📄 axpy_sse2.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"#define STACK	16#define ARGS	 0	#define STACK_M		 4 + STACK + ARGS(%esp)#define ALPHA		16 + STACK + ARGS(%esp)#define STACK_X		24 + STACK + ARGS(%esp)#define STACK_INCX	28 + STACK + ARGS(%esp)#define STACK_Y		32 + STACK + ARGS(%esp)#define STACK_INCY	36 + STACK + ARGS(%esp)#define M	 %ebx#define X	 %esi#define Y	 %edi#define INCX	 %ecx#define INCY	 %edx	#ifdef PENTIUM4#define PREFETCHSIZE	128#endif	PROLOGUE	pushl	%edi	pushl	%esi	pushl	%ebx	pushl	%ebp	PROFCODE	movsd	ALPHA, %xmm7	unpcklpd %xmm7, %xmm7	movl	STACK_M,    M	movl	STACK_X,    X	movl	STACK_INCX, INCX	movl	STACK_Y,    Y	movl	STACK_INCY, INCY	leal	(, INCX, SIZE), INCX	leal	(, INCY, SIZE), INCY	cmpl	$SIZE, INCX	jne	.L100	cmpl	$SIZE, INCY	jne	.L100	testl	$SIZE, Y	je	.L00	movsd	0 * SIZE(X), %xmm0	mulsd	%xmm7, %xmm0	addsd	0 * SIZE(Y), %xmm0	movsd	%xmm0, 0 * SIZE(Y)	addl	$1 * SIZE, X	addl	$1 * SIZE, Y	decl	M	jle	.L999	ALIGN_3.L00:	testl	$SIZE, X	jne	.L20	movl	M,  %eax	sarl	$4,  %eax	jle	.L15	ALIGN_3.L11:	movapd	 0 * SIZE(X), %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 2 * SIZE(X), %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	movapd	 4 * SIZE(X), %xmm2	mulpd	%xmm7, %xmm2	addpd	 4 * SIZE(Y), %xmm2	movapd	%xmm2,  4 * SIZE(Y)	movapd	 6 * SIZE(X), %xmm3	mulpd	%xmm7, %xmm3	addpd	 6 * SIZE(Y), %xmm3	movapd	%xmm3,  6 * SIZE(Y)	movapd	 8 * SIZE(X), %xmm0	mulpd	%xmm7, %xmm0	addpd	 8 * SIZE(Y), %xmm0	movapd	%xmm0, 8 * SIZE(Y)#ifdef PENTIUM4	prefetcht0     (PREFETCHSIZE + 0) * SIZE(X)#endif	movapd	10 * SIZE(X), %xmm1	mulpd	%xmm7, %xmm1	addpd	10 * SIZE(Y), %xmm1	movapd	%xmm1, 10 * SIZE(Y)	movapd	12 * SIZE(X), %xmm2	mulpd	%xmm7, %xmm2	addpd	12 * SIZE(Y), %xmm2	movapd	%xmm2, 12 * SIZE(Y)	movapd	14 * SIZE(X), %xmm3	mulpd	%xmm7, %xmm3	addpd	14 * SIZE(Y), %xmm3	movapd	%xmm3, 14 * SIZE(Y)#ifdef PENTIUM4	prefetcht0     (PREFETCHSIZE + 0) * SIZE(Y)#endif	addl	$16 * SIZE, X	addl	$16 * SIZE, Y	decl	%eax	jg	.L11	ALIGN_3.L15:	movl	M,  %eax	testl	$8, %eax	jle	.L16	movapd	 0 * SIZE(X), %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 2 * SIZE(X), %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	movapd	 4 * SIZE(X), %xmm2	mulpd	%xmm7, %xmm2	addpd	 4 * SIZE(Y), %xmm2	movapd	%xmm2,  4 * SIZE(Y)	movapd	 6 * SIZE(X), %xmm3	mulpd	%xmm7, %xmm3	addpd	 6 * SIZE(Y), %xmm3	movapd	%xmm3,  6 * SIZE(Y)	addl	$8 * SIZE, X	addl	$8 * SIZE, Y	ALIGN_3.L16:	testl	$4, %eax	jle	.L17	movapd	 0 * SIZE(X), %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 2 * SIZE(X), %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	addl	$4 * SIZE, X	addl	$4 * SIZE, Y	ALIGN_3.L17:	testl	$2, %eax	jle	.L18	movapd	 0 * SIZE(X), %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	addl	$2 * SIZE, X	addl	$2 * SIZE, Y	ALIGN_3.L18:	testl	$1, %eax	jle	.L99	movsd	0 * SIZE(X), %xmm0	mulsd	%xmm7, %xmm0	addsd	0 * SIZE(Y), %xmm0	movsd	%xmm0, 	0 * SIZE(Y)	jmp	.L99	ALIGN_3.L20:	movapd	-1 * SIZE(X), %xmm0	movl	M, %eax	sarl	$4,   %eax	jle	.L25	ALIGN_4.L21:	movapd	 1 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 3 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	movapd	 5 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 4 * SIZE(Y), %xmm0	movapd	%xmm0,  4 * SIZE(Y)	movapd	 7 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	 6 * SIZE(Y), %xmm1	movapd	%xmm1,  6 * SIZE(Y)	movapd	 9 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 8 * SIZE(Y), %xmm0	movapd	%xmm0,  8 * SIZE(Y)	movapd	11 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	10 * SIZE(Y), %xmm1	movapd	%xmm1, 10 * SIZE(Y)	movapd	13 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	12 * SIZE(Y), %xmm0	movapd	%xmm0, 12 * SIZE(Y)	movapd	15 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	14 * SIZE(Y), %xmm1	movapd	%xmm1, 14 * SIZE(Y)#ifdef PENTIUM4	prefetcht0     (PREFETCHSIZE + 0) * SIZE(X)	prefetcht0     (PREFETCHSIZE + 0) * SIZE(Y)#endif	addl	$16 * SIZE, X	addl	$16 * SIZE, Y	decl	%eax	jg	.L21	ALIGN_3.L25:	movl	M,  %eax	testl	$8, %eax	jle	.L26	movapd	 1 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 3 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	movapd	 5 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 4 * SIZE(Y), %xmm0	movapd	%xmm0,  4 * SIZE(Y)	movapd	 7 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	 6 * SIZE(Y), %xmm1	movapd	%xmm1,  6 * SIZE(Y)	addl	$8 * SIZE, X	addl	$8 * SIZE, Y	ALIGN_3.L26:	testl	$4, %eax	jle	.L27	movapd	 1 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	 3 * SIZE(X), %xmm0	SHUFPD_1 %xmm0, %xmm1	mulpd	%xmm7, %xmm1	addpd	 2 * SIZE(Y), %xmm1	movapd	%xmm1,  2 * SIZE(Y)	addl	$4 * SIZE, X	addl	$4 * SIZE, Y	ALIGN_3.L27:	testl	$2, %eax	jle	.L28	movapd	 1 * SIZE(X), %xmm1	SHUFPD_1 %xmm1, %xmm0	mulpd	%xmm7, %xmm0	addpd	 0 * SIZE(Y), %xmm0	movapd	%xmm0,  0 * SIZE(Y)	movapd	%xmm1, %xmm0	addl	$2 * SIZE, X	addl	$2 * SIZE, Y	ALIGN_3.L28:	testl	$1, %eax	jle	.L99	unpckhpd %xmm0, %xmm0	mulsd	%xmm7, %xmm0	addsd	0 * SIZE(Y), %xmm0	movsd	%xmm0, 	0 * SIZE(Y)	ALIGN_3.L99:	xorl	%eax,%eax	popl	%ebp	popl	%ebx	popl	%esi	popl	%edi	ret	ALIGN_3.L100:	movl	M, %eax	movl	Y, %ebp	sarl	$3,   %eax	jle	.L114	ALIGN_3.L110:	movsd	0 * SIZE(X), %xmm0	addl	INCX, X	movhpd	0 * SIZE(X), %xmm0	addl	INCX, X	mulpd	%xmm7, %xmm0	movsd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	movhpd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	addpd	%xmm6, %xmm0	movsd	0 * SIZE(X), %xmm1	addl	INCX, X	movhpd	0 * SIZE(X), %xmm1	addl	INCX, X	mulpd	%xmm7, %xmm1	movsd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	movhpd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	addpd	%xmm6, %xmm1	movsd	0 * SIZE(X), %xmm2	addl	INCX, X	movhpd	0 * SIZE(X), %xmm2	addl	INCX, X	mulpd	%xmm7, %xmm2	movsd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	movhpd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	addpd	%xmm6, %xmm2	movsd	0 * SIZE(X), %xmm3	addl	INCX, X	movhpd	0 * SIZE(X), %xmm3	addl	INCX, X	mulpd	%xmm7, %xmm3	movsd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	movhpd	0 * SIZE(%ebp), %xmm6	addl	INCY, %ebp	addpd	%xmm6, %xmm3	movsd	%xmm0, 0 * SIZE(Y)	addl	INCY, Y	movhpd	%xmm0, 0 * SIZE(Y)	addl	INCY, Y	movsd	%xmm1, 0 * SIZE(Y)	addl	INCY, Y	movhpd	%xmm1, 0 * SIZE(Y)	addl	INCY, Y	movsd	%xmm2, 0 * SIZE(Y)	addl	INCY, Y	movhpd	%xmm2, 0 * SIZE(Y)	addl	INCY, Y	movsd	%xmm3, 0 * SIZE(Y)	addl	INCY, Y	movhpd	%xmm3, 0 * SIZE(Y)	addl	INCY, Y	decl	%eax	jg	.L110	ALIGN_3.L114:	movl	M, %eax	andl	$7,   %eax	jle	.L999	ALIGN_3.L115:	movsd	(X), %xmm0	addl	INCX, X	mulsd	%xmm7, %xmm0	addsd	(Y), %xmm0	movsd	%xmm0, (Y)	addl	INCY, Y	decl	%eax	jg	.L115	ALIGN_3.L999:	xorl	%eax,%eax	popl	%ebp	popl	%ebx	popl	%esi	popl	%edi	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 + -