📄 zdot.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"#if defined(F_INTERFACE) && defined(F_INTERFACE_F2C)#define RESULT ARG1 /* rdi */#define N ARG2 /* rsi */#define X ARG3 /* rdx */#define INCX ARG4 /* rcx */#define Y ARG5 /* r8 */#define INCY ARG6 /* r9 */#else#define N ARG1 /* rdi */#define X ARG2 /* rsi */#define INCX ARG3 /* rdx */#define Y ARG4 /* rcx */#define INCY ARG5 /* r8 */#endif PROLOGUE PROFCODE #ifdef F_INTERFACE#ifndef USE64BITINT movslq (N), N # N movslq (INCX),INCX # INCX movslq (INCY),INCY # INCY#else movq (N), N # N movq (INCX),INCX # INCX movq (INCY),INCY # INCY#endif#endif testq N, N jle .L88 salq $ZBASE_SHIFT, INCX salq $ZBASE_SHIFT, INCY fldz fldz fldz fldz cmpq $2 * SIZE, INCX jne .L14 cmpq $2 * SIZE, INCY jne .L14 movq N, %rax sarq $1, %rax 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) addq $4 * SIZE, X addq $4 * SIZE, Y decq %rax jg .L16 ALIGN_3.L15: movq N, %rax andq $1, %rax 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 testq INCX, INCX # if (incx < 0) jge .L28 movq N, %rax decq %rax imulq INCX, %rax subq %rax, X ALIGN_3.L28: testq INCY, INCY # if (incy < 0) jge .L29 movq N, %rax decq %rax imulq INCY, %rax subq %rax, Y ALIGN_3.L29:#endif movq N, %rax sarq $1, %rax 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) addq INCX, X FLD 0 * SIZE(X) addq 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) addq INCX, X addq INCY, Y decq %rax jg .L31 ALIGN_3.L30: movq N, %rax andq $1, %rax 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)) movq RESULT, %rax#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(%rax) FST 0 * SIZE(%rax)#elif defined(F_INTERFACE) && defined(F_INTERFACE_GFORT) && !defined(DOUBLE) subq $2 * SIZE, %rsp FST 1 * SIZE(%rsp) FST 0 * SIZE(%rsp) movq 0 * SIZE(%rsp), %rax movq 1 * SIZE(%rsp), %rdx addq $2 * SIZE, %rsp#else fxch %st(1)#endif ret ALIGN_3.L88:#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) movq RESULT, %rax#endif fldz fldz#if (defined(F_INTERFACE) && defined(F_INTERFACE_F2C)) FST 1 * SIZE(%rax) FST 0 * SIZE(%rax)#elif defined(F_INTERFACE) && defined(F_INTERFACE_GFORT) && !defined(DOUBLE) xor %rax, %rax xor %edx, %edx#endif 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 + -