📄 rot_sse.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#undef MOVSD#if defined(CORE2) || defined(PENRYN)#define MOVSD		movsd#endif#ifdef PENTIUM4#define PREFETCH	prefetcht0#define PREFETCH_SIZE	144#define MOVSD		movsd#endif#if defined(OPTERON) || defined(BARCELONA)#define PREFETCH	prefetchw#define PREFETCH_SIZE	(16 * 16)#ifdef OPTERON#define MOVSD		movlps#else#define MOVSD		movsd#endif#endif#define C	%xmm14#define S	%xmm15	PROLOGUE	PROFCODE#ifdef WINDOWS_ABI	movq	40(%rsp), INCY	movss	48(%rsp), %xmm0	movss	56(%rsp), %xmm1#endif	SAVEREGISTERS	leaq	(, INCX, SIZE), INCX		leaq	(, INCY, SIZE), INCY		pshufd	$0x0, %xmm0, C	pshufd	$0x0, %xmm1, S	cmpq	$0, N	jle	.L999	cmpq	$SIZE, INCX	jne	.L50	cmpq	$SIZE, INCY	jne	.L50	testq	$SIZE, X	je	.L05	movss	0 * SIZE(Y), %xmm1	movss	0 * SIZE(X), %xmm0	movaps	%xmm1, %xmm2	movaps	%xmm0, %xmm3	mulss	C, %xmm0	mulss	S, %xmm1	mulss	C, %xmm2	mulss	S, %xmm3	addss	%xmm1, %xmm0	subss	%xmm3, %xmm2	movss	%xmm0, 0 * SIZE(X)	movss	%xmm2, 0 * SIZE(Y)	addq	$1 * SIZE, X	addq	$1 * SIZE, Y	decq	N	jle	.L999.L05:	testq	$2 * SIZE, X	je	.L10	cmpq	$1, N	je	.L17	movsd	0 * SIZE(Y), %xmm1	movsd	0 * SIZE(X), %xmm0	movaps	%xmm1, %xmm2	movaps	%xmm0, %xmm3	mulps	C, %xmm0	mulps	S, %xmm1	mulps	C, %xmm2	mulps	S, %xmm3	addps	%xmm1, %xmm0	subps	%xmm3, %xmm2	movlps	%xmm0, 0 * SIZE(X)	movlps	%xmm2, 0 * SIZE(Y)	addq	$2 * SIZE, X	addq	$2 * SIZE, Y	subq	$2, N	jle	.L999	ALIGN_2.L10:	testq	$3 * SIZE, Y	jne	.L20	movq	N,  %rax	sarq	$5, %rax	jle	.L14	movaps	 0 * SIZE(Y), %xmm1	movaps	 4 * SIZE(Y), %xmm3	movaps	 8 * SIZE(Y), %xmm9	movaps	12 * SIZE(Y), %xmm11	movaps	 0 * SIZE(X), %xmm0	movaps	 4 * SIZE(X), %xmm2	movaps	 8 * SIZE(X), %xmm8	movaps	12 * SIZE(X), %xmm10		decq	%rax	jle	.L12	ALIGN_3.L11:#ifdef PENTIUM4	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(X)#endif	movaps	%xmm1, %xmm4	mulps	S, %xmm1#if defined(OPTERON) || defined(BARCELONA)	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(X)#endif	movaps	%xmm3, %xmm6	mulps	S, %xmm3	movaps	%xmm0, %xmm5	mulps	C, %xmm0	movaps	%xmm2, %xmm7	mulps	C, %xmm2	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	movaps	16 * SIZE(Y), %xmm1	addps	%xmm3, %xmm2	movaps	20 * SIZE(Y), %xmm3	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 0 * SIZE(X)	movaps	16 * SIZE(X), %xmm0#if defined(OPTERON) || defined(BARCELONA)	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(Y)#endif	movaps	%xmm2, 4 * SIZE(X)	movaps	20 * SIZE(X), %xmm2	movaps	%xmm4, 0 * SIZE(Y)	movaps	%xmm6, 4 * SIZE(Y)	movaps	%xmm9, %xmm4	mulps	S, %xmm9	movaps	%xmm8, %xmm5	mulps	C, %xmm8	movaps	%xmm11, %xmm6	mulps	S, %xmm11	movaps	%xmm10, %xmm7	mulps	C, %xmm10	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm9,  %xmm8	movaps	24 * SIZE(Y), %xmm9	addps	%xmm11, %xmm10	movaps	28 * SIZE(Y), %xmm11	subps	%xmm5,  %xmm4	subps	%xmm7,  %xmm6	movaps	%xmm8,  8 * SIZE(X)	movaps	24 * SIZE(X), %xmm8	movaps	%xmm10,12 * SIZE(X)	movaps	28 * SIZE(X), %xmm10	movaps	%xmm4,  8 * SIZE(Y)	movaps	%xmm6, 12 * SIZE(Y)#ifdef PENTIUM4	PREFETCH	(PREFETCH_SIZE + 0) * SIZE(Y)#endif#if defined(OPTERON) || defined(BARCELONA)	PREFETCH	(PREFETCH_SIZE + 16) * SIZE(X)#endif	movaps	%xmm1, %xmm4	mulps	S, %xmm1	movaps	%xmm3, %xmm6	mulps	S, %xmm3	movaps	%xmm0, %xmm5	mulps	C, %xmm0	movaps	%xmm2, %xmm7	mulps	C, %xmm2	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	movaps	32 * SIZE(Y), %xmm1	addps	%xmm3, %xmm2	movaps	36 * SIZE(Y), %xmm3	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 16 * SIZE(X)	movaps	32 * SIZE(X), %xmm0	movaps	%xmm2, 20 * SIZE(X)	movaps	36 * SIZE(X), %xmm2	movaps	%xmm4, 16 * SIZE(Y)	movaps	%xmm6, 20 * SIZE(Y)#if defined(OPTERON) || defined(BARCELONA)	PREFETCH	(PREFETCH_SIZE + 16) * SIZE(Y)#endif	movaps	%xmm9, %xmm4	mulps	S, %xmm9	movaps	%xmm8, %xmm5	mulps	C, %xmm8	movaps	%xmm11, %xmm6	mulps	S, %xmm11	movaps	%xmm10, %xmm7	mulps	C, %xmm10	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm9,  %xmm8	movaps	40 * SIZE(Y), %xmm9	addps	%xmm11, %xmm10	movaps	44 * SIZE(Y), %xmm11	subps	%xmm5,  %xmm4	subps	%xmm7,  %xmm6	movaps	%xmm8,  24 * SIZE(X)	movaps	40 * SIZE(X), %xmm8	movaps	%xmm10, 28 * SIZE(X)	movaps	44 * SIZE(X), %xmm10	movaps	%xmm4,  24 * SIZE(Y)	movaps	%xmm6,  28 * SIZE(Y)	addq	$32 * SIZE, X	addq	$32 * SIZE, Y	decq	%rax	jg	.L11	ALIGN_3.L12:	movaps	%xmm1, %xmm4	mulps	S, %xmm1	movaps	%xmm3, %xmm6	mulps	S, %xmm3	movaps	%xmm0, %xmm5	mulps	C, %xmm0	movaps	%xmm2, %xmm7	mulps	C, %xmm2	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	movaps	16 * SIZE(Y), %xmm1	addps	%xmm3, %xmm2	movaps	20 * SIZE(Y), %xmm3	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 0 * SIZE(X)	movaps	16 * SIZE(X), %xmm0	movaps	%xmm2, 4 * SIZE(X)	movaps	20 * SIZE(X), %xmm2	movaps	%xmm4, 0 * SIZE(Y)	movaps	%xmm6, 4 * SIZE(Y)	movaps	%xmm9, %xmm4	mulps	S, %xmm9	movaps	%xmm8, %xmm5	mulps	C, %xmm8	movaps	%xmm11, %xmm6	mulps	S, %xmm11	movaps	%xmm10, %xmm7	mulps	C, %xmm10	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm9,  %xmm8	movaps	24 * SIZE(Y), %xmm9	addps	%xmm11, %xmm10	movaps	28 * SIZE(Y), %xmm11	subps	%xmm5,  %xmm4	subps	%xmm7,  %xmm6	movaps	%xmm8,  8 * SIZE(X)	movaps	24 * SIZE(X), %xmm8	movaps	%xmm10,12 * SIZE(X)	movaps	28 * SIZE(X), %xmm10	movaps	%xmm4,  8 * SIZE(Y)	movaps	%xmm6, 12 * SIZE(Y)	movaps	%xmm1, %xmm4	mulps	S, %xmm1	movaps	%xmm3, %xmm6	mulps	S, %xmm3	movaps	%xmm0, %xmm5	mulps	C, %xmm0	movaps	%xmm2, %xmm7	mulps	C, %xmm2	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	addps	%xmm3, %xmm2	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 16 * SIZE(X)	movaps	%xmm2, 20 * SIZE(X)	movaps	%xmm4, 16 * SIZE(Y)	movaps	%xmm6, 20 * SIZE(Y)	movaps	%xmm9, %xmm4	mulps	S, %xmm9	movaps	%xmm8, %xmm5	mulps	C, %xmm8	movaps	%xmm11, %xmm6	mulps	S, %xmm11	movaps	%xmm10, %xmm7	mulps	C, %xmm10	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm9,  %xmm8	addps	%xmm11, %xmm10	subps	%xmm5,  %xmm4	subps	%xmm7,  %xmm6	movaps	%xmm8,  24 * SIZE(X)	movaps	%xmm10, 28 * SIZE(X)	movaps	%xmm4,  24 * SIZE(Y)	movaps	%xmm6,  28 * SIZE(Y)	addq	$32 * SIZE, X	addq	$32 * SIZE, Y	ALIGN_3.L14:	testq	$31, N	jle	.L999	testq	$16, N	jle	.L15	movaps	0 * SIZE(Y), %xmm1	movaps	0 * SIZE(X), %xmm0	movaps	4 * SIZE(Y), %xmm3	movaps	4 * SIZE(X), %xmm2	movaps	%xmm1, %xmm4	movaps	%xmm0, %xmm5	movaps	%xmm3, %xmm6	movaps	%xmm2, %xmm7	mulps	C, %xmm0	mulps	S, %xmm1	mulps	C, %xmm2	mulps	S, %xmm3	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	addps	%xmm3, %xmm2	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 0 * SIZE(X)	movaps	%xmm2, 4 * SIZE(X)	movaps	%xmm4, 0 * SIZE(Y)	movaps	%xmm6, 4 * SIZE(Y)	movaps	 8 * SIZE(Y), %xmm1	movaps	 8 * SIZE(X), %xmm0	movaps	12 * SIZE(Y), %xmm3	movaps	12 * SIZE(X), %xmm2	movaps	%xmm1, %xmm4	movaps	%xmm0, %xmm5	movaps	%xmm3, %xmm6	movaps	%xmm2, %xmm7	mulps	C, %xmm0	mulps	S, %xmm1	mulps	C, %xmm2	mulps	S, %xmm3	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	addps	%xmm3, %xmm2	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0,  8 * SIZE(X)	movaps	%xmm2, 12 * SIZE(X)	movaps	%xmm4,  8 * SIZE(Y)	movaps	%xmm6, 12 * SIZE(Y)	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	ALIGN_3.L15:	testq	$8, N	jle	.L16	movaps	0 * SIZE(Y), %xmm1	movaps	0 * SIZE(X), %xmm0	movaps	4 * SIZE(Y), %xmm3	movaps	4 * SIZE(X), %xmm2	movaps	%xmm1, %xmm4	movaps	%xmm0, %xmm5	movaps	%xmm3, %xmm6	movaps	%xmm2, %xmm7	mulps	C, %xmm0	mulps	S, %xmm1	mulps	C, %xmm2	mulps	S, %xmm3	mulps	C, %xmm4	mulps	S, %xmm5	mulps	C, %xmm6	mulps	S, %xmm7	addps	%xmm1, %xmm0	addps	%xmm3, %xmm2	subps	%xmm5, %xmm4	subps	%xmm7, %xmm6	movaps	%xmm0, 0 * SIZE(X)	movaps	%xmm2, 4 * SIZE(X)	movaps	%xmm4, 0 * SIZE(Y)	movaps	%xmm6, 4 * SIZE(Y)	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L16:	testq	$4, N	jle	.L17	movaps	0 * SIZE(Y), %xmm1	movaps	0 * SIZE(X), %xmm0	movaps	%xmm1, %xmm2	movaps	%xmm0, %xmm3	mulps	C, %xmm0	mulps	S, %xmm1	mulps	C, %xmm2	mulps	S, %xmm3	addps	%xmm1, %xmm0	subps	%xmm3, %xmm2	movaps	%xmm0, 0 * SIZE(X)	movaps	%xmm2, 0 * SIZE(Y)	addq	$4 * SIZE, X	addq	$4 * SIZE, Y	ALIGN_3.L17:	testq	$2, N	jle	.L18
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 + -