📄 zgemv_n_sse2.s

📁 Optimized GotoBLAS libraries
💻 S
📖 第 1 页 / 共 5 页
字号:
#ifdef PENTIUM4	PREFETCHW	PREFETCHSIZE * SIZE(Y1)#endif	addpd	 %xmm8,  %xmm4	MOVDDUP( 8 * SIZE, A1, %xmm8)	mulpd	 %xmm3, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(10 * SIZE, A1, %xmm10)	mulpd	 %xmm3, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(12 * SIZE, A1, %xmm12)	mulpd	 %xmm3, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(14 * SIZE, A1, %xmm14)	mulpd	 %xmm3, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP( 9 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(11 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(13 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(15 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 8 * SIZE(Y1), %xmm4	movapd	 %xmm5,  2 * SIZE(Y1)	movapd	10 * SIZE(Y1), %xmm5	movapd	 %xmm6,  4 * SIZE(Y1)	movapd	12 * SIZE(Y1), %xmm6	movapd	 %xmm7,  6 * SIZE(Y1)	movapd	14 * SIZE(Y1), %xmm7#if defined(OPTERON) || defined(CORE2) || defined(PENRYN)	PREFETCH	(PREFETCHSIZE + 8) * SIZE(A1)#endif#ifdef PENTIUM4	PREFETCH	(PREFETCHSIZE +  0) * SIZE(A1, LDA, 1)#endif	addpd	 %xmm8,  %xmm4	MOVDDUP2( 8 * SIZE, (A1, LDA, 1), %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP2(10 * SIZE, (A1, LDA, 1), %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP2(12 * SIZE, (A1, LDA, 1), %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP2(14 * SIZE, (A1, LDA, 1), %xmm14)	mulpd	 %xmm1, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP2( 9 * SIZE, (A1, LDA, 1), %xmm9)	mulpd	 %xmm2, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP2(11 * SIZE, (A1, LDA, 1), %xmm11)	mulpd	 %xmm2, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP2(13 * SIZE, (A1, LDA, 1), %xmm13)	mulpd	 %xmm2, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP2(15 * SIZE, (A1, LDA, 1), %xmm15)	mulpd	 %xmm2, %xmm14	addpd	 %xmm8,  %xmm4	MOVDDUP(16 * SIZE, A1, %xmm8)	mulpd	 %xmm3, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(18 * SIZE, A1, %xmm10)	mulpd	 %xmm3, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(20 * SIZE, A1, %xmm12)	mulpd	 %xmm3, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(22 * SIZE, A1, %xmm14)	mulpd	 %xmm3, %xmm15#if defined(OPTERON) || defined(CORE2) || defined(PENRYN)	PREFETCH	(PREFETCHSIZE + 8) * SIZE(A1, LDA, 1)#endif	ADD	 %xmm9,  %xmm4	MOVDDUP(17 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(19 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(21 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(23 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  8 * SIZE(Y1)	movapd	16 * SIZE(Y1), %xmm4	movapd	 %xmm5, 10 * SIZE(Y1)	movapd	18 * SIZE(Y1), %xmm5	movapd	 %xmm6, 12 * SIZE(Y1)	movapd	20 * SIZE(Y1), %xmm6	movapd	 %xmm7, 14 * SIZE(Y1)	movapd	22 * SIZE(Y1), %xmm7	subq	$-16 * SIZE, A1	subq	$-16 * SIZE, Y1	decq	 I	jg	.L24	ALIGN_3.L25:	addpd	 %xmm8,  %xmm4	MOVDDUP2( 0 * SIZE, (A1, LDA, 1), %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP2( 2 * SIZE, (A1, LDA, 1), %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP2( 4 * SIZE, (A1, LDA, 1), %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP2( 6 * SIZE, (A1, LDA, 1), %xmm14)	mulpd	 %xmm1, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP2( 1 * SIZE, (A1, LDA, 1), %xmm9)	mulpd	 %xmm2, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP2( 3 * SIZE, (A1, LDA, 1), %xmm11)	mulpd	 %xmm2, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP2( 5 * SIZE, (A1, LDA, 1), %xmm13)	mulpd	 %xmm2, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP2( 7 * SIZE, (A1, LDA, 1), %xmm15)	mulpd	 %xmm2, %xmm14	addpd	 %xmm8,  %xmm4	MOVDDUP( 8 * SIZE, A1, %xmm8)	mulpd	 %xmm3, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(10 * SIZE, A1, %xmm10)	mulpd	 %xmm3, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(12 * SIZE, A1, %xmm12)	mulpd	 %xmm3, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(14 * SIZE, A1, %xmm14)	mulpd	 %xmm3, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP( 9 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(11 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(13 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(15 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 8 * SIZE(Y1), %xmm4	movapd	 %xmm5,  2 * SIZE(Y1)	movapd	10 * SIZE(Y1), %xmm5	movapd	 %xmm6,  4 * SIZE(Y1)	movapd	12 * SIZE(Y1), %xmm6	movapd	 %xmm7,  6 * SIZE(Y1)	movapd	14 * SIZE(Y1), %xmm7	addpd	 %xmm8,  %xmm4	MOVDDUP2( 8 * SIZE, (A1, LDA, 1), %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP2(10 * SIZE, (A1, LDA, 1), %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP2(12 * SIZE, (A1, LDA, 1), %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP2(14 * SIZE, (A1, LDA, 1), %xmm14)	mulpd	 %xmm1, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP2( 9 * SIZE, (A1, LDA, 1), %xmm9)	mulpd	 %xmm2, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP2(11 * SIZE, (A1, LDA, 1), %xmm11)	mulpd	 %xmm2, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP2(13 * SIZE, (A1, LDA, 1), %xmm13)	mulpd	 %xmm2, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP2(15 * SIZE, (A1, LDA, 1), %xmm15)	mulpd	 %xmm2, %xmm14	addpd	 %xmm8,  %xmm4	mulpd	 %xmm3, %xmm9	addpd	 %xmm10, %xmm5	mulpd	 %xmm3, %xmm11	addpd	 %xmm12, %xmm6	mulpd	 %xmm3, %xmm13	addpd	 %xmm14, %xmm7	mulpd	 %xmm3, %xmm15	ADD	 %xmm9,  %xmm4	ADD	 %xmm11, %xmm5	ADD	 %xmm13, %xmm6	ADD	 %xmm15, %xmm7	movapd	 %xmm4,  8 * SIZE(Y1)	movapd	 %xmm5, 10 * SIZE(Y1)	movapd	 %xmm6, 12 * SIZE(Y1)	movapd	 %xmm7, 14 * SIZE(Y1)	subq	$-16 * SIZE, A1	subq	$-16 * SIZE, Y1	ALIGN_3.L26:	testq	$4, M	je	.L27	MOVDDUP( 0 * SIZE, A1, %xmm8)	MOVDDUP( 1 * SIZE, A1, %xmm9)	MOVDDUP( 2 * SIZE, A1, %xmm10)	MOVDDUP( 3 * SIZE, A1, %xmm11)	MOVDDUP( 4 * SIZE, A1, %xmm12)	MOVDDUP( 5 * SIZE, A1, %xmm13)	MOVDDUP( 6 * SIZE, A1, %xmm14)	MOVDDUP( 7 * SIZE, A1, %xmm15)	movapd	 0 * SIZE(Y1), %xmm4	movapd	 2 * SIZE(Y1), %xmm5	movapd	 4 * SIZE(Y1), %xmm6	movapd	 6 * SIZE(Y1), %xmm7	mulpd	 %xmm0, %xmm8	mulpd	 %xmm0, %xmm10	mulpd	 %xmm0, %xmm12	mulpd	 %xmm0, %xmm14	addpd	 %xmm8,  %xmm4	MOVDDUP2( 0 * SIZE, (A1, LDA, 1), %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP2( 2 * SIZE, (A1, LDA, 1), %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP2( 4 * SIZE, (A1, LDA, 1), %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP2( 6 * SIZE, (A1, LDA, 1), %xmm14)	mulpd	 %xmm1, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP2( 1 * SIZE, (A1, LDA, 1), %xmm9)	mulpd	 %xmm2, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP2( 3 * SIZE, (A1, LDA, 1), %xmm11)	mulpd	 %xmm2, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP2( 5 * SIZE, (A1, LDA, 1), %xmm13)	mulpd	 %xmm2, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP2( 7 * SIZE, (A1, LDA, 1), %xmm15)	mulpd	 %xmm2, %xmm14	addpd	 %xmm8,  %xmm4	mulpd	 %xmm3, %xmm9	addpd	 %xmm10, %xmm5	mulpd	 %xmm3, %xmm11	addpd	 %xmm12, %xmm6	mulpd	 %xmm3, %xmm13	addpd	 %xmm14, %xmm7	mulpd	 %xmm3, %xmm15	ADD	 %xmm9,  %xmm4	ADD	 %xmm11, %xmm5	ADD	 %xmm13, %xmm6	ADD	 %xmm15, %xmm7	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 %xmm5,  2 * SIZE(Y1)	movapd	 %xmm6,  4 * SIZE(Y1)	movapd	 %xmm7,  6 * SIZE(Y1)	addq	 $8 * SIZE, A1	addq	 $8 * SIZE, Y1	ALIGN_3.L27:	testq	$2, M	je	.L28	MOVDDUP( 0 * SIZE, A1, %xmm8)	MOVDDUP( 1 * SIZE, A1, %xmm9)	MOVDDUP( 2 * SIZE, A1, %xmm10)	MOVDDUP( 3 * SIZE, A1, %xmm11)	MOVDDUP2( 0 * SIZE, (A1, LDA, 1), %xmm12)	MOVDDUP2( 1 * SIZE, (A1, LDA, 1), %xmm13)	MOVDDUP2( 2 * SIZE, (A1, LDA, 1), %xmm14)	MOVDDUP2( 3 * SIZE, (A1, LDA, 1), %xmm15)	movapd	 0 * SIZE(Y1), %xmm4	movapd	 2 * SIZE(Y1), %xmm5	mulpd	 %xmm0, %xmm8	mulpd	 %xmm1, %xmm9	mulpd	 %xmm0, %xmm10	mulpd	 %xmm1, %xmm11	mulpd	 %xmm2, %xmm12	addpd	 %xmm8,  %xmm4	mulpd	 %xmm3, %xmm13	addpd	 %xmm10, %xmm5	mulpd	 %xmm2, %xmm14	ADD	 %xmm9,  %xmm4	mulpd	 %xmm3, %xmm15	ADD	 %xmm11, %xmm5	addpd	 %xmm12, %xmm4	addpd	 %xmm14, %xmm5	ADD	 %xmm13, %xmm4	ADD	 %xmm15, %xmm5	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 %xmm5,  2 * SIZE(Y1)	addq	 $4 * SIZE, A1	addq	 $4 * SIZE, Y1	ALIGN_3.L28:	testq	$1, M	je	.L30	MOVDDUP( 0 * SIZE, A1, %xmm8)	MOVDDUP( 1 * SIZE, A1, %xmm9)	MOVDDUP2( 0 * SIZE, (A1, LDA, 1), %xmm12)	MOVDDUP2( 1 * SIZE, (A1, LDA, 1), %xmm13)	movapd	 0 * SIZE(Y1), %xmm4	mulpd	 %xmm0, %xmm8	mulpd	 %xmm1, %xmm9	mulpd	 %xmm2, %xmm12	mulpd	 %xmm3, %xmm13	addpd	 %xmm8,  %xmm4	ADD	 %xmm9,  %xmm4	addpd	 %xmm12, %xmm4	ADD	 %xmm13, %xmm4	movapd	 %xmm4,  0 * SIZE(Y1)	ALIGN_3.L30:	testq	$1, N	je	.L995	movq	YY, Y1	movq	A,  A1	MOVDDUP(0 * SIZE, X, %xmm0)	MOVDDUP(1 * SIZE, X, %xmm1)	addq	INCX, X	movlpd	  0 + ALPHA_R,   %xmm13	movhpd	  8 + ALPHA_R,   %xmm13	movlpd	  0 + ALPHA_I,   %xmm14	movhpd	  8 + ALPHA_I,   %xmm14	movlpd	  0 + COMP_MASK, %xmm15	movhpd	  8 + COMP_MASK, %xmm15	mulpd	  %xmm13, %xmm0	mulpd	  %xmm14, %xmm1	ADDX	  %xmm1,  %xmm0	movapd	  %xmm0, %xmm1	SHUFPD_1 %xmm1, %xmm1	xorpd	  %xmm15, %xmm1	movq	M,  I	sarq	$3, I	jle	.L36	MOVDDUP( 0 * SIZE, A1, %xmm8)	MOVDDUP( 2 * SIZE, A1, %xmm10)	MOVDDUP( 4 * SIZE, A1, %xmm12)	MOVDDUP( 6 * SIZE, A1, %xmm14)	MOVDDUP( 1 * SIZE, A1, %xmm9)	MOVDDUP( 3 * SIZE, A1, %xmm11)	MOVDDUP( 5 * SIZE, A1, %xmm13)	MOVDDUP( 7 * SIZE, A1, %xmm15)	movapd	 0 * SIZE(Y1), %xmm4	movapd	 2 * SIZE(Y1), %xmm5	movapd	 4 * SIZE(Y1), %xmm6	movapd	 6 * SIZE(Y1), %xmm7	mulpd	 %xmm0, %xmm8	mulpd	 %xmm0, %xmm10	mulpd	 %xmm0, %xmm12	mulpd	 %xmm0, %xmm14	decq	 I	jle	 .L35	ALIGN_3.L34:#if defined(OPTERON) || defined(CORE2) || defined(PENRYN)	PREFETCH	PREFETCHSIZE * SIZE(A1)#endif#ifdef PENTIUM4	PREFETCH	PREFETCHSIZE * SIZE(A1)#endif	addpd	 %xmm8,  %xmm4	MOVDDUP( 8 * SIZE, A1, %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(10 * SIZE, A1, %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(12 * SIZE, A1, %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(14 * SIZE, A1, %xmm14)	mulpd	 %xmm1, %xmm15#ifdef OPTERON	PREFETCHW	PREFETCHSIZE * SIZE(Y1)#endif	ADD	 %xmm9,  %xmm4	MOVDDUP( 9 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(11 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(13 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(15 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 8 * SIZE(Y1), %xmm4	movapd	 %xmm5,  2 * SIZE(Y1)	movapd	10 * SIZE(Y1), %xmm5	movapd	 %xmm6,  4 * SIZE(Y1)	movapd	12 * SIZE(Y1), %xmm6	movapd	 %xmm7,  6 * SIZE(Y1)	movapd	14 * SIZE(Y1), %xmm7#if defined(OPTERON) || defined(CORE2) || defined(PENRYN)	PREFETCH	(PREFETCHSIZE + 8) * SIZE(A1)#endif#ifdef PENTIUM4	PREFETCHW	PREFETCHSIZE * SIZE(Y1)#endif	addpd	 %xmm8,  %xmm4	MOVDDUP(16 * SIZE, A1, %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(18 * SIZE, A1, %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(20 * SIZE, A1, %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(22 * SIZE, A1, %xmm14)	mulpd	 %xmm1, %xmm15#ifdef OPTERON	PREFETCHW	(PREFETCHSIZE + 8) * SIZE(Y1)#endif	ADD	 %xmm9,  %xmm4	MOVDDUP(17 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(19 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(21 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(23 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  8 * SIZE(Y1)	movapd	16 * SIZE(Y1), %xmm4	movapd	 %xmm5, 10 * SIZE(Y1)	movapd	18 * SIZE(Y1), %xmm5	movapd	 %xmm6, 12 * SIZE(Y1)	movapd	20 * SIZE(Y1), %xmm6	movapd	 %xmm7, 14 * SIZE(Y1)	movapd	22 * SIZE(Y1), %xmm7	subq	$-16 * SIZE, A1	subq	$-16 * SIZE, Y1	decq	 I	jg	.L34	ALIGN_3.L35:	addpd	 %xmm8,  %xmm4	MOVDDUP( 8 * SIZE, A1, %xmm8)	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	MOVDDUP(10 * SIZE, A1, %xmm10)	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	MOVDDUP(12 * SIZE, A1, %xmm12)	mulpd	 %xmm1, %xmm13	addpd	 %xmm14, %xmm7	MOVDDUP(14 * SIZE, A1, %xmm14)	mulpd	 %xmm1, %xmm15	ADD	 %xmm9,  %xmm4	MOVDDUP( 9 * SIZE, A1, %xmm9)	mulpd	 %xmm0, %xmm8	ADD	 %xmm11, %xmm5	MOVDDUP(11 * SIZE, A1, %xmm11)	mulpd	 %xmm0, %xmm10	ADD	 %xmm13, %xmm6	MOVDDUP(13 * SIZE, A1, %xmm13)	mulpd	 %xmm0, %xmm12	ADD	 %xmm15, %xmm7	MOVDDUP(15 * SIZE, A1, %xmm15)	mulpd	 %xmm0, %xmm14	movapd	 %xmm4,  0 * SIZE(Y1)	movapd	 8 * SIZE(Y1), %xmm4	movapd	 %xmm5,  2 * SIZE(Y1)	movapd	10 * SIZE(Y1), %xmm5	movapd	 %xmm6,  4 * SIZE(Y1)	movapd	12 * SIZE(Y1), %xmm6	movapd	 %xmm7,  6 * SIZE(Y1)	movapd	14 * SIZE(Y1), %xmm7	addpd	 %xmm8,  %xmm4	mulpd	 %xmm1, %xmm9	addpd	 %xmm10, %xmm5	mulpd	 %xmm1, %xmm11	addpd	 %xmm12, %xmm6	mulpd	 %xmm1, %xmm13
⌨️ 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 + -