📄 zgemv_n.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 32#endif#if defined(PENTIUM4) || defined(ATHLON)#define P (DTB_ENTRIES / 2)#endif	#ifndef P#define P DTB_ENTRIES#endif#define STACK	16#define ARGS	16	#define PLDA_M	  0 + STACK(%esp)#define XP	  4 + STACK(%esp)#define MIN_N	  8 + STACK(%esp)#define IS	 12 + 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	movl	LDA,  %ebx	addl	%ebx, %ebx		# lda *= 2	leal	0(,%ebx,SIZE),%ebx	# EBX : lda	movl	$0, IS	movl	M, %ecx	movl	N, %esi	test	%ecx, %ecx	jle	.L79			# goto END	test	%esi, %esi	jle	.L79			# goto END	movl	INCY, %eax	addl	%eax, %eax		# incy *= 2	leal	(,%eax,SIZE),%eax	movl	%eax, INCY	movl	LDA, %eax	imull	$P,  %eax		# P * lda	subl	M   ,%eax		# P * lda - m	leal	(, %eax, SIZE), %eax	addl	%eax, %eax	movl	%eax, PLDA_M	ALIGN_2.L32:	movl	IS, %esi	movl	$P, %edx	movl	N,   %eax	subl	%esi,%eax		# n - is	cmpl	%edx,  %eax#ifdef PENTIUM	jle	.L33	movl	%edx,  %eax.L33:#else	cmovg	%edx,  %eax#endif	movl	%eax, MIN_N	movl	INCX, %edx	addl	%edx, %edx	addl	%esi, %esi	leal	(%edi, %esi, SIZE), %esi # xp = x + is	movl	%esi, XP	cmpl	$2, %edx	je	.L34			# if incx == 1 goto L34	movl	BUFFER, %esi	leal	(, %edx, SIZE), %edx	movl	%esi, XP		# xp = buffer	sarl	$1,%eax	jle	.L35	ALIGN_2.L36:	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%edx,%edi		# x += incx	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%edx,%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_N, %eax	andl	$1,    %eax	jle	.L34	FLD	0 * SIZE(%edi)	FLD	1 * SIZE(%edi)	addl	%edx,%edi		# x += incx	FST	1 * SIZE(%esi)	FST	0 * SIZE(%esi)	ALIGN_3/* Main Routine */.L34:	movl	 Y, %ecx		# c_offset	movl	 M, %ebp		# j = m	ALIGN_3.L61:	movl	A, %edx			# a_offset = a	fldz	addl	$2 * SIZE, A		# a++	fldz	movl	XP,%esi	fldz	movl	MIN_N,%eax	fldz	FLD	(%esi)			#  bt1 = *(b_offset + 0)	sarl	$1, %eax	jle	.L64	ALIGN_3.L65:#ifdef PENTIUM4       prefetchnta	16 * SIZE(%esi)#endif	FLD	 0 * SIZE(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 1 * SIZE(%edx)		# 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(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 1 * SIZE(%edx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	FLD	 2 * SIZE(%esi)		# bt1  = *(b_offset + 2)	addl	$2 * SIZE, %esi		# b_offset += 2	addl	%ebx,  %edx		# a_offset += lda	FLD	 0 * SIZE(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 1 * SIZE(%edx)		# 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(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 1 * SIZE(%edx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	FLD	 2 * SIZE(%esi)		# bt1  = *(b_offset + 2)	addl	$2 * SIZE, %esi		# b_offset += 2	addl	%ebx,  %edx		# a_offset += lda	decl	%eax	jg	.L65.L64:	movl	MIN_N, %eax	andl	$1, %eax	jle	.L70	ALIGN_2.L71:	FLD	 0 * SIZE(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(2)		# ct1 += at1	FMUL	 1 * SIZE(%edx)		# 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(%edx)		# at1  = *(a_offset + 0)	fmul	%st(1)			# at1 *= bt1	faddp	%st, %st(4)		# ct3 += at1	FMUL	 1 * SIZE(%edx)		# bt1 *= *(a_offset + 1)	faddp	%st, %st(4)		# ct4 += bt1	fldz	ALIGN_2.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)	movl	INCY,  %eax	FADD	0 * SIZE(%ecx)	FST	0 * SIZE(%ecx)	fmul	%st(2)	fxch	%st(1)	fmul	%st(3)	faddp	%st, %st(1)	FADD	1 * SIZE(%ecx)	FST	1 * SIZE(%ecx)	addl	%eax, %ecx	decl	%ebp	jg	.L61.L60:	movl	PLDA_M, %esi	addl	%esi, A		# a += P * lda - m	addl	$P, IS	movl	N, %esi	cmpl	%esi,IS	jl	.L32.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 + -