	shufps	$221, %xmm3, %xmm4
-	shufps	$136, 48(%esp), %xmm0
-	mulps	16(%esp), %xmm1
+	shufps	$136, 32(%esp), %xmm0
+	mulps	(%esp), %xmm1
 	movaps	%xmm2, %xmm3
-	movaps	80(%esp), %xmm5
-	mulps	32(%esp), %xmm0
-	shufps	$221, 48(%esp), %xmm5
-	mulps	16(%esp), %xmm6
+	movaps	64(%esp), %xmm5
+	mulps	16(%esp), %xmm0
+	shufps	$221, 32(%esp), %xmm5
+	mulps	(%esp), %xmm6
 	addps	%xmm1, %xmm0
-	movaps	96(%esp), %xmm1
-	shufps	$221, 64(%esp), %xmm1
-	mulps	16(%esp), %xmm4
-	mulps	32(%esp), %xmm1
-	mulps	32(%esp), %xmm5
+	movaps	80(%esp), %xmm1
+	shufps	$221, 48(%esp), %xmm1
+	mulps	(%esp), %xmm4
+	mulps	16(%esp), %xmm1
+	mulps	16(%esp), %xmm5
 	addps	%xmm6, %xmm1
 	addps	%xmm4, %xmm5
 	movaps	%xmm0, %xmm4
@@ -348,63 +342,53 @@ alignedBufWetDryMixSplittedSSE:
 	addl	$32, %eax
 	cmpl	%edi, %ebp
 	ja	.L37
-	movl	124(%esp), %edi
-	movl	120(%esp), %eax
+	movl	108(%esp), %edi
+	movl	104(%esp), %eax
 	addl	%edi, %edi
-	cmpl	%eax, 124(%esp)
-	je	.L44
+	cmpl	%eax, 108(%esp)
+	je	.L39
 .L36:
-	leal	(%edx,%edi,8), %ebx
+	movss	156(%esp), %xmm0
 	xorl	%ebp, %ebp
-	leal	8(%edx,%edi,8), %edx
+	movss	160(%esp), %xmm1
 	movl	%edi, %eax
+	leal	(%edx,%edi,8), %ebx
+	leal	8(%edx,%edi,8), %edx
 	.p2align 4,,7
 	.p2align 3
 .L38:
-	flds	(%ebx)
+	movss	(%esi,%eax,4), %xmm3
 	addl	$2, %ebp
-	fmul	%st(2), %st
-	flds	(%esi,%eax,4)
-	fmul	%st(2), %st
-	faddp	%st, %st(1)
-	fstps	(%ebx)
-	flds	4(%ebx)
-	fmul	%st(2), %st
-	flds	(%ecx,%eax,4)
-	fmul	%st(2), %st
-	faddp	%st, %st(1)
-	fstps	4(%ebx)
+	movss	(%ebx), %xmm2
+	mulss	%xmm0, %xmm3
+	mulss	%xmm1, %xmm2
+	addss	%xmm3, %xmm2
+	movss	%xmm2, (%ebx)
+	movss	4(%ebx), %xmm2
+	movss	(%ecx,%eax,4), %xmm3
+	mulss	%xmm1, %xmm2
+	mulss	%xmm0, %xmm3
+	addss	%xmm3, %xmm2
+	movss	%xmm2, 4(%ebx)
 	addl	$16, %ebx
-	flds	(%edx)
-	fmul	%st(2), %st
-	flds	4(%esi,%eax,4)
-	fmul	%st(2), %st
-	faddp	%st, %st(1)
-	fstps	(%edx)
-	flds	4(%edx)
-	fmul	%st(2), %st
-	flds	4(%ecx,%eax,4)
+	movss	4(%esi,%eax,4), %xmm3
+	movss	(%edx), %xmm2
+	mulss	%xmm0, %xmm3
+	mulss	%xmm1, %xmm2
+	addss	%xmm3, %xmm2
+	movss	%xmm2, (%edx)
+	movss	4(%edx), %xmm2
+	movss	4(%ecx,%eax,4), %xmm3
+	mulss	%xmm1, %xmm2
 	leal	(%edi,%ebp), %eax
-	fmul	%st(2), %st
-	faddp	%st, %st(1)
-	fstps	4(%edx)
+	mulss	%xmm0, %xmm3
+	addss	%xmm3, %xmm2
+	movss	%xmm2, 4(%edx)
 	addl	$16, %edx
-	cmpl	%eax, 180(%esp)
+	cmpl	%eax, 164(%esp)
 	jg	.L38
-	fstp	%st(0)
-	fstp	%st(0)
-	jmp	.L39
-.L43:
-	fstp	%st(0)
-	fstp	%st(0)
-	jmp	.L39
-.L44:
-	fstp	%st(0)
-	fstp	%st(0)
-	.p2align 4,,7
-	.p2align 3
 .L39:
-	addl	$140, %esp
+	addl	$124, %esp
 	popl	%ebx
 	popl	%esi
 	popl	%edi
@@ -417,39 +401,34 @@ alignedBufWetDryMixSplittedSSE:
 unalignedBufMixLRCoeffSSE:
 	pushl	%esi
 	pushl	%ebx
-	subl	$4, %esp
-	movl	32(%esp), %esi
-	flds	24(%esp)
-	movl	16(%esp), %eax
-	movl	20(%esp), %edx
-	movl	%esi, %ebx
-	flds	28(%esp)
-	shrl	$31, %ebx
-	leal	(%esi,%ebx), %ecx
+	movl	28(%esp), %ebx
+	movl	12(%esp), %eax
+	movl	16(%esp), %edx
+	movss	20(%esp), %xmm1
+	movl	%ebx, %esi
+	shrl	$31, %esi
+	leal	(%ebx,%esi), %ecx
 	andl	$1, %ecx
-	cmpl	%ebx, %ecx
-	jne	.L54
-.L46:
-	testl	%esi, %esi
-	jle	.L55
-	leal	-1(%esi), %ebx
-	shrl	%ebx
+	cmpl	%esi, %ecx
+	movss	24(%esp), %xmm3
+	jne	.L52
+.L44:
+	testl	%ebx, %ebx
+	jle	.L49
 	testb	$15, %al
-	jne	.L48
-	fxch	%st(1)
-	fstps	(%esp)
+	jne	.L46
+	movaps	%xmm1, %xmm0
+	subl	$1, %ebx
+	unpcklps	%xmm3, %xmm0
+	shrl	%ebx
 	xorps	%xmm2, %xmm2
-	movss	(%esp), %xmm0
-	addl	$1, %ebx
-	fstps	(%esp)
-	xorl	%ecx, %ecx
-	movss	(%esp), %xmm1
-	unpcklps	%xmm1, %xmm0
 	movaps	%xmm0, %xmm3
+	addl	$1, %ebx
 	movlhps	%xmm0, %xmm3
+	xorl	%ecx, %ecx
 	.p2align 4,,7
 	.p2align 3
-.L49:
+.L47:
 	movaps	%xmm2, %xmm1
 	addl	$1, %ecx
 	movlps	(%edx), %xmm1
@@ -463,65 +442,53 @@ unalignedBufMixLRCoeffSSE:
 	movaps	%xmm0, (%eax)
 	addl	$16, %eax
 	cmpl	%ebx, %ecx
-	jb	.L49
-	jmp	.L51
-	.p2align 4,,7
-	.p2align 3
-.L55:
-	fstp	%st(0)
-	fstp	%st(0)
-	.p2align 4,,7
-	.p2align 3
-.L51:
-	addl	$4, %esp
+	jb	.L47
+.L49:
 	popl	%ebx
 	popl	%esi
 	ret
 	.p2align 4,,7
 	.p2align 3
-.L48:
+.L46:
 	xorl	%ecx, %ecx
 	.p2align 4,,7
 	.p2align 3
-.L50:
-	flds	(%edx,%ecx,8)
-	fmul	%st(2), %st
-	fadds	(%eax,%ecx,8)
-	fstps	(%eax,%ecx,8)
-	flds	4(%edx,%ecx,8)
-	fmul	%st(1), %st
-	fadds	4(%eax,%ecx,8)
-	fstps	4(%eax,%ecx,8)
-	flds	8(%edx,%ecx,8)
-	fmul	%st(2), %st
-	fadds	8(%eax,%ecx,8)
-	fstps	8(%eax,%ecx,8)
-	flds	12(%edx,%ecx,8)
-	fmul	%st(1), %st
-	fadds	12(%eax,%ecx,8)
-	fstps	12(%eax,%ecx,8)
+.L48:
+	movss	(%edx,%ecx,8), %xmm0
+	mulss	%xmm1, %xmm0
+	addss	(%eax,%ecx,8), %xmm0
+	movss	%xmm0, (%eax,%ecx,8)
+	movss	4(%edx,%ecx,8), %xmm0
+	mulss	%xmm3, %xmm0
+	addss	4(%eax,%ecx,8), %xmm0
+	movss	%xmm0, 4(%eax,%ecx,8)
+	movss	8(%edx,%ecx,8), %xmm0
+	mulss	%xmm1, %xmm0
+	addss	8(%eax,%ecx,8), %xmm0
+	movss	%xmm0, 8(%eax,%ecx,8)
+	movss	12(%edx,%ecx,8), %xmm0
+	mulss	%xmm3, %xmm0
+	addss	12(%eax,%ecx,8), %xmm0
+	movss	%xmm0, 12(%eax,%ecx,8)
 	addl	$2, %ecx
-	cmpl	%ecx, %esi
-	jg	.L50
-	fstp	%st(0)
-	fstp	%st(0)
-	addl	$4, %esp
+	cmpl	%ecx, %ebx
+	jg	.L48
 	popl	%ebx
 	popl	%esi
 	ret
-.L54:
-	flds	(%edx)
-	subl	$1, %esi
-	fmul	%st(2), %st
-	fadds	(%eax)
-	fstps	(%eax)
-	flds	4(%edx)
+.L52:
+	movss	(%edx), %xmm0
+	subl	$1, %ebx
+	mulss	%xmm1, %xmm0
+	addss	(%eax), %xmm0
+	movss	%xmm0, (%eax)
+	movss	4(%edx), %xmm0
 	addl	$8, %edx
-	fmul	%st(1), %st
-	fadds	4(%eax)
-	fstps	4(%eax)
+	mulss	%xmm3, %xmm0
+	addss	4(%eax), %xmm0
+	movss	%xmm0, 4(%eax)
 	addl	$8, %eax
-	jmp	.L46
+	jmp	.L44
 	.size	unalignedBufMixLRCoeffSSE, .-unalignedBufMixLRCoeffSSE
-	.ident	"GCC: (GNU) 4.4.0 20090304 (experimental)"
+	.ident	"GCC: (Ubuntu 4.4.0-0ubuntu2) 4.4.0"
 	.section	.note.GNU-stack,"",@progbits

src/core/basic_ops_x86_sse2.s

@@ -70,19 +70,19 @@ alignedConvertToS16SSE2:
 	pushl	%ebx
 	subl	$8, %esp
 	movl	36(%esp), %eax
+	movss	.LC0, %xmm6
 	cmpb	$0, 44(%esp)
-	flds	.LC0
 	movl	28(%esp), %edx
 	movl	32(%esp), %ebx
 	movl	%eax, %esi
-	fmuls	40(%esp)
+	mulss	40(%esp), %xmm6
 	jne	.L13
 	testw	%ax, %ax
-	jle	.L35
+	jle	.L15
 	movl	%eax, %edi
 	shrw	$2, %di
 	cmpw	$3, %ax
-	movw	%ax, 4(%esp)
+	movw	%ax, 2(%esp)
 	leal	0(,%edi,4), %ebp
 	ja	.L33
 .L28:
@@ -98,19 +98,17 @@ alignedConvertToS16SSE2:
 	.p2align 4,,7
 	.p2align 3
 .L25:
-	flds	(%edx)
-	fmul	%st(1), %st
-	fstps	4(%esp)
-	cvttss2si	4(%esp), %ecx
-	flds	4(%edx)
-	fmul	%st(1), %st
+	movss	(%edx), %xmm0
+	mulss	%xmm6, %xmm0
+	cvttss2si	%xmm0, %ecx
+	movss	4(%edx), %xmm0
 	cmpl	$-32768, %ecx
+	mulss	%xmm6, %xmm0
 	cmovl	%edi, %ecx
 	cmpl	$32767, %ecx
 	cmovg	%ebx, %ecx
-	fstps	4(%esp)
 	movw	%cx, (%eax)
-	cvttss2si	4(%esp), %ecx
+	cvttss2si	%xmm0, %ecx
 	cmpl	$-32768, %ecx
 	cmovl	%edi, %ecx
 	cmpl	$32767, %ecx
@@ -121,15 +119,6 @@ alignedConvertToS16SSE2:
 	addl	$4, %eax
 	cmpw	%bp, %si
 	jg	.L25
-	fstp	%st(0)
-	jmp	.L15
-.L35:
-	fstp	%st(0)
-	jmp	.L15
-.L36:
-	fstp	%st(0)
-	.p2align 4,,7
-	.p2align 3
 .L15:
 	movswl	%si,%esi
 	addl	$8, %esp
@@ -143,11 +132,11 @@ alignedConvertToS16SSE2:
 	.p2align 3
 .L13:
 	testw	%ax, %ax
-	jle	.L36
+	jle	.L15
 	movl	%eax, %ebp
 	shrw	$2, %bp
 	cmpw	$3, %si
-	movw	%ax, 4(%esp)
+	movw	%ax, 2(%esp)
 	leal	0(,%ebp,4), %eax
 	ja	.L34
 .L27:
@@ -162,24 +151,22 @@ alignedConvertToS16SSE2:
 	.p2align 4,,7
 	.p2align 3
 .L20:
-	flds	(%ecx)
+	movss	(%ecx), %xmm0
 	movl	$32767, %ebp
-	fmul	%st(1), %st
-	fstps	4(%esp)
-	cvttss2si	4(%esp), %ebx
-	flds	4(%ecx)
-	fmul	%st(1), %st
+	mulss	%xmm6, %xmm0
+	cvttss2si	%xmm0, %ebx
+	movss	4(%ecx), %xmm0
 	cmpl	$-32768, %ebx
 	cmovl	%edi, %ebx
 	cmpl	$32767, %ebx
+	mulss	%xmm6, %xmm0
 	cmovg	%ebp, %ebx
 	movzbl	%bh, %ebp
 	sall	$8, %ebx
 	orl	%ebp, %ebx
 	movl	$32767, %ebp
-	fstps	4(%esp)
 	movw	%bx, (%edx)
-	cvttss2si	4(%esp), %ebx
+	cvttss2si	%xmm0, %ebx
 	cmpl	$-32768, %ebx
 	cmovl	%edi, %ebx
 	cmpl	$32767, %ebx
@@ -193,107 +180,105 @@ alignedConvertToS16SSE2:
 	addl	$4, %edx
 	cmpw	%ax, %si
 	jg	.L20
-	fstp	%st(0)
 	jmp	.L15
 	.p2align 4,,7
 	.p2align 3
 .L34:
 	testw	%ax, %ax
 	je	.L27
-	fsts	(%esp)
+	movaps	%xmm6, %xmm0
 	xorl	%ecx, %ecx
 	movdqa	.LC1, %xmm3
-	movss	(%esp), %xmm0
-	xorl	%edi, %edi
-	movdqa	.LC2, %xmm2
 	shufps	$0, %xmm0, %xmm0
+	movdqa	.LC2, %xmm2
+	movss	%xmm6, 4(%esp)
+	xorl	%edi, %edi
 	movaps	%xmm0, %xmm7
 	.p2align 4,,7
 	.p2align 3
 .L19:
 	movaps	(%edx,%ecx,2), %xmm0
-	movdqa	%xmm2, %xmm4
+	movdqa	%xmm2, %xmm5
 	movdqa	%xmm2, %xmm6
 	addl	$1, %edi
-	movaps	16(%edx,%ecx,2), %xmm5
+	movaps	16(%edx,%ecx,2), %xmm4
 	mulps	%xmm7, %xmm0
-	mulps	%xmm7, %xmm5
+	mulps	%xmm7, %xmm4
 	cvttps2dq	%xmm0, %xmm0
 	movdqa	%xmm0, %xmm1
 	pcmpgtd	%xmm3, %xmm1
 	pand	%xmm1, %xmm0
 	pandn	%xmm3, %xmm1
 	por	%xmm0, %xmm1
-	cvttps2dq	%xmm5, %xmm5
+	cvttps2dq	%xmm4, %xmm4
 	movdqa	%xmm1, %xmm0
 	pcmpgtd	%xmm2, %xmm0
-	pand	%xmm0, %xmm4
+	pand	%xmm0, %xmm5
 	pandn	%xmm1, %xmm0
 	movdqa	%xmm0, %xmm1
-	movdqa	%xmm5, %xmm0
-	por	%xmm4, %xmm1
+	movdqa	%xmm4, %xmm0
+	por	%xmm5, %xmm1
 	pcmpgtd	%xmm3, %xmm0
-	movdqa	.LC3, %xmm4
-	pand	%xmm0, %xmm5
-	pand	%xmm1, %xmm4
-	pandn	%xmm3, %xmm0
-	psrad	$8, %xmm4
-	por	%xmm5, %xmm0
-	pslld	$8, %xmm1
-	movdqa	%xmm0, %xmm5
-	pcmpgtd	%xmm2, %xmm5
-	pand	%xmm5, %xmm6
-	pandn	%xmm0, %xmm5
-	movdqa	%xmm5, %xmm0
 	movdqa	.LC3, %xmm5
-	por	%xmm6, %xmm0
-	pand	%xmm0, %xmm5
-	pslld	$8, %xmm0
+	pand	%xmm0, %xmm4
+	pand	%xmm1, %xmm5
+	pandn	%xmm3, %xmm0
 	psrad	$8, %xmm5
-	movdqa	%xmm4, %xmm6
-	punpcklwd	%xmm5, %xmm4
-	punpckhwd	%xmm5, %xmm6
-	movdqa	%xmm4, %xmm5
-	punpcklwd	%xmm6, %xmm4
-	punpckhwd	%xmm6, %xmm5
-	punpcklwd	%xmm5, %xmm4
-	movdqa	%xmm1, %xmm5
+	por	%xmm4, %xmm0
+	pslld	$8, %xmm1
+	movdqa	%xmm0, %xmm4
+	pcmpgtd	%xmm2, %xmm4
+	pand	%xmm4, %xmm6
+	pandn	%xmm0, %xmm4
+	movdqa	%xmm4, %xmm0
+	movdqa	.LC3, %xmm4
+	por	%xmm6, %xmm0
+	pand	%xmm0, %xmm4
+	pslld	$8, %xmm0
+	psrad	$8, %xmm4
+	movdqa	%xmm5, %xmm6
+	punpcklwd	%xmm4, %xmm5
+	punpckhwd	%xmm4, %xmm6
+	movdqa	%xmm5, %xmm4
+	punpcklwd	%xmm6, %xmm5
+	punpckhwd	%xmm6, %xmm4
+	punpcklwd	%xmm4, %xmm5
+	movdqa	%xmm1, %xmm4
 	punpcklwd	%xmm0, %xmm1
-	punpckhwd	%xmm0, %xmm5
-	movdqa	%xmm1, %xmm0
-	punpcklwd	%xmm5, %xmm1
-	punpckhwd	%xmm5, %xmm0
-	punpcklwd	%xmm0, %xmm1
-	por	%xmm1, %xmm4
-	movdqa	%xmm4, (%ebx,%ecx)
+	punpckhwd	%xmm0, %xmm4
+	movdqa	%xmm1, %xmm6
+	punpcklwd	%xmm4, %xmm1
+	punpckhwd	%xmm4, %xmm6
+	punpcklwd	%xmm6, %xmm1
+	por	%xmm1, %xmm5
+	movdqa	%xmm5, (%ebx,%ecx)
 	addl	$16, %ecx
 	cmpw	%di, %bp
 	ja	.L19
-	cmpw	4(%esp), %ax
+	cmpw	2(%esp), %ax
+	movss	4(%esp), %xmm6
 	jne	.L18
-	fstp	%st(0)
 	jmp	.L15
 	.p2align 4,,7
 	.p2align 3
 .L33:
 	testw	%bp, %bp
-	.p2align 4,,4
+	.p2align 4,,3
 	.p2align 3
 	je	.L28
-	fsts	(%esp)
+	movaps	%xmm6, %xmm0
 	xorl	%eax, %eax
 	movdqa	.LC1, %xmm3
-	movss	(%esp), %xmm0
-	xorl	%ecx, %ecx
-	movdqa	.LC2, %xmm2
 	shufps	$0, %xmm0, %xmm0
+	movdqa	.LC2, %xmm2
+	xorl	%ecx, %ecx
 	movaps	%xmm0, %xmm5
 	.p2align 4,,7
 	.p2align 3
 .L24:
 	movaps	(%edx,%eax,2), %xmm0
 	addl	$1, %ecx
-	movdqa	%xmm2, %xmm6
+	movdqa	%xmm2, %xmm7
 	movaps	16(%edx,%eax,2), %xmm4
 	mulps	%xmm5, %xmm0
 	mulps	%xmm5, %xmm4
@@ -306,23 +291,23 @@ alignedConvertToS16SSE2:
 	cvttps2dq	%xmm4, %xmm4
 	movdqa	%xmm1, %xmm0
 	pcmpgtd	%xmm2, %xmm0
-	pand	%xmm0, %xmm6
+	pand	%xmm0, %xmm7
 	pandn	%xmm1, %xmm0
 	movdqa	%xmm0, %xmm1
 	movdqa	%xmm4, %xmm0
-	por	%xmm6, %xmm1
+	por	%xmm7, %xmm1
 	pcmpgtd	%xmm3, %xmm0
-	movdqa	%xmm2, %xmm6
+	movdqa	%xmm2, %xmm7
 	pand	%xmm0, %xmm4
 	pandn	%xmm3, %xmm0
 	por	%xmm4, %xmm0
 	movdqa	%xmm0, %xmm4
 	pcmpgtd	%xmm2, %xmm4
-	pand	%xmm4, %xmm6
+	pand	%xmm4, %xmm7
 	pandn	%xmm0, %xmm4
 	movdqa	%xmm4, %xmm0
 	movdqa	%xmm1, %xmm4
-	por	%xmm6, %xmm0
+	por	%xmm7, %xmm0
 	punpckhwd	%xmm0, %xmm4
 	punpcklwd	%xmm0, %xmm1
 	movdqa	%xmm1, %xmm0
@@ -333,9 +318,8 @@ alignedConvertToS16SSE2:
 	addl	$16, %eax
 	cmpw	%cx, %di
 	ja	.L24
-	cmpw	%bp, 4(%esp)
+	cmpw	%bp, 2(%esp)
 	jne	.L23
-	fstp	%st(0)
 	jmp	.L15
 	.size	alignedConvertToS16SSE2, .-alignedConvertToS16SSE2
 	.section	.rodata.cst4,"aM",@progbits,4
@@ -361,5 +345,5 @@ alignedConvertToS16SSE2:
 	.long	65280
 	.long	65280
 	.long	65280
-	.ident	"GCC: (GNU) 4.4.0 20090304 (experimental)"
+	.ident	"GCC: (Ubuntu 4.4.0-0ubuntu2) 4.4.0"
 	.section	.note.GNU-stack,"",@progbits