📄 dot_sse_core2.s

📁 Optimized GotoBLAS libraries
💻 S
📖 第 1 页 / 共 2 页
字号:
12 下一页
/*********************************************************************//*                                                                   *//*             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 N	ARG1	/* rdi */#define X	ARG2	/* rsi */#define INCX	ARG3	/* rdx */#define Y	ARG4	/* rcx */#ifndef WINDOWS_ABI#define INCY	ARG5	/* r8  */#else#define INCY	%r10#endif	PROLOGUE	PROFCODE#ifdef WINDOWS_ABI	movq	40(%rsp), INCY#endif	SAVEREGISTERS#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	leaq	(, INCX, SIZE), INCX		leaq	(, INCY, SIZE), INCY		pxor	%xmm0, %xmm0	pxor	%xmm1, %xmm1	pxor	%xmm2, %xmm2	pxor	%xmm3, %xmm3	cmpq	$0, N	jle	.L999	cmpq	$SIZE, INCX	jne	.L50	cmpq	$SIZE, INCY	jne	.L50	subq	$-32 * SIZE, X	subq	$-32 * SIZE, Y	cmpq	$3, N	jle	.L17	testq	$SIZE, Y	je	.L05	movss	-32 * SIZE(X), %xmm0	mulss	-32 * SIZE(Y), %xmm0	addq	$1 * SIZE, X	addq	$1 * SIZE, Y	decq	N	ALIGN_2.L05:	testq	$2 * SIZE, Y	je	.L10	movsd	-32 * SIZE(X), %xmm4	movsd	-32 * SIZE(Y), %xmm1	mulps	%xmm4, %xmm1	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	subq	$2, N	jle	.L999	ALIGN_2.L10:	testq	$3 * SIZE, X	jne	.L20	movq	N,  %rax	sarq	$5, %rax	jle	.L14	movaps	-32 * SIZE(X), %xmm12	pxor	%xmm4, %xmm4	movaps	-28 * SIZE(X), %xmm13	pxor	%xmm5, %xmm5	movaps	-24 * SIZE(X), %xmm14	pxor	%xmm6, %xmm6	movaps	-20 * SIZE(X), %xmm15	pxor	%xmm7, %xmm7	movaps	-32 * SIZE(Y), %xmm8	movaps	-28 * SIZE(Y), %xmm9	movaps	-24 * SIZE(Y), %xmm10	movaps	-20 * SIZE(Y), %xmm11	subq	$1, %rax	jle	.L12	ALIGN_3.L11:	addps	%xmm4,  %xmm0	movaps	-16 * SIZE(X), %xmm4	mulps	%xmm8,  %xmm12	movaps	-16 * SIZE(Y), %xmm8	addps	%xmm5,  %xmm1	movaps	-12 * SIZE(X), %xmm5	mulps	%xmm9,  %xmm13	movaps	-12 * SIZE(Y), %xmm9	addps	%xmm6,  %xmm2	movaps	 -8 * SIZE(X), %xmm6	mulps	%xmm10, %xmm14	movaps	 -8 * SIZE(Y), %xmm10	addps	%xmm7,  %xmm3	movaps	 -4 * SIZE(X), %xmm7	mulps	%xmm11, %xmm15	movaps	 -4 * SIZE(Y), %xmm11	addps	%xmm12,  %xmm0	movaps	  0 * SIZE(X), %xmm12	mulps	%xmm8,  %xmm4	movaps	  0 * SIZE(Y), %xmm8	addps	%xmm13,  %xmm1	movaps	  4 * SIZE(X), %xmm13	mulps	%xmm9,  %xmm5	movaps	  4 * SIZE(Y), %xmm9	addps	%xmm14,  %xmm2	movaps	  8 * SIZE(X), %xmm14	mulps	%xmm10, %xmm6	movaps	  8 * SIZE(Y), %xmm10	addps	%xmm15,  %xmm3	movaps	 12 * SIZE(X), %xmm15	mulps	%xmm11, %xmm7	subq	$-32 * SIZE, X	movaps	 12 * SIZE(Y), %xmm11	subq	$-32 * SIZE, Y	subq	$1, %rax	jg,pt	.L11	ALIGN_3.L12:	addps	%xmm4,  %xmm0	movaps	-16 * SIZE(X), %xmm4	mulps	%xmm8,  %xmm12	movaps	-16 * SIZE(Y), %xmm8	addps	%xmm5,  %xmm1	movaps	-12 * SIZE(X), %xmm5	mulps	%xmm9,  %xmm13	movaps	-12 * SIZE(Y), %xmm9	addps	%xmm6,  %xmm2	movaps	 -8 * SIZE(X), %xmm6	mulps	%xmm10, %xmm14	movaps	 -8 * SIZE(Y), %xmm10	addps	%xmm7,  %xmm3	movaps	 -4 * SIZE(X), %xmm7	mulps	%xmm11, %xmm15	movaps	 -4 * SIZE(Y), %xmm11	addps	%xmm12,  %xmm0	mulps	%xmm8,  %xmm4	addps	%xmm13,  %xmm1	mulps	%xmm9,  %xmm5	addps	%xmm14,  %xmm2	mulps	%xmm10, %xmm6	addps	%xmm15,  %xmm3	mulps	%xmm11, %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	subq	$-32 * SIZE, X	addps	%xmm7,  %xmm3	subq	$-32 * SIZE, Y	ALIGN_3.L14:	movq	N,  %rax	andq	$16, %rax	jle	.L15	movaps	-32 * SIZE(X), %xmm4	movaps	-28 * SIZE(X), %xmm5	movaps	-24 * SIZE(X), %xmm6	movaps	-20 * SIZE(X), %xmm7	mulps	-32 * SIZE(Y), %xmm4	mulps	-28 * SIZE(Y), %xmm5	mulps	-24 * SIZE(Y), %xmm6	mulps	-20 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	addps	%xmm7,  %xmm3	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	ALIGN_3.L15:	movq	N,  %rax	andq	$8, %rax	jle	.L16	movaps	-32 * SIZE(X), %xmm4	movaps	-28 * SIZE(X), %xmm5	mulps	-32 * SIZE(Y), %xmm4	mulps	-28 * SIZE(Y), %xmm5	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L16:	movq	N,  %rax	andq	$4, %rax	jle	.L17	movaps	-32 * SIZE(X), %xmm4	mulps	-32 * SIZE(Y), %xmm4	addps	%xmm4,  %xmm2	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L17:	movq	N,  %rax	andq	$2, %rax	jle	.L18	movsd	-32 * SIZE(X), %xmm4	movsd	-32 * SIZE(Y), %xmm8	mulps	%xmm8,  %xmm4	addps	%xmm4,  %xmm3	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	ALIGN_3.L18:	movq	N,  %rax	andq	$1, %rax	jle	.L999	movss	-32 * SIZE(X), %xmm4	mulss	-32 * SIZE(Y), %xmm4	addss	%xmm4,  %xmm0	jmp	.L999	ALIGN_3.L20:	testq	$2 * SIZE, X	jne	.L30	movaps	-33 * SIZE(X), %xmm4	movq	N,  %rax	sarq	$5, %rax	jle	.L23	movaps	-29 * SIZE(X), %xmm5	pxor	%xmm12, %xmm12	movaps	-25 * SIZE(X), %xmm6	pxor	%xmm13, %xmm13	movaps	-21 * SIZE(X), %xmm7	pxor	%xmm14, %xmm14	movaps	-17 * SIZE(X), %xmm8	pxor	%xmm15, %xmm15	decq	%rax	jle .L22	ALIGN_3.L21:	addps	%xmm12, %xmm0	movaps	-13 * SIZE(X), %xmm12	addps	%xmm13, %xmm1	movaps	 -9 * SIZE(X), %xmm13	addps	%xmm14, %xmm2	movaps	 -5 * SIZE(X), %xmm14	addps	%xmm15, %xmm3	movaps	 -1 * SIZE(X), %xmm15	movaps	%xmm5, %xmm9	palignr	$4, %xmm4, %xmm5	mulps	-32 * SIZE(Y), %xmm5	movaps	%xmm6, %xmm10	palignr	$4, %xmm9, %xmm6	mulps	-28 * SIZE(Y), %xmm6	movaps	%xmm7, %xmm11	palignr	$4, %xmm10, %xmm7	mulps	-24 * SIZE(Y), %xmm7	movaps	%xmm8, %xmm4	palignr	$4, %xmm11, %xmm8	mulps	-20 * SIZE(Y), %xmm8	addps	%xmm5, %xmm0	movaps	  3 * SIZE(X), %xmm5	addps	%xmm6, %xmm1	movaps	  7 * SIZE(X), %xmm6	addps	%xmm7, %xmm2	movaps	 11 * SIZE(X), %xmm7	addps	%xmm8, %xmm3	movaps	 15 * SIZE(X), %xmm8	movaps	%xmm12, %xmm9	palignr	$4, %xmm4, %xmm12	mulps	-16 * SIZE(Y), %xmm12	movaps	%xmm13, %xmm10	palignr	$4, %xmm9, %xmm13	mulps	-12 * SIZE(Y), %xmm13	movaps	%xmm14, %xmm11	palignr	$4, %xmm10, %xmm14	mulps	 -8 * SIZE(Y), %xmm14	subq	$-32 * SIZE, X	movaps	%xmm15, %xmm4	palignr	$4, %xmm11, %xmm15	mulps	 -4 * SIZE(Y), %xmm15	subq	$-32 * SIZE, Y	subq	$1, %rax	jg,pt	.L21	ALIGN_3.L22:	addps	%xmm12, %xmm0	movaps	-13 * SIZE(X), %xmm12	addps	%xmm13, %xmm1	movaps	 -9 * SIZE(X), %xmm13	addps	%xmm14, %xmm2	movaps	 -5 * SIZE(X), %xmm14	addps	%xmm15, %xmm3	movaps	 -1 * SIZE(X), %xmm15	movaps	%xmm5, %xmm9	palignr	$4, %xmm4, %xmm5	movaps	%xmm6, %xmm10	palignr	$4, %xmm9, %xmm6	movaps	%xmm7, %xmm11	palignr	$4, %xmm10, %xmm7	movaps	%xmm8, %xmm4	palignr	$4, %xmm11, %xmm8	mulps	-32 * SIZE(Y), %xmm5	mulps	-28 * SIZE(Y), %xmm6	mulps	-24 * SIZE(Y), %xmm7	mulps	-20 * SIZE(Y), %xmm8	addps	%xmm5, %xmm0	addps	%xmm6, %xmm1	addps	%xmm7, %xmm2	addps	%xmm8, %xmm3	movaps	%xmm12, %xmm9	palignr	$4, %xmm4, %xmm12	movaps	%xmm13, %xmm10	palignr	$4, %xmm9, %xmm13	movaps	%xmm14, %xmm11	palignr	$4, %xmm10, %xmm14	movaps	%xmm15, %xmm4	palignr	$4, %xmm11, %xmm15	mulps	-16 * SIZE(Y), %xmm12	mulps	-12 * SIZE(Y), %xmm13	mulps	 -8 * SIZE(Y), %xmm14	mulps	 -4 * SIZE(Y), %xmm15	addps	%xmm12, %xmm0	addps	%xmm13, %xmm1	addps	%xmm14, %xmm2	addps	%xmm15, %xmm3	subq	$-32 * SIZE, X	subq	$-32 * SIZE, Y	ALIGN_3.L23:	testq	$16, N	jle	.L24	ALIGN_3	movaps	-29 * SIZE(X), %xmm5	movaps	-25 * SIZE(X), %xmm6	movaps	-21 * SIZE(X), %xmm7	movaps	-17 * SIZE(X), %xmm8	movaps	%xmm5, %xmm9	movaps	%xmm6, %xmm10	palignr	$4, %xmm4, %xmm5	palignr	$4, %xmm9, %xmm6	movaps	%xmm7, %xmm11	movaps	%xmm8, %xmm4	palignr	$4, %xmm10, %xmm7	palignr	$4, %xmm11, %xmm8	mulps	-32 * SIZE(Y), %xmm5	mulps	-28 * SIZE(Y), %xmm6	mulps	-24 * SIZE(Y), %xmm7	mulps	-20 * SIZE(Y), %xmm8	addps	%xmm5, %xmm0	addps	%xmm6, %xmm1	addps	%xmm7, %xmm2	addps	%xmm8, %xmm3	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	ALIGN_3.L24:	testq	$8, N	jle	.L25	ALIGN_3	movaps	-29 * SIZE(X), %xmm5	movaps	-25 * SIZE(X), %xmm6	movaps	%xmm5, %xmm7	movaps	%xmm6, %xmm8	palignr	$4, %xmm4, %xmm5	palignr	$4, %xmm7, %xmm6	movaps	%xmm8, %xmm4	mulps	-32 * SIZE(Y), %xmm5	mulps	-28 * SIZE(Y), %xmm6	addps	%xmm5,  %xmm0	addps	%xmm6,  %xmm1	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L25:	testq	$4, N	jle	.L26	ALIGN_3	movaps	-29 * SIZE(X), %xmm5	palignr	$4, %xmm4, %xmm5	mulps	-32 * SIZE(Y), %xmm5	addps	%xmm5,  %xmm2	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L26:	testq	$2, N	jle	.L27	ALIGN_3	movsd	-32 * SIZE(X), %xmm4
12 下一页
⌨️ 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 + -