📄 axpy_sse2_opteron.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#define PREFETCH	prefetch#define PREFETCHW	prefetchw#ifdef BARCELONA#define PREFETCHSIZE	(8 * 8 + 4)#else#define PREFETCHSIZE	(8 * 8 + 4)#endif	PROLOGUE	PROFCODE#ifndef WINDOWS_ABI#ifndef XDOUBLE	movq	 8(%rsp), INCY#else	movq	24(%rsp), INCY#endif	movaps	%xmm0,  ALPHA#else	movaps	%xmm3,  ALPHA	movq	40(%rsp), X	movq	48(%rsp), INCX	movq	56(%rsp), Y	movq	64(%rsp), INCY#endif	SAVEREGISTERS	unpcklpd ALPHA, ALPHA	leaq	(, INCX, SIZE), INCX	leaq	(, INCY, SIZE), INCY	cmpq	$SIZE, INCX	jne	.L40	cmpq	$SIZE, INCY	jne	.L40	subq	$-16 * SIZE, X	subq	$-16 * SIZE, Y	testq	$SIZE, Y	je	.L10	movsd	-16 * SIZE(X), %xmm0	mulsd	ALPHA, %xmm0	addsd	-16 * SIZE(Y), %xmm0	movlpd	%xmm0, -16 * SIZE(Y)	addq	$1 * SIZE, X	addq	$1 * SIZE, Y	decq	M	jle	.L19	ALIGN_3.L10:	testq	$SIZE, X	jne	.L20	movq	M,  %rax	sarq	$4, %rax	jle	.L13	movapd	-16 * SIZE(X), %xmm0	movapd	-14 * SIZE(X), %xmm1	movapd	-12 * SIZE(X), %xmm2	movapd	-10 * SIZE(X), %xmm3	movapd	 -8 * SIZE(X), %xmm4	movapd	 -6 * SIZE(X), %xmm5	movapd	 -4 * SIZE(X), %xmm6	movapd	 -2 * SIZE(X), %xmm7	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	mulpd	ALPHA, %xmm2	mulpd	ALPHA, %xmm3	decq	%rax	jle .L12	ALIGN_3.L11:	addpd	-16 * SIZE(Y), %xmm0	mulpd	ALPHA, %xmm4	PREFETCH	(PREFETCHSIZE + 0) * SIZE(X)	addpd	-14 * SIZE(Y), %xmm1	mulpd	ALPHA, %xmm5	addpd	-12 * SIZE(Y), %xmm2	mulpd	ALPHA, %xmm6	addpd	-10 * SIZE(Y), %xmm3	mulpd	ALPHA, %xmm7	PREFETCHW	(PREFETCHSIZE + 0) * SIZE(Y)	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	movapd	 0 * SIZE(X), %xmm0	movapd	 2 * SIZE(X), %xmm1	movapd	 4 * SIZE(X), %xmm2	movapd	 6 * SIZE(X), %xmm3	addpd	-8 * SIZE(Y), %xmm4	mulpd	ALPHA, %xmm0	PREFETCH	(PREFETCHSIZE + 8) * SIZE(X)	addpd	-6 * SIZE(Y), %xmm5	mulpd	ALPHA, %xmm1	addpd	-4 * SIZE(Y), %xmm6	mulpd	ALPHA, %xmm2	addpd	-2 * SIZE(Y), %xmm7	mulpd	ALPHA, %xmm3	PREFETCHW	(PREFETCHSIZE + 8) * SIZE(Y)	movapd	%xmm4,  -8 * SIZE(Y)	movapd	%xmm5,  -6 * SIZE(Y)	movapd	%xmm6,  -4 * SIZE(Y)	movapd	%xmm7,  -2 * SIZE(Y)	movapd	 8 * SIZE(X), %xmm4	movapd	10 * SIZE(X), %xmm5	movapd	12 * SIZE(X), %xmm6	movapd	14 * SIZE(X), %xmm7	subq	$-16 * SIZE, X	subq	$-16 * SIZE, Y	decq	%rax	jg	.L11	ALIGN_3.L12:	addpd	-16 * SIZE(Y), %xmm0	mulpd	ALPHA, %xmm4	addpd	-14 * SIZE(Y), %xmm1	mulpd	ALPHA, %xmm5	addpd	-12 * SIZE(Y), %xmm2	mulpd	ALPHA, %xmm6	addpd	-10 * SIZE(Y), %xmm3	mulpd	ALPHA, %xmm7	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	addpd	-8 * SIZE(Y), %xmm4	addpd	-6 * SIZE(Y), %xmm5	addpd	-4 * SIZE(Y), %xmm6	addpd	-2 * SIZE(Y), %xmm7	movapd	%xmm4,  -8 * SIZE(Y)	movapd	%xmm5,  -6 * SIZE(Y)	movapd	%xmm6,  -4 * SIZE(Y)	movapd	%xmm7,  -2 * SIZE(Y)	subq	$-16 * SIZE, X	subq	$-16 * SIZE, Y	ALIGN_3.L13:	testq	$15, M	jle	.L19	testq	$8, M	jle	.L14	movapd	-16 * SIZE(X), %xmm0	movapd	-14 * SIZE(X), %xmm1	movapd	-12 * SIZE(X), %xmm2	movapd	-10 * SIZE(X), %xmm3	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	mulpd	ALPHA, %xmm2	mulpd	ALPHA, %xmm3	addpd	-16 * SIZE(Y), %xmm0	addpd	-14 * SIZE(Y), %xmm1	addpd	-12 * SIZE(Y), %xmm2	addpd	-10 * SIZE(Y), %xmm3	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L14:	testq	$4, M	jle	.L15	movapd	-16 * SIZE(X), %xmm0	movapd	-14 * SIZE(X), %xmm1	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	addpd	-16 * SIZE(Y), %xmm0	addpd	-14 * SIZE(Y), %xmm1	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L15:	testq	$2, M	jle	.L16	movapd	-16 * SIZE(X), %xmm0	mulpd	ALPHA, %xmm0	addpd	-16 * SIZE(Y), %xmm0	movapd	%xmm0, -16 * SIZE(Y)	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	ALIGN_3.L16:	testq	$1, M	jle	.L19	movlpd	-16 * SIZE(X), %xmm0	mulsd	ALPHA, %xmm0	addsd	-16 * SIZE(Y), %xmm0	movlpd	%xmm0, 	-16 * SIZE(Y)	addq	$SIZE, Y	ALIGN_3.L19:	xorq	%rax, %rax	RESTOREREGISTERS	ret	ALIGN_3.L20:	movq	M,  %rax	sarq	$4, %rax	jle	.L23	movlpd	-16 * SIZE(X), %xmm0	movhpd	-15 * SIZE(X), %xmm0	movlpd	-14 * SIZE(X), %xmm1	movhpd	-13 * SIZE(X), %xmm1	movlpd	-12 * SIZE(X), %xmm2	movhpd	-11 * SIZE(X), %xmm2	movlpd	-10 * SIZE(X), %xmm3	movhpd	 -9 * SIZE(X), %xmm3	movlpd	 -8 * SIZE(X), %xmm4	movhpd	 -7 * SIZE(X), %xmm4	movlpd	 -6 * SIZE(X), %xmm5	movhpd	 -5 * SIZE(X), %xmm5	movlpd	 -4 * SIZE(X), %xmm6	movhpd	 -3 * SIZE(X), %xmm6	movlpd	 -2 * SIZE(X), %xmm7	movhpd	 -1 * SIZE(X), %xmm7	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	mulpd	ALPHA, %xmm2	mulpd	ALPHA, %xmm3	decq	%rax	jle .L22	ALIGN_3.L21:	PREFETCH	(PREFETCHSIZE + 0) * SIZE(X)	addpd	-16 * SIZE(Y), %xmm0	mulpd	ALPHA, %xmm4	addpd	-14 * SIZE(Y), %xmm1	mulpd	ALPHA, %xmm5	addpd	-12 * SIZE(Y), %xmm2	mulpd	ALPHA, %xmm6	addpd	-10 * SIZE(Y), %xmm3	mulpd	ALPHA, %xmm7	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	PREFETCHW	(PREFETCHSIZE + 0) * SIZE(Y)	movlpd	 0 * SIZE(X), %xmm0	movhpd	 1 * SIZE(X), %xmm0	movlpd	 2 * SIZE(X), %xmm1	movhpd	 3 * SIZE(X), %xmm1	movlpd	 4 * SIZE(X), %xmm2	movhpd	 5 * SIZE(X), %xmm2	movlpd	 6 * SIZE(X), %xmm3	movhpd	 7 * SIZE(X), %xmm3	PREFETCH	(PREFETCHSIZE + 8) * SIZE(X)	addpd	-8 * SIZE(Y), %xmm4	mulpd	ALPHA, %xmm0	addpd	-6 * SIZE(Y), %xmm5	mulpd	ALPHA, %xmm1	addpd	-4 * SIZE(Y), %xmm6	mulpd	ALPHA, %xmm2	addpd	-2 * SIZE(Y), %xmm7	mulpd	ALPHA, %xmm3	movapd	%xmm4,  -8 * SIZE(Y)	movapd	%xmm5,  -6 * SIZE(Y)	movapd	%xmm6, -4 * SIZE(Y)	movapd	%xmm7, -2 * SIZE(Y)	PREFETCHW	(PREFETCHSIZE + 8) * SIZE(Y)	movlpd	 8 * SIZE(X), %xmm4	movhpd	 9 * SIZE(X), %xmm4	movlpd	10 * SIZE(X), %xmm5	movhpd	11 * SIZE(X), %xmm5	movlpd	12 * SIZE(X), %xmm6	movhpd	13 * SIZE(X), %xmm6	movlpd	14 * SIZE(X), %xmm7	movhpd	15 * SIZE(X), %xmm7	addq	$16 * SIZE, Y	addq	$16 * SIZE, X	decq	%rax	jg	.L21	ALIGN_3.L22:	addpd	-16 * SIZE(Y), %xmm0	mulpd	ALPHA, %xmm4	addpd	-14 * SIZE(Y), %xmm1	mulpd	ALPHA, %xmm5	addpd	-12 * SIZE(Y), %xmm2	mulpd	ALPHA, %xmm6	addpd	-10 * SIZE(Y), %xmm3	mulpd	ALPHA, %xmm7	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	addpd	-8 * SIZE(Y), %xmm4	addpd	-6 * SIZE(Y), %xmm5	addpd	-4 * SIZE(Y), %xmm6	addpd	-2 * SIZE(Y), %xmm7	movapd	%xmm4,  -8 * SIZE(Y)	movapd	%xmm5,  -6 * SIZE(Y)	movapd	%xmm6, -4 * SIZE(Y)	movapd	%xmm7, -2 * SIZE(Y)	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	ALIGN_3.L23:	movq	M,  %rax	andq	$8, %rax	jle	.L24	ALIGN_3	movlpd	-16 * SIZE(X), %xmm0	movhpd	-15 * SIZE(X), %xmm0	movlpd	-14 * SIZE(X), %xmm1	movhpd	-13 * SIZE(X), %xmm1	movlpd	-12 * SIZE(X), %xmm2	movhpd	-11 * SIZE(X), %xmm2	movlpd	-10 * SIZE(X), %xmm3	movhpd	 -9 * SIZE(X), %xmm3	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	mulpd	ALPHA, %xmm2	mulpd	ALPHA, %xmm3	addpd	-16 * SIZE(Y), %xmm0	addpd	-14 * SIZE(Y), %xmm1	addpd	-12 * SIZE(Y), %xmm2	addpd	-10 * SIZE(Y), %xmm3	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	movapd	%xmm2, -12 * SIZE(Y)	movapd	%xmm3, -10 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L24:	movq	M,  %rax	andq	$4, %rax	jle	.L25	ALIGN_3	movlpd	-16 * SIZE(X), %xmm0	movhpd	-15 * SIZE(X), %xmm0	movlpd	-14 * SIZE(X), %xmm1	movhpd	-13 * SIZE(X), %xmm1	mulpd	ALPHA, %xmm0	mulpd	ALPHA, %xmm1	addpd	-16 * SIZE(Y), %xmm0	addpd	-14 * SIZE(Y), %xmm1	movapd	%xmm0, -16 * SIZE(Y)	movapd	%xmm1, -14 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L25:	movq	M,  %rax	andq	$2, %rax	jle	.L26	ALIGN_3	movlpd	-16 * SIZE(X), %xmm0	movhpd	-15 * SIZE(X), %xmm0	mulpd	ALPHA, %xmm0	addpd	-16 * SIZE(Y), %xmm0	movapd	%xmm0, -16 * SIZE(Y)	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	ALIGN_3.L26:	movq	M,  %rax	andq	$1, %rax	jle	.L29	ALIGN_3	movlpd	-16 * SIZE(X), %xmm0	mulsd	ALPHA, %xmm0	addsd	-16 * SIZE(Y), %xmm0	movlpd	%xmm0, 	-16 * SIZE(Y)	addq	$SIZE, Y	ALIGN_3.L29:	xorq	%rax, %rax	RESTOREREGISTERS	ret	ALIGN_3.L40:	movq	Y, YY	movq	M,  %rax	sarq	$3, %rax	jle	.L45	ALIGN_3.L41:	movlpd	0 * SIZE(X), %xmm0	addq	INCX, X	movhpd	0 * SIZE(X), %xmm0	addq	INCX, X	mulpd	ALPHA, %xmm0	movlpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	movhpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	addpd	%xmm6, %xmm0	movlpd	0 * SIZE(X), %xmm1	addq	INCX, X	movhpd	0 * SIZE(X), %xmm1	addq	INCX, X	mulpd	ALPHA, %xmm1	movlpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	movhpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	addpd	%xmm6, %xmm1	movlpd	0 * SIZE(X), %xmm2	addq	INCX, X	movhpd	0 * SIZE(X), %xmm2	addq	INCX, X	mulpd	ALPHA, %xmm2	movlpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	movhpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	addpd	%xmm6, %xmm2	movlpd	0 * SIZE(X), %xmm3	addq	INCX, X	movhpd	0 * SIZE(X), %xmm3	addq	INCX, X	mulpd	ALPHA, %xmm3	movlpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	movhpd	0 * SIZE(YY), %xmm6	addq	INCY, YY	addpd	%xmm6, %xmm3	movlpd	%xmm0, 0 * SIZE(Y)	addq	INCY, Y	movhpd	%xmm0, 0 * SIZE(Y)	addq	INCY, Y	movlpd	%xmm1, 0 * SIZE(Y)	addq	INCY, Y	movhpd	%xmm1, 0 * SIZE(Y)	addq	INCY, Y	movlpd	%xmm2, 0 * SIZE(Y)	addq	INCY, Y	movhpd	%xmm2, 0 * SIZE(Y)	addq	INCY, Y	movlpd	%xmm3, 0 * SIZE(Y)	addq	INCY, Y	movhpd	%xmm3, 0 * SIZE(Y)	addq	INCY, Y	decq	%rax	jg	.L41	ALIGN_3.L45:	movq	M,  %rax	andq	$7, %rax	jle	.L47	ALIGN_3.L46:	movlpd	(X), %xmm0	addq	INCX, X	mulsd	%xmm15, %xmm0	addsd	(Y), %xmm0	movlpd	%xmm0, (Y)	addq	INCY, Y	decq	%rax	jg	.L46	ALIGN_3.L47:	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 + -