📄 ztrsm_kernel_lt_2x2_core2.s

📁 Optimized GotoBLAS libraries
💻 S
📖 第 1 页 / 共 3 页
字号:
	xorpd	 %xmm7, %xmm12	xorpd	 %xmm7, %xmm14	mulpd	 %xmm2, %xmm8	mulpd	 %xmm2, %xmm10	mulpd	 %xmm3, %xmm12	mulpd	 %xmm3, %xmm14	subpd	 %xmm8, %xmm13	subpd	 %xmm10, %xmm15	subpd	 %xmm12, %xmm13	subpd	 %xmm14, %xmm15	pshufd	$0x4e, %xmm13, %xmm12	pshufd	$0x4e, %xmm15, %xmm14	xorpd	 %xmm7, %xmm12	xorpd	 %xmm7, %xmm14	mulpd	 %xmm4, %xmm13	mulpd	 %xmm5, %xmm12	mulpd	 %xmm4, %xmm15	mulpd	 %xmm5, %xmm14	addpd	 %xmm12, %xmm13	addpd	 %xmm14, %xmm15#endif#ifdef RN	movddup	-16 * SIZE(B), %xmm0	movddup	-15 * SIZE(B), %xmm1	movddup	-14 * SIZE(B), %xmm2	movddup	-13 * SIZE(B), %xmm3	movddup	-10 * SIZE(B), %xmm4	movddup	 -9 * SIZE(B), %xmm5	pshufd	$0x4e, %xmm9, %xmm8	pshufd	$0x4e, %xmm13, %xmm12	xorpd	 %xmm7, %xmm8	xorpd	 %xmm7, %xmm12	mulpd	 %xmm0, %xmm9	mulpd	 %xmm1, %xmm8	mulpd	 %xmm0, %xmm13	mulpd	 %xmm1, %xmm12	addpd	 %xmm8, %xmm9	addpd	 %xmm12, %xmm13	movapd	 %xmm9, %xmm8	movapd	 %xmm13, %xmm10	pshufd	 $0x4e, %xmm9, %xmm12	pshufd	 $0x4e, %xmm13, %xmm14	xorpd	 %xmm7, %xmm12	xorpd	 %xmm7, %xmm14	mulpd	 %xmm2, %xmm8	mulpd	 %xmm2, %xmm10	mulpd	 %xmm3, %xmm12	mulpd	 %xmm3, %xmm14	subpd	 %xmm8, %xmm11	subpd	 %xmm10, %xmm15	subpd	 %xmm12, %xmm11	subpd	 %xmm14, %xmm15	pshufd	$0x4e, %xmm11, %xmm10	pshufd	$0x4e, %xmm15, %xmm14	xorpd	 %xmm7, %xmm10	xorpd	 %xmm7, %xmm14	mulpd	 %xmm4, %xmm11	mulpd	 %xmm5, %xmm10	mulpd	 %xmm4, %xmm15	mulpd	 %xmm5, %xmm14	addpd	 %xmm10, %xmm11	addpd	 %xmm14, %xmm15#endif#ifdef RT	movddup	-10 * SIZE(B), %xmm0	movddup	 -9 * SIZE(B), %xmm1	movddup	-12 * SIZE(B), %xmm2	movddup	-11 * SIZE(B), %xmm3	movddup	-16 * SIZE(B), %xmm4	movddup	-15 * SIZE(B), %xmm5	pshufd	$0x4e, %xmm11, %xmm10	pshufd	$0x4e, %xmm15, %xmm14	xorpd	 %xmm7, %xmm10	xorpd	 %xmm7, %xmm14	mulpd	 %xmm0, %xmm11	mulpd	 %xmm1, %xmm10	mulpd	 %xmm0, %xmm15	mulpd	 %xmm1, %xmm14	addpd	 %xmm10, %xmm11	addpd	 %xmm14, %xmm15	movapd	 %xmm11, %xmm8	movapd	 %xmm15, %xmm10	pshufd	 $0x4e, %xmm11, %xmm12	pshufd	 $0x4e, %xmm15, %xmm14	xorpd	 %xmm7, %xmm12	xorpd	 %xmm7, %xmm14	mulpd	 %xmm2, %xmm8	mulpd	 %xmm2, %xmm10	mulpd	 %xmm3, %xmm12	mulpd	 %xmm3, %xmm14	subpd	 %xmm8, %xmm9	subpd	 %xmm10, %xmm13	subpd	 %xmm12, %xmm9	subpd	 %xmm14, %xmm13	pshufd	$0x4e, %xmm9, %xmm8	pshufd	$0x4e, %xmm13, %xmm12	xorpd	 %xmm7, %xmm8	xorpd	 %xmm7, %xmm12	mulpd	 %xmm4, %xmm9	mulpd	 %xmm5, %xmm8	mulpd	 %xmm4, %xmm13	mulpd	 %xmm5, %xmm12	addpd	 %xmm8, %xmm9	addpd	 %xmm12, %xmm13#endif#ifdef LN	subq	$4 * SIZE, CO1	subq	$4 * SIZE, CO2#endif	movsd	%xmm9,    0 * SIZE(CO1)	movhpd	%xmm9,    1 * SIZE(CO1)	movsd	%xmm13,   2 * SIZE(CO1)	movhpd	%xmm13,   3 * SIZE(CO1)	movsd	%xmm11,   0 * SIZE(CO2)	movhpd	%xmm11,   1 * SIZE(CO2)	movsd	%xmm15,   2 * SIZE(CO2)	movhpd	%xmm15,   3 * SIZE(CO2)#if defined(LN) || defined(LT)	movapd	%xmm9,  -16 * SIZE(B)	movapd	%xmm11, -14 * SIZE(B)	movapd	%xmm13, -12 * SIZE(B)	movapd	%xmm15, -10 * SIZE(B)	movddup	%xmm9,  %xmm8	unpckhpd %xmm9,  %xmm9	movddup	%xmm11, %xmm10	unpckhpd %xmm11, %xmm11	movddup	%xmm13, %xmm12	unpckhpd %xmm13, %xmm13	movddup	%xmm15, %xmm14	unpckhpd %xmm15, %xmm15	movapd	%xmm8,  -16 * SIZE(BO)	movapd	%xmm9,  -14 * SIZE(BO)	movapd	%xmm10, -12 * SIZE(BO)	movapd	%xmm11, -10 * SIZE(BO)	movapd	%xmm12,  -8 * SIZE(BO)	movapd	%xmm13,  -6 * SIZE(BO)	movapd	%xmm14,  -4 * SIZE(BO)	movapd	%xmm15,  -2 * SIZE(BO)#else	movapd	%xmm9,  -16 * SIZE(AO)	movapd	%xmm13, -14 * SIZE(AO)	movapd	%xmm11, -12 * SIZE(AO)	movapd	%xmm15, -10 * SIZE(AO)#endif#ifndef LN	addq	$4 * SIZE, CO1	addq	$4 * SIZE, CO2#endif#if defined(LT) || defined(RN)	movq	K,  %rax	subq	KK, %rax	salq	$ZBASE_SHIFT, %rax	leaq	(AO, %rax, 2), AO#ifdef LT	addq	$8 * SIZE, B#endif#endif#ifdef LN	subq	$2, KK	movq	BORIG, B#endif#ifdef LT	addq	$2, KK#endif#ifdef RT	movq	K, %rax	movq	BORIG, B	salq	$1 + ZBASE_SHIFT, %rax	addq	%rax, AORIG#endif	decq	I			# i --	jg	.L10	ALIGN_4	.L30:	testq	$1, M	jle	.L99#ifdef LN       movq	K, %rax       salq	$0 + ZBASE_SHIFT, %rax       subq	%rax, AORIG#endif#if defined(LN) || defined(RT)	movq	KK, %rax	movq	AORIG, AO	salq	$ZBASE_SHIFT, %rax	addq	%rax, AO#endif	leaq	16 * SIZE + BUFFER, BO#if defined(LN) || defined(RT)	movq	KK, %rax	salq	$1 + ZBASE_SHIFT, %rax	leaq	(BO, %rax, 2), BO#endif		pxor	%xmm8, %xmm8	pxor	%xmm9, %xmm9	pxor	%xmm10, %xmm10	pxor	%xmm11, %xmm11#if defined(LT) || defined(RN)	movq	KK, %rax#else	movq	K, %rax	subq	KK, %rax#endif	sarq	$2, %rax	je	.L42.L41:	PREFETCH (PREFETCHSIZE +  0) * SIZE(AO)	movapd	-16 * SIZE(AO), %xmm0	movapd	-16 * SIZE(BO), %xmm2	movapd	-14 * SIZE(BO), %xmm3	movapd	-12 * SIZE(BO), %xmm4	movapd	-10 * SIZE(BO), %xmm5	mulpd	%xmm0, %xmm2	mulpd	%xmm0, %xmm3	mulpd	%xmm0, %xmm4	mulpd	%xmm0, %xmm5	ADD1	%xmm2, %xmm8	ADD2	%xmm3, %xmm9	ADD1	%xmm4, %xmm10	ADD2	%xmm5, %xmm11	movapd	-14 * SIZE(AO), %xmm0	movapd	 -8 * SIZE(BO), %xmm2	movapd	 -6 * SIZE(BO), %xmm3	movapd	 -4 * SIZE(BO), %xmm4	movapd	 -2 * SIZE(BO), %xmm5	mulpd	%xmm0, %xmm2	mulpd	%xmm0, %xmm3	mulpd	%xmm0, %xmm4	mulpd	%xmm0, %xmm5	ADD1	%xmm2, %xmm8	ADD2	%xmm3, %xmm9	ADD1	%xmm4, %xmm10	ADD2	%xmm5, %xmm11	movapd	-12 * SIZE(AO), %xmm0	movapd	  0 * SIZE(BO), %xmm2	movapd	  2 * SIZE(BO), %xmm3	movapd	  4 * SIZE(BO), %xmm4	movapd	  6 * SIZE(BO), %xmm5	mulpd	%xmm0, %xmm2	mulpd	%xmm0, %xmm3	mulpd	%xmm0, %xmm4	mulpd	%xmm0, %xmm5	ADD1	%xmm2, %xmm8	ADD2	%xmm3, %xmm9	ADD1	%xmm4, %xmm10	ADD2	%xmm5, %xmm11	movapd	-10 * SIZE(AO), %xmm0	movapd	  8 * SIZE(BO), %xmm2	movapd	 10 * SIZE(BO), %xmm3	movapd	 12 * SIZE(BO), %xmm4	movapd	 14 * SIZE(BO), %xmm5	mulpd	%xmm0, %xmm2	mulpd	%xmm0, %xmm3	mulpd	%xmm0, %xmm4	mulpd	%xmm0, %xmm5	ADD1	%xmm2, %xmm8	ADD2	%xmm3, %xmm9	ADD1	%xmm4, %xmm10	ADD2	%xmm5, %xmm11	subq	$ -8 * SIZE, AO	subq	$-32 * SIZE, BO	subq	$1, %rax	jne    .L41.L42:#if defined(LT) || defined(RN)	movq	KK, %rax#else	movq	K, %rax	subq	KK, %rax#endif	movapd	POSINV,  %xmm7	andq	$3, %rax		# if (k & 1)	BRANCH	jle .L44.L43:	movapd	-16 * SIZE(AO), %xmm0	movapd	-16 * SIZE(BO), %xmm2	movapd	-14 * SIZE(BO), %xmm3	movapd	-12 * SIZE(BO), %xmm4	movapd	-10 * SIZE(BO), %xmm5	mulpd	%xmm0, %xmm2	mulpd	%xmm0, %xmm3	mulpd	%xmm0, %xmm4	mulpd	%xmm0, %xmm5	ADD1	%xmm2, %xmm8	ADD2	%xmm3, %xmm9	ADD1	%xmm4, %xmm10	ADD2	%xmm5, %xmm11	addq	$2 * SIZE, AO	addq	$8 * SIZE, BO	subq	$1, %rax	jg	.L43	ALIGN_4.L44:#if defined(LN) || defined(RT)	movq	KK, %rax#ifdef LN	subq	$1, %rax#else	subq	$2, %rax#endif	movq	AORIG, AO	movq	BORIG, B	leaq	16 * SIZE + BUFFER, BO	salq	$ZBASE_SHIFT, %rax	leaq	(AO, %rax, 1), AO	leaq	(B,  %rax, 2), B	leaq	(BO, %rax, 4), BO#endif	SHUFPD_1 %xmm9, %xmm9	SHUFPD_1 %xmm11, %xmm11#if defined(NN) || defined(NT) || defined(TN) || defined(TT) || \    defined(NR) || defined(NC) || defined(TR) || defined(TC)	xorpd	%xmm7, %xmm9	xorpd	%xmm7, %xmm11#else	xorpd	%xmm7, %xmm8	xorpd	%xmm7, %xmm10#endif#if defined(NN) || defined(NT) || defined(TN) || defined(TT) || \    defined(RR) || defined(RC) || defined(CR) || defined(CC)	subpd	%xmm9, %xmm8	subpd	%xmm11, %xmm10#else	addpd	%xmm9, %xmm8	addpd	%xmm11, %xmm10#endif#if defined(LN) || defined(LT)	movapd	-16 * SIZE(B), %xmm9	movapd	-14 * SIZE(B), %xmm11	subpd	%xmm8,  %xmm9	subpd	%xmm10,  %xmm11#else	movapd	-16 * SIZE(AO), %xmm9	movapd	-14 * SIZE(AO), %xmm11	subpd	%xmm8,  %xmm9	subpd	%xmm10,  %xmm11#endif#ifndef CONJ	SHUFPD_1 %xmm7, %xmm7#endif#if defined(LN) || defined(LT)	movddup	-16 * SIZE(AO), %xmm0	movddup	-15 * SIZE(AO), %xmm1	pshufd	$0x4e, %xmm9, %xmm8	pshufd	$0x4e, %xmm11, %xmm10	xorpd	 %xmm7, %xmm8	xorpd	 %xmm7, %xmm10	mulpd	 %xmm0, %xmm9	mulpd	 %xmm1, %xmm8	mulpd	 %xmm0, %xmm11	mulpd	 %xmm1, %xmm10	addpd	 %xmm8, %xmm9	addpd	 %xmm10, %xmm11#endif#ifdef RN	movddup	-16 * SIZE(B), %xmm0	movddup	-15 * SIZE(B), %xmm1	movddup	-14 * SIZE(B), %xmm2	movddup	-13 * SIZE(B), %xmm3	movddup	-10 * SIZE(B), %xmm4	movddup	 -9 * SIZE(B), %xmm5	pshufd	$0x4e, %xmm9, %xmm8	xorpd	 %xmm7, %xmm8	mulpd	 %xmm0, %xmm9	mulpd	 %xmm1, %xmm8	addpd	 %xmm8, %xmm9	movapd	 %xmm9, %xmm8	pshufd	 $0x4e, %xmm9, %xmm12	xorpd	 %xmm7, %xmm12	mulpd	 %xmm2, %xmm8	mulpd	 %xmm3, %xmm12	subpd	 %xmm8, %xmm11	subpd	 %xmm12, %xmm11	pshufd	$0x4e, %xmm11, %xmm10	xorpd	 %xmm7, %xmm10	mulpd	 %xmm4, %xmm11	mulpd	 %xmm5, %xmm10	addpd	 %xmm10, %xmm11#endif#ifdef RT	movddup	-10 * SIZE(B), %xmm0	movddup	 -9 * SIZE(B), %xmm1	movddup	-12 * SIZE(B), %xmm2	movddup	-11 * SIZE(B), %xmm3	movddup	-16 * SIZE(B), %xmm4	movddup	-15 * SIZE(B), %xmm5	pshufd	$0x4e, %xmm11, %xmm10	xorpd	 %xmm7, %xmm10	mulpd	 %xmm0, %xmm11	mulpd	 %xmm1, %xmm10	addpd	 %xmm10, %xmm11	movapd	 %xmm11, %xmm8	pshufd	 $0x4e, %xmm11, %xmm12	xorpd	 %xmm7, %xmm12	mulpd	 %xmm2, %xmm8	mulpd	 %xmm3, %xmm12	subpd	 %xmm8, %xmm9	subpd	 %xmm12, %xmm9	pshufd	$0x4e, %xmm9, %xmm8	xorpd	 %xmm7, %xmm8	mulpd	 %xmm4, %xmm9	mulpd	 %xmm5, %xmm8	addpd	 %xmm8, %xmm9#endif#ifdef LN	subq	$2 * SIZE, CO1	subq	$2 * SIZE, CO2#endif	movsd	%xmm9,   0 * SIZE(CO1)	movhpd	%xmm9,   1 * SIZE(CO1)	movsd	%xmm11,  0 * SIZE(CO2)	movhpd	%xmm11,  1 * SIZE(CO2)#if defined(LN) || defined(LT)	movapd	%xmm9,  -16 * SIZE(B)	movapd	%xmm11, -14 * SIZE(B)	movddup	%xmm9,  %xmm8	unpckhpd %xmm9,  %xmm9	movddup	%xmm11, %xmm10	unpckhpd %xmm11, %xmm11	movapd	%xmm8,  -16 * SIZE(BO)	movapd	%xmm9,  -14 * SIZE(BO)	movapd	%xmm10, -12 * SIZE(BO)	movapd	%xmm11, -10 * SIZE(BO)#else	movapd	%xmm9,  -16 * SIZE(AO)	movapd	%xmm11, -14 * SIZE(AO)#endif#ifndef LN	addq	$2 * SIZE, CO1	addq	$2 * SIZE, CO2#endif#if defined(LT) || defined(RN)	movq	K,  %rax	subq	KK, %rax	salq	$ZBASE_SHIFT, %rax	leaq	(AO, %rax, 1), AO#ifdef LT	addq	$4 * SIZE, B#endif#endif#ifdef LN	subq	$1, KK	movq	BORIG, B#endif#ifdef LT	addq	$1, KK#endif#ifdef RT	movq	K, %rax	movq	BORIG, B	salq	$0 + ZBASE_SHIFT, %rax	addq	%rax, AORIG#endif	ALIGN_4	.L99:#ifdef LN       leaq	(, K, SIZE), %rax       leaq	(B, %rax, 4), B#endif#if defined(LT) || defined(RN)	movq	K,  %rax	subq	KK, %rax	leaq	(,%rax, SIZE), %rax	leaq	(B,  %rax, 2 * COMPSIZE), B#endif#ifdef RN	addq	$2, KK#endif#ifdef RT	subq	$2, KK#endif	decq	J			# j --	jg	.L01.L100:	testq	$1, N	jle	.L999.L101:#ifdef LN	movq	OFFSET, %rax	addq	M, %rax	movq	%rax, KK#endif		leaq	BUFFER, BO#ifdef RT       movq	K, %rax       salq	$0 + ZBASE_SHIFT, %rax       subq	%rax, B#endif#if defined(LN) || defined(RT)	movq	KK, %rax	movq	B, BORIG	salq	$ZBASE_SHIFT, %rax	leaq	(B,  %rax, 1), B	leaq	(BO, %rax, 2), BO#endif	#if defined(LT)	movq	OFFSET, %rax	movq	%rax, KK#endif#if defined(LT) || defined(RN)	movq	KK, %rax#else	movq	K, %rax	subq	KK, %rax#endif	sarq	$2, %rax	jle	.L103	ALIGN_4	.L102:	movddup	 -16 * SIZE(B), %xmm8	movddup	 -15 * SIZE(B), %xmm9	movddup	 -14 * SIZE(B), %xmm10	movddup	 -13 * SIZE(B), %xmm11	movddup	 -12 * SIZE(B), %xmm12	movddup	 -11 * SIZE(B), %xmm13	movddup	 -10 * SIZE(B), %xmm14	movddup	  -9 * SIZE(B), %xmm15	movapd	%xmm8,  0 * SIZE(BO)	movapd	%xmm9,  2 * SIZE(BO)	movapd	%xmm10,  4 * SIZE(BO)	movapd	%xmm11,  6 * SIZE(BO)	movapd	%xmm12,  8 * SIZE(BO)	movapd	%xmm13, 10 * SIZE(BO)	movapd	%xmm14, 12 * SIZE(BO)	movapd	%xmm15, 14 * SIZE(BO)	addq	$  8 * SIZE, B	subq	$-16 * SIZE, BO	decq	%rax	jne	.L102	ALIGN_4.L103:#if defined(LT) || defined(RN)	movq	KK, %rax#else	movq	K, %rax	subq	KK, %rax#endif	andq	$3, %rax	BRANCH	jle	.L105	ALIGN_4.L104:	movddup	 -16 * SIZE(B), %xmm8	movddup	 -15 * SIZE(B), %xmm9	movapd	%xmm8,  0 * SIZE(BO)	movapd	%xmm9,  2 * SIZE(BO)	addq	$4 * SIZE, BO	addq	$2 * SIZE, B	decq	%rax	jne	.L104	ALIGN_4	.L105:#if defined(LT) || defined(RN)	movq	A, AO#else	movq	A, AORIG#endif#ifdef RT       subq	LDC, C#endif	movq	C, CO1#ifndef RT	addq	LDC, C#endif	movq	M,  I	sarq	$1, I		# i = (m >> 2)
⌨️ 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 + -