📄 izamax_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 STACK_X		 8 + STACK + ARGS(%esp)#define STACK_INCX	12 + STACK + ARGS(%esp)#define RET	%eax#define	M	%ebx#define X	%ecx#define INCX	%edx#define I	%esi#define MM	%ebp#define XX	%edi#define TEMP	%ebx	#ifdef MIN#define maxpd	minpd#define maxsd	minsd#endif	PROLOGUE	pushl	%ebp	pushl	%edi	pushl	%esi	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	xor	RET, RET	testl	M, M	jle	.L999	testl	INCX, INCX	jle	.L999	sall	$ZBASE_SHIFT, INCX	movl	M, MM	movl	X, XX	cmpeqpd	%xmm7, %xmm7	psrlq	$1, %xmm7	movsd	0 * SIZE(XX), %xmm0	movsd	1 * SIZE(XX), %xmm1	addl	INCX, XX	decl	MM	andpd	 %xmm7, %xmm0	andpd	 %xmm7, %xmm1	addpd	 %xmm1, %xmm0	unpcklpd %xmm0, %xmm0	cmpl	$2 * SIZE, INCX	jne	.L60#define PREFETCHSIZE 256	movl	MM,  I	sarl	$3, I	jle	.L25	ALIGN_4.L21:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	 0 * SIZE(XX), %xmm1	movsd	 1 * SIZE(XX), %xmm2	movhpd	 2 * SIZE(XX), %xmm1	movhpd	 3 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	 4 * SIZE(XX), %xmm3	movsd	 5 * SIZE(XX), %xmm4	movhpd	 6 * SIZE(XX), %xmm3	movhpd	 7 * SIZE(XX), %xmm4	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	prefetcht0	(PREFETCHSIZE + 8) * SIZE(XX)	movsd	 8 * SIZE(XX), %xmm1	movsd	 9 * SIZE(XX), %xmm2	movhpd	10 * SIZE(XX), %xmm1	movhpd	11 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	12 * SIZE(XX), %xmm3	movsd	13 * SIZE(XX), %xmm4	movhpd	14 * SIZE(XX), %xmm3	movhpd	15 * SIZE(XX), %xmm4	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	addl	$16 * SIZE, XX	decl	I	jg	.L21	ALIGN_4.L25:	andl	$7,  MM	jle	.L30	testl	$4, MM	je	.L26	movsd	 0 * SIZE(XX), %xmm1	movsd	 1 * SIZE(XX), %xmm2	movhpd	 2 * SIZE(XX), %xmm1	movhpd	 3 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	 4 * SIZE(XX), %xmm3	movsd	 5 * SIZE(XX), %xmm4	movhpd	 6 * SIZE(XX), %xmm3	movhpd	 7 * SIZE(XX), %xmm4	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	addl	$8 * SIZE, XX	ALIGN_3.L26:	testl	$2, MM	je	.L27	movsd	 0 * SIZE(XX), %xmm1	movsd	 1 * SIZE(XX), %xmm2	movhpd	 2 * SIZE(XX), %xmm1	movhpd	 3 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	addl	$4 * SIZE, XX	ALIGN_3	.L27:	testl	$1, MM	je	.L30	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxsd	%xmm1, %xmm0	ALIGN_4.L30:	movl	X, XX	movl	M, MM	movapd	 %xmm0, %xmm1	unpckhpd %xmm0, %xmm0	maxsd	 %xmm1, %xmm0	unpcklpd %xmm0, %xmm0	movl	MM,  I	sarl	$2, I	jle	.L35	ALIGN_4.L31:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	movhpd	2 * SIZE(XX), %xmm1	movhpd	3 * SIZE(XX), %xmm2	movsd	4 * SIZE(XX), %xmm3	movsd	5 * SIZE(XX), %xmm4	movhpd	6 * SIZE(XX), %xmm3	movhpd	7 * SIZE(XX), %xmm4	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	cmpeqpd	%xmm0, %xmm1	cmpeqpd	%xmm0, %xmm3	orpd	%xmm3, %xmm1	movmskpd %xmm1, TEMP	testl	 $3, TEMP	jne	 .L33	addl	$8 * SIZE, XX	addl	$4, RET	decl	I	jg	.L31	jmp	.L35	ALIGN_4.L33:	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	movsd	2 * SIZE(XX), %xmm3	movsd	3 * SIZE(XX), %xmm4	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	movsd	4 * SIZE(XX), %xmm1	movsd	5 * SIZE(XX), %xmm2	movsd	6 * SIZE(XX), %xmm3	movsd	7 * SIZE(XX), %xmm4	addl	$8 * SIZE, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	ALIGN_3.L35:	testl	$2, MM	je	.L36	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	movsd	2 * SIZE(XX), %xmm3	movsd	3 * SIZE(XX), %xmm4	addl	$4 * SIZE, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	ALIGN_3	.L36:	incl	RET	jmp	.L999	ALIGN_3.L60:	movl	MM,  I	sarl	$3, I	jle	.L65	ALIGN_4	.L61:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm1	movhpd	1 * SIZE(XX), %xmm2	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm3	movhpd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm1	movhpd	1 * SIZE(XX), %xmm2	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm3	movhpd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	decl	I	jg	.L61	ALIGN_4.L65:	andl	$7,  MM	jle	.L70	testl	$4, MM	je	.L66	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm1	movhpd	1 * SIZE(XX), %xmm2	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm3	movhpd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm4, %xmm3	maxpd	%xmm3, %xmm0	ALIGN_3.L66:	testl	$2, MM	je	.L67	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm1	movhpd	1 * SIZE(XX), %xmm2	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxpd	%xmm1, %xmm0	ALIGN_3	.L67:	testl	$1, MM	je	.L70	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	addpd	%xmm2, %xmm1	maxsd	%xmm1, %xmm0	ALIGN_3.L70:	movl	X, XX	movl	M, MM	movapd	%xmm0, %xmm1	unpckhpd %xmm0, %xmm0	maxsd	%xmm1, %xmm0	unpcklpd %xmm0, %xmm0	movl	MM,  I	sarl	$2, I	jle	.L75	ALIGN_4.L71:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm1	movhpd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	movhpd	0 * SIZE(XX), %xmm3	movhpd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	cmpeqpd	%xmm0, %xmm1	cmpeqpd	%xmm0, %xmm3	orpd	%xmm3, %xmm1	movmskpd %xmm1, TEMP	testl	 $3, TEMP	jne	 .L73	addl	$4, RET	decl	I	jg	.L71	jmp	.L75	ALIGN_4.L73:	leal	(, INCX, 4), TEMP	subl	TEMP, XX	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	ALIGN_3.L75:	testl	$2, MM	je	.L76	movsd	0 * SIZE(XX), %xmm1	movsd	1 * SIZE(XX), %xmm2	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm3	movsd	1 * SIZE(XX), %xmm4	addl	INCX, XX	andpd	%xmm7, %xmm1	andpd	%xmm7, %xmm2	andpd	%xmm7, %xmm3	andpd	%xmm7, %xmm4	addpd	%xmm2,  %xmm1	addpd	%xmm4,  %xmm3	incl	RET	comisd	%xmm0, %xmm1	je	.L999	incl	RET	comisd	%xmm0, %xmm3	je	.L999	ALIGN_3	.L76:	incl	RET	ALIGN_4.L999:	popl	%ebx	popl	%esi	popl	%edi	popl	%ebp	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 + -