symv_l_sse2.s

Optimized GotoBLAS libraries

	movhpd	 yy2, 3 * SIZE(YY)
	movsd	 6 * SIZE(YY), yy2
	movhpd	 7 * SIZE(YY), yy2

	movapd	 xtemp1, xt1
	mulpd	 a3,     xt1
	mulpd	 atemp1, a3
	addpd	 xt1,    xsum1
	addpd	 a3,     yy1
	movsd	 6 * SIZE(A1, LDA, 1), a3
	movhpd	 7 * SIZE(A1, LDA, 1), a3

	PREFETCH	PREFETCHSIZE(A2)

	movapd	 xtemp2, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp1, a1
	addpd	 xt1,    xsum1
	addpd	 a1,     yy2
	movsd	 4 * SIZE(A2), a1
	movhpd	 5 * SIZE(A2), a1

	movapd	 xtemp1, xt1
	mulpd	 a2,     xt1
	mulpd	 atemp2, a2
	addpd	 xt1,    xsum2
	addpd	 a2,     yy1
	movsd	 6 * SIZE(A2), a2
	movhpd	 7 * SIZE(A2), a2

#if !defined(CORE2) && !defined(PENRYN)
	PREFETCHW	PREFETCHSIZE(YY)
#endif

	movapd	 xtemp2, xt1
	mulpd	 a3,     xt1
	mulpd	 atemp2, a3
	addpd	 xt1,    xsum2
	addpd	 a3,     yy2
	movsd	 4 * SIZE(A2, LDA, 1), a3
	movhpd	 5 * SIZE(A2, LDA, 1), a3

	movapd	 xtemp1, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp3, a1
	addpd	 xt1,    xsum3
	addpd	 a1,     yy1
	movsd	 6 * SIZE(A2, LDA, 1), a1
	movhpd	 7 * SIZE(A2, LDA, 1), a1

	PREFETCH	PREFETCHSIZE(A2, LDA, 1)

	movapd	 xtemp2, xt1
	mulpd	 a2,     xt1
	mulpd	 atemp3, a2
	addpd	 xt1,    xsum3
	addpd	 a2,     yy2
	movsd	10 * SIZE(A1), a2
	movhpd	11 * SIZE(A1), a2

	movapd	 xtemp1, xt1
	movapd	 8 * SIZE(XX), xtemp1
	mulpd	 a3,     xt1
	mulpd	 atemp4, a3
	addpd	 xt1,    xsum4
	addpd	 a3,     yy1
	movsd	 8 * SIZE(A1, LDA, 1), a3
	movhpd	 9 * SIZE(A1, LDA, 1), a3

	movapd	 xtemp2, xt1
	movapd	10 * SIZE(XX), xtemp2
	mulpd	 a1,     xt1
	mulpd	 atemp4, a1
	addpd	 xt1,    xsum4
	addpd	 a1,     yy2
	movsd	 8 * SIZE(A1), a1
	movhpd	 9 * SIZE(A1), a1

	movsd	 yy1, 4 * SIZE(YY)
	movhpd	 yy1, 5 * SIZE(YY)
	movsd	 8 * SIZE(YY), yy1
	movhpd	 9 * SIZE(YY), yy1

	movsd	 yy2, 6 * SIZE(YY)
	movhpd	 yy2, 7 * SIZE(YY)
	movsd	10 * SIZE(YY), yy2
	movhpd	11 * SIZE(YY), yy2

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

	decq	 I
	jg	 .L12
	ALIGN_3

.L15:
	movq	M,  I
	subq	IS, I
	subq	$4, I
	test	$4, I
	jle	.L17

	movapd	 xtemp1, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp1, a1
	addpd	 xt1,    xsum1
	addpd	 a1,     yy1
	movsd	 2 * SIZE(A1, LDA, 1), a1
	movhpd	 3 * SIZE(A1, LDA, 1), a1

	movapd	 xtemp2, xt1
	mulpd	 a2,     xt1
	mulpd	 atemp1, a2
	addpd	 xt1,    xsum1
	addpd	 a2,     yy2
	movsd	 0 * SIZE(A2), a2
	movhpd	 1 * SIZE(A2), a2

	movapd	 xtemp1, xt1
	mulpd	 a3,     xt1
	mulpd	 atemp2, a3
	addpd	 xt1,    xsum2
	addpd	 a3,     yy1
	movsd	 2 * SIZE(A2), a3
	movhpd	 3 * SIZE(A2), a3

	movapd	 xtemp2, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp2, a1
	addpd	 xt1,    xsum2
	addpd	 a1,     yy2
	movsd	 0 * SIZE(A2, LDA, 1), a1
	movhpd	 1 * SIZE(A2, LDA, 1), a1

	movapd	 xtemp1, xt1
	mulpd	 a2,     xt1
	mulpd	 atemp3, a2
	addpd	 xt1,    xsum3
	addpd	 a2,     yy1
	movsd	 2 * SIZE(A2, LDA, 1), a2
	movhpd	 3 * SIZE(A2, LDA, 1), a2

	movapd	 xtemp2, xt1
	mulpd	 a3,     xt1
	mulpd	 atemp3, a3
	addpd	 xt1,    xsum3
	addpd	 a3,     yy2
	movsd	 4 * SIZE(A1, LDA, 1), a3
	movhpd	 5 * SIZE(A1, LDA, 1), a3

	movapd	 xtemp1, xt1
	movapd	 4 * SIZE(XX), xtemp1
	mulpd	 a1,     xt1
	mulpd	 atemp4, a1
	addpd	 xt1,    xsum4
	addpd	 a1,     yy1
	movsd	 4 * SIZE(A1), a1
	movhpd	 5 * SIZE(A1), a1

	movapd	 xtemp2, xt1
	movapd	 6 * SIZE(XX), xtemp2
	mulpd	 a2,     xt1
	mulpd	 atemp4, a2
	addpd	 xt1,    xsum4
	addpd	 a2,     yy2
	movsd	 6 * SIZE(A1), a2
	movhpd	 7 * SIZE(A1), a2

	movsd	 yy1, 0 * SIZE(YY)
	movhpd	 yy1, 1 * SIZE(YY)
	movsd	 4 * SIZE(YY), yy1
	movhpd	 5 * SIZE(YY), yy1

	movsd	 yy2, 2 * SIZE(YY)
	movhpd	 yy2, 3 * SIZE(YY)
	movsd	 6 * SIZE(YY), yy2
	movhpd	 7 * SIZE(YY), yy2

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

.L17:
	testq	$2, M
	jle	.L18

	movapd	 xtemp1, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp1, a1
	addpd	 xt1,    xsum1
	addpd	 a1,     yy1
	movsd	 0 * SIZE(A1, LDA, 1), a1
	movhpd	 1 * SIZE(A1, LDA, 1), a1

	movapd	 xtemp1, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp2, a1
	addpd	 xt1,    xsum2
	addpd	 a1,     yy1
	movsd	 0 * SIZE(A2), a1
	movhpd	 1 * SIZE(A2), a1

	movapd	 xtemp1, xt1
	mulpd	 a1,     xt1
	mulpd	 atemp3, a1
	addpd	 xt1,    xsum3
	addpd	 a1,     yy1
	movsd	 0 * SIZE(A2, LDA, 1), a1
	movhpd	 1 * SIZE(A2, LDA, 1), a1

	movapd	 xtemp1, xt1
	movapd	 2 * SIZE(XX), xtemp1
	mulpd	 a1,     xt1
	mulpd	 atemp4, a1
	addpd	 xt1,    xsum4
	addpd	 a1,     yy1
	movsd	 2 * SIZE(A1), a1

	movsd	 yy1, 0 * SIZE(YY)
	movhpd	 yy1, 1 * SIZE(YY)
	movsd	 2 * SIZE(YY), yy1

	addq	 $2 * SIZE, XX
	addq	 $2 * SIZE, YY
	addq	 $2 * SIZE, A1
	addq	 $2 * SIZE, A2
	ALIGN_3

.L18:
	testq	$1, M
	jle	.L19

	movapd	 xtemp1, xt1
	mulsd	 a1,     xt1
	mulsd	 atemp1, a1
	addsd	 xt1,    xsum1
	addpd	 a1,     yy1
	movsd	 0 * SIZE(A1, LDA, 1), a1

	movapd	 xtemp1, xt1
	mulsd	 a1,     xt1
	mulsd	 atemp2, a1
	addsd	 xt1,    xsum2
	addsd	 a1,     yy1
	movsd	 0 * SIZE(A2), a1

	movapd	 xtemp1, xt1
	mulsd	 a1,     xt1
	mulsd	 atemp3, a1
	addsd	 xt1,    xsum3
	addsd	 a1,     yy1
	movsd	 0 * SIZE(A2, LDA, 1), a1

	movapd	 xtemp1, xt1
	mulsd	 a1,     xt1
	mulsd	 atemp4, a1
	addsd	 xt1,    xsum4
	addsd	 a1,     yy1

	movsd	 yy1, 0 * SIZE(YY)
	ALIGN_3

.L19:
#ifndef HAVE_SSE3
	movapd	xsum1, atemp1
	movapd	xsum3, atemp3

	unpcklpd xsum2, xsum1
	unpcklpd xsum4, xsum3

	unpckhpd xsum2, atemp1
	unpckhpd xsum4, atemp3

	addpd	 atemp1, xsum1
	addpd	 atemp3, xsum3
#else
	haddpd	 xsum2, xsum1
	haddpd	 xsum4, xsum3
#endif

	movsd	 0 * SIZE(NEW_Y, IS, SIZE), yy1
	movhpd	 1 * SIZE(NEW_Y, IS, SIZE), yy1
	movsd	 2 * SIZE(NEW_Y, IS, SIZE), yy2
	movhpd	 3 * SIZE(NEW_Y, IS, SIZE), yy2

	addpd	 xsum1, yy1
	addpd	 xsum3, yy2

	movsd	 yy1, 0 * SIZE(NEW_Y, IS, SIZE)
	movhpd	 yy1, 1 * SIZE(NEW_Y, IS, SIZE)
	movsd	 yy2, 2 * SIZE(NEW_Y, IS, SIZE)
	movhpd	 yy2, 3 * SIZE(NEW_Y, IS, SIZE)

	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	2 * SIZE(A, LDA, 2), A

	movapd		0 * SIZE(NEW_X, IS, SIZE), atemp2

	movsd	 0 * SIZE(A1), xsum1
	movhpd	 1 * SIZE(A1), xsum1
	mulpd	 atemp2, xsum1

	movsd	 1 * SIZE(A1), xsum2
	movhpd	 1 * SIZE(A1, LDA, 1), xsum2
	mulpd	 atemp2, xsum2

#ifndef HAVE_SSE3
	movapd	 atemp2, atemp1
	unpcklpd atemp1, atemp1
#else
	movddup	 atemp2, atemp1
#endif
	unpckhpd atemp2, atemp2

	testq	$1, M
	jle	.L29

	movsd	 2 * SIZE(A1), a1
	movsd	 2 * SIZE(A1, LDA, 1), a2
	movsd	 2 * SIZE(NEW_X, IS, SIZE), xtemp1
	movsd	 2 * SIZE(NEW_Y, IS, SIZE), yy1

	movapd	 xtemp1, xt1
	mulsd	 a1,     xt1
	mulsd	 atemp1, a1
	addsd	 xt1,    xsum1
	addpd	 a1,     yy1

	movapd	 xtemp1, xt1
	mulsd	 a2,     xt1
	mulsd	 atemp2, a2
	addsd	 xt1,    xsum2
	addsd	 a2,     yy1

	movsd	 yy1, 2 * SIZE(NEW_Y, IS, SIZE)
	ALIGN_3

.L29:
#ifndef HAVE_SSE3
	movapd	xsum1, atemp1
	unpcklpd xsum2, xsum1
	unpckhpd xsum2, atemp1
	addpd	 atemp1, xsum1
#else
	haddpd	 xsum2, xsum1
#endif

	movsd	 0 * SIZE(NEW_Y, IS, SIZE), yy1
	movhpd	 1 * SIZE(NEW_Y, IS, SIZE), yy1

	addpd	 xsum1, yy1

	movsd	 yy1, 0 * SIZE(NEW_Y, IS, SIZE)
	movhpd	 yy1, 1 * SIZE(NEW_Y, IS, SIZE)

	addq	 $2, IS
	ALIGN_3

.L30:
	testq	$1, M
	jle	.L990

	movsd	 0 * SIZE(A), xsum1
	movsd	 0 * SIZE(NEW_X, IS, SIZE), atemp1
	movsd	 0 * SIZE(NEW_Y, IS, SIZE), yy1

	mulsd	 atemp1, xsum1
	addsd	 xsum1, yy1
	movsd	 yy1, 0 * SIZE(NEW_Y, IS, SIZE)
	ALIGN_3

.L990:
	cmpq   $SIZE, INCY
	je    .L999

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

.L996:
	movapd	 0 * SIZE(NEW_Y), %xmm0
	movapd	 2 * SIZE(NEW_Y), %xmm1
	movapd	 4 * SIZE(NEW_Y), %xmm2
	movapd	 6 * SIZE(NEW_Y), %xmm3

	movsd	%xmm0,  0 * SIZE(Y)
	addq	INCY, Y
	movhpd	%xmm0,  0 * SIZE(Y)
	addq	INCY, Y
	movsd	%xmm1,  0 * SIZE(Y)
	addq	INCY, Y
	movhpd	%xmm1,  0 * SIZE(Y)
	addq	INCY, Y
	movsd	%xmm2,  0 * SIZE(Y)
	addq	INCY, Y
	movhpd	%xmm2,  0 * SIZE(Y)
	addq	INCY, Y
	movsd	%xmm3,  0 * SIZE(Y)
	addq	INCY, Y
	movhpd	%xmm3,  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:
	movsd	0 * SIZE(NEW_Y), %xmm0

	movsd	%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