📄 zaxpy_sse.s

📁 Optimized GotoBLAS libraries
💻 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	16#define ARGS	 0	#define STACK_M	 4 + STACK + ARGS(%esp)#ifdef DOUBLE#define ALPHA_R	16 + STACK + ARGS(%esp)#define ALPHA_I	24 + STACK + ARGS(%esp)#define STACK_X	32 + STACK + ARGS(%esp)#define STACK_INCX	36 + STACK + ARGS(%esp)#define STACK_Y	40 + STACK + ARGS(%esp)#define STACK_INCY	44 + STACK + ARGS(%esp)#else#define ALPHA_R	16 + STACK + ARGS(%esp)#define ALPHA_I	20 + STACK + ARGS(%esp)#define STACK_X	24 + STACK + ARGS(%esp)#define STACK_INCX	28 + STACK + ARGS(%esp)#define STACK_Y	32 + STACK + ARGS(%esp)#define STACK_INCY	36 + STACK + ARGS(%esp)#endif#ifndef HAVE_SSE2#define pxor	xorps#define movsd	movlps#endif#ifdef PENTIUM4#define PREFETCHSIZE	128#endif#ifdef OPTERON#define PREFETCHSIZE	176#define movsd		movlps#endif	PROLOGUE	pushl	%edi	pushl	%esi	pushl	%ebx	pushl	%ebp#define M	%ebx#define X	%esi#define INCX	%ecx#define Y	%edi#define INCY	%edx#define YY	%ebp	PROFCODE#ifndef CONJ	movss	ALPHA_R, %xmm6	movss	ALPHA_I, %xmm5	shufps	$0x00, %xmm6, %xmm6	shufps	$0x00, %xmm5, %xmm5	pxor	%xmm7, %xmm7	subps	%xmm5, %xmm7	unpcklps %xmm5, %xmm7#else	movss	ALPHA_R, %xmm6	movss	ALPHA_I, %xmm7	shufps	$0x00, %xmm6, %xmm6	shufps	$0x00, %xmm7, %xmm7	pxor	%xmm5, %xmm5	subps	%xmm6, %xmm5	unpcklps %xmm5, %xmm6#endif	movl	STACK_M,    M	movl	STACK_X,    X	movl	STACK_INCX, INCX	movl	STACK_Y,    Y	movl	STACK_INCY, INCY	leal	(, INCX, SIZE), INCX	leal	(, INCY, SIZE), INCY	addl	INCX, INCX	addl	INCY, INCY	cmpl	$2 * SIZE, INCX	jne	.L20	cmpl	$2 * SIZE, INCY	jne	.L20	movl	M,  %eax	sarl	$3, %eax	jle	.L15	ALIGN_3.L12:#ifdef OPTERON	prefetcht0     (PREFETCHSIZE + 0) * SIZE(X)	prefetchw      (PREFETCHSIZE + 0) * SIZE(Y)#endif	movsd	 0 * SIZE(X), %xmm1	movhps	 2 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm2	movhps	 2 * SIZE(Y), %xmm2	movsd	 4 * SIZE(X), %xmm4	movhps	 6 * SIZE(X), %xmm4	movsd	 4 * SIZE(Y), %xmm5	movhps	 6 * SIZE(Y), %xmm5	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  0 * SIZE(Y)	movhps	%xmm2,  2 * SIZE(Y)	movsd	%xmm5,  4 * SIZE(Y)	movhps	%xmm5,  6 * SIZE(Y)	movsd	 8 * SIZE(X), %xmm1	movhps	10 * SIZE(X), %xmm1	movsd	 8 * SIZE(Y), %xmm2	movhps	10 * SIZE(Y), %xmm2	movsd	12 * SIZE(X), %xmm4	movhps	14 * SIZE(X), %xmm4	movsd	12 * SIZE(Y), %xmm5	movhps	14 * SIZE(Y), %xmm5	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  8 * SIZE(Y)	movhps	%xmm2, 10 * SIZE(Y)	movsd	%xmm5, 12 * SIZE(Y)	movhps	%xmm5, 14 * SIZE(Y)	addl	$16 * SIZE, X	addl	$16 * SIZE, Y	decl	%eax	jg	.L12	ALIGN_3.L15:	testl	$4, M	jle	.L16	movsd	 0 * SIZE(X), %xmm1	movhps	 2 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm2	movhps	 2 * SIZE(Y), %xmm2	movsd	 4 * SIZE(X), %xmm4	movhps	 6 * SIZE(X), %xmm4	movsd	 4 * SIZE(Y), %xmm5	movhps	 6 * SIZE(Y), %xmm5	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  0 * SIZE(Y)	movhps	%xmm2,  2 * SIZE(Y)	movsd	%xmm5,  4 * SIZE(Y)	movhps	%xmm5,  6 * SIZE(Y)	addl	$8 * SIZE, X	addl	$8 * SIZE, Y	ALIGN_3.L16:	testl	$2, M	jle	.L17	movsd	 0 * SIZE(X), %xmm1	movhps	 2 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm2	movhps	 2 * SIZE(Y), %xmm2	movaps	 %xmm1, %xmm0	shufps	 $0xb1, %xmm1, %xmm1	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	addps	%xmm1, %xmm0	addps	%xmm0, %xmm2	movsd	%xmm2,  0 * SIZE(Y)	movhps	%xmm2,  2 * SIZE(Y)	addl	$4 * SIZE, X	addl	$4 * SIZE, Y	ALIGN_3.L17:	testl	$1, M	jle	.L99	movsd	 0 * SIZE(X), %xmm1	movsd	 0 * SIZE(Y), %xmm2	movaps	 %xmm1, %xmm0	shufps	 $0xb1, %xmm1, %xmm1	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	addps	%xmm1, %xmm0	addps	%xmm0, %xmm2	movsd	%xmm2,  0 * SIZE(Y)	jmp	.L99	ALIGN_3.L20:	movl	Y, YY	movl	M,  %eax	sarl	$3, %eax	jle	.L25	ALIGN_3.L22:	movsd	 (X), %xmm1	addl	 INCX, X	movhps	 (X), %xmm1	addl	 INCX, X	movsd	 (Y), %xmm2	addl	 INCY, Y	movhps	 (Y), %xmm2	addl	 INCY, Y	movsd	 (X), %xmm4	addl	 INCX, X	movhps	 (X), %xmm4	addl	 INCX, X	movsd	 (Y), %xmm5	addl	 INCY, Y	movhps	 (Y), %xmm5	addl	 INCY, Y	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  (YY)	addl	 INCY, YY	movhps	%xmm2,  (YY)	addl	 INCY, YY	movsd	%xmm5,  (YY)	addl	 INCY, YY	movhps	%xmm5,  (YY)	addl	 INCY, YY	movsd	 (X), %xmm1	addl	 INCX, X	movhps	 (X), %xmm1	addl	 INCX, X	movsd	 (Y), %xmm2	addl	 INCY, Y	movhps	 (Y), %xmm2	addl	 INCY, Y	movsd	 (X), %xmm4	addl	 INCX, X	movhps	 (X), %xmm4	addl	 INCX, X	movsd	 (Y), %xmm5	addl	 INCY, Y	movhps	 (Y), %xmm5	addl	 INCY, Y	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  (YY)	addl	 INCY, YY	movhps	%xmm2,  (YY)	addl	 INCY, YY	movsd	%xmm5,  (YY)	addl	 INCY, YY	movhps	%xmm5,  (YY)	addl	 INCY, YY	decl	%eax	jg	.L22	ALIGN_3.L25:	testl	$4, M	jle	.L26	movsd	 (X), %xmm1	addl	 INCX, X	movhps	 (X), %xmm1	addl	 INCX, X	movsd	 (Y), %xmm2	addl	 INCY, Y	movhps	 (Y), %xmm2	addl	 INCY, Y	movsd	 (X), %xmm4	addl	 INCX, X	movhps	 (X), %xmm4	addl	 INCX, X	movsd	 (Y), %xmm5	addl	 INCY, Y	movhps	 (Y), %xmm5	addl	 INCY, Y	movaps	 %xmm1, %xmm0	movaps	 %xmm4, %xmm3	shufps	 $0xb1, %xmm1, %xmm1	shufps	 $0xb1, %xmm4, %xmm4	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	mulps	%xmm6, %xmm3	mulps	%xmm7, %xmm4	addps	%xmm1, %xmm0	addps	%xmm4, %xmm3	addps	%xmm0, %xmm2	addps	%xmm3, %xmm5	movsd	%xmm2,  (YY)	addl	 INCY, YY	movhps	%xmm2,  (YY)	addl	 INCY, YY	movsd	%xmm5,  (YY)	addl	 INCY, YY	movhps	%xmm5,  (YY)	addl	 INCY, YY	ALIGN_3.L26:	testl	$2, M	jle	.L27	movsd	 (X), %xmm1	addl	 INCX, X	movhps	 (X), %xmm1	addl	 INCX, X	movsd	 (Y), %xmm2	addl	 INCY, Y	movhps	 (Y), %xmm2	addl	 INCY, Y	movaps	 %xmm1, %xmm0	shufps	 $0xb1, %xmm1, %xmm1	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	addps	%xmm1, %xmm0	addps	%xmm0, %xmm2	movsd	%xmm2,  (YY)	addl	 INCY, YY	movhps	%xmm2,  (YY)	addl	 INCY, YY	ALIGN_3.L27:	testl	$1, M	jle	.L99	movsd	 (X), %xmm1	movsd	 (Y), %xmm2	movaps	 %xmm1, %xmm0	shufps	 $0xb1, %xmm1, %xmm1	mulps	%xmm6, %xmm0	mulps	%xmm7, %xmm1	addps	%xmm1, %xmm0	addps	%xmm0, %xmm2	movsd	%xmm2,  0 * SIZE(YY)	ALIGN_3.L99:	xorl	%eax,%eax	popl	%ebp	popl	%ebx	popl	%esi	popl	%edi	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 + -