📄 zgemv_t.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"#ifdef PENTIUM#define P 88#endif#ifndef P#define P 400#endif#define STACK	16#define ARGS	24	#define NLDA	  0 + STACK(%esp)#define XP	  4 + STACK(%esp)#define MIN_M	  8 + STACK(%esp)#define J	 12 + STACK(%esp)#define IS	 16 + STACK(%esp)#define M	 4 + STACK + ARGS(%esp)#define N	 8 + STACK + ARGS(%esp)#define K	12 + STACK + ARGS(%esp)#ifdef DOUBLE#define ALPHA_R	16 + STACK + ARGS(%esp)#define ALPHA_I	24 + STACK + ARGS(%esp)#define A	32 + STACK + ARGS(%esp)#define LDA	36 + STACK + ARGS(%esp)#define X	40 + STACK + ARGS(%esp)#define INCX	44 + STACK + ARGS(%esp)#define Y	48 + STACK + ARGS(%esp)#define INCY	52 + STACK + ARGS(%esp)#define BUFFER	56 + STACK + ARGS(%esp)#else#define ALPHA_R	16 + STACK + ARGS(%esp)#define ALPHA_I	20 + STACK + ARGS(%esp)#define A	24 + STACK + ARGS(%esp)#define LDA	28 + STACK + ARGS(%esp)#define X	32 + STACK + ARGS(%esp)#define INCX	36 + STACK + ARGS(%esp)#define Y	40 + STACK + ARGS(%esp)#define INCY	44 + STACK + ARGS(%esp)#define BUFFER	48 + STACK + ARGS(%esp)#endif	PROLOGUE	subl	$ARGS,	%esp	pushl	%ebp	pushl	%edi	pushl	%esi	pushl	%ebx	PROFCODE	FLD	ALPHA_I	FLD	ALPHA_R	movl	X, %edi			# X	movl	$0, IS	movl	M, %ebx	movl	N, %ecx	testl	%ebx, %ebx	jle	.L79	testl	%ecx, %ecx	jle	.L79	movl	INCX, %esi	addl	%esi, %esi	leal	(,%esi,SIZE), %esi	movl	%esi, INCX	movl	INCY, %esi	addl	%esi, %esi	leal	(, %esi, SIZE), %esi	movl	%esi, INCY	movl	LDA,  %ebx	movl	N,    %eax	imull	%ebx, %eax	movl	$P,   %esi	subl	%eax, %esi	leal	(, %esi, SIZE), %esi	addl	%esi, %esi	movl	%esi, NLDA	leal	(,%ebx,SIZE), %esi	addl	%esi, %esi	movl	%esi, LDA	ALIGN_2.L32:	movl	IS,   %esi	movl	$P,   %edx	movl	M,    %eax	subl	%esi, %eax	cmpl	%edx, %eax#ifdef PENTIUM	jle	.L33	movl	%edx, %eax.L33:#else	cmovg	%edx, %eax#endif	movl	%eax, MIN_M	movl	IS,   %ecx	addl	%ecx, %ecx	leal	(%edi,%ecx,SIZE), %ecx		# xp = x + is	movl	INCX, %ebx	movl	%ecx, XP	cmpl	$2 * SIZE, %ebx	je	.L34	movl	BUFFER, %esi	movl	MIN_M, %eax	movl	%esi, XP	sarl	$1, %eax	jle	.L35	ALIGN_3.L36:	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%ebx,%edi		# x += incx	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%ebx,%edi		# x += incx	FST	3 * SIZE(%esi)	FST	2 * SIZE(%esi)	FST	1 * SIZE(%esi)	FST	0 * SIZE(%esi)	addl	$4 * SIZE, %esi		# xp += 4	decl	%eax	jg	.L36	ALIGN_3.L35:	movl	MIN_M, %eax	andl	$1,%eax	jle	.L34	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%ebx,%edi		# x += incx	FST	1 * SIZE(%esi)	FST	0 * SIZE(%esi)	ALIGN_3/* Main Routine */.L34:	movl	Y, %ebp			# coffset = y	movl	N, %ecx	testl	%ecx, %ecx	jle	.L60	ALIGN_2.L61:	movl	A, %ebx			# a_offset = a	fldz				# ct1 = ZERO	movl	LDA, %edx	fldz				# ct1 = ZERO	addl	%ebx, %edx	fldz				# ct1 = ZERO	movl	%edx, A	fldz				# ct1 = ZERO	movl	XP, %esi	FLD	(%esi)			#  bt1 = *(b_offset + 0)	movl	MIN_M, %eax	sarl	$1,    %eax	jle	.L64	ALIGN_3#define PRESIZE 8.L65:#ifdef HAS_PREFETCH       prefetcht0	PRESIZE * SIZE(%ebx)       prefetcht0	PRESIZE * SIZE(%esi)#endif	FLD	 0 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 1 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)#ifndef CONJ	faddp	%st, %st(2)		# ct2 += bt1#else	fsubrp	%st, %st(2)		# ct2 -= bt1#endif	FLD	 1 * SIZE(%esi)		# bt1  = *(b_offset + 1)	FLD	 0 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 1 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	FLD	 2 * SIZE(%esi)		# bt1  = *(b_offset + 1)	FLD	 2 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 3 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)#ifndef CONJ	faddp	%st, %st(2)		# ct2 += bt1#else	fsubrp	%st, %st(2)		# ct2 -= bt1#endif	FLD	 3 * SIZE(%esi)		# bt1  = *(b_offset + 1)	FLD	 2 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 3 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	FLD	 4 * SIZE(%esi)		# bt1  = *(b_offset + 1)	addl	$4 * SIZE, %esi	addl	$4 * SIZE, %ebx	decl	%eax	jg	.L65	ALIGN_3.L64:	movl	MIN_M, %eax	andl	$1, %eax	jle	.L70	ALIGN_3.L71:	FLD	 0 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 1 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)#ifndef CONJ	faddp	%st, %st(2)		# ct2 += bt1#else	fsubrp	%st, %st(2)		# ct2 -= bt1#endif	FLD	 1 * SIZE(%esi)		# bt1  = *(b_offset + 1)	FLD	 0 * SIZE(%ebx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 1 * SIZE(%ebx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	fldz	ALIGN_3.L70:	ffreep	%st(0)#ifndef XCONJ#ifndef CONJ	fsubp	%st, %st(3)	faddp	%st, %st(1)#else	faddp	%st, %st(3)	faddp	%st, %st(1)#endif#else#ifndef CONJ	faddp	%st, %st(3)	fsubp	%st, %st(1)#else	fsubp	%st, %st(3)	fsubp	%st, %st(1)#endif#endif	fld	%st(0)		# ct4 = ct2	fmul	%st(4)	fld	%st(2)	fmul	%st(4)	fsubp	%st, %st(1)	FADD	0 * SIZE(%ebp)	FST	0 * SIZE(%ebp)	fmul	%st(2)	fxch	%st(1)	fmul	%st(3)	faddp	%st, %st(1)	FADD	1 * SIZE(%ebp)	FST	1 * SIZE(%ebp)	addl	INCY, %ebp	decl	%ecx	jg	.L61	ALIGN_3.L60:	movl	A, %ebx	addl	NLDA, %ebx	movl	%ebx, A	addl	$P,  IS	movl	M, %esi	cmpl	%esi, IS	jl	.L32	ALIGN_3.L79:	ffreep	%st(0)	ffreep	%st(0)	popl	%ebx	popl	%esi	popl	%edi	popl	%ebp	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 + -