rot_sse2.s

Optimized GotoBLAS libraries

	mulpd	C, %xmm0
	mulpd	S, %xmm1

	mulpd	C, %xmm2
	mulpd	S, %xmm3

	addpd	%xmm1, %xmm0
	subpd	%xmm3, %xmm2

	movapd	%xmm0, 0 * SIZE(X)
	movapd	%xmm2, 0 * SIZE(Y)

	addq	$2 * SIZE, X
	addq	$2 * SIZE, Y
	ALIGN_3

.L17:
	testq	$1, N
	jle	.L999

	movsd	0 * SIZE(Y), %xmm1
	movsd	0 * SIZE(X), %xmm0

	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulsd	C, %xmm0
	mulsd	S, %xmm1

	mulsd	C, %xmm2
	mulsd	S, %xmm3

	addsd	%xmm1, %xmm0
	subsd	%xmm3, %xmm2

	movsd	%xmm0, 0 * SIZE(X)
	movsd	%xmm2, 0 * SIZE(Y)
	jmp	.L999
	ALIGN_3

.L20:
	movapd	-1 * SIZE(Y), %xmm1

	movq	N,  %rax
	sarq	$4, %rax
	jle	.L24
	ALIGN_3

.L21:
#ifdef PENTIUM4
	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(X)
#endif

	movapd	1 * SIZE(Y), %xmm3
	movapd	3 * SIZE(Y), %xmm8
#if defined(OPTERON) || defined(BARCELONA)
	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(X)
#endif
	movapd	0 * SIZE(X), %xmm0
	movapd	2 * SIZE(X), %xmm2

	SHUFPD_1 %xmm3, %xmm1
	SHUFPD_1 %xmm8, %xmm3

	movapd	%xmm1, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm3, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm1
	mulpd	C, %xmm2
	mulpd	S, %xmm3

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm1, %xmm0
	addpd	%xmm3, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 0 * SIZE(X)
	movapd	%xmm2, 2 * SIZE(X)

	movlpd	%xmm4, 0 * SIZE(Y)
	movhpd	%xmm4, 1 * SIZE(Y)
	movlpd	%xmm6, 2 * SIZE(Y)
	movhpd	%xmm6, 3 * SIZE(Y)


	movapd	5 * SIZE(Y), %xmm9
	movapd	7 * SIZE(Y), %xmm1
#if defined(OPTERON) || defined(BARCELONA)
	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(Y)
#endif
	movapd	4 * SIZE(X), %xmm0
	movapd	6 * SIZE(X), %xmm2

	SHUFPD_1 %xmm9, %xmm8
	SHUFPD_1 %xmm1, %xmm9

	movapd	%xmm8, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm9, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm8
	mulpd	C, %xmm2
	mulpd	S, %xmm9

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm8, %xmm0
	addpd	%xmm9, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 4 * SIZE(X)
	movapd	%xmm2, 6 * SIZE(X)
	movlpd	%xmm4, 4 * SIZE(Y)
	movhpd	%xmm4, 5 * SIZE(Y)
	movlpd	%xmm6, 6 * SIZE(Y)
	movhpd	%xmm6, 7 * SIZE(Y)

#ifdef PENTIUM4
	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(Y)
#endif
#if defined(OPTERON) || defined(BARCELONA)
	PREFETCH	(PREFETCH_SIZE + 8) * SIZE(X)
#endif

	movapd	 9 * SIZE(Y), %xmm3
	movapd	11 * SIZE(Y), %xmm8
	movapd	 8 * SIZE(X), %xmm0
	movapd	10 * SIZE(X), %xmm2

	SHUFPD_1 %xmm3, %xmm1
	SHUFPD_1 %xmm8, %xmm3

	movapd	%xmm1, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm3, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm1
	mulpd	C, %xmm2
	mulpd	S, %xmm3

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm1, %xmm0
	addpd	%xmm3, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0,  8 * SIZE(X)
	movapd	%xmm2, 10 * SIZE(X)
	movlpd	%xmm4,  8 * SIZE(Y)
	movhpd	%xmm4,  9 * SIZE(Y)
	movlpd	%xmm6, 10 * SIZE(Y)
	movhpd	%xmm6, 11 * SIZE(Y)

#if defined(OPTERON) || defined(BARCELONA)
	PREFETCH	(PREFETCH_SIZE + 8) * SIZE(Y)
#endif

	movapd	13 * SIZE(Y), %xmm9
	movapd	15 * SIZE(Y), %xmm1
	movapd	12 * SIZE(X), %xmm0
	movapd	14 * SIZE(X), %xmm2

	SHUFPD_1 %xmm9, %xmm8
	SHUFPD_1 %xmm1, %xmm9

	movapd	%xmm8, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm9, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm8
	mulpd	C, %xmm2
	mulpd	S, %xmm9

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm8, %xmm0
	addpd	%xmm9, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 12 * SIZE(X)
	movapd	%xmm2, 14 * SIZE(X)
	movlpd	%xmm4, 12 * SIZE(Y)
	movhpd	%xmm4, 13 * SIZE(Y)
	movlpd	%xmm6, 14 * SIZE(Y)
	movhpd	%xmm6, 15 * SIZE(Y)

	addq	$16 * SIZE, X
	addq	$16 * SIZE, Y
	decq	%rax
	jg	.L21
	ALIGN_3

.L24:
	testq	$15, N
	jle	.L999

	testq	$8, N
	jle	.L25

	movapd	1 * SIZE(Y), %xmm3
	movapd	3 * SIZE(Y), %xmm8
	movapd	0 * SIZE(X), %xmm0
	movapd	2 * SIZE(X), %xmm2

	SHUFPD_1 %xmm3, %xmm1
	SHUFPD_1 %xmm8, %xmm3

	movapd	%xmm1, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm3, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm1
	mulpd	C, %xmm2
	mulpd	S, %xmm3

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm1, %xmm0
	addpd	%xmm3, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 0 * SIZE(X)
	movapd	%xmm2, 2 * SIZE(X)
	movlpd	%xmm4, 0 * SIZE(Y)
	movhpd	%xmm4, 1 * SIZE(Y)
	movlpd	%xmm6, 2 * SIZE(Y)
	movhpd	%xmm6, 3 * SIZE(Y)

	movapd	5 * SIZE(Y), %xmm9
	movapd	7 * SIZE(Y), %xmm1
	movapd	4 * SIZE(X), %xmm0
	movapd	6 * SIZE(X), %xmm2

	SHUFPD_1 %xmm9, %xmm8
	SHUFPD_1 %xmm1, %xmm9

	movapd	%xmm8, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm9, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm8
	mulpd	C, %xmm2
	mulpd	S, %xmm9

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm8, %xmm0
	addpd	%xmm9, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 4 * SIZE(X)
	movapd	%xmm2, 6 * SIZE(X)
	movlpd	%xmm4, 4 * SIZE(Y)
	movhpd	%xmm4, 5 * SIZE(Y)
	movlpd	%xmm6, 6 * SIZE(Y)
	movhpd	%xmm6, 7 * SIZE(Y)

	addq	$8 * SIZE, X
	addq	$8 * SIZE, Y
	ALIGN_3

.L25:
	testq	$4, N
	jle	.L26

	movapd	1 * SIZE(Y), %xmm3
	movapd	3 * SIZE(Y), %xmm8
	movapd	0 * SIZE(X), %xmm0
	movapd	2 * SIZE(X), %xmm2

	SHUFPD_1 %xmm3, %xmm1
	SHUFPD_1 %xmm8, %xmm3

	movapd	%xmm1, %xmm4
	movapd	%xmm0, %xmm5
	movapd	%xmm3, %xmm6
	movapd	%xmm2, %xmm7

	mulpd	C, %xmm0
	mulpd	S, %xmm1
	mulpd	C, %xmm2
	mulpd	S, %xmm3

	mulpd	C, %xmm4
	mulpd	S, %xmm5
	mulpd	C, %xmm6
	mulpd	S, %xmm7

	addpd	%xmm1, %xmm0
	addpd	%xmm3, %xmm2
	subpd	%xmm5, %xmm4
	subpd	%xmm7, %xmm6

	movapd	%xmm0, 0 * SIZE(X)
	movapd	%xmm2, 2 * SIZE(X)
	movlpd	%xmm4, 0 * SIZE(Y)
	movhpd	%xmm4, 1 * SIZE(Y)
	movlpd	%xmm6, 2 * SIZE(Y)
	movhpd	%xmm6, 3 * SIZE(Y)
	movapd	%xmm8, %xmm1

	addq	$4 * SIZE, X
	addq	$4 * SIZE, Y
	ALIGN_3

.L26:
	testq	$2, N
	jle	.L27

	movapd	1 * SIZE(Y), %xmm4
	movapd	0 * SIZE(X), %xmm0

	SHUFPD_1 %xmm4, %xmm1
	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulpd	C, %xmm0
	mulpd	S, %xmm1

	mulpd	C, %xmm2
	mulpd	S, %xmm3

	addpd	%xmm1, %xmm0
	subpd	%xmm3, %xmm2

	movapd	%xmm0, 0 * SIZE(X)
	movlpd	%xmm2, 0 * SIZE(Y)
	movhpd	%xmm2, 1 * SIZE(Y)
	movapd	%xmm4, %xmm1

	addq	$2 * SIZE, X
	addq	$2 * SIZE, Y
	ALIGN_3

.L27:
	testq	$1, N
	jle	.L999

	unpckhpd %xmm1, %xmm1
	movsd	0 * SIZE(X), %xmm0

	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulsd	C, %xmm0
	mulsd	S, %xmm1

	mulsd	C, %xmm2
	mulsd	S, %xmm3

	addsd	%xmm1, %xmm0
	subsd	%xmm3, %xmm2

	movsd	%xmm0, 0 * SIZE(X)
	movsd	%xmm2, 0 * SIZE(Y)
	jmp	.L999
	ALIGN_3


.L50:
	movq	N,  %rax
	sarq	$2, %rax
	jle	.L55
	ALIGN_3

.L53:
	MOVSD	(Y), %xmm1
	movhpd	(Y, INCY), %xmm1
	MOVSD	(X), %xmm0
	movhpd	(X, INCX), %xmm0

	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulpd	C, %xmm0
	mulpd	S, %xmm1

	mulpd	C, %xmm2
	mulpd	S, %xmm3

	addpd	%xmm1, %xmm0
	subpd	%xmm3, %xmm2

	movlpd	%xmm0, (X)
	movhpd	%xmm0, (X, INCX)
	movlpd	%xmm2, (Y)
	movhpd	%xmm2, (Y, INCY)

	leaq	(X, INCX, 2), X
	leaq	(Y, INCY, 2), Y

	MOVSD	(Y), %xmm1
	movhpd	(Y, INCY), %xmm1
	MOVSD	(X), %xmm0
	movhpd	(X, INCX), %xmm0

	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulpd	C, %xmm0
	mulpd	S, %xmm1

	mulpd	C, %xmm2
	mulpd	S, %xmm3

	addpd	%xmm1, %xmm0
	subpd	%xmm3, %xmm2

	movlpd	%xmm0, (X)
	movhpd	%xmm0, (X, INCX)
	movlpd	%xmm2, (Y)
	movhpd	%xmm2, (Y, INCY)

	leaq	(X, INCX, 2), X
	leaq	(Y, INCY, 2), Y

	decq	%rax
	jg	.L53
	ALIGN_3

.L55:
	movq	N,  %rax
	andq	$3, %rax
	jle	.L999
	ALIGN_3

.L56:
	movsd	(Y), %xmm1
	movsd	(X), %xmm0

	movapd	%xmm1, %xmm2
	movapd	%xmm0, %xmm3

	mulsd	C, %xmm0
	mulsd	S, %xmm1

	mulsd	C, %xmm2
	mulsd	S, %xmm3

	addsd	%xmm1, %xmm0
	subsd	%xmm3, %xmm2

	movsd	%xmm0, (X)
	movsd	%xmm2, (Y)

	addq	INCX, X
	addq	INCY, Y

	decq	%rax
	jg	.L56
	ALIGN_3

.L999:
	RESTOREREGISTERS

	ret

	EPILOGUE