📄 dot_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	12#define ARGS     0	#define STACK_N		 4 + STACK + ARGS(%esp)#define STACK_X		 8 + STACK + ARGS(%esp)#define STACK_INCX	12 + STACK + ARGS(%esp)#define STACK_Y		16 + STACK + ARGS(%esp)#define STACK_INCY	20 + STACK + ARGS(%esp)#define N	%ecx#define X	%esi#define INCX	%ebx#define Y	%edi#define INCY	%edx	PROLOGUE	pushl	%edi	pushl	%esi	pushl	%ebx	PROFCODE	movl	STACK_N,     N	movl	STACK_X,     X	movl	STACK_INCX,  INCX	movl	STACK_Y,     Y	movl	STACK_INCY,  INCY#ifdef F_INTERFACE	movl	(N),   N		# N	movl	(INCX),INCX		# INCX	movl	(INCY),INCY		# INCY#endif	leal	(, INCX, SIZE), INCX		leal	(, INCY, SIZE), INCY	#ifndef	HAVE_SSE2	xorps	%xmm0, %xmm0	xorps	%xmm1, %xmm1	xorps	%xmm2, %xmm2	xorps	%xmm3, %xmm3#else	pxor	%xmm0, %xmm0	pxor	%xmm1, %xmm1	pxor	%xmm2, %xmm2	pxor	%xmm3, %xmm3#endif	cmpl	$0, N	jle	.L999	cmpl	$SIZE, INCX	jne	.L50	cmpl	$SIZE, INCY	jne	.L50	cmpl	$3, N	jle	.L17	testl	$SIZE, Y	je	.L05	movss	0 * SIZE(X), %xmm0	mulss	0 * SIZE(Y), %xmm0	addl	$1 * SIZE, X	addl	$1 * SIZE, Y	decl	N	ALIGN_2.L05:	testl	$2 * SIZE, Y	je	.L10	movss	 0 * SIZE(X), %xmm4	movss	 1 * SIZE(X), %xmm5	mulss	 0 * SIZE(Y), %xmm4	mulss	 1 * SIZE(Y), %xmm5	addss	%xmm4,  %xmm1	addss	%xmm5,  %xmm2	addl	$2 * SIZE, X	addl	$2 * SIZE, Y	subl	$2, N	jle	.L999	ALIGN_2.L10:	testl	$3 * SIZE, X	jne	.L20	movl	N,  %eax	sarl	$5, %eax	jle	.L14	movaps	 0 * SIZE(X), %xmm4	movaps	 4 * SIZE(X), %xmm5	movaps	 8 * SIZE(X), %xmm6	movaps	12 * SIZE(X), %xmm7	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	mulps	 8 * SIZE(Y), %xmm6	mulps	12 * SIZE(Y), %xmm7	decl	%eax	jle	.L12	ALIGN_3.L11:	addps	%xmm4,  %xmm0	movaps	16 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movaps	20 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movaps	24 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movaps	28 * SIZE(X), %xmm7	mulps	16 * SIZE(Y), %xmm4	mulps	20 * SIZE(Y), %xmm5	mulps	24 * SIZE(Y), %xmm6	mulps	28 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	movaps	32 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movaps	36 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movaps	40 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movaps	44 * SIZE(X), %xmm7	mulps	32 * SIZE(Y), %xmm4	mulps	36 * SIZE(Y), %xmm5	mulps	40 * SIZE(Y), %xmm6	mulps	44 * SIZE(Y), %xmm7	addl	$32 * SIZE, X	addl	$32 * SIZE, Y	decl	%eax	jg	.L11	ALIGN_3.L12:	addps	%xmm4,  %xmm0	movaps	16 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movaps	20 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movaps	24 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movaps	28 * SIZE(X), %xmm7	mulps	16 * SIZE(Y), %xmm4	mulps	20 * SIZE(Y), %xmm5	mulps	24 * SIZE(Y), %xmm6	mulps	28 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	addps	%xmm7,  %xmm3	addl	$32 * SIZE, X	addl	$32 * SIZE, Y	ALIGN_3.L14:	testl	$31, N	jle	.L999	testl	$16, N	jle	.L15	movaps	 0 * SIZE(X), %xmm4	movaps	 4 * SIZE(X), %xmm5	movaps	 8 * SIZE(X), %xmm6	movaps	12 * SIZE(X), %xmm7	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	mulps	 8 * SIZE(Y), %xmm6	mulps	12 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	addps	%xmm7,  %xmm3	addl	$16 * SIZE, X	addl	$16 * SIZE, Y	ALIGN_3.L15:	testl	$8, N	jle	.L16	movaps	 0 * SIZE(X), %xmm4	movaps	 4 * SIZE(X), %xmm5	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addl	$8 * SIZE, X	addl	$8 * SIZE, Y	ALIGN_3.L16:	testl	$4, N	jle	.L17	movaps	 0 * SIZE(X), %xmm4	mulps	 0 * SIZE(Y), %xmm4	addps	%xmm4,  %xmm0	addl	$4 * SIZE, X	addl	$4 * SIZE, Y	ALIGN_3.L17:	testl	$2, N	jle	.L18	movss	 0 * SIZE(X), %xmm4	movss	 1 * SIZE(X), %xmm5	mulss	 0 * SIZE(Y), %xmm4	mulss	 1 * SIZE(Y), %xmm5	addss	%xmm4,  %xmm0	addss	%xmm5,  %xmm1	addl	$2 * SIZE, X	addl	$2 * SIZE, Y	ALIGN_3.L18:	testl	$1, N	jle	.L999	movss	 0 * SIZE(X), %xmm4	mulss	 0 * SIZE(Y), %xmm4	addss	%xmm4,  %xmm0	jmp	.L999	ALIGN_3.L20:	movl	N,  %eax	sarl	$5, %eax	jle	.L24	movlps	 0 * SIZE(X), %xmm4	movhps	 2 * SIZE(X), %xmm4	movlps	 4 * SIZE(X), %xmm5	movhps	 6 * SIZE(X), %xmm5	movlps	 8 * SIZE(X), %xmm6	movhps	10 * SIZE(X), %xmm6	movlps	12 * SIZE(X), %xmm7	movhps	14 * SIZE(X), %xmm7	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	mulps	 8 * SIZE(Y), %xmm6	mulps	12 * SIZE(Y), %xmm7	decl	%eax	jle	.L22	ALIGN_3.L21:	addps	%xmm4,  %xmm0	movlps	16 * SIZE(X), %xmm4	movhps	18 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movlps	20 * SIZE(X), %xmm5	movhps	22 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movlps	24 * SIZE(X), %xmm6	movhps	26 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movlps	28 * SIZE(X), %xmm7	movhps	30 * SIZE(X), %xmm7	mulps	16 * SIZE(Y), %xmm4	mulps	20 * SIZE(Y), %xmm5	mulps	24 * SIZE(Y), %xmm6	mulps	28 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	movlps	32 * SIZE(X), %xmm4	movhps	34 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movlps	36 * SIZE(X), %xmm5	movhps	38 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movlps	40 * SIZE(X), %xmm6	movhps	42 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movlps	44 * SIZE(X), %xmm7	movhps	46 * SIZE(X), %xmm7	mulps	32 * SIZE(Y), %xmm4	mulps	36 * SIZE(Y), %xmm5	mulps	40 * SIZE(Y), %xmm6	mulps	44 * SIZE(Y), %xmm7	addl	$32 * SIZE, X	addl	$32 * SIZE, Y	decl	%eax	jg	.L21	ALIGN_3.L22:	addps	%xmm4,  %xmm0	movlps	16 * SIZE(X), %xmm4	movhps	18 * SIZE(X), %xmm4	addps	%xmm5,  %xmm1	movlps	20 * SIZE(X), %xmm5	movhps	22 * SIZE(X), %xmm5	addps	%xmm6,  %xmm2	movlps	24 * SIZE(X), %xmm6	movhps	26 * SIZE(X), %xmm6	addps	%xmm7,  %xmm3	movlps	28 * SIZE(X), %xmm7	movhps	30 * SIZE(X), %xmm7	mulps	16 * SIZE(Y), %xmm4	mulps	20 * SIZE(Y), %xmm5	mulps	24 * SIZE(Y), %xmm6	mulps	28 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	addps	%xmm7,  %xmm3	addl	$32 * SIZE, X	addl	$32 * SIZE, Y	ALIGN_3.L24:	testl	$31, N	jle	.L999	testl	$16, N	jle	.L25	movlps	 0 * SIZE(X), %xmm4	movhps	 2 * SIZE(X), %xmm4	movlps	 4 * SIZE(X), %xmm5	movhps	 6 * SIZE(X), %xmm5	movlps	 8 * SIZE(X), %xmm6	movhps	10 * SIZE(X), %xmm6	movlps	12 * SIZE(X), %xmm7	movhps	14 * SIZE(X), %xmm7	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	mulps	 8 * SIZE(Y), %xmm6	mulps	12 * SIZE(Y), %xmm7	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addps	%xmm6,  %xmm2	addps	%xmm7,  %xmm3	addl	$16 * SIZE, X	addl	$16 * SIZE, Y	ALIGN_3.L25:	testl	$8, N	jle	.L26	movlps	 0 * SIZE(X), %xmm4	movhps	 2 * SIZE(X), %xmm4	movlps	 4 * SIZE(X), %xmm5	movhps	 6 * SIZE(X), %xmm5	mulps	 0 * SIZE(Y), %xmm4	mulps	 4 * SIZE(Y), %xmm5	addps	%xmm4,  %xmm0	addps	%xmm5,  %xmm1	addl	$8 * SIZE, X	addl	$8 * SIZE, Y	ALIGN_3.L26:	testl	$4, N	jle	.L27	movlps	 0 * SIZE(X), %xmm4	movhps	 2 * SIZE(X), %xmm4	mulps	 0 * SIZE(Y), %xmm4	addps	%xmm4,  %xmm0	addl	$4 * SIZE, X	addl	$4 * SIZE, Y	ALIGN_3.L27:	testl	$2, N	jle	.L28	movss	 0 * SIZE(X), %xmm4	movss	 1 * SIZE(X), %xmm5	mulss	 0 * SIZE(Y), %xmm4	mulss	 1 * SIZE(Y), %xmm5	addss	%xmm4,  %xmm0	addss	%xmm5,  %xmm1	addl	$2 * SIZE, X	addl	$2 * SIZE, Y	ALIGN_3.L28:	testl	$1, N	jle	.L999	movss	 0 * SIZE(X), %xmm4	mulss	 0 * SIZE(Y), %xmm4	addss	%xmm4,  %xmm0	jmp	.L999	ALIGN_3.L50:#ifdef F_INTERFACE	testl	INCX, INCX	jge	.L51	movl	N, %eax	decl	%eax	imull	INCX, %eax	subl	%eax, X	ALIGN_3.L51:	testl	INCY, INCY	jge	.L52	movl	N, %eax	decl	%eax	imull	INCY, %eax	subl	%eax, Y	ALIGN_3.L52:#endif	movl	N,  %eax	sarl	$2, %eax	jle	.L55	ALIGN_3.L53:	movss	0 * SIZE(X), %xmm4	addl	INCX, X	mulss	0 * SIZE(Y), %xmm4	addl	INCY, Y	movss	0 * SIZE(X), %xmm5	addl	INCX, X	mulss	0 * SIZE(Y), %xmm5	addl	INCY, Y	movss	0 * SIZE(X), %xmm6	addl	INCX, X	mulss	0 * SIZE(Y), %xmm6	addl	INCY, Y	movss	0 * SIZE(X), %xmm7	addl	INCX, X	mulss	0 * SIZE(Y), %xmm7	addl	INCY, Y	addss	%xmm4, %xmm0	addss	%xmm5, %xmm1	addss	%xmm6, %xmm2	addss	%xmm7, %xmm3	decl	%eax	jg	.L53	ALIGN_3.L55:	movl	N, %eax	andl	$3,   %eax	jle	.L999	ALIGN_3.L56:	movss	0 * SIZE(X), %xmm4	addl	INCX, X	mulss	0 * SIZE(Y), %xmm4	addl	INCY, Y	addss	%xmm4, %xmm0	decl	%eax	jg	.L56	ALIGN_3.L999:	addps	%xmm1, %xmm0	addps	%xmm3, %xmm2	addps	%xmm2, %xmm0#if defined(HAVE_SSE3) && !defined(__INTERIX)	haddps	%xmm0, %xmm0	haddps	%xmm0, %xmm0#elif defined(HAVE_SSE2)	movhlps	%xmm0, %xmm1	addps	%xmm1, %xmm0		pshufd	$1, %xmm0, %xmm1	addss	 %xmm1, %xmm0#else	movhlps	%xmm0, %xmm1	addps	%xmm1, %xmm0		movaps	%xmm0, %xmm1	shufps	$1, %xmm0, %xmm0	addss	 %xmm1, %xmm0#endif	movss	%xmm0, STACK_N	flds	STACK_N	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 + -