axpy_sse2_core2.s

Optimized GotoBLAS libraries

	movaps	-16 * SIZE(Y), %xmm4
	mulpd	ALPHA, %xmm0
	addpd	%xmm4, %xmm0
	movaps	%xmm0, -16 * SIZE(Y)

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

.L18:
	testq	$1, M
	jle	.L19

	movsd	-16 * SIZE(X), %xmm0
	movsd	-16 * SIZE(Y), %xmm4
	mulsd	ALPHA, %xmm0
	addsd	%xmm4, %xmm0
	movsd	%xmm0, -16 * SIZE(Y)
	ALIGN_3

.L19:
	xorq	%rax,%rax

	RESTOREREGISTERS

	ret
	ALIGN_3

.L20:
	movaps	-1 * SIZE(X), %xmm0
	movq	M,  %rax
	sarq	$4, %rax
	jle	.L23

	movaps	 1 * SIZE(X), %xmm1
	movaps	 3 * SIZE(X), %xmm2
	movaps	 5 * SIZE(X), %xmm3
	movaps	 7 * SIZE(X), %xmm8

	movaps	 9 * SIZE(X), %xmm9
	movaps	11 * SIZE(X), %xmm10
	movaps	13 * SIZE(X), %xmm11
	movaps	15 * SIZE(X), %xmm12

	movaps	0 * SIZE(Y), %xmm4
	movaps	2 * SIZE(Y), %xmm5
	movaps	4 * SIZE(Y), %xmm6
	movaps	6 * SIZE(Y), %xmm14
	decq	%rax
	jle .L22
	ALIGN_4

.L21:
#ifdef PENRYN
	shufps	$0x4e, %xmm1, %xmm0
	shufps	$0x4e, %xmm2, %xmm1
	shufps	$0x4e, %xmm3, %xmm2
	shufps	$0x4e, %xmm8, %xmm3

	shufps	$0x4e, %xmm9,  %xmm8
	shufps	$0x4e, %xmm10, %xmm9
	shufps	$0x4e, %xmm11, %xmm10
	shufps	$0x4e, %xmm12, %xmm11
#else
	SHUFPD_1 %xmm1, %xmm0
	SHUFPD_1 %xmm2, %xmm1
	SHUFPD_1 %xmm3, %xmm2
	SHUFPD_1 %xmm8, %xmm3

	SHUFPD_1 %xmm9,  %xmm8
	SHUFPD_1 %xmm10, %xmm9
	SHUFPD_1 %xmm11, %xmm10
	SHUFPD_1 %xmm12, %xmm11
#endif

	mulpd	ALPHA, %xmm0
	mulpd	ALPHA, %xmm1
	mulpd	ALPHA, %xmm2
	mulpd	ALPHA, %xmm3

	mulpd	ALPHA, %xmm8
	mulpd	ALPHA, %xmm9
	mulpd	ALPHA, %xmm10
	mulpd	ALPHA, %xmm11

	addpd	%xmm4, %xmm0
	movaps	 8 * SIZE(Y), %xmm4
	addpd	%xmm5, %xmm1
	movaps	10 * SIZE(Y), %xmm5
	addpd	%xmm6, %xmm2
	movaps	12 * SIZE(Y), %xmm6
	addpd	%xmm14, %xmm3
	movaps	14 * SIZE(Y), %xmm14

	movaps	%xmm0, 0 * SIZE(Y)
	movaps	%xmm1, 2 * SIZE(Y)
	movaps	%xmm2, 4 * SIZE(Y)
	movaps	%xmm3, 6 * SIZE(Y)

	movaps	%xmm12, %xmm0
	movaps	17 * SIZE(X), %xmm1
	movaps	19 * SIZE(X), %xmm2
	movaps	21 * SIZE(X), %xmm3
	movaps	31 * SIZE(X), %xmm12

	addpd	%xmm4, %xmm8
	addpd	%xmm5, %xmm9
	addpd	%xmm6, %xmm10
	addpd	%xmm14, %xmm11

	movaps	16 * SIZE(Y), %xmm4
	movaps	18 * SIZE(Y), %xmm5
	movaps	20 * SIZE(Y), %xmm6
	movaps	22 * SIZE(Y), %xmm14

	movaps	%xmm8,   8 * SIZE(Y)
	movaps	%xmm9,  10 * SIZE(Y)
	movaps	%xmm10, 12 * SIZE(Y)
	movaps	%xmm11, 14 * SIZE(Y)

	movaps	23 * SIZE(X), %xmm8
	movaps	25 * SIZE(X), %xmm9
	movaps	27 * SIZE(X), %xmm10
	movaps	29 * SIZE(X), %xmm11

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

.L22:
#ifdef PENRYN
	shufps	$0x4e, %xmm1, %xmm0
	shufps	$0x4e, %xmm2, %xmm1
	shufps	$0x4e, %xmm3, %xmm2
	shufps	$0x4e, %xmm8, %xmm3

	shufps	$0x4e, %xmm9,  %xmm8
	shufps	$0x4e, %xmm10, %xmm9
	shufps	$0x4e, %xmm11, %xmm10
	shufps	$0x4e, %xmm12, %xmm11
#else
	SHUFPD_1 %xmm1, %xmm0
	SHUFPD_1 %xmm2, %xmm1
	SHUFPD_1 %xmm3, %xmm2
	SHUFPD_1 %xmm8, %xmm3

	SHUFPD_1 %xmm9,  %xmm8
	SHUFPD_1 %xmm10, %xmm9
	SHUFPD_1 %xmm11, %xmm10
	SHUFPD_1 %xmm12, %xmm11
#endif

	mulpd	ALPHA, %xmm0
	mulpd	ALPHA, %xmm1
	mulpd	ALPHA, %xmm2
	mulpd	ALPHA, %xmm3

	mulpd	ALPHA, %xmm8
	mulpd	ALPHA, %xmm9
	mulpd	ALPHA, %xmm10
	mulpd	ALPHA, %xmm11

	addpd	%xmm4, %xmm0
	movaps	 8 * SIZE(Y), %xmm4
	addpd	%xmm5, %xmm1
	movaps	10 * SIZE(Y), %xmm5
	addpd	%xmm6, %xmm2
	movaps	12 * SIZE(Y), %xmm6
	addpd	%xmm14, %xmm3
	movaps	14 * SIZE(Y), %xmm14

	movaps	%xmm0, 0 * SIZE(Y)
	movaps	%xmm1, 2 * SIZE(Y)
	movaps	%xmm2, 4 * SIZE(Y)
	movaps	%xmm3, 6 * SIZE(Y)

	addpd	%xmm4, %xmm8
	addpd	%xmm5, %xmm9
	addpd	%xmm6, %xmm10
	addpd	%xmm14, %xmm11

	movaps	%xmm8,  8 * SIZE(Y)
	movaps	%xmm9, 10 * SIZE(Y)
	movaps	%xmm10, 12 * SIZE(Y)
	movaps	%xmm11, 14 * SIZE(Y)

	movaps	%xmm12, %xmm0
	addq	$16 * SIZE, X
	addq	$16 * SIZE, Y
	ALIGN_3

.L23:
	movq	M,  %rax
	andq	$8, %rax
	jle	.L24
	ALIGN_3

	movaps	1 * SIZE(X), %xmm1
	movaps	3 * SIZE(X), %xmm2
	movaps	5 * SIZE(X), %xmm3
	movaps	7 * SIZE(X), %xmm8

	movaps	0 * SIZE(Y), %xmm4
	movaps	2 * SIZE(Y), %xmm5
	movaps	4 * SIZE(Y), %xmm6
	movaps	6 * SIZE(Y), %xmm14

#ifdef PENRYN
	shufps	$0x4e, %xmm1, %xmm0
	shufps	$0x4e, %xmm2, %xmm1
	shufps	$0x4e, %xmm3, %xmm2
	shufps	$0x4e, %xmm8, %xmm3
#else
	SHUFPD_1 %xmm1, %xmm0
	SHUFPD_1 %xmm2, %xmm1
	SHUFPD_1 %xmm3, %xmm2
	SHUFPD_1 %xmm8, %xmm3
#endif

	mulpd	ALPHA, %xmm0
	mulpd	ALPHA, %xmm1
	mulpd	ALPHA, %xmm2
	mulpd	ALPHA, %xmm3

	addpd	%xmm4, %xmm0
	addpd	%xmm5, %xmm1
	addpd	%xmm6, %xmm2
	addpd	%xmm14, %xmm3

	movaps	%xmm0, 0 * SIZE(Y)
	movaps	%xmm1, 2 * SIZE(Y)
	movaps	%xmm2, 4 * SIZE(Y)
	movaps	%xmm3, 6 * SIZE(Y)

	movaps	%xmm8, %xmm0

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

.L24:
	movq	M,  %rax
	andq	$4, %rax
	jle	.L25
	ALIGN_3

	movaps	1 * SIZE(X), %xmm1
	movaps	3 * SIZE(X), %xmm2

	movaps	0 * SIZE(Y), %xmm4
	movaps	2 * SIZE(Y), %xmm5

#ifdef PENRYN
	shufps	$0x4e, %xmm1, %xmm0
	shufps	$0x4e, %xmm2, %xmm1
#else
	SHUFPD_1 %xmm1, %xmm0
	SHUFPD_1 %xmm2, %xmm1
#endif

	mulpd	ALPHA, %xmm0
	mulpd	ALPHA, %xmm1

	addpd	%xmm4, %xmm0
	addpd	%xmm5, %xmm1

	movaps	%xmm0, 0 * SIZE(Y)
	movaps	%xmm1, 2 * SIZE(Y)
	movaps	%xmm2, %xmm0

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

.L25:
	movq	M,  %rax
	andq	$2, %rax
	jle	.L26
	ALIGN_3

	movaps	1 * SIZE(X), %xmm1
	movaps	0 * SIZE(Y), %xmm4

#ifdef PENRYN
	shufps	$0x4e, %xmm1, %xmm0
#else
	SHUFPD_1 %xmm1, %xmm0
#endif
	mulpd	ALPHA, %xmm0
	addpd	%xmm4, %xmm0

	movaps	%xmm0, 0 * SIZE(Y)

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

.L26:
	movq	M,  %rax
	andq	$1, %rax
	jle	.L29
	ALIGN_3

	movsd	0 * SIZE(X), %xmm0
	mulsd	ALPHA, %xmm0
	addsd	0 * SIZE(Y), %xmm0

	movsd	%xmm0, 	0 * SIZE(Y)
	addq	$SIZE, Y
	ALIGN_3

.L29:
	xorq	%rax,%rax

	RESTOREREGISTERS

	ret
	ALIGN_3

.L40:
	movq	Y, YY
	movq	M,  %rax
	sarq	$3, %rax
	jle	.L45
	ALIGN_3

.L41:
	movsd	0 * SIZE(X), %xmm0
	addq	INCX, X
	movhpd	0 * SIZE(X), %xmm0
	addq	INCX, X
	mulpd	ALPHA, %xmm0

	movsd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	movhpd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	addpd	%xmm6, %xmm0

	movsd	0 * SIZE(X), %xmm1
	addq	INCX, X
	movhpd	0 * SIZE(X), %xmm1
	addq	INCX, X
	mulpd	ALPHA, %xmm1

	movsd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	movhpd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	addpd	%xmm6, %xmm1

	movsd	0 * SIZE(X), %xmm2
	addq	INCX, X
	movhpd	0 * SIZE(X), %xmm2
	addq	INCX, X
	mulpd	ALPHA, %xmm2

	movsd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	movhpd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	addpd	%xmm6, %xmm2

	movsd	0 * SIZE(X), %xmm3
	addq	INCX, X
	movhpd	0 * SIZE(X), %xmm3
	addq	INCX, X
	mulpd	ALPHA, %xmm3

	movsd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	movhpd	0 * SIZE(YY), %xmm6
	addq	INCY, YY
	addpd	%xmm6, %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

	decq	%rax
	jg	.L41
	ALIGN_3

.L45:
	movq	M,  %rax
	andq	$7, %rax
	jle	.L47
	ALIGN_3

.L46:
	movsd	(X), %xmm0
	addq	INCX, X
	mulsd	ALPHA, %xmm0
	addsd	(Y), %xmm0
	movsd	%xmm0, (Y)
	addq	INCY, Y
	decq	%rax
	jg	.L46
	ALIGN_3

.L47:
	xorq	%rax, %rax

	RESTOREREGISTERS

	ret

	EPILOGUE