📄 gemv_t_sse2_core2.s

📁 Optimized GotoBLAS libraries
💻 S
📖 第 1 页 / 共 5 页
字号:
上一页 1 2 3 45
	addpd	%xmm9, %xmm1	movapd	 5 * SIZE(AO2), %xmm9	mulpd	%xmm13, %xmm10	addpd	%xmm10, %xmm2	movapd	 6 * SIZE(AO1, LDA, 2), %xmm10	SHUFPD_1 %xmm7,  %xmm11	mulpd	%xmm13, %xmm11	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm11, %xmm3	movapd	 5 * SIZE(AO2, LDA, 2), %xmm11	mulpd	%xmm12, %xmm4	addpd	%xmm4, %xmm0	SHUFPD_1 %xmm9,  %xmm5	mulpd	%xmm12, %xmm5	addpd	%xmm5, %xmm1	movapd	 7 * SIZE(AO2), %xmm5	mulpd	%xmm12, %xmm6	addpd	%xmm6, %xmm2	SHUFPD_1 %xmm11, %xmm7	mulpd	%xmm12, %xmm7	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm7, %xmm3	movapd	 7 * SIZE(AO2, LDA, 2), %xmm7	mulpd	%xmm13, %xmm8	addpd	%xmm8, %xmm0	SHUFPD_1 %xmm5,  %xmm9	mulpd	%xmm13, %xmm9	addpd	%xmm9, %xmm1	mulpd	%xmm13, %xmm10	addpd	%xmm10, %xmm2	SHUFPD_1 %xmm7,  %xmm11	mulpd	%xmm13, %xmm11	addpd	%xmm11, %xmm3	movapd	10 * SIZE(BO), %xmm13	subq	$-8 * SIZE, AO1	subq	$-8 * SIZE, AO2	subq	$-8 * SIZE, BO.L125:	movq	MIN_M, I	andq	$4,  I	je	.L126	movapd	 0 * SIZE(AO1), %xmm4	movapd	 0 * SIZE(AO1, LDA, 2), %xmm6	movapd	 2 * SIZE(AO1), %xmm8	movapd	 1 * SIZE(AO2), %xmm9	movapd	 2 * SIZE(AO1, LDA, 2), %xmm10	movapd	 1 * SIZE(AO2, LDA, 2), %xmm11	mulpd	%xmm12, %xmm4	addpd	%xmm4, %xmm0	SHUFPD_1 %xmm9,  %xmm5	mulpd	%xmm12, %xmm5	addpd	%xmm5, %xmm1	movapd	 3 * SIZE(AO2), %xmm5	mulpd	%xmm12, %xmm6	addpd	%xmm6, %xmm2	SHUFPD_1 %xmm11, %xmm7	mulpd	%xmm12, %xmm7	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm7, %xmm3	movapd	 3 * SIZE(AO2, LDA, 2), %xmm7	mulpd	%xmm13, %xmm8	addpd	%xmm8, %xmm0	SHUFPD_1 %xmm5,  %xmm9	mulpd	%xmm13, %xmm9	addpd	%xmm9, %xmm1	mulpd	%xmm13, %xmm10	addpd	%xmm10, %xmm2	SHUFPD_1 %xmm7, %xmm11	mulpd	%xmm13, %xmm11	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm11, %xmm3	subq	$-4 * SIZE, AO1	subq	$-4 * SIZE, AO2	subq	$-4 * SIZE, BO.L126:	movq	MIN_M, I	andq	$2,  I	je	.L127	movapd	 0 * SIZE(AO1), %xmm4	movapd	 1 * SIZE(AO2), %xmm9	movapd	 0 * SIZE(AO1, LDA, 2), %xmm6	movapd	 1 * SIZE(AO2, LDA, 2), %xmm11	mulpd	%xmm12, %xmm4	addpd	%xmm4, %xmm0	SHUFPD_1 %xmm9,  %xmm5	mulpd	%xmm12, %xmm5	addpd	%xmm5, %xmm1	mulpd	%xmm12, %xmm6	addpd	%xmm6, %xmm2	SHUFPD_1 %xmm11, %xmm7	mulpd	%xmm12, %xmm7	movapd	 2 * SIZE(BO), %xmm12	addpd	%xmm7, %xmm3	movapd	%xmm9, %xmm5	movapd	%xmm7, %xmm11	subq	$-2 * SIZE, AO1	subq	$-2 * SIZE, AO2	subq	$-2 * SIZE, BO.L127:	movq	MIN_M, I	andq	$1,  I	je	.L129	movsd	0 * SIZE(AO1), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm0	movsd	0 * SIZE(AO2), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm1	movsd	0 * SIZE(AO1, LDA, 2), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm2	movsd	0 * SIZE(AO2, LDA, 2), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm3.L129:#ifdef HAVE_SSE3	haddpd	%xmm0, %xmm0	haddpd	%xmm1, %xmm1	haddpd	%xmm2, %xmm2	haddpd	%xmm3, %xmm3#else	movapd	%xmm0, %xmm9	movapd	%xmm1, %xmm10	movapd	%xmm2, %xmm11	movapd	%xmm3, %xmm12	unpckhpd %xmm0, %xmm0	unpckhpd %xmm1, %xmm1	unpckhpd %xmm2, %xmm2	unpckhpd %xmm3, %xmm3	addsd	 %xmm9,  %xmm0	addsd	 %xmm10, %xmm1	addsd	 %xmm11, %xmm2	addsd	 %xmm12, %xmm3#endif	mulsd	ALPHA, %xmm0	mulsd	ALPHA, %xmm1	mulsd	ALPHA, %xmm2	mulsd	ALPHA, %xmm3	movq	CO, TEMP	addsd	(TEMP), %xmm0	addq	INCY, TEMP	addsd	(TEMP), %xmm1	addq	INCY, TEMP	addsd	(TEMP), %xmm2	addq	INCY, TEMP	addsd	(TEMP), %xmm3	movsd	%xmm0, (CO)	addq	INCY, CO	movsd	%xmm1, (CO)	addq	INCY, CO	movsd	%xmm2, (CO)	addq	INCY, CO	movsd	%xmm3, (CO)	addq	INCY, CO	decq	J	jg	.L121	ALIGN_3.L130:	movq	N,  J	andq	$2, J	jle	.L140	ALIGN_3.L131:	movq	A, AO1	leaq	(A, LDA, 1), AO2	leaq	(A, LDA, 2), A	movq	BUFFER, BO	movapd	 0 * SIZE(BO), %xmm12	pxor	%xmm0, %xmm0	movapd	 2 * SIZE(BO), %xmm13	pxor	%xmm1, %xmm1	pxor	%xmm2, %xmm2	pxor	%xmm3, %xmm3#if defined(CORE2) || defined(PENRYN)	PREFETCHW	4 * SIZE(CO)#endif	movapd	-1 * SIZE(AO2), %xmm5	movq	MIN_M, I	sarq	$4,  I	jle	.L134	movapd	 0 * SIZE(AO1), %xmm4	movapd	 2 * SIZE(AO1), %xmm8	movapd	 1 * SIZE(AO2), %xmm9	decq	I	jle	.L133	ALIGN_3.L132:	PREFETCH	PREFETCHSIZE * SIZE(AO1)	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 3 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	 5 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	 8 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 7 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	mulpd	%xmm13, %xmm9	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	10 * SIZE(AO1), %xmm8	movapd	 9 * SIZE(AO2), %xmm9	PREFETCH	PREFETCHSIZE * SIZE(AO2)	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	12 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	12 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	11 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	14 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	14 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	13 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	16 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	16 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	15 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	18 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	18 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	17 * SIZE(AO2), %xmm9	subq	$-16 * SIZE, AO1	subq	$-16 * SIZE, AO2	subq	$-16 * SIZE, BO	decq	I	jg	.L132	ALIGN_3.L133:	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 3 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	 5 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	 8 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 7 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	mulpd	%xmm13, %xmm9	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	10 * SIZE(AO1), %xmm8	movapd	 9 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	12 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	12 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	11 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	14 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	14 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	13 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	mulpd	%xmm12, %xmm5	movapd	16 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	15 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	mulpd	%xmm13, %xmm9	movapd	18 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	subq	$-16 * SIZE, AO1	subq	$-16 * SIZE, AO2	subq	$-16 * SIZE, BO.L134:	movq	MIN_M, I	andq	$8,  I	je	.L135	movapd	 0 * SIZE(AO1), %xmm4	movapd	 1 * SIZE(AO2), %xmm9	movapd	 2 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm12, %xmm5	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 3 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm13, %xmm9	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	movapd	 5 * SIZE(AO2), %xmm9	SHUFPD_1 %xmm9, %xmm5	mulpd	%xmm12, %xmm4	addpd	%xmm4, %xmm0	mulpd	%xmm12, %xmm5	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 7 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	mulpd	%xmm13, %xmm9	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	subq	$-8 * SIZE, AO1	subq	$-8 * SIZE, AO2	subq	$-8 * SIZE, BO.L135:	movq	MIN_M, I	andq	$4,  I	je	.L136	movapd	 0 * SIZE(AO1), %xmm4	movapd	 1 * SIZE(AO2), %xmm9	movapd	 2 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	mulpd	%xmm12, %xmm5	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	movapd	 3 * SIZE(AO2), %xmm5	mulpd	%xmm13, %xmm8	SHUFPD_1 %xmm5, %xmm9	addpd	%xmm8, %xmm2	mulpd	%xmm13, %xmm9	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm9, %xmm3	subq	$-4 * SIZE, AO1	subq	$-4 * SIZE, AO2	subq	$-4 * SIZE, BO.L136:	movq	MIN_M, I	andq	$2,  I	je	.L137	movapd	 0 * SIZE(AO1), %xmm4	movapd	 1 * SIZE(AO2), %xmm9	mulpd	%xmm12, %xmm4	SHUFPD_1 %xmm9, %xmm5	addpd	%xmm4, %xmm0	mulpd	%xmm12, %xmm5	movapd	 2 * SIZE(BO), %xmm12	addpd	%xmm5, %xmm1	subq	$-2 * SIZE, AO1	subq	$-2 * SIZE, AO2	subq	$-2 * SIZE, BO.L137:	movq	MIN_M, I	andq	$1,  I	je	.L139	movsd	0 * SIZE(AO1), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm0	movsd	0 * SIZE(AO2), %xmm8	mulsd	%xmm12, %xmm8	addsd	%xmm8, %xmm1.L139:	addpd	%xmm2, %xmm0	addpd	%xmm3, %xmm1#ifdef HAVE_SSE3	haddpd	%xmm0, %xmm0	haddpd	%xmm1, %xmm1#else	movapd	%xmm0, %xmm9	movapd	%xmm1, %xmm10	unpckhpd %xmm0, %xmm0	unpckhpd %xmm1, %xmm1	addsd	 %xmm9,  %xmm0	addsd	 %xmm10, %xmm1#endif	mulsd	ALPHA, %xmm0	mulsd	ALPHA, %xmm1	movq	CO, TEMP	addsd	(TEMP), %xmm0	addq	INCY, TEMP	addsd	(TEMP), %xmm1	addq	INCY, TEMP	movsd	%xmm0, (CO)	addq	INCY, CO	movsd	%xmm1, (CO)	addq	INCY, CO	ALIGN_3.L140:	movq	N,  J	andq	$1, J	jle	.L149	ALIGN_3.L141:	movq	A, AO1	addq	LDA, A	movq	BUFFER, BO	movapd	 0 * SIZE(BO), %xmm12	pxor	%xmm0, %xmm0	movapd	 2 * SIZE(BO), %xmm13	pxor	%xmm1, %xmm1	pxor	%xmm2, %xmm2	pxor	%xmm3, %xmm3#if defined(CORE2) || defined(PENRYN)	PREFETCHW	4 * SIZE(CO)#endif	movq	MIN_M, I	sarq	$4,  I	jle	.L144	movapd	 0 * SIZE(AO1), %xmm4	movapd	 2 * SIZE(AO1), %xmm8	decq	I	jle	.L143	ALIGN_3.L142:	PREFETCH	PREFETCHSIZE * SIZE(AO1)	mulpd	%xmm12, %xmm4	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm0	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	 8 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm0	movapd	10 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	12 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	12 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	14 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm0	movapd	14 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	16 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	16 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	18 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm0	movapd	18 * SIZE(AO1), %xmm8	subq	$-16 * SIZE, AO1	subq	$-16 * SIZE, BO	decq	I	jg	.L142	ALIGN_3.L143:	mulpd	%xmm12, %xmm4	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm1	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm2	movapd	 8 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm3	movapd	10 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	12 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	12 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	14 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm1	movapd	14 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	16 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm2	subq	$-16 * SIZE, AO1	mulpd	%xmm13, %xmm8	movapd	18 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm3	subq	$-16 * SIZE, BO.L144:	movq	MIN_M, I	andq	$8,  I	je	.L145	movapd	 0 * SIZE(AO1), %xmm4	movapd	 2 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	 4 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm0	movapd	 4 * SIZE(AO1), %xmm4	mulpd	%xmm13, %xmm8	movapd	 6 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm1	movapd	 6 * SIZE(AO1), %xmm8	mulpd	%xmm12, %xmm4	movapd	 8 * SIZE(BO), %xmm12	addpd	%xmm4, %xmm2	subq	$-8 * SIZE, AO1	mulpd	%xmm13, %xmm8	movapd	10 * SIZE(BO), %xmm13	addpd	%xmm8, %xmm3	subq	$-8 * SIZE, BO.L145:	movq	MIN_M, I	andq	$4,  I	je	.L146	movapd	 0 * SIZE(AO1), %xmm4	movapd	 2 * SIZE(AO1), %xmm8	subq	$-4 * SIZE, AO1	mulpd	%xmm12, %xmm4
上一页 1 2 3 45
⌨️ 快捷键说明

复制代码 Ctrl + C
搜索代码 Ctrl + F
全屏模式 F11
切换主题 Ctrl + Shift + D
显示快捷键 ?
增大字号 Ctrl + =
减小字号 Ctrl + -