📄 xdot.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 #if defined(DOUBLE) && defined(F_INTERFACE_GFORT)#define F_INTERFACE_F2C#endif#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C))#define RESULT 4 + STACK + ARGS(%esp)#define STACK_N 8 + STACK + ARGS(%esp)#define STACK_X 12 + STACK + ARGS(%esp)#define STACK_INCX 16 + STACK + ARGS(%esp)#define STACK_Y 20 + STACK + ARGS(%esp)#define STACK_INCY 24 + STACK + ARGS(%esp)#else#define STACK_N 4 + STACK + ARGS(%esp)#define STACK_X 8 + STACK + ARGS(%esp)#define STACK_INCX 12 + STACK + ARGS(%esp)#define STACK_Y 16 + STACK + ARGS(%esp)#define STACK_INCY 20 + STACK + ARGS(%esp)#endif PROLOGUE pushl %edi pushl %esi pushl %ebx PROFCODE#define N %ebx#define X %esi#define INCX %ecx#define Y %edi#define INCY %edx movl STACK_N, N movl STACK_X, X movl STACK_INCX, INCX movl STACK_Y, Y movl STACK_INCY, INCY#ifdef F_INTERFACE movl (N),N movl (INCX),INCX movl (INCY),INCY#endif testl N, N jle .L88 sall $ZBASE_SHIFT, INCX sall $ZBASE_SHIFT, INCY fldz fldz fldz fldz cmpl $2 * SIZE, INCX jne .L14 cmpl $2 * SIZE, INCY jne .L14 movl N, %eax sarl $1, %eax jle .L15 ALIGN_3.L16: FLD 0 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 1 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) FLD 2 * SIZE(X) FLD 2 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 3 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 3 * SIZE(X) FLD 2 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 3 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) addl $4 * SIZE, X addl $4 * SIZE, Y decl %eax jg .L16 ALIGN_3.L15: movl N, %eax andl $1, %eax jle .L27 ALIGN_3.L22: FLD 0 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 1 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) jmp .L27 ALIGN_3.L14:#ifdef F_INTERFACE testl INCX, INCX # if (incx < 0) jge .L28 movl N, %eax decl %eax imull INCX, %eax subl %eax, X ALIGN_3.L28: testl INCY, INCY # if (incy < 0) jge .L29 movl N, %eax decl %eax imull INCY, %eax subl %eax, Y ALIGN_3.L29:#endif movl N, %eax sarl $1, %eax jle .L30 ALIGN_3.L31: FLD 0 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 1 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) addl INCX, X FLD 0 * SIZE(X) addl INCY, Y FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 1 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) addl INCX, X addl INCY, Y decl %eax jg .L31 ALIGN_3.L30: movl N, %eax andl $1, %eax jle .L27 ALIGN_3.L37: FLD 0 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(2) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(2) FLD 1 * SIZE(X) FLD 0 * SIZE(Y) fmul %st(1) faddp %st, %st(4) FLD 1 * SIZE(Y) fmulp %st, %st(1) faddp %st, %st(4) ALIGN_3.L27:#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) movl RESULT, %eax#endif#ifndef CONJ fsubp %st, %st(3) faddp %st, %st(1)#else faddp %st, %st(3) fsubp %st, %st(1)#endif#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) FST 1 * SIZE(%eax) FST 0 * SIZE(%eax)#elif defined(F_INTERFACE) && defined(F_INTERFACE_GFORT) && !defined(DOUBLE) subl $2 * SIZE, %esp FST 1 * SIZE(%esp) FST 0 * SIZE(%esp) movl 0 * SIZE(%esp), %eax movl 1 * SIZE(%esp), %edx addl $2 * SIZE, %esp#else fxch %st(1)#endif popl %ebx popl %esi popl %edi#if defined(F_INTERFACE) && defined(F_PATHSCALE) ret $0x4#else ret#endif ALIGN_3.L88:#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) movl RESULT, %eax#endif fldz fldz#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) FST 1 * SIZE(%eax) FST 0 * SIZE(%eax)#elif defined(F_INTERFACE) && defined(F_INTERFACE_GFORT) && !defined(DOUBLE) xor %eax, %eax xor %edx, %edx#endif popl %ebx popl %esi popl %edi#if defined(F_INTERFACE) && defined(F_PATHSCALE) ret $0x4#else ret#endif 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 + -