📄 zaxpy_sse2_core2.s

📁 Optimized GotoBLAS libraries
💻 S
📖 第 1 页 / 共 3 页
字号:
12 3 下一页
/*********************************************************************//*                                                                   *//*             Optimized BLAS libraries                              *//*                     By Kazushige Goto <kgoto@tacc.utexas.edu>     *//*                                                                   *//* Copyright (c) The University of Texas, 2005. All rights reserved. *//* UNIVERSITY EXPRESSLY DISCLAIMS ANY AND ALL WARRANTIES CONCERNING  *//* THIS SOFTWARE AND DOCUMENTATION, INCLUDING ANY WARRANTIES OF      *//* MERCHANTABILITY, FITNESS FOR ANY PARTICULAR PURPOSE,              *//* NON-INFRINGEMENT AND WARRANTIES OF PERFORMANCE, AND ANY WARRANTY  *//* THAT MIGHT OTHERWISE ARISE FROM COURSE OF DEALING OR USAGE OF     *//* TRADE. NO WARRANTY IS EITHER EXPRESS OR IMPLIED WITH RESPECT TO   *//* THE USE OF THE SOFTWARE OR DOCUMENTATION.                         *//* Under no circumstances shall University be liable for incidental, *//* special, indirect, direct or consequential damages or loss of     *//* profits, interruption of business, or related expenses which may  *//* arise from use of Software or Documentation, including but not    *//* limited to those resulting from defects in Software and/or        *//* Documentation, or loss or inaccuracy of data of any kind.         *//*********************************************************************/#define ASSEMBLER#include "common.h"#ifndef WINDOWS_ABI#define M	ARG1#define X	ARG4#define INCX	ARG5#define Y	ARG6#define INCY	ARG2#else#define M	ARG1#define X	ARG2#define INCX	ARG3#define Y	ARG4#define INCY	%r10#endif#define	YY	%r11#define ALPHA	%xmm15#ifdef HAVE_SSE3#define MOVDDUP(a, b, c)	movddup	a(b), c#define MOVDDUP2(a, b, c)	movddup	a##b, c#else#define MOVDDUP(a, b, c)	movsd	a(b), c;movhpd	a(b), c#define MOVDDUP2(a, b, c)	movsd	a##b, c;movhpd	a##b, c#endif	PROLOGUE	PROFCODE#ifndef WINDOWS_ABI#ifndef XDOUBLE	movq	 8(%rsp), INCY#else	movq	40(%rsp), INCY#endif#else	movaps	%xmm3, %xmm0	movsd	40(%rsp), %xmm1		movq	48(%rsp), X	movq	56(%rsp), INCX	movq	64(%rsp), Y	movq	72(%rsp), INCY#endif	SAVEREGISTERS	#ifndef CONJ	movapd	%xmm0, %xmm14		#  a   0	pxor	%xmm15, %xmm15		#  0   0	subsd	%xmm1,  %xmm15		# -b   0	unpcklpd %xmm14, %xmm15		# -b   a	unpcklpd %xmm1,  %xmm14		#  a   b#else	movapd	%xmm0, %xmm14		#  a   0	movapd	%xmm1, %xmm15		#  b   0	pxor	%xmm13, %xmm13		#  0   0	subsd	%xmm0,  %xmm13		# -a   0	unpcklpd %xmm13, %xmm15		#  b   -a	unpcklpd %xmm1,  %xmm14		#  a   b#endif	salq	$ZBASE_SHIFT, INCX	salq	$ZBASE_SHIFT, INCY	testq	$SIZE, Y	jne	.L50	cmpq	$2 * SIZE, INCX	jne	.L30	cmpq	$2 * SIZE, INCY	jne	.L30#ifndef CONJ	movddup	 %xmm0,  %xmm14	pxor	 %xmm15, %xmm15	subsd	 %xmm1,  %xmm15	unpcklpd %xmm1,  %xmm15#else	movddup	 %xmm1,  %xmm15	movapd	 %xmm0,  %xmm14	pxor	 %xmm13, %xmm13	subsd	 %xmm0,  %xmm13	unpcklpd %xmm13, %xmm14#endif	testq	$SIZE, X	jne	.L20	movq	M,  %rax	sarq	$3, %rax	jle	.L15	movapd	 0 * SIZE(X), %xmm0	movapd	 2 * SIZE(X), %xmm2	movapd	 4 * SIZE(X), %xmm6	movapd	 6 * SIZE(X), %xmm7	decq	%rax	jle	.L12	ALIGN_3.L11:	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	%xmm14, %xmm8	movapd	%xmm14, %xmm9#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	shufps	 $0x4e, %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	shufps	 $0x4e, %xmm7, %xmm7#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	SHUFPD_1 %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	SHUFPD_1 %xmm7, %xmm7#endif		movapd	 0 * SIZE(Y), %xmm4	addpd	%xmm1, %xmm4	mulpd	%xmm15, %xmm0	movapd	 2 * SIZE(Y), %xmm5	addpd	%xmm3, %xmm5	mulpd	%xmm15, %xmm2	movapd	 4 * SIZE(Y), %xmm10	addpd	%xmm8, %xmm10	mulpd	%xmm15, %xmm6	movapd	 6 * SIZE(Y), %xmm11	addpd	%xmm9, %xmm11	mulpd	%xmm15, %xmm7	addpd	%xmm0, %xmm4	movapd	 8 * SIZE(X), %xmm0	addpd	%xmm2, %xmm5	movapd	10 * SIZE(X), %xmm2	addpd	%xmm6, %xmm10	movapd	12 * SIZE(X), %xmm6	addpd	%xmm7, %xmm11	movapd	14 * SIZE(X), %xmm7	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	%xmm14, %xmm8	movapd	%xmm14, %xmm9	movapd	%xmm4,   0 * SIZE(Y)	movapd	%xmm5,   2 * SIZE(Y)	movapd	%xmm10,  4 * SIZE(Y)	movapd	%xmm11,  6 * SIZE(Y)#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	shufps	 $0x4e, %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	shufps	 $0x4e, %xmm7, %xmm7#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	SHUFPD_1 %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	SHUFPD_1 %xmm7, %xmm7#endif	movapd	 8 * SIZE(Y), %xmm4	addpd	%xmm1, %xmm4	mulpd	%xmm15, %xmm0	movapd	10 * SIZE(Y), %xmm5	addpd	%xmm3, %xmm5	mulpd	%xmm15, %xmm2	movapd	12 * SIZE(Y), %xmm10	addpd	%xmm8, %xmm10	mulpd	%xmm15, %xmm6	movapd	14 * SIZE(Y), %xmm11	addpd	%xmm9, %xmm11	mulpd	%xmm15, %xmm7	addpd	%xmm0, %xmm4	movapd	16 * SIZE(X), %xmm0	addpd	%xmm2, %xmm5	movapd	18 * SIZE(X), %xmm2	addpd	%xmm6, %xmm10	movapd	20 * SIZE(X), %xmm6	addpd	%xmm7, %xmm11	movapd	22 * SIZE(X), %xmm7	movapd	%xmm4,   8 * SIZE(Y)	movapd	%xmm5,  10 * SIZE(Y)	movapd	%xmm10, 12 * SIZE(Y)	movapd	%xmm11, 14 * SIZE(Y)	subq	$-16 * SIZE, Y	subq	$-16 * SIZE, X	decq	%rax	jg	.L11	ALIGN_3.L12:	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	%xmm14, %xmm8	movapd	%xmm14, %xmm9#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	shufps	 $0x4e, %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	shufps	 $0x4e, %xmm7, %xmm7#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	SHUFPD_1 %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	SHUFPD_1 %xmm7, %xmm7#endif	movapd	 0 * SIZE(Y), %xmm4	addpd	%xmm1, %xmm4	mulpd	%xmm15, %xmm0	movapd	 2 * SIZE(Y), %xmm5	addpd	%xmm3, %xmm5	mulpd	%xmm15, %xmm2	movapd	 4 * SIZE(Y), %xmm10	addpd	%xmm8, %xmm10	mulpd	%xmm15, %xmm6	movapd	 6 * SIZE(Y), %xmm11	addpd	%xmm9, %xmm11	mulpd	%xmm15, %xmm7	addpd	%xmm0, %xmm4	movapd	 8 * SIZE(X), %xmm0	addpd	%xmm2, %xmm5	movapd	10 * SIZE(X), %xmm2	addpd	%xmm6, %xmm10	movapd	12 * SIZE(X), %xmm6	addpd	%xmm7, %xmm11	movapd	14 * SIZE(X), %xmm7	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	%xmm14, %xmm8	movapd	%xmm14, %xmm9	movapd	%xmm4,   0 * SIZE(Y)	movapd	%xmm5,   2 * SIZE(Y)	movapd	%xmm10,  4 * SIZE(Y)	movapd	%xmm11,  6 * SIZE(Y)#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	shufps	 $0x4e, %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	shufps	 $0x4e, %xmm7, %xmm7#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	SHUFPD_1 %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	SHUFPD_1 %xmm7, %xmm7#endif	movapd	 8 * SIZE(Y), %xmm4	addpd	%xmm1, %xmm4	mulpd	%xmm15, %xmm0	movapd	10 * SIZE(Y), %xmm5	addpd	%xmm3, %xmm5	mulpd	%xmm15, %xmm2	movapd	12 * SIZE(Y), %xmm10	addpd	%xmm8, %xmm10	mulpd	%xmm15, %xmm6	movapd	14 * SIZE(Y), %xmm11	addpd	%xmm9, %xmm11	mulpd	%xmm15, %xmm7	addpd	%xmm0, %xmm4	addpd	%xmm2, %xmm5	addpd	%xmm6, %xmm10	addpd	%xmm7, %xmm11	movapd	%xmm4,   8 * SIZE(Y)	movapd	%xmm5,  10 * SIZE(Y)	movapd	%xmm10, 12 * SIZE(Y)	movapd	%xmm11, 14 * SIZE(Y)	subq	$-16 * SIZE, X	subq	$-16 * SIZE, Y	ALIGN_3.L15:	movq	M,  %rax	andq	$4, %rax	jle	.L16	movapd	 0 * SIZE(X), %xmm0	movapd	 2 * SIZE(X), %xmm2	movapd	 4 * SIZE(X), %xmm6	movapd	 6 * SIZE(X), %xmm7	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	%xmm14, %xmm8	movapd	%xmm14, %xmm9	movapd	 0 * SIZE(Y), %xmm4	movapd	 2 * SIZE(Y), %xmm5	movapd	 4 * SIZE(Y), %xmm10	movapd	 6 * SIZE(Y), %xmm11#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	shufps	 $0x4e, %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	shufps	 $0x4e, %xmm7, %xmm7#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2	mulpd	%xmm6,  %xmm8	SHUFPD_1 %xmm6, %xmm6	mulpd	%xmm7,  %xmm9	SHUFPD_1 %xmm7, %xmm7#endif	mulpd	%xmm15, %xmm0	mulpd	%xmm15, %xmm2	mulpd	%xmm15, %xmm6	mulpd	%xmm15, %xmm7	addpd	%xmm1, %xmm4	addpd	%xmm3, %xmm5	addpd	%xmm8, %xmm10	addpd	%xmm9, %xmm11	addpd	%xmm0, %xmm4	addpd	%xmm2, %xmm5	addpd	%xmm6, %xmm10	addpd	%xmm7, %xmm11	movapd	%xmm4,   0 * SIZE(Y)	movapd	%xmm5,   2 * SIZE(Y)	movapd	%xmm10,  4 * SIZE(Y)	movapd	%xmm11,  6 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L16:	movq	M,  %rax	andq	$2, %rax	jle	.L17	movapd	 0 * SIZE(X), %xmm0	movapd	 2 * SIZE(X), %xmm2	movapd	%xmm14, %xmm1	movapd	%xmm14, %xmm3	movapd	 0 * SIZE(Y), %xmm4	movapd	 2 * SIZE(Y), %xmm5#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	shufps	 $0x4e, %xmm2, %xmm2#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0	mulpd	%xmm2,  %xmm3	SHUFPD_1 %xmm2, %xmm2#endif	mulpd	%xmm15, %xmm0	mulpd	%xmm15, %xmm2	addpd	%xmm1, %xmm4	addpd	%xmm3, %xmm5	addpd	%xmm0, %xmm4	addpd	%xmm2, %xmm5	movapd	%xmm4,   0 * SIZE(Y)	movapd	%xmm5,   2 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L17:	movq	M,  %rax	andq	$1, %rax	jle	.L999	movapd	 0 * SIZE(X), %xmm0	movapd	%xmm14, %xmm1	movapd	 0 * SIZE(Y), %xmm4#ifdef PENRYN	mulpd	%xmm0,  %xmm1	shufps	 $0x4e, %xmm0, %xmm0#else	mulpd	%xmm0,  %xmm1	SHUFPD_1 %xmm0, %xmm0#endif	mulpd	%xmm15, %xmm0	addpd	%xmm0, %xmm4	addpd	%xmm1, %xmm4	movapd	%xmm4,   0 * SIZE(Y)	jmp	.L999	ALIGN_3.L20:	movq	M,  %rax	sarq	$3, %rax	jle	.L25	movsd	 0 * SIZE(X), %xmm0	movhpd	 1 * SIZE(X), %xmm0	movsd	 2 * SIZE(X), %xmm2	movhpd	 3 * SIZE(X), %xmm2	movsd	 4 * SIZE(X), %xmm4	movhpd	 5 * SIZE(X), %xmm4	movsd	 6 * SIZE(X), %xmm6	movhpd	 7 * SIZE(X), %xmm6	pshufd	 $0x4e, %xmm0, %xmm1	pshufd	 $0x4e, %xmm2, %xmm3	pshufd	 $0x4e, %xmm4, %xmm5	pshufd	 $0x4e, %xmm6, %xmm7	subq	 $1, %rax	jle	 .L22	ALIGN_3.L21:	movapd	 0 * SIZE(Y), %xmm8	movapd	 2 * SIZE(Y), %xmm9	movapd	 4 * SIZE(Y), %xmm10	movapd	 6 * SIZE(Y), %xmm11	mulpd	%xmm14, %xmm0	mulpd	%xmm14, %xmm2	mulpd	%xmm14, %xmm4	mulpd	%xmm14, %xmm6	addpd	%xmm0, %xmm8	addpd	%xmm2, %xmm9	addpd	%xmm4, %xmm10	addpd	%xmm6, %xmm11	mulpd	%xmm15, %xmm1	mulpd	%xmm15, %xmm3	mulpd	%xmm15, %xmm5	mulpd	%xmm15, %xmm7	addpd	%xmm1, %xmm8	addpd	%xmm3, %xmm9	addpd	%xmm5, %xmm10	addpd	%xmm7, %xmm11	movapd	%xmm8,   0 * SIZE(Y)	movapd	%xmm9,   2 * SIZE(Y)	movapd	%xmm10,  4 * SIZE(Y)	movapd	%xmm11,  6 * SIZE(Y)	movsd	 8 * SIZE(X), %xmm0	movhpd	 9 * SIZE(X), %xmm0	movsd	10 * SIZE(X), %xmm2	movhpd	11 * SIZE(X), %xmm2	movsd	12 * SIZE(X), %xmm4	movhpd	13 * SIZE(X), %xmm4	movsd	14 * SIZE(X), %xmm6	movhpd	15 * SIZE(X), %xmm6	pshufd	 $0x4e, %xmm0, %xmm1	pshufd	 $0x4e, %xmm2, %xmm3	pshufd	 $0x4e, %xmm4, %xmm5	pshufd	 $0x4e, %xmm6, %xmm7	movapd	 8 * SIZE(Y), %xmm8	movapd	10 * SIZE(Y), %xmm9	movapd	12 * SIZE(Y), %xmm10	movapd	14 * SIZE(Y), %xmm11	mulpd	%xmm14, %xmm0	mulpd	%xmm14, %xmm2	mulpd	%xmm14, %xmm4	mulpd	%xmm14, %xmm6	addpd	%xmm0, %xmm8	addpd	%xmm2, %xmm9	addpd	%xmm4, %xmm10	addpd	%xmm6, %xmm11	mulpd	%xmm15, %xmm1	mulpd	%xmm15, %xmm3	mulpd	%xmm15, %xmm5	mulpd	%xmm15, %xmm7	addpd	%xmm1, %xmm8	addpd	%xmm3, %xmm9	addpd	%xmm5, %xmm10	addpd	%xmm7, %xmm11	movapd	%xmm8,   8 * SIZE(Y)	movapd	%xmm9,  10 * SIZE(Y)	movapd	%xmm10, 12 * SIZE(Y)	movapd	%xmm11, 14 * SIZE(Y)	movsd	16 * SIZE(X), %xmm0	movhpd	17 * SIZE(X), %xmm0	movsd	18 * SIZE(X), %xmm2	movhpd	19 * SIZE(X), %xmm2	movsd	20 * SIZE(X), %xmm4	movhpd	21 * SIZE(X), %xmm4	movsd	22 * SIZE(X), %xmm6	movhpd	23 * SIZE(X), %xmm6	pshufd	 $0x4e, %xmm0, %xmm1	pshufd	 $0x4e, %xmm2, %xmm3	pshufd	 $0x4e, %xmm4, %xmm5	pshufd	 $0x4e, %xmm6, %xmm7	subq	$-16 * SIZE, X	subq	$-16 * SIZE, Y	subq	$1, %rax	jg	.L21	ALIGN_3.L22:	movapd	 0 * SIZE(Y), %xmm8	movapd	 2 * SIZE(Y), %xmm9	movapd	 4 * SIZE(Y), %xmm10	movapd	 6 * SIZE(Y), %xmm11	mulpd	%xmm14, %xmm0	mulpd	%xmm14, %xmm2	mulpd	%xmm14, %xmm4	mulpd	%xmm14, %xmm6	addpd	%xmm0, %xmm8	addpd	%xmm2, %xmm9	addpd	%xmm4, %xmm10	addpd	%xmm6, %xmm11	mulpd	%xmm15, %xmm1	mulpd	%xmm15, %xmm3
12 3 下一页
⌨️ 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 + -