📄 zasum_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	12#define ARGS     8#define V	 0 + STACK(%esp)#define STACK_M		 4 + STACK + ARGS(%esp)#define STACK_X		 8 + STACK + ARGS(%esp)#define STACK_INCX	12 + STACK + ARGS(%esp)#define I	%eax#define M	%ecx#define X	%esi#define INCX	%ebx#define FLAG	%edi		PROLOGUE	subl	$ARGS, %esp	pushl	%esi	pushl	%edi	pushl	%ebx	PROFCODE	movl	STACK_M, M	movl	STACK_X, X	movl	STACK_INCX, INCX#ifdef F_INTERFACE	movl	(M), M	movl	(INCX), INCX#endif	pxor	%xmm0, %xmm0	pxor	%xmm7, %xmm7	testl	M, M	jle	.L999	testl	INCX, INCX	jle	.L999	pxor	%xmm1, %xmm1	cmpeqpd	%xmm7, %xmm7	psrlq	$1, %xmm7	xor	FLAG, FLAG		sall	$ZBASE_SHIFT, INCX	cmpl	$2 * SIZE, INCX	jne	.L40	/* Analigned Check */	testl	$7, X	jne	.L30		# Unaligned Mode	testl	$15, X	je	.L05	subl	$1, M	movsd	0 * SIZE(X), %xmm0	addl	$SIZE, X	andpd	%xmm7, %xmm0	mov	$1, FLAG	ALIGN_3.L05:	movl	M,  I	sarl	$4, I	jle	.L19		movapd	 0 * SIZE(X), %xmm2	movapd	 8 * SIZE(X), %xmm3	movapd	16 * SIZE(X), %xmm4	movapd	24 * SIZE(X), %xmm5	decl	I	jle	.L12	ALIGN_4#ifdef PENTIUM3#define PREFETCHSIZE	128#endif#ifdef CORE2#define PREFETCHSIZE	168#endif#ifdef PENRYN#define PREFETCHSIZE	168#endif#ifdef PENTIUM4#define PREFETCHSIZE	168#endif#ifdef PENTIUMM#define PREFETCHSIZE	168#endif#if defined(OPTERON) || defined(BARCELONA)#define PREFETCHSIZE	168#endif	ALIGN_3.L10:#if defined(OPTERON) || defined(BARCELONA)       prefetcht0	(PREFETCHSIZE +  0) * SIZE(X)#endif	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	2 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	movapd	4 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	6 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1#if defined(OPTERON) || defined(BARCELONA)       prefetcht0	(PREFETCHSIZE +  8) * SIZE(X)#endif	movapd	32 * SIZE(X), %xmm2	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	10 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1	movapd	12 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	14 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1#ifdef PENTIUM4       prefetcht0	(PREFETCHSIZE +  0) * SIZE(X)#endif#if defined(OPTERON) || defined(BARCELONA)       prefetcht0	(PREFETCHSIZE + 16) * SIZE(X)#endif	movapd	40 * SIZE(X), %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	18 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movapd	20 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	22 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1#if defined(OPTERON) || defined(BARCELONA)       prefetcht0	(PREFETCHSIZE + 24) * SIZE(X)#endif	movapd	48 * SIZE(X), %xmm4	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm0	movapd	26 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm1	movapd	28 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm0	movapd	30 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm1	movapd	56 * SIZE(X), %xmm5#ifdef PENTIUM4	prefetcht0	(PREFETCHSIZE + 16) * SIZE(X)#endif	addl	$32 * SIZE, X	subl	$1, I	jg	.L10	ALIGN_4.L12:	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	2 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	movapd	4 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	6 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	10 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1	movapd	12 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	14 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	18 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movapd	20 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	22 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm0	movapd	26 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm1	movapd	28 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm0	movapd	30 * SIZE(X), %xmm5	andpd	%xmm7, %xmm5	addpd	%xmm5, %xmm1	addl	$32 * SIZE, X	ALIGN_4.L19:	testl	$15,  M	je	.L29	testl	$8, M	je	.L20	movapd	0 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	2 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	movapd	4 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm0	movapd	6 * SIZE(X), %xmm2	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	movapd	8 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	10 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1	movapd	12 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm0	movapd	14 * SIZE(X), %xmm3	andpd	%xmm7, %xmm3	addpd	%xmm3, %xmm1	addl	$16 * SIZE, X	ALIGN_2.L20:	testl	$4, M	je	.L21	movapd	0 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	2 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movapd	4 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	6 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	addl	$8 * SIZE, X	ALIGN_3.L21:	testl	$2, M	je	.L22	movapd	0 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movapd	2 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	addl	$4 * SIZE, X	ALIGN_3	.L22:	testl	$1, M	je	.L29	movapd	0 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	addl	$2 * SIZE, X	ALIGN_3.L29:	testl	$1, FLAG	je	.L998	movsd	0 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addsd	%xmm4, %xmm0	jmp	.L998	ALIGN_3/* Unaligned Mode */.L30:	movl	M,  I	sarl	$3, I	jle	.L35	ALIGN_4	.L31:	prefetcht0	PREFETCHSIZE * SIZE(X)	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	2 * SIZE(X), %xmm4	movhpd	3 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movsd	4 * SIZE(X), %xmm4	movhpd	5 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	6 * SIZE(X), %xmm4	movhpd	7 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	prefetcht0	(PREFETCHSIZE + 8) * SIZE(X)	movsd	8 * SIZE(X), %xmm4	movhpd	9 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	10 * SIZE(X), %xmm4	movhpd	11 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movsd	12 * SIZE(X), %xmm4	movhpd	13 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	14 * SIZE(X), %xmm4	movhpd	15 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	addl	$16 * SIZE, X	decl	I	jg	.L31	ALIGN_4.L35:	andl	$7,  M	jle	.L998	ALIGN_4.L36:	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	addl	$2 * SIZE, X	decl	M	jg	.L36	jmp	.L998	ALIGN_4.L40:	movl	M,  I	sarl	$2, I	jle	.L60	ALIGN_4	.L50:	prefetcht0	PREFETCHSIZE * SIZE(X)	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	addl	INCX, X	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	addl	INCX, X	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	addl	INCX, X	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	addl	INCX, X	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm1	decl	I	jg	.L50	ALIGN_4.L60:	andl	$3,  M	jle	.L998	ALIGN_4.L61:	movsd	0 * SIZE(X), %xmm4	movhpd	1 * SIZE(X), %xmm4	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm0	addl	INCX, X	decl	M	jg	.L61	ALIGN_4.L998:	addpd	%xmm1, %xmm0	MOVAPD	%xmm0, %xmm1	UNPCKHPD %xmm0, %xmm0	ADDSD	 %xmm1, %xmm0	ALIGN_4.L999:	movsd	%xmm0, V	fldl	V	popl	%ebx	popl	%edi	popl	%esi	addl	$ARGS, %esp	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 + -