symv_u_sse.s

Optimized GotoBLAS libraries

	addps	 a3,     yy1
	movsd	16 * SIZE(A2), a3
	movhps	18 * SIZE(A2), a3

	movaps	 xtemp2, xt1
	movaps	20 * SIZE(XX), xtemp2
	mulps	 a4,     xt1
	mulps	 atemp4, a4
	addps	 xt1,    xsum4
	addps	 a4,     yy1
	movsd	16 * SIZE(A2, LDA, 1), a4
	movhps	18 * SIZE(A2, LDA, 1), a4

	movlps	 yy1, 12 * SIZE(YY)
	movhps	 yy1, 14 * SIZE(YY)
	movsd	16 * SIZE(YY), yy1
	movhps	18 * SIZE(YY), yy1

	addq	 $16 * SIZE, XX
	addq	 $16 * SIZE, YY
	addq	 $16 * SIZE, A1
	addq	 $16 * SIZE, A2

	decq	 I
	jg	 .L12
	ALIGN_3

.L14:
	testq	$8, IS
	jle	.L15

	movaps	 xtemp1, xt1
	mulps	 a1,     xt1
	mulps	 atemp1, a1
	addps	 xt1,    xsum1
	addps	 a1,     yy1
	movsd	 4 * SIZE(A1), a1
	movhps	 6 * SIZE(A1), a1

	movaps	 xtemp1, xt1
	mulps	 a2,     xt1
	mulps	 atemp2, a2
	addps	 xt1,    xsum2
	addps	 a2,     yy1
	movsd	 4 * SIZE(A1, LDA, 1), a2
	movhps	 6 * SIZE(A1, LDA, 1), a2

	movaps	 xtemp1, xt1
	mulps	 a3,     xt1
	mulps	 atemp3, a3
	addps	 xt1,    xsum3
	addps	 a3,     yy1
	movsd	 4 * SIZE(A2), a3
	movhps	 6 * SIZE(A2), a3

	movaps	 xtemp1, xt1
	movaps	 8 * SIZE(XX), xtemp1
	mulps	 a4,     xt1
	mulps	 atemp4, a4
	addps	 xt1,    xsum4
	addps	 a4,     yy1
	movsd	 4 * SIZE(A2, LDA, 1), a4
	movhps	 6 * SIZE(A2, LDA, 1), a4

	movlps	 yy1, 0 * SIZE(YY)
	movhps	 yy1, 2 * SIZE(YY)
	movsd	 4 * SIZE(YY), yy1
	movhps	 6 * SIZE(YY), yy1

	movaps	 xtemp2, xt1
	mulps	 a1,     xt1
	mulps	 atemp1, a1
	addps	 xt1,    xsum1
	addps	 a1,     yy1
	movsd	 8 * SIZE(A1), a1
	movhps	10 * SIZE(A1), a1

	movaps	 xtemp2, xt1
	mulps	 a2,     xt1
	mulps	 atemp2, a2
	addps	 xt1,    xsum2
	addps	 a2,     yy1
	movsd	 8 * SIZE(A1, LDA, 1), a2
	movhps	10 * SIZE(A1, LDA, 1), a2

	movaps	 xtemp2, xt1
	mulps	 a3,     xt1
	mulps	 atemp3, a3
	addps	 xt1,    xsum3
	addps	 a3,     yy1
	movsd	 8 * SIZE(A2), a3
	movhps	10 * SIZE(A2), a3

	movaps	 xtemp2, xt1
	movaps	12 * SIZE(XX), xtemp2
	mulps	 a4,     xt1
	mulps	 atemp4, a4
	addps	 xt1,    xsum4
	addps	 a4,     yy1
	movsd	 8 * SIZE(A2, LDA, 1), a4
	movhps	10 * SIZE(A2, LDA, 1), a4

	movlps	 yy1, 4 * SIZE(YY)
	movhps	 yy1, 6 * SIZE(YY)
	movsd	 8 * SIZE(YY), yy1
	movhps	10 * SIZE(YY), yy1

	addq	 $8 * SIZE, XX
	addq	 $8 * SIZE, YY
	addq	 $8 * SIZE, A1
	addq	 $8 * SIZE, A2
	ALIGN_3

.L15:
	testq	$4, IS
	jle	.L18

	movaps	 xtemp1, xt1
	mulps	 a1,     xt1
	mulps	 atemp1, a1
	addps	 xt1,    xsum1
	addps	 a1,     yy1

	movaps	 xtemp1, xt1
	mulps	 a2,     xt1
	mulps	 atemp2, a2
	addps	 xt1,    xsum2
	addps	 a2,     yy1

	movaps	 xtemp1, xt1
	mulps	 a3,     xt1
	mulps	 atemp3, a3
	addps	 xt1,    xsum3
	addps	 a3,     yy1

	movaps	 xtemp1, xt1
	mulps	 a4,     xt1
	mulps	 atemp4, a4
	addps	 xt1,    xsum4
	addps	 a4,     yy1

	movlps	 yy1, 0 * SIZE(YY)
	movhps	 yy1, 2 * SIZE(YY)
	movsd	 4 * SIZE(YY), yy1
	movhps	 6 * SIZE(YY), yy1

	addq	 $4 * SIZE, XX
	addq	 $4 * SIZE, YY
	addq	 $4 * SIZE, A1
	addq	 $4 * SIZE, A2
	ALIGN_3

.L18:
	movaps		0 * SIZE(NEW_X, IS, SIZE), atemp1

	movss	 0 * SIZE(A1), a1
	movss	 0 * SIZE(A1, LDA, 1), a2
	movss	 0 * SIZE(A2), a3
	movss	 0 * SIZE(A2, LDA, 1), a4

	unpcklps a3, a1
	unpcklps a4, a2
	unpcklps a2, a1

	mulps	 atemp1, a1
	addps	 a1, xsum1

	movsd	 0 * SIZE(A1, LDA, 1), a1
	movss	 1 * SIZE(A2), a2
	movhps	 1 * SIZE(A2, LDA, 1), a2

	shufps	 $0x84, a2, a1

	mulps	 atemp1, a1
	addps	 a1, xsum2

	movsd	 0 * SIZE(A2), a1
	movss	 2 * SIZE(A2), a2
	movhps	 2 * SIZE(A2, LDA, 1), a2

	shufps	 $0x84, a2, a1

	mulps	 atemp1, a1
	addps	 a1, xsum3

	movsd	 0 * SIZE(A2, LDA, 1), a1
	movhps	 2 * SIZE(A2, LDA, 1), a1

	mulps	 atemp1, a1
	addps	 a1, xsum4


#ifndef HAVE_SSE3
	movaps	 xsum1,  xtemp1
	unpcklps xsum3,  xsum1
	unpckhps xsum3,  xtemp1

	movaps	 xsum2,  xtemp2
	unpcklps xsum4,  xsum2
	unpckhps xsum4,  xtemp2

	movaps	 xsum1,  xsum3
	unpcklps xsum2,  xsum1
	unpckhps xsum2,  xsum3

	movaps	 xtemp1, xsum4
	unpcklps xtemp2, xtemp1
	unpckhps xtemp2, xsum4

	addps	 xsum3,  xsum1
	addps	 xtemp1, xsum4
	addps	 xsum4,  xsum1
#else
	haddps	 xsum2, xsum1
	haddps	 xsum4, xsum3

	haddps	 xsum3, xsum1
#endif

	addps	 xsum1, yy1

	movlps	 yy1, 0 * SIZE(YY)
	movhps	 yy1, 2 * SIZE(YY)

	addq	 $4, IS

	movq	 IS, I
	addq	 $4, I
	cmpq	 M, I
	jle	 .L11
	ALIGN_3

.L20:
	testq	$2, M
	jle	.L30

	movq	A,  A1
	leaq	(A, LDA, 2), A

	movsd		0 * SIZE(NEW_X, IS, SIZE), atemp4

	pshufd	$0x00, atemp4, atemp1
	pshufd	$0x55, atemp4, atemp2

	pxor		xsum1, xsum1
	pxor		xsum2, xsum2

	movaps	 0 * SIZE(NEW_X), xtemp1

	movsd	 0 * SIZE(A1), a1
	movhps	 2 * SIZE(A1), a1
	movsd	 0 * SIZE(A1, LDA, 1), a2
	movhps	 2 * SIZE(A1, LDA, 1), a2

	movsd	 0 * SIZE(NEW_Y), yy1
	movhps	 2 * SIZE(NEW_Y), yy1

	movq		NEW_X, XX
	movq		NEW_Y, YY

	movq	IS,  I
	sarq	$2,  I
	jle	.L28
	ALIGN_3

.L22:
	movaps	 xtemp1, xt1
	mulps	 a1,     xt1
	mulps	 atemp1, a1
	addps	 xt1,    xsum1
	addps	 a1,     yy1
	movsd	 4 * SIZE(A1), a1
	movhps	 6 * SIZE(A1), a1

	movaps	 xtemp1, xt1
	movaps	 4 * SIZE(XX), xtemp1
	mulps	 a2,     xt1
	mulps	 atemp2, a2
	addps	 xt1,    xsum2
	addps	 a2,     yy1
	movsd	 4 * SIZE(A1, LDA, 1), a2
	movhps	 6 * SIZE(A1, LDA, 1), a2

	movlps	 yy1, 0 * SIZE(YY)
	movhps	 yy1, 2 * SIZE(YY)
	movsd	 4 * SIZE(YY), yy1
	movhps	 6 * SIZE(YY), yy1

	addq	 $4 * SIZE, XX
	addq	 $4 * SIZE, YY
	addq	 $4 * SIZE, A1

	decq	 I
	jg	 .L22
	ALIGN_3

.L28:
	movsd		0 * SIZE(NEW_X, IS, SIZE), atemp1

	movss	 0 * SIZE(A1), a1
	movss	 0 * SIZE(A1, LDA, 1), a2

	unpcklps a2, a1

	mulps	 atemp1, a1
	addps	 a1, xsum1

	movsd	 0 * SIZE(A1, LDA, 1), a1
	mulps	 atemp1, a1
	addps	 a1, xsum2

#ifndef HAVE_SSE3
	movhlps	 xsum1, xsum3
	movhlps	 xsum2, xsum4
	addps	 xsum3, xsum1
	addps	 xsum4, xsum2

	unpcklps xsum2, xsum1
	movhlps	 xsum1, xsum2

	addps	 xsum2, xsum1
#else
	haddps	 xsum2, xsum1
	haddps	 xsum1, xsum1
#endif

	addps	 xsum1, yy1

	movlps	 yy1, 0 * SIZE(YY)

	addq	 $2, IS
	ALIGN_3

.L30:
	testq	$1, M
	jle	.L990

	movq	A,  A1

	movss		0 * SIZE(NEW_X, IS, SIZE), atemp1

	pshufd	$0x00, atemp1, atemp1

	pxor		xsum1, xsum1
	pxor		xsum2, xsum2

	movss	 0 * SIZE(NEW_Y), yy1

	movss	 0 * SIZE(NEW_X), xtemp1
	movss	 1 * SIZE(NEW_X), xtemp2

	movss	 0 * SIZE(A1), a1
	movss	 1 * SIZE(A1), a2

	movq		NEW_X, XX
	movq		NEW_Y, YY

	movq	IS,  I
	sarq	$1,  I
	jle	.L38
	ALIGN_3

.L32:
	movaps	 xtemp1, xt1
	movss	 2 * SIZE(XX), xtemp1
	mulps	 a1,     xt1
	mulps	 atemp1, a1
	addps	 xt1,    xsum1
	addps	 a1,     yy1
	movss	 2 * SIZE(A1), a1

	movss	 yy1, 0 * SIZE(YY)
	movss	 1 * SIZE(YY), yy1

	movaps	 xtemp2, xt1
	movss	 3 * SIZE(XX), xtemp2
	mulps	 a2,     xt1
	mulps	 atemp1, a2
	addps	 xt1,    xsum1
	addps	 a2,     yy1
	movss	 3 * SIZE(A1), a2

	movss	 yy1, 1 * SIZE(YY)
	movss	 2 * SIZE(YY), yy1

	addq	 $2 * SIZE, XX
	addq	 $2 * SIZE, YY
	addq	 $2 * SIZE, A1

	decq	 I
	jg	 .L32
	ALIGN_3

.L38:
	movsd		0 * SIZE(NEW_X, IS, SIZE), atemp1

	movss	 0 * SIZE(A1), a1
	mulss	 atemp1, a1
	addss	 a1, xsum1

#ifndef HAVE_SSE3
	movhlps	 xsum1, xsum3
	movhlps	 xsum2, xsum4
	addps	 xsum3, xsum1
	addps	 xsum4, xsum2

	unpcklps xsum2, xsum1
	movhlps	 xsum1, xsum2

	addps	 xsum2, xsum1
#else
	addss	 xsum2, xsum1
#endif

	addss	 xsum1, yy1

	movss	 yy1, 0 * SIZE(YY)

	addq	 $2, IS
	ALIGN_3

.L990:
	cmpq   $SIZE, INCY
	je    .L999

	movq	M,  %rax
	sarq	$3, %rax
	jle	.L997
	ALIGN_3

.L996:
	movss	 0 * SIZE(NEW_Y), %xmm0
	movss	 1 * SIZE(NEW_Y), %xmm1
	movss	 2 * SIZE(NEW_Y), %xmm2
	movss	 3 * SIZE(NEW_Y), %xmm3
	movss	 4 * SIZE(NEW_Y), %xmm4
	movss	 5 * SIZE(NEW_Y), %xmm5
	movss	 6 * SIZE(NEW_Y), %xmm6
	movss	 7 * SIZE(NEW_Y), %xmm7

	movss	%xmm0,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm1,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm2,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm3,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm4,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm5,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm6,  0 * SIZE(Y)
	addq	INCY, Y
	movss	%xmm7,  0 * SIZE(Y)
	addq	INCY, Y

	addq	$8 * SIZE, NEW_Y
	decq	%rax
	jg	.L996
	ALIGN_3

.L997:
	movq	M, %rax
	andq	$7, %rax
	jle	.L999
	ALIGN_3

.L998:
	movss	0 * SIZE(NEW_Y), %xmm0

	movss	%xmm0,  0 * SIZE(Y)
	addq	INCY, Y

	addq	$1 * SIZE, NEW_Y

	decq	%rax
	jg	.L998
	ALIGN_3


.L999:
	movq	  0(%rsp), %rbx
	movq	  8(%rsp), %rbp
	movq	 16(%rsp), %r12
	movq	 24(%rsp), %r13
	movq	 32(%rsp), %r14
	movq	 40(%rsp), %r15

#ifdef WINDOWS_ABI
	movq	 48(%rsp), %rdi
	movq	 56(%rsp), %rsi
	movups	 64(%rsp), %xmm6
	movups	 80(%rsp), %xmm7
	movups	 96(%rsp), %xmm8
	movups	112(%rsp), %xmm9
	movups	128(%rsp), %xmm10
	movups	144(%rsp), %xmm11
	movups	160(%rsp), %xmm12
	movups	176(%rsp), %xmm13
	movups	192(%rsp), %xmm14
	movups	208(%rsp), %xmm15
#endif

	addq	$STACKSIZE, %rsp
	ret
	EPILOGUE