trsm_kernel_lt_8x2_sse.s

Optimized GotoBLAS libraries

	movl	BORIG, B
	sall	$1 + BASE_SHIFT, %eax
	addl	%eax, AORIG
#endif
	ALIGN_2	

.L70:
	testl	$1, M
	jle	.L99

#ifdef LN
       movl	K, %eax
       sall	$BASE_SHIFT, %eax
       subl	%eax, AORIG
#endif

#if defined(LN) || defined(RT)
	movl	KK, %eax
	movl	AORIG, AA
	sall	$BASE_SHIFT, %eax
	addl	%eax, AA
#endif

	leal	BUFFER, BB

#if defined(LN) || defined(RT)
	movl	KK, %eax
	sall	$1 + BASE_SHIFT, %eax
	leal	(BB, %eax, 4), BB
#endif	

	movss	 0 * SIZE(BB), %xmm2
	xorps	%xmm4, %xmm4
	movss	 0 * SIZE(AA), %xmm0
	xorps	%xmm5, %xmm5
	movss	 16 * SIZE(BB), %xmm3
	xorps	%xmm6, %xmm6
	movss	 4 * SIZE(AA), %xmm1
	xorps	%xmm7, %xmm7

#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K, %eax
	subl	KK, %eax
#endif
	sarl	$3, %eax
	je	.L72
	ALIGN_2

.L71:
	mulss	%xmm0, %xmm2
	mulss	 4 * SIZE(BB), %xmm0
	addss	%xmm2, %xmm4
	movss	 8 * SIZE(BB), %xmm2
	addss	%xmm0, %xmm5
	movss	 1 * SIZE(AA), %xmm0
	mulss	%xmm0, %xmm2
	mulss	12 * SIZE(BB), %xmm0
	addss	%xmm2, %xmm6
	movss	32 * SIZE(BB), %xmm2
	addss	%xmm0, %xmm7
	movss	 2 * SIZE(AA), %xmm0
	mulss	%xmm0, %xmm3
	mulss	20 * SIZE(BB), %xmm0
	addss	%xmm3, %xmm4
	movss	24 * SIZE(BB), %xmm3
	addss	%xmm0, %xmm5
	movss	 3 * SIZE(AA), %xmm0
	mulss	%xmm0, %xmm3
	mulss	28 * SIZE(BB), %xmm0
	addss	%xmm3, %xmm6
	movss	48 * SIZE(BB), %xmm3
	addss	%xmm0, %xmm7
	movss	 8 * SIZE(AA), %xmm0
	mulss	%xmm1, %xmm2
	mulss	36 * SIZE(BB), %xmm1
	addss	%xmm2, %xmm4
	movss	40 * SIZE(BB), %xmm2
	addss	%xmm1, %xmm5
	movss	 5 * SIZE(AA), %xmm1
	mulss	%xmm1, %xmm2
	mulss	44 * SIZE(BB), %xmm1
	addss	%xmm2, %xmm6
	movss	64 * SIZE(BB), %xmm2
	addss	%xmm1, %xmm7
	movss	 6 * SIZE(AA), %xmm1
	mulss	%xmm1, %xmm3
	mulss	52 * SIZE(BB), %xmm1
	addss	%xmm3, %xmm4
	movss	56 * SIZE(BB), %xmm3
	addss	%xmm1, %xmm5
	movss	 7 * SIZE(AA), %xmm1
	mulss	%xmm1, %xmm3
	mulss	60 * SIZE(BB), %xmm1
	addss	%xmm3, %xmm6
	movss	80 * SIZE(BB), %xmm3
	addss	%xmm1, %xmm7
	movss	12 * SIZE(AA), %xmm1

	addl   $ 8 * SIZE, AA
	addl   $64 * SIZE, BB
	decl   %eax
	jne    .L71
	ALIGN_2

.L72:
#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K, %eax
	subl	KK, %eax
#endif
	andl	$7, %eax		# if (k & 1)
	BRANCH
	je .L74

.L73:
	mulss	%xmm0, %xmm2
	mulss	 4 * SIZE(BB), %xmm0
	addss	%xmm2, %xmm4
	movss	 8 * SIZE(BB), %xmm2
	addss	%xmm0, %xmm5
	movss	 1 * SIZE(AA), %xmm0

	addl	$1 * SIZE, AA
	addl	$8 * SIZE, BB
	decl	%eax
	jg	.L73
	ALIGN_4

.L74:
	addss	%xmm6, %xmm4
	addss	%xmm7, %xmm5

#if defined(LN) || defined(RT)
	movl	KK, %eax
#ifdef LN
	subl	$1, %eax
#else
	subl	$2, %eax
#endif

	movl	AORIG, AA
	movl	BORIG, B
	leal	BUFFER, BB

	sall	$BASE_SHIFT, %eax
	leal	(AA, %eax, 1), AA
	leal	(B,  %eax, 2), B
	leal	(BB, %eax, 8), BB
#endif

#if defined(LN) || defined(LT)
	unpcklps %xmm5, %xmm4

	movsd	 0 * SIZE(B), %xmm2

	subps	%xmm4,  %xmm2
#else
	movss	 0 * SIZE(AA), %xmm0
	movss	 1 * SIZE(AA), %xmm2

	subss	%xmm4, %xmm0
	subss	%xmm5, %xmm2
#endif

#if defined(LN) || defined(LT)
	movaps	MASK, %xmm6
#endif

#if defined(LN) || defined(LT)
	movss	 0 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm2
#endif

#ifdef RN
	movss	 0 * SIZE(B), %xmm6
	mulss	%xmm6, %xmm0

	movss	 1 * SIZE(B), %xmm6
	movaps	%xmm6, %xmm5

	mulss	%xmm0, %xmm5
	subss	%xmm5, %xmm2

	movss	 3 * SIZE(B), %xmm6
	mulss	%xmm6, %xmm2
#endif

#ifdef RT
	movss	 3 * SIZE(B), %xmm6
	mulss	%xmm6, %xmm2

	movss	 2 * SIZE(B), %xmm6
	movaps	%xmm6, %xmm5

	mulss	%xmm2, %xmm5
	subss	%xmm5, %xmm0

	movss	 0 * SIZE(B), %xmm6
	mulss	%xmm6, %xmm0
#endif

#if defined(LN) || defined(LT)
	movlps	%xmm2,   0 * SIZE(B)

	pshufd	$0x00, %xmm2, %xmm0
	pshufd	$0x55, %xmm2, %xmm1
	movaps	%xmm0,   0 * SIZE(BB)
	movaps	%xmm1,   4 * SIZE(BB)
#else
	movss	%xmm0,   0 * SIZE(AA)
	movss	%xmm2,   1 * SIZE(AA)
#endif

#ifdef LN
	subl	$1 * SIZE, CO1
#endif

#if defined(LN) || defined(LT)
	movaps	 %xmm2, %xmm0
	shufps	 $0x88, %xmm3, %xmm2
	shufps	 $0xdd, %xmm3, %xmm0

	movss	%xmm2, 0 * SIZE(CO1)
	movss	%xmm0, 0 * SIZE(CO1, LDC)
#else
	movss	%xmm0, 0 * SIZE(CO1)
	movss	%xmm2, 0 * SIZE(CO1, LDC)
#endif

#ifndef LN
	addl	$1 * SIZE, CO1
#endif

#if defined(LT) || defined(RN)
	movl	K,  %eax
	subl	KK, %eax
	leal	(,%eax, SIZE), %eax
	leal	(AA, %eax, 1), AA
#ifdef LT
	addl	$2 * SIZE, B
#endif
#endif

#ifdef LN
	subl	$1, KK
	movl	BORIG, B
#endif

#ifdef LT
	addl	$1, KK
#endif

#ifdef RT
	movl	K, %eax
	movl	BORIG, B
	sall	$BASE_SHIFT, %eax
	addl	%eax, AORIG
#endif
	ALIGN_2	

.L99:
#ifdef LN
       movl	K, %eax
       leal	(, %eax, SIZE), %eax
       leal 	(B, %eax, 2), B
#endif

#if defined(LT) || defined(RN)
	movl	K,  %eax
	subl	KK, %eax
	leal	(,%eax, SIZE), %eax
	leal	(B,  %eax, 2), B
#endif

#ifdef RN
	addl	$2, KK
#endif

#ifdef RT
	subl	$2, KK
#endif

	decl	J			# j --
	jg	.L01
	ALIGN_2

.L100:
	testl	$1, N
	jle	.L999
	ALIGN_2
	
#ifdef LN
	movl	OFFSET, %eax
	addl	M, %eax
	movl	%eax, KK
#endif	

	leal	BUFFER, BB

#ifdef RT
       movl	K, %eax
       sall	$BASE_SHIFT, %eax
       subl	%eax, B
#endif

#if defined(LN) || defined(RT)
	movl	KK, %eax
	movl	B, BORIG
        sall	$BASE_SHIFT, %eax
	leal	(B,  %eax, 1), B
	leal	(BB, %eax, 4), BB
#endif	

#ifdef LT
	movl	OFFSET, %eax
	movl	%eax, KK
#endif

#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K,  %eax
	subl	KK, %eax
#endif
	sarl	$3, %eax
	jle	.L103
	ALIGN_4
	
.L102:
	movaps	 0 * SIZE(B), %xmm3
	movaps	 4 * SIZE(B), %xmm7

	pshufd	 $0x00, %xmm3, %xmm0
	pshufd	 $0x55, %xmm3, %xmm1
	pshufd	 $0xaa, %xmm3, %xmm2
	pshufd	 $0xff, %xmm3, %xmm3

	pshufd	 $0x00, %xmm7, %xmm4
	pshufd	 $0x55, %xmm7, %xmm5
	pshufd	 $0xaa, %xmm7, %xmm6
	pshufd	 $0xff, %xmm7, %xmm7

	movaps	%xmm0,  0 * SIZE(BB)
	movaps	%xmm1,  4 * SIZE(BB)
	movaps	%xmm2,  8 * SIZE(BB)
	movaps	%xmm3, 12 * SIZE(BB)
	movaps	%xmm4, 16 * SIZE(BB)
	movaps	%xmm5, 20 * SIZE(BB)
	movaps	%xmm6, 24 * SIZE(BB)
	movaps	%xmm7, 28 * SIZE(BB)

	addl	$ 8 * SIZE, B
	addl	$32 * SIZE, BB

	decl	%eax
	BRANCH
	jne	.L102
	ALIGN_2

.L103:
#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K,  %eax
	subl	KK, %eax
#endif
	andl	$7, %eax
	BRANCH
	jle	.L105
	ALIGN_2

.L104:
	movss	 0 * SIZE(B), %xmm3

	pshufd	 $0x00, %xmm3, %xmm0

	movaps	%xmm0,  0 * SIZE(BB)

	addl	$1 * SIZE, B
	addl	$4 * SIZE, BB

	decl	%eax
	jne	.L104
	ALIGN_4

.L105:
#if defined(LT) || defined(RN)
	movl	A, AA
#else
	movl	A, %eax
	movl	%eax, AORIG
#endif

#ifdef RT
	subl	LDC, C
#endif
	movl	C, CO1
#ifndef RT
	addl	LDC, C
#endif

	movl	M,  %ebx
	sarl	$3, %ebx	# i = (m >> 2)
	jle	.L130
	ALIGN_4

.L110:
#ifdef LN
       movl	K, %eax
       sall	$3 + BASE_SHIFT, %eax
       subl	%eax, AORIG
#endif

#if defined(LN) || defined(RT)
	movl	KK, %eax
	movl	AORIG, AA
	sall	$3 + BASE_SHIFT, %eax
	addl	%eax, AA
#endif

	leal	BUFFER, BB

#if defined(LN) || defined(RT)
	movl	KK, %eax
	sall	$BASE_SHIFT, %eax
	leal	(BB, %eax, 4), BB
#endif	

	movaps	 0 * SIZE(BB), %xmm2
	xorps	%xmm4, %xmm4
	movaps	 0 * SIZE(AA), %xmm0
	xorps	%xmm5, %xmm5
	movaps	 16 * SIZE(BB), %xmm3
	xorps	%xmm6, %xmm6
	movaps	 16 * SIZE(AA), %xmm1
	xorps	%xmm7, %xmm7

	PREFETCHW      7 * SIZE(CO1)

#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K, %eax
	subl	KK, %eax
#endif
	sarl	$3, %eax
	je	.L112
	ALIGN_2

.L111:
	mulps	%xmm2, %xmm0
	mulps	 4 * SIZE(AA), %xmm2
	addps	%xmm0, %xmm4
	movaps	 8 * SIZE(AA), %xmm0
	addps	%xmm2, %xmm6
	movaps	 4 * SIZE(BB), %xmm2
	mulps	%xmm2, %xmm0
	mulps	12 * SIZE(AA), %xmm2
	addps	%xmm0, %xmm5
	movaps	32 * SIZE(AA), %xmm0
	addps	%xmm2, %xmm7
	movaps	 8 * SIZE(BB), %xmm2
	mulps	%xmm2, %xmm1
	mulps	20 * SIZE(AA), %xmm2
	addps	%xmm1, %xmm4
	movaps	24 * SIZE(AA), %xmm1
	addps	%xmm2, %xmm6
	movaps	12 * SIZE(BB), %xmm2
	mulps	%xmm2, %xmm1
	mulps	28 * SIZE(AA), %xmm2
	addps	%xmm1, %xmm5
	movaps	48 * SIZE(AA), %xmm1
	addps	%xmm2, %xmm7
	movaps	32 * SIZE(BB), %xmm2
	mulps	%xmm3, %xmm0
	mulps	36 * SIZE(AA), %xmm3
	addps	%xmm0, %xmm4
	movaps	40 * SIZE(AA), %xmm0
	addps	%xmm3, %xmm6
	movaps	20 * SIZE(BB), %xmm3
	mulps	%xmm3, %xmm0
	mulps	44 * SIZE(AA), %xmm3
	addps	%xmm0, %xmm5
	movaps	64 * SIZE(AA), %xmm0
	addps	%xmm3, %xmm7
	movaps	24 * SIZE(BB), %xmm3
	mulps	%xmm3, %xmm1
	mulps	52 * SIZE(AA), %xmm3
	addps	%xmm1, %xmm4
	movaps	56 * SIZE(AA), %xmm1
	addps	%xmm3, %xmm6
	movaps	28 * SIZE(BB), %xmm3
	mulps	%xmm3, %xmm1
	mulps	60 * SIZE(AA), %xmm3
	addps	%xmm1, %xmm5
	movaps	80 * SIZE(AA), %xmm1
	addps	%xmm3, %xmm7
	movaps	48 * SIZE(BB), %xmm3

	addl   $64 * SIZE, AA
	addl   $32 * SIZE, BB
	decl   %eax
	jne    .L111
	ALIGN_2

.L112:
#if defined(LT) || defined(RN)
	movl	KK, %eax
#else
	movl	K, %eax
	subl	KK, %eax
#endif
	andl	$7, %eax		# if (k & 1)
	BRANCH
	je .L114

.L113:
	movaps	 0 * SIZE(BB), %xmm2
	movaps	 0 * SIZE(AA), %xmm0
	mulps	%xmm2, %xmm0
	addps	%xmm0, %xmm4
	mulps	 4 * SIZE(AA), %xmm2
	addps	%xmm2, %xmm6

	addl	$8 * SIZE, AA
	addl	$4 * SIZE, BB
	subl	$1, %eax
	jg	.L113
	ALIGN_4

.L114:
	addps	%xmm5, %xmm4
	addps	%xmm7, %xmm6

#if defined(LN) || defined(RT)
	movl	KK, %eax
#ifdef LN
	subl	$8, %eax
#else
	subl	$1, %eax
#endif

	movl	AORIG, AA
	movl	BORIG, B
	leal	BUFFER, BB

	sall	$BASE_SHIFT, %eax
	leal	(AA, %eax, 8), AA
	leal	(B,  %eax, 1), B
	leal	(BB, %eax, 4), BB
#endif

#if defined(LN) || defined(LT)
	movaps	 0 * SIZE(B), %xmm0
	movaps	 4 * SIZE(B), %xmm1

	subps	%xmm4,  %xmm0
	subps	%xmm6,  %xmm1
#else
	movaps	 0 * SIZE(AA), %xmm0
	movaps	 4 * SIZE(AA), %xmm1

	subps	%xmm4, %xmm0
	subps	%xmm6, %xmm1
#endif

#if defined(LN) || defined(LT)
	movaps	MASK, %xmm6
#endif

#ifdef LN
#if 0
	movss	63 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm7

	pshufd	$0xee, %xmm7, %xmm1

	movss	62 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movsd	60 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	58 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	56 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	54 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm7

	pshufd	$0x44, %xmm7, %xmm1

	movsd	52 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	50 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	48 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2


	movss	45 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm5

	pshufd	$0xee, %xmm5, %xmm1

	movss	44 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	42 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	40 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	36 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm5

	pshufd	$0x44, %xmm5, %xmm1

	movsd	34 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	32 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	27 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm3

	pshufd	$0xee, %xmm3, %xmm1

	movss	26 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	24 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	18 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm3

	pshufd	$0x44, %xmm3, %xmm1

	movsd	16 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	 9 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm2

	pshufd	$0xee, %xmm2, %xmm1

	movss	 8 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm2

	movss	 0 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm2
#endif
#endif

#ifdef LT
	movaps	 0 * SIZE(AA), %xmm2
	movaps	 4 * SIZE(AA), %xmm3

	mulss	 %xmm2, %xmm0

	pshufd	$0x00, %xmm0,  %xmm4

	xorps	%xmm0


#if 0


	movsd	 2 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	 4 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	 6 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movss	 9 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm2

	pshufd	$0xee, %xmm2, %xmm1

	movsd	10 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	12 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	14 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movss	18 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm3

	pshufd	$0x44, %xmm3, %xmm1

	movss	19 * SIZE(AA), %xmm0
	pshufd	$0x05, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm3

	movsd	20 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	22 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movss	27 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm3

	pshufd	$0xee, %xmm3, %xmm1

	movsd	28 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	30 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movss	36 * SIZE(AA), %xmm0
	shufps	$0x00, %xmm6,  %xmm0
	mulps	%xmm0, %xmm5

	pshufd	$0x44, %xmm5, %xmm1

	movss	37 * SIZE(AA), %xmm0
	pshufd	$0x05, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm5

	movsd	38 * SIZE(AA), %xmm0
	pshufd	$0x50, %xmm0, %xmm0
	mulps	%xmm1, %xmm0
	subps	%xmm0, %xmm7

	movss	45 * SIZE(AA), %xmm0
	movaps	 %xmm6, %xmm1
	shufps	$0x00, %xmm0,  %xmm1
	mulps	%xmm1, %xmm5

	pshufd	$0xee, %xmm5, %xmm1