📄 zaxpy_sse.s

📁 Optimized GotoBLAS libraries
💻 S
字号:
/*********************************************************************//*                                                                   *//*             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	PROLOGUE	PROFCODE#ifndef WINDOWS_ABI#ifndef XDOUBLE	movq	 8(%rsp), INCY#else	movq	40(%rsp), INCY#endif#else	movaps	%xmm3, %xmm0	movss	40(%rsp), %xmm1		movq	48(%rsp), X	movq	56(%rsp), INCX	movq	64(%rsp), Y	movq	72(%rsp), INCY#endif	SAVEREGISTERS	#ifndef CONJ	pshufd	$0, %xmm0, %xmm14	pshufd	$0, %xmm1, %xmm15	pxor	%xmm13, %xmm13	subps	%xmm15, %xmm13	unpcklps %xmm14, %xmm13	unpcklps %xmm15, %xmm14	movaps	 %xmm13, %xmm15#else	pshufd	$0, %xmm0, %xmm14	pshufd	$0, %xmm1, %xmm15	pxor	%xmm13, %xmm13	subps	%xmm14, %xmm13	unpcklps %xmm15, %xmm14	unpcklps %xmm13, %xmm15#endif	leaq	(, INCX, SIZE * 2), INCX	leaq	(, INCY, SIZE * 2), INCY	cmpq	$2 * SIZE, INCX	jne	.L50	cmpq	$2 * SIZE, INCY	jne	.L50	testq	$2 * SIZE, X	je	.L05	movsd	 0 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm8	movsldup	 %xmm1, %xmm0	movshdup	 %xmm1, %xmm1	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,  0 * SIZE(Y)	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	decq	M	jle	.L999	ALIGN_2.L05:	testq	$1 * SIZE, X	jne	.L30	testq	$3 * SIZE, Y	jne	.L20		movq	M,  %rax	sarq	$3, %rax	jle	.L15	ALIGN_3.L12:	movsldup	 0 * SIZE(X), %xmm0	movsldup	 4 * SIZE(X), %xmm2	movsldup	 8 * SIZE(X), %xmm4	movsldup	12 * SIZE(X), %xmm6	movshdup	 0 * SIZE(X), %xmm1	movshdup	 4 * SIZE(X), %xmm3	movshdup	 8 * SIZE(X), %xmm5	movshdup	12 * SIZE(X), %xmm7	movaps	 0 * SIZE(Y), %xmm8	movaps	 4 * SIZE(Y), %xmm9	movaps	 8 * SIZE(Y), %xmm10	movaps	12 * SIZE(Y), %xmm11	mulps	%xmm14, %xmm0	mulps	%xmm14, %xmm2	mulps	%xmm14, %xmm4	mulps	%xmm14, %xmm6	mulps	%xmm15, %xmm1	mulps	%xmm15, %xmm3	mulps	%xmm15, %xmm5	mulps	%xmm15, %xmm7	addps	%xmm0, %xmm8	addps	%xmm2, %xmm9	addps	%xmm4, %xmm10	addps	%xmm6, %xmm11	addps	%xmm1, %xmm8	addps	%xmm3, %xmm9	addps	%xmm5, %xmm10	addps	%xmm7, %xmm11	movaps	%xmm8,   0 * SIZE(Y)	movaps	%xmm9,   4 * SIZE(Y)	movaps	%xmm10,  8 * SIZE(Y)	movaps	%xmm11, 12 * SIZE(Y)	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	decq	%rax	jg	.L12	ALIGN_3.L15:	testq	$4, M	jle	.L16	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movsldup	 4 * SIZE(X), %xmm2	movshdup	 4 * SIZE(X), %xmm3	movaps	 0 * SIZE(Y), %xmm8	movaps	 4 * SIZE(Y), %xmm9	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	movaps	%xmm8,   0 * SIZE(Y)	movaps	%xmm9,   4 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_2.L16:	testq	$2, M	jle	.L17	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movaps	 0 * SIZE(Y), %xmm8	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movaps	%xmm8,   0 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_2.L17:	testq	$1, M	jle	.L999	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movaps	 0 * SIZE(Y), %xmm8	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,  0 * SIZE(Y)	jmp	.L999	ALIGN_3.L20:	movq	M,  %rax	sarq	$3, %rax	jle	.L25	ALIGN_3.L22:	movsldup	 0 * SIZE(X), %xmm0	movsldup	 4 * SIZE(X), %xmm2	movsldup	 8 * SIZE(X), %xmm4	movsldup	12 * SIZE(X), %xmm6	movshdup	 0 * SIZE(X), %xmm1	movshdup	 4 * SIZE(X), %xmm3	movshdup	 8 * SIZE(X), %xmm5	movshdup	12 * SIZE(X), %xmm7	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	movsd	 4 * SIZE(Y), %xmm9	movhps	 6 * SIZE(Y), %xmm9	movsd	 8 * SIZE(Y), %xmm10	movhps	10 * SIZE(Y), %xmm10	movsd	12 * SIZE(Y), %xmm11	movhps	14 * SIZE(Y), %xmm11	mulps	%xmm14, %xmm0	mulps	%xmm14, %xmm2	mulps	%xmm14, %xmm4	mulps	%xmm14, %xmm6	mulps	%xmm15, %xmm1	mulps	%xmm15, %xmm3	mulps	%xmm15, %xmm5	mulps	%xmm15, %xmm7	addps	%xmm0, %xmm8	addps	%xmm2, %xmm9	addps	%xmm4, %xmm10	addps	%xmm6, %xmm11	addps	%xmm1, %xmm8	addps	%xmm3, %xmm9	addps	%xmm5, %xmm10	addps	%xmm7, %xmm11	movlpd	%xmm8,   0 * SIZE(Y)	movhpd	%xmm8,   2 * SIZE(Y)	movlpd	%xmm9,   4 * SIZE(Y)	movhpd	%xmm9,   6 * SIZE(Y)	movlpd	%xmm10,  8 * SIZE(Y)	movhpd	%xmm10, 10 * SIZE(Y)	movlpd	%xmm11, 12 * SIZE(Y)	movhpd	%xmm11, 14 * SIZE(Y)	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	decq	%rax	jg	.L22	ALIGN_3.L25:	testq	$4, M	jle	.L26	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movsldup	 4 * SIZE(X), %xmm2	movshdup	 4 * SIZE(X), %xmm3	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	movsd	 4 * SIZE(Y), %xmm9	movhps	 6 * SIZE(Y), %xmm9	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	movlpd	%xmm8,   0 * SIZE(Y)	movhpd	%xmm8,   2 * SIZE(Y)	movlpd	%xmm9,   4 * SIZE(Y)	movhpd	%xmm9,   6 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_2.L26:	testq	$2, M	jle	.L27	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movlpd	%xmm8,   0 * SIZE(Y)	movhpd	%xmm8,   2 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_2.L27:	testq	$1, M	jle	.L999	movsldup	 0 * SIZE(X), %xmm0	movshdup	 0 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm8	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,  0 * SIZE(Y)	jmp	.L999	ALIGN_3.L30:		movq	M,  %rax	sarq	$3, %rax	jle	.L35	ALIGN_3.L32:	movsd	 0 * SIZE(X), %xmm0	movhps	 2 * SIZE(X), %xmm0	movsd	 4 * SIZE(X), %xmm2	movhps	 6 * SIZE(X), %xmm2	movsd	 8 * SIZE(X), %xmm4	movhps	10 * SIZE(X), %xmm4	movsd	12 * SIZE(X), %xmm6	movhps	14 * SIZE(X), %xmm6	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	movshdup	 %xmm2, %xmm3	movsldup	 %xmm2, %xmm2	movshdup	 %xmm4, %xmm5	movsldup	 %xmm4, %xmm4	movshdup	 %xmm6, %xmm7	movsldup	 %xmm6, %xmm6	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	mulps	%xmm14, %xmm4	mulps	%xmm15, %xmm5	mulps	%xmm14, %xmm6	mulps	%xmm15, %xmm7	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	movsd	 4 * SIZE(Y), %xmm9	movhps	 6 * SIZE(Y), %xmm9	movsd	 8 * SIZE(Y), %xmm10	movhps	10 * SIZE(Y), %xmm10	movsd	12 * SIZE(Y), %xmm11	movhps	14 * SIZE(Y), %xmm11	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	addps	%xmm4, %xmm10	addps	%xmm5, %xmm10	addps	%xmm6, %xmm11	addps	%xmm7, %xmm11	movsd	%xmm8,   0 * SIZE(Y)	movhps	%xmm8,   2 * SIZE(Y)	movsd	%xmm9,   4 * SIZE(Y)	movhps	%xmm9,   6 * SIZE(Y)	movsd	%xmm10,  8 * SIZE(Y)	movhps	%xmm10, 10 * SIZE(Y)	movsd	%xmm11, 12 * SIZE(Y)	movhps	%xmm11, 14 * SIZE(Y)	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	decq	%rax	jg	.L32	ALIGN_3.L35:	testq	$4, M	jle	.L36	movsd	 0 * SIZE(X), %xmm0	movhps	 2 * SIZE(X), %xmm0	movsd	 4 * SIZE(X), %xmm2	movhps	 6 * SIZE(X), %xmm2	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	movshdup	 %xmm2, %xmm3	movsldup	 %xmm2, %xmm2	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	movsd	 4 * SIZE(Y), %xmm9	movhps	 6 * SIZE(Y), %xmm9	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	movsd	%xmm8,   0 * SIZE(Y)	movhps	%xmm8,   2 * SIZE(Y)	movsd	%xmm9,   4 * SIZE(Y)	movhps	%xmm9,   6 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L36:	testq	$2, M	jle	.L37	movsd	 0 * SIZE(X), %xmm0	movhps	 2 * SIZE(X), %xmm0	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	movsd	 0 * SIZE(Y), %xmm8	movhps	 2 * SIZE(Y), %xmm8	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,   0 * SIZE(Y)	movhps	%xmm8,   2 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L37:	testq	$1, M	jle	.L999	movsd	 0 * SIZE(X), %xmm0	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	movsd	 0 * SIZE(Y), %xmm8	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,   0 * SIZE(Y)	jmp	.L999	ALIGN_3.L50:	movq	Y, YY	movq	M,  %rax	sarq	$3, %rax	jle	.L55	ALIGN_3.L52:	movsd	 (X), %xmm0	addq	 INCX, X	movhps	 (X), %xmm0	addq	 INCX, X	movsd	 (X), %xmm2	addq	 INCX, X	movhps	 (X), %xmm2	addq	 INCX, X	movsd	 (X), %xmm4	addq	 INCX, X	movhps	 (X), %xmm4	addq	 INCX, X	movsd	 (X), %xmm6	addq	 INCX, X	movhps	 (X), %xmm6	addq	 INCX, X	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	movshdup	 %xmm2, %xmm3	movsldup	 %xmm2, %xmm2	movshdup	 %xmm4, %xmm5	movsldup	 %xmm4, %xmm4	movshdup	 %xmm6, %xmm7	movsldup	 %xmm6, %xmm6	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	mulps	%xmm14, %xmm4	mulps	%xmm15, %xmm5	mulps	%xmm14, %xmm6	mulps	%xmm15, %xmm7	movsd	(Y), %xmm8	addq	 INCY, Y	movhps	(Y), %xmm8	addq	 INCY, Y	movsd	(Y), %xmm9	addq	 INCY, Y	movhps	(Y), %xmm9	addq	 INCY, Y	movsd	(Y), %xmm10	addq	 INCY, Y	movhps	(Y), %xmm10	addq	 INCY, Y	movsd	(Y), %xmm11	addq	 INCY, Y	movhps	(Y), %xmm11	addq	 INCY, Y	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	addps	%xmm4, %xmm10	addps	%xmm5, %xmm10	addps	%xmm6, %xmm11	addps	%xmm7, %xmm11	movsd	%xmm8,  (YY)	addq	 INCY, YY	movhps	%xmm8,  (YY)	addq	 INCY, YY	movsd	%xmm9,  (YY)	addq	 INCY, YY	movhps	%xmm9,  (YY)	addq	 INCY, YY	movsd	%xmm10, (YY)	addq	 INCY, YY	movhps	%xmm10, (YY)	addq	 INCY, YY	movsd	%xmm11, (YY)	addq	 INCY, YY	movhps	%xmm11, (YY)	addq	 INCY, YY	decq	%rax	jg	.L52	ALIGN_3.L55:	testq	$4, M	jle	.L56	movsd	 (X), %xmm0	addq	 INCX, X	movhps	 (X), %xmm0	addq	 INCX, X	movsd	 (X), %xmm2	addq	 INCX, X	movhps	 (X), %xmm2	addq	 INCX, X	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	movshdup	 %xmm2, %xmm3	movsldup	 %xmm2, %xmm2	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	mulps	%xmm14, %xmm2	mulps	%xmm15, %xmm3	movsd	(Y), %xmm8	addq	 INCY, Y	movhps	(Y), %xmm8	addq	 INCY, Y	movsd	(Y), %xmm9	addq	 INCY, Y	movhps	(Y), %xmm9	addq	 INCY, Y	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	addps	%xmm2, %xmm9	addps	%xmm3, %xmm9	movsd	%xmm8,  (YY)	addq	 INCY, YY	movhps	%xmm8,  (YY)	addq	 INCY, YY	movsd	%xmm9,  (YY)	addq	 INCY, YY	movhps	%xmm9,  (YY)	addq	 INCY, YY	ALIGN_3.L56:	testq	$2, M	jle	.L57	movsd	 (X), %xmm0	addq	 INCX, X	movhps	 (X), %xmm0	addq	 INCX, X	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	movsd	(Y), %xmm8	addq	 INCY, Y	movhps	(Y), %xmm8	addq	 INCY, Y	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,  (YY)	addq	 INCY, YY	movhps	%xmm8,  (YY)	addq	 INCY, YY	ALIGN_3.L57:	testq	$1, M	jle	.L999	movsd	 (X), %xmm0	movshdup	 %xmm0, %xmm1	movsldup	 %xmm0, %xmm0	mulps	%xmm14, %xmm0	mulps	%xmm15, %xmm1	movsd	 (Y), %xmm8	addps	%xmm0, %xmm8	addps	%xmm1, %xmm8	movsd	%xmm8,   (Y)	ALIGN_3.L999:	xorq	%rax, %rax	RESTOREREGISTERS	ret	EPILOGUE
⌨️ 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 + -