📄 axpy_sse2_core2.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"#ifndef WINDOWS_ABI#define M ARG1#define X ARG4#define INCX ARG5#define Y ARG6#define INCY ARG2#else#define M ARG1#define X ARG2#define INCX ARG3#define Y ARG4#define INCY %r10#endif#define YY %r11#define ALPHA %xmm15#define PREFETCH prefetcht0#define PREFETCH_SIZE (8 * 10) PROLOGUE PROFCODE#ifndef WINDOWS_ABI#ifndef XDOUBLE movq 8(%rsp), INCY#else movq 24(%rsp), INCY#endif movaps %xmm0, ALPHA#else movaps %xmm3, ALPHA movq 40(%rsp), X movq 48(%rsp), INCX movq 56(%rsp), Y movq 64(%rsp), INCY#endif SAVEREGISTERS unpcklpd ALPHA, ALPHA leaq (, INCX, SIZE), INCX leaq (, INCY, SIZE), INCY cmpq $SIZE, INCX jne .L40 cmpq $SIZE, INCY jne .L40 testq $SIZE, Y je .L10 movsd 0 * SIZE(X), %xmm0 mulsd ALPHA, %xmm0 addsd 0 * SIZE(Y), %xmm0 movsd %xmm0, 0 * SIZE(Y) addq $1 * SIZE, X addq $1 * SIZE, Y decq M jle .L19 ALIGN_4.L10: testq $SIZE, X jne .L20 subq $-16 * SIZE, X subq $-16 * SIZE, Y movq M, %rax sarq $5, %rax jle .L13 movaps -16 * SIZE(X), %xmm0 movaps -14 * SIZE(X), %xmm1 movaps -12 * SIZE(X), %xmm2 movaps -10 * SIZE(X), %xmm3 mulpd ALPHA, %xmm0 mulpd ALPHA, %xmm1 mulpd ALPHA, %xmm2 mulpd ALPHA, %xmm3 movaps -16 * SIZE(Y), %xmm8 movaps -14 * SIZE(Y), %xmm9 movaps -12 * SIZE(Y), %xmm10 movaps -10 * SIZE(Y), %xmm11 movaps -8 * SIZE(X), %xmm4 movaps -6 * SIZE(X), %xmm5 movaps -4 * SIZE(X), %xmm6 movaps -2 * SIZE(X), %xmm7 addpd %xmm8, %xmm0 movaps -8 * SIZE(Y), %xmm8 mulpd ALPHA, %xmm4 addpd %xmm9, %xmm1 movaps -6 * SIZE(Y), %xmm9 decq %rax jle .L12 ALIGN_3.L11: movaps %xmm0, -16 * SIZE(Y) movaps 0 * SIZE(X), %xmm0 mulpd ALPHA, %xmm5 addpd %xmm10, %xmm2 movaps -4 * SIZE(Y), %xmm10 movaps %xmm1, -14 * SIZE(Y) movaps 2 * SIZE(X), %xmm1 mulpd ALPHA, %xmm6 addpd %xmm11, %xmm3 movaps -2 * SIZE(Y), %xmm11 movaps %xmm2, -12 * SIZE(Y) movaps 4 * SIZE(X), %xmm2 mulpd ALPHA, %xmm7 addpd %xmm8, %xmm4 movaps 0 * SIZE(Y), %xmm8 movaps %xmm3, -10 * SIZE(Y) movaps 6 * SIZE(X), %xmm3 mulpd ALPHA, %xmm0 addpd %xmm9, %xmm5 movaps 2 * SIZE(Y), %xmm9 movaps %xmm4, -8 * SIZE(Y) movaps 8 * SIZE(X), %xmm4 mulpd ALPHA, %xmm1 addpd %xmm10, %xmm6 movaps 4 * SIZE(Y), %xmm10 movaps %xmm5, -6 * SIZE(Y) movaps 10 * SIZE(X), %xmm5 mulpd ALPHA, %xmm2 addpd %xmm11, %xmm7 movaps 6 * SIZE(Y), %xmm11 movaps %xmm6, -4 * SIZE(Y) movaps 12 * SIZE(X), %xmm6 mulpd ALPHA, %xmm3 addpd %xmm8, %xmm0 movaps 8 * SIZE(Y), %xmm8 movaps %xmm7, -2 * SIZE(Y) movaps 14 * SIZE(X), %xmm7 mulpd ALPHA, %xmm4 addpd %xmm9, %xmm1 movaps 10 * SIZE(Y), %xmm9 movaps %xmm0, 0 * SIZE(Y) movaps 16 * SIZE(X), %xmm0 mulpd ALPHA, %xmm5 addpd %xmm10, %xmm2 movaps 12 * SIZE(Y), %xmm10 movaps %xmm1, 2 * SIZE(Y) movaps 18 * SIZE(X), %xmm1 mulpd ALPHA, %xmm6 addpd %xmm11, %xmm3 movaps 14 * SIZE(Y), %xmm11 movaps %xmm2, 4 * SIZE(Y) movaps 20 * SIZE(X), %xmm2 mulpd ALPHA, %xmm7 addpd %xmm8, %xmm4 movaps 16 * SIZE(Y), %xmm8 movaps %xmm3, 6 * SIZE(Y) movaps 22 * SIZE(X), %xmm3 mulpd ALPHA, %xmm0 addpd %xmm9, %xmm5 movaps 18 * SIZE(Y), %xmm9 movaps %xmm4, 8 * SIZE(Y) movaps 24 * SIZE(X), %xmm4 mulpd ALPHA, %xmm1 addpd %xmm10, %xmm6 movaps 20 * SIZE(Y), %xmm10 movaps %xmm5, 10 * SIZE(Y) movaps 26 * SIZE(X), %xmm5 mulpd ALPHA, %xmm2 addpd %xmm11, %xmm7 movaps 22 * SIZE(Y), %xmm11 movaps %xmm6, 12 * SIZE(Y) movaps 28 * SIZE(X), %xmm6 mulpd ALPHA, %xmm3 addpd %xmm8, %xmm0 movaps 24 * SIZE(Y), %xmm8 movaps %xmm7, 14 * SIZE(Y) movaps 30 * SIZE(X), %xmm7 mulpd ALPHA, %xmm4 addpd %xmm9, %xmm1 movaps 26 * SIZE(Y), %xmm9 subq $-32 * SIZE, Y subq $-32 * SIZE, X decq %rax jg,pt .L11 ALIGN_3.L12: movaps %xmm0, -16 * SIZE(Y) movaps 0 * SIZE(X), %xmm0 mulpd ALPHA, %xmm5 addpd %xmm10, %xmm2 movaps -4 * SIZE(Y), %xmm10 movaps %xmm1, -14 * SIZE(Y) movaps 2 * SIZE(X), %xmm1 mulpd ALPHA, %xmm6 addpd %xmm11, %xmm3 movaps -2 * SIZE(Y), %xmm11 movaps %xmm2, -12 * SIZE(Y) movaps 4 * SIZE(X), %xmm2 mulpd ALPHA, %xmm7 addpd %xmm8, %xmm4 movaps 0 * SIZE(Y), %xmm8 movaps %xmm3, -10 * SIZE(Y) movaps 6 * SIZE(X), %xmm3 mulpd ALPHA, %xmm0 addpd %xmm9, %xmm5 movaps 2 * SIZE(Y), %xmm9 movaps %xmm4, -8 * SIZE(Y) movaps 8 * SIZE(X), %xmm4 mulpd ALPHA, %xmm1 addpd %xmm10, %xmm6 movaps 4 * SIZE(Y), %xmm10 movaps %xmm5, -6 * SIZE(Y) movaps 10 * SIZE(X), %xmm5 mulpd ALPHA, %xmm2 addpd %xmm11, %xmm7 movaps 6 * SIZE(Y), %xmm11 movaps %xmm6, -4 * SIZE(Y) movaps 12 * SIZE(X), %xmm6 mulpd ALPHA, %xmm3 addpd %xmm8, %xmm0 movaps 8 * SIZE(Y), %xmm8 movaps %xmm7, -2 * SIZE(Y) movaps 14 * SIZE(X), %xmm7 mulpd ALPHA, %xmm4 addpd %xmm9, %xmm1 movaps 10 * SIZE(Y), %xmm9 movaps %xmm0, 0 * SIZE(Y) mulpd ALPHA, %xmm5 addpd %xmm10, %xmm2 movaps 12 * SIZE(Y), %xmm10 movaps %xmm1, 2 * SIZE(Y) mulpd ALPHA, %xmm6 addpd %xmm11, %xmm3 movaps 14 * SIZE(Y), %xmm11 movaps %xmm2, 4 * SIZE(Y) mulpd ALPHA, %xmm7 addpd %xmm9, %xmm5 movaps %xmm3, 6 * SIZE(Y) addpd %xmm8, %xmm4 movaps %xmm4, 8 * SIZE(Y) addpd %xmm10, %xmm6 movaps %xmm5, 10 * SIZE(Y) addpd %xmm11, %xmm7 movaps %xmm6, 12 * SIZE(Y) movaps %xmm7, 14 * SIZE(Y) subq $-32 * SIZE, X subq $-32 * SIZE, Y ALIGN_3.L13: testq $16, M jle .L15 movaps -16 * SIZE(X), %xmm0 movaps -14 * SIZE(X), %xmm1 movaps -12 * SIZE(X), %xmm2 movaps -10 * SIZE(X), %xmm3 movaps -16 * SIZE(Y), %xmm8 movaps -14 * SIZE(Y), %xmm9 movaps -12 * SIZE(Y), %xmm10 movaps -10 * SIZE(Y), %xmm11 mulpd ALPHA, %xmm0 mulpd ALPHA, %xmm1 mulpd ALPHA, %xmm2 mulpd ALPHA, %xmm3 movaps -8 * SIZE(X), %xmm4 movaps -6 * SIZE(X), %xmm5 movaps -4 * SIZE(X), %xmm6 movaps -2 * SIZE(X), %xmm7 addpd %xmm8, %xmm0 movaps -8 * SIZE(Y), %xmm8 mulpd ALPHA, %xmm4 addpd %xmm9, %xmm1 movaps -6 * SIZE(Y), %xmm9 movaps %xmm0, -16 * SIZE(Y) mulpd ALPHA, %xmm5 addpd %xmm10, %xmm2 movaps -4 * SIZE(Y), %xmm10 movaps %xmm1, -14 * SIZE(Y) mulpd ALPHA, %xmm6 addpd %xmm11, %xmm3 movaps -2 * SIZE(Y), %xmm11 movaps %xmm2, -12 * SIZE(Y) mulpd ALPHA, %xmm7 addpd %xmm9, %xmm5 movaps %xmm3, -10 * SIZE(Y) addpd %xmm8, %xmm4 movaps %xmm4, -8 * SIZE(Y) addpd %xmm10, %xmm6 movaps %xmm5, -6 * SIZE(Y) addpd %xmm11, %xmm7 movaps %xmm6, -4 * SIZE(Y) movaps %xmm7, -2 * SIZE(Y) subq $-16 * SIZE, X subq $-16 * SIZE, Y ALIGN_3.L15: testq $8, M jle .L16 movaps -16 * SIZE(X), %xmm0 mulpd ALPHA, %xmm0 movaps -16 * SIZE(Y), %xmm4 movaps -14 * SIZE(X), %xmm1 mulpd ALPHA, %xmm1 movaps -14 * SIZE(Y), %xmm5 movaps -12 * SIZE(X), %xmm2 mulpd ALPHA, %xmm2 movaps -12 * SIZE(Y), %xmm6 movaps -10 * SIZE(X), %xmm3 mulpd ALPHA, %xmm3 movaps -10 * SIZE(Y), %xmm7 addpd %xmm4, %xmm0 addpd %xmm5, %xmm1 addpd %xmm6, %xmm2 addpd %xmm7, %xmm3 movaps %xmm0, -16 * SIZE(Y) movaps %xmm1, -14 * SIZE(Y) movaps %xmm2, -12 * SIZE(Y) movaps %xmm3, -10 * SIZE(Y) addq $8 * SIZE, X addq $8 * SIZE, Y ALIGN_3.L16: testq $4, M jle .L17 movaps -16 * SIZE(X), %xmm0 mulpd ALPHA, %xmm0 movaps -16 * SIZE(Y), %xmm4 movaps -14 * SIZE(X), %xmm1 mulpd ALPHA, %xmm1 movaps -14 * SIZE(Y), %xmm5 addpd %xmm4, %xmm0 addpd %xmm5, %xmm1 movaps %xmm0, -16 * SIZE(Y) movaps %xmm1, -14 * SIZE(Y) addq $4 * SIZE, X addq $4 * SIZE, Y ALIGN_3.L17: testq $2, M jle .L18 movaps -16 * SIZE(X), %xmm0⌨️ 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 + -