📄 asum.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 M ARG1#define X ARG2#define INCX ARG3#define I %rax PROLOGUE PROFCODE#ifdef F_INTERFACE movslq (M), M movslq (INCX), INCX#endif fldz testq M, M jle .L999 testq INCX, INCX jle .L999 salq $BASE_SHIFT, INCX fldz fldz fldz cmpq $SIZE, INCX jne .L40 movq M, I sarq $3, I jle .L20 ALIGN_4 .L10: FLD 0 * SIZE(X) fabs FLD 1 * SIZE(X) fabs FLD 2 * SIZE(X) fabs FLD 3 * SIZE(X) fabs faddp %st, %st(7) faddp %st, %st(5) faddp %st, %st(3) faddp %st, %st(1) FLD 4 * SIZE(X) fabs FLD 5 * SIZE(X) fabs FLD 6 * SIZE(X) fabs FLD 7 * SIZE(X) fabs addq $8 * SIZE, X faddp %st, %st(7) faddp %st, %st(5) faddp %st, %st(3) faddp %st, %st(1) decq I jg .L10 ALIGN_4.L20: andq $7, M jle .L998 ALIGN_4.L21: FLD (X) fabs faddp %st,%st(1) addq $1 * SIZE, X decq M jg .L21 jmp .L998 ALIGN_4.L40: movq M, I sarq $3, I jle .L60 ALIGN_4 .L50: FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs faddp %st, %st(7) faddp %st, %st(5) faddp %st, %st(3) faddp %st, %st(1) FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs FLD (X) addq INCX, X fabs faddp %st, %st(7) faddp %st, %st(5) faddp %st, %st(3) faddp %st, %st(1) decq I jg .L50 ALIGN_4.L60: andq $7, M jle .L998 ALIGN_4.L61: FLD (X) addq INCX, X fabs faddp %st,%st(1) decq M jg .L61 ALIGN_4.L998: faddp %st,%st(2) faddp %st,%st(1) faddp %st,%st(1) ALIGN_4.L999: 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 + -