📄 zcopy.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 0 #define M 4 + STACK + ARGS(%esp)#define X 8 + STACK + ARGS(%esp)#define INCX 12 + STACK + ARGS(%esp)#define Y 16 + STACK + ARGS(%esp)#define INCY 20 + STACK + ARGS(%esp) PROLOGUE pushl %edi pushl %esi pushl %ebx PROFCODE#ifdef F_INTERFACE_GFORT EMMS#endif movl M, %ebx movl X, %ecx movl INCX, %esi movl Y, %edx movl INCY, %edi testl %ebx, %ebx # if m == 0 goto End jle .L999 sall $ZBASE_SHIFT, %esi sall $ZBASE_SHIFT, %edi cmpl $2 * SIZE, %esi # if incx != 1 jne .L100 cmpl $2 * SIZE, %edi # if incy != 1 jne .L100 movl %ebx, %eax # i = m sarl $2, %eax jle .L20 ALIGN_2.L11:#if defined(DOUBLE) || defined(XDOUBLE) FLD 7 * SIZE(%ecx) FLD 6 * SIZE(%ecx) FLD 5 * SIZE(%ecx) FLD 4 * SIZE(%ecx) FLD 3 * SIZE(%ecx) FLD 2 * SIZE(%ecx) FLD 1 * SIZE(%ecx) FLD 0 * SIZE(%ecx) FST 0 * SIZE(%edx) FST 1 * SIZE(%edx) FST 2 * SIZE(%edx) FST 3 * SIZE(%edx) FST 4 * SIZE(%edx) FST 5 * SIZE(%edx) FST 6 * SIZE(%edx) FST 7 * SIZE(%edx)#else fldl 6 * SIZE(%ecx) fldl 4 * SIZE(%ecx) fldl 2 * SIZE(%ecx) fldl 0 * SIZE(%ecx) fstpl 0 * SIZE(%edx) fstpl 2 * SIZE(%edx) fstpl 4 * SIZE(%edx) fstpl 6 * SIZE(%edx)#endif addl $8 * SIZE, %ecx addl $8 * SIZE, %edx decl %eax jg .L11 ALIGN_2.L20: movl %ebx, %eax # i = m andl $3, %eax jle .L99 ALIGN_2.L21:#if defined(DOUBLE) || defined(XDOUBLE) FLD 1 * SIZE(%ecx) FLD 0 * SIZE(%ecx) FST 0 * SIZE(%edx) FST 1 * SIZE(%edx)#else fldl 0 * SIZE(%ecx) fstpl 0 * SIZE(%edx)#endif addl $2 * SIZE, %ecx addl $2 * SIZE, %edx decl %eax jg .L21.L99: xorl %eax,%eax popl %ebx popl %esi popl %edi ret ALIGN_3.L100: movl %ebx, %eax sarl $2, %eax jle .L120 ALIGN_2.L111:#if defined(DOUBLE) || defined(XDOUBLE) FLD 0 * SIZE(%ecx) FLD 1 * SIZE(%ecx) addl %esi, %ecx FLD 0 * SIZE(%ecx) FLD 1 * SIZE(%ecx) addl %esi, %ecx FLD 0 * SIZE(%ecx) FLD 1 * SIZE(%ecx) addl %esi, %ecx FLD 0 * SIZE(%ecx) FLD 1 * SIZE(%ecx) addl %esi, %ecx fxch %st(7) FST 0 * SIZE(%edx) fxch %st(5) FST 1 * SIZE(%edx) addl %edi, %edx fxch %st(3) FST 0 * SIZE(%edx) fxch %st(1) FST 1 * SIZE(%edx) addl %edi, %edx FST 0 * SIZE(%edx) FST 1 * SIZE(%edx) addl %edi, %edx FST 0 * SIZE(%edx) FST 1 * SIZE(%edx) addl %edi, %edx#else fldl 0 * SIZE(%ecx) addl %esi, %ecx fldl 0 * SIZE(%ecx) addl %esi, %ecx fldl 0 * SIZE(%ecx) addl %esi, %ecx fldl 0 * SIZE(%ecx) addl %esi, %ecx fxch %st(3) fstpl 0 * SIZE(%edx) addl %edi, %edx fxch %st(1) fstpl 0 * SIZE(%edx) addl %edi, %edx fstpl 0 * SIZE(%edx) addl %edi, %edx fstpl 0 * SIZE(%edx) addl %edi, %edx#endif decl %eax jg .L111.L120: movl %ebx, %eax andl $3, %eax jle .L999 ALIGN_2.L121: FLD 0 * SIZE(%ecx) FLD 1 * SIZE(%ecx) addl %esi, %ecx fxch %st(1) FST 0 * SIZE(%edx) FST 1 * SIZE(%edx) addl %edi, %edx decl %eax jg .L121.L999: xorl %eax,%eax popl %ebx popl %esi popl %edi 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 + -