📄 izamax_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"	#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 maxps	minps#define maxss	minss#endif	#ifndef HAVE_SSE2#define pxor	xorps#define movsd	movlps#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#ifdef ABS#ifndef HAVE_SSE2	subl	$8, %esp	movl	$0x7fffffff, (%esp)	movss	(%esp), %xmm7	shufps	$0, %xmm7, %xmm7	addl	$8, %esp#else	cmpeqps	%xmm7, %xmm7	psrld	$1, %xmm7		/* Generate ABS */#endif#endif	movss	0 * SIZE(XX), %xmm0	movss	1 * SIZE(XX), %xmm1	addl	INCX, XX	decl	MM	andps	%xmm7, %xmm0	andps	%xmm7, %xmm1	addps	%xmm1,  %xmm0	shufps	$0, %xmm0, %xmm0	cmpl	$2 * SIZE, INCX	jne	.L70.L30:	movl	MM,  I	sarl	$3, I	jle	.L35	ALIGN_4	#define PREFETCHSIZE 80.L31:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	 0 * SIZE(XX), %xmm1	movhps	 2 * SIZE(XX), %xmm1	movsd	 4 * SIZE(XX), %xmm2	movhps	 6 * SIZE(XX), %xmm2	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	movsd	 8 * SIZE(XX), %xmm1	movhps	10 * SIZE(XX), %xmm1	movsd	12 * SIZE(XX), %xmm2	movhps	14 * SIZE(XX), %xmm2	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	addl	$16 * SIZE, XX	decl	I	jg	.L31	ALIGN_4.L35:	andl	$7,  MM	jle	.L40	testl	$4, MM	je	.L36	movsd	 0 * SIZE(XX), %xmm1	movhps	 2 * SIZE(XX), %xmm1	movsd	 4 * SIZE(XX), %xmm2	movhps	 6 * SIZE(XX), %xmm2	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	addl	$8 * SIZE, XX	ALIGN_3.L36:	testl	$2, MM	je	.L37	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	movss	2 * SIZE(XX), %xmm3	movss	3 * SIZE(XX), %xmm4	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	maxss	%xmm1,  %xmm0	maxss	%xmm3,  %xmm0	addl	$4 * SIZE, XX	ALIGN_3	.L37:	testl	$1, MM	je	.L40	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	addps	%xmm2,  %xmm1	maxss	%xmm1,  %xmm0	ALIGN_4.L40:	movl	X, XX	movl	M, MM	movaps	%xmm0, %xmm1	movhlps %xmm0, %xmm0	maxps	%xmm1, %xmm0	movaps	%xmm0, %xmm1	shufps	$1, %xmm0, %xmm0	maxss	%xmm1, %xmm0	shufps	$0, %xmm0, %xmm0	movl	MM,  I	sarl	$2, I	jle	.L45	ALIGN_4	.L41:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	movhps	2 * SIZE(XX), %xmm1	movsd	4 * SIZE(XX), %xmm2	movhps	6 * SIZE(XX), %xmm2	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	cmpeqps	%xmm0, %xmm1	movmskps %xmm1, TEMP	testl	 $15, TEMP	jne	 .L43	addl	$8 * SIZE, XX	addl	$4, RET	decl	I	jg	.L41	jmp	.L45	ALIGN_4.L43:	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	movss	2 * SIZE(XX), %xmm3	movss	3 * SIZE(XX), %xmm4	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	movss	4 * SIZE(XX), %xmm1	movss	5 * SIZE(XX), %xmm2	movss	6 * SIZE(XX), %xmm3	movss	7 * SIZE(XX), %xmm4	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	addl	$8 * SIZE, XX	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	ALIGN_3.L45:	testl	$2, MM	je	.L47	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	movss	2 * SIZE(XX), %xmm3	movss	3 * SIZE(XX), %xmm4	addl	$4 * SIZE, XX	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	ALIGN_3	.L47:	incl	RET	jmp	.L999	ALIGN_3.L70:	movl	MM,  I	sarl	$3, I	jle	.L75	ALIGN_4	.L71:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm1	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm2	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm2	addl	INCX, XX	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	movsd	0 * SIZE(XX), %xmm1	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm1	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm2	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm2	addl	INCX, XX	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	decl	I	jg	.L71	ALIGN_4.L75:	andl	$7,  MM	jle	.L80	testl	$4, MM	je	.L76	movsd	0 * SIZE(XX), %xmm1	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm1	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm2	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm2	addl	INCX, XX	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	maxps	%xmm1,  %xmm0	ALIGN_3	.L76:	testl	$2, MM	je	.L77	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	addl	INCX, XX	movss	0 * SIZE(XX), %xmm3	movss	1 * SIZE(XX), %xmm4	addl	INCX, XX	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	maxss	%xmm1,  %xmm0	maxss	%xmm3,  %xmm0	ALIGN_3	.L77:	testl	$1, MM	je	.L80	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	addps	%xmm2,  %xmm1	maxss	%xmm1,  %xmm0	ALIGN_4.L80:	movl	X, XX	movl	M, MM	movaps	%xmm0, %xmm1	movhlps %xmm0, %xmm0	maxps	%xmm1, %xmm0	movaps	%xmm0, %xmm1	shufps	$1, %xmm0, %xmm0	maxss	%xmm1, %xmm0	shufps	$0, %xmm0, %xmm0	movl	MM,  I	sarl	$2, I	jle	.L85	ALIGN_4	.L81:	prefetcht0	PREFETCHSIZE * SIZE(XX)	movsd	0 * SIZE(XX), %xmm1	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm1	addl	INCX, XX	movsd	0 * SIZE(XX), %xmm2	addl	INCX, XX	movhps	0 * SIZE(XX), %xmm2	addl	INCX, XX	movaps	%xmm1, %xmm3	shufps	$0x88, %xmm2, %xmm1	shufps	$0xdd, %xmm2, %xmm3	andps	%xmm7, %xmm1	andps	%xmm7, %xmm3	addps	%xmm3,  %xmm1	cmpeqps	%xmm0, %xmm1	movmskps %xmm1, TEMP	testl	 $15, TEMP	jne	 .L83	addl	$4, RET	decl	I	jg	.L81	jmp	.L85	ALIGN_4.L83:	leal	(, INCX, 4), TEMP	subl	TEMP, XX	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	addl	INCX, XX	movss	0 * SIZE(XX), %xmm3	movss	1 * SIZE(XX), %xmm4	addl	INCX, XX	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	addl	INCX, XX	movss	0 * SIZE(XX), %xmm3	movss	1 * SIZE(XX), %xmm4	addl	INCX, XX	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2,  %xmm1	addps	%xmm4,  %xmm3	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	ALIGN_3.L85:	testl	$2, MM	je	.L87	movss	0 * SIZE(XX), %xmm1	movss	1 * SIZE(XX), %xmm2	addl	INCX, XX	movss	0 * SIZE(XX), %xmm3	movss	1 * SIZE(XX), %xmm4	addl	INCX, XX	andps	%xmm7, %xmm1	andps	%xmm7, %xmm2	andps	%xmm7, %xmm3	andps	%xmm7, %xmm4	addps	%xmm2, %xmm1	addps	%xmm4, %xmm3	incl	RET	comiss	%xmm0, %xmm1	je	.L999	incl	RET	comiss	%xmm0, %xmm3	je	.L999	ALIGN_3	.L87:	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 + -