📄 zscal_sse2.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)#define STACK_ALPHA_R	16 + STACK + ARGS(%esp)#define STACK_ALPHA_I	24 + STACK + ARGS(%esp)#define STACK_X		32 + STACK + ARGS(%esp)#define STACK_INCX	36 + STACK + ARGS(%esp)#define M	%ebx#define X	%ecx#define INCX	%edx#define I	%esi#define XX	%edi#define FLAG	%ebp#ifdef PENTIUM4#define PREFETCHSIZE 88#endif#ifdef OPTERON#define PREFETCHSIZE 80#define movsd	     movlpd#endif	PROLOGUE	pushl	%edi	pushl	%esi	pushl	%ebx	pushl	%ebp	PROFCODE	movl	STACK_M,     M	movl	STACK_X,     X	movl	STACK_INCX,  INCX	movsd	STACK_ALPHA_R, %xmm0	movsd	STACK_ALPHA_I, %xmm1	sall	$ZBASE_SHIFT, INCX	xor	FLAG, FLAG	pxor	%xmm7, %xmm7	comisd	%xmm0, %xmm7	jne	.L100		# Alpha_r != ZERO	comisd	%xmm1, %xmm7	jne	.L100		# Alpha_i != ZERO/* Alpha == ZERO */	cmpl	$2 * SIZE, INCX	jne	.L50/* INCX == 1 */	cmpl	$4, M	jle	.L20	testl	$7, X	jne	.L20		# Unaligned Mode	testl	$15, X		# aligned for quad word?	je	.L05	movsd	%xmm7, 0 * SIZE(X)	addl	$SIZE, X	movl	$1, FLAG	decl	M	ALIGN_3.L05:/* Aligned Mode */	movl	M,  I	# rcx = n	sarl	$3, I	jle	.L12	ALIGN_4.L11:#ifdef HAVE_3DNOW	prefetchw	88 * SIZE(X)#endif	movapd	%xmm7,  0 * SIZE(X)	movapd	%xmm7,  2 * SIZE(X)	movapd	%xmm7,  4 * SIZE(X)	movapd	%xmm7,  6 * SIZE(X)#ifdef HAVE_3DNOW	prefetchw	96 * SIZE(X)#endif	movapd	%xmm7,  8 * SIZE(X)	movapd	%xmm7, 10 * SIZE(X)	movapd	%xmm7, 12 * SIZE(X)	movapd	%xmm7, 14 * SIZE(X)	addl	$16 * SIZE, X	decl	I	jg	.L11	ALIGN_4.L12:	testl	$7, M	je	.L19	testl	$4, M	je	.L13	movapd	%xmm7,  0 * SIZE(X)	movapd	%xmm7,  2 * SIZE(X)	movapd	%xmm7,  4 * SIZE(X)	movapd	%xmm7,  6 * SIZE(X)	addl	$8 * SIZE, X	ALIGN_3.L13:	testl  $2, M	je    .L14	movapd	%xmm7,  0 * SIZE(X)	movapd	%xmm7,  2 * SIZE(X)	addl	$4 * SIZE, X	ALIGN_3.L14:	testl  $1, M	je    .L19	movapd	%xmm7,  0 * SIZE(X)	addl	$2 * SIZE, X	ALIGN_3.L19:	testl  $1, FLAG	je    .L999	movsd	%xmm7, 0 * SIZE(X)	jmp	.L999	ALIGN_4/* Unaligned Mode */.L20:	movl	M,  I	# rcx = n	sarl	$2, I	jle	.L22	ALIGN_4.L21:#ifdef HAVE_3DNOW	prefetchw	88 * SIZE(X)#endif	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	movsd	%xmm7, 2 * SIZE(X)	movsd	%xmm7, 3 * SIZE(X)	movsd	%xmm7, 4 * SIZE(X)	movsd	%xmm7, 5 * SIZE(X)	movsd	%xmm7, 6 * SIZE(X)	movsd	%xmm7, 7 * SIZE(X)	addl	$8 * SIZE, X	decl	I	jg	.L21	ALIGN_4.L22:	testl	$3, M	je	.L999	testl	$2, M	je	.L23	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	movsd	%xmm7, 2 * SIZE(X)	movsd	%xmm7, 3 * SIZE(X)	addl	$4 * SIZE, X	ALIGN_3.L23:	testl	$1, M	je	.L999	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	$2 * SIZE, X	jmp	.L999	ALIGN_4/* incx != 1 */.L50:	testl	$15, X	jne	.L60/* Aligned Mode */	movl	M,  I		# rcx = n	sarl	$2, I	jle	.L52	ALIGN_4.L51:	#ifdef HAVE_3DNOW       prefetchw	88 * SIZE(X)#endif	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	decl	I	jg	.L51	ALIGN_4.L52:	testl	$3, M	je	.L999	testl	$2, M	je	.L53	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	movapd	%xmm7, 0 * SIZE(X)	addl	INCX, X	ALIGN_3.L53:	testl	$1, M	je	.L999	movapd	%xmm7, 0 * SIZE(X)	jmp	.L999	ALIGN_4/* Unaligned Mode */.L60:	movl	M,  I		# rcx = n	sarl	$2, I	jle	.L62	ALIGN_4.L61:#ifdef HAVE_3DNOW	prefetchw	88 * SIZE(X)#endif	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	decl	I	jg	.L61	ALIGN_4.L62:	testl	$3, M	je	.L999	testl	$2, M	je	.L63	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	addl	INCX, X	ALIGN_3.L63:	testl	$1, M	je	.L999	movsd	%xmm7, 0 * SIZE(X)	movsd	%xmm7, 1 * SIZE(X)	jmp	.L999	ALIGN_4/* Alpha != ZERO */.L100:	unpcklpd  %xmm0, %xmm0	subpd	  %xmm1, %xmm7	unpcklpd  %xmm1, %xmm7	movapd	  %xmm7, %xmm1	cmpl	$2 * SIZE, INCX	jne	.L150	cmpl	$4, M	jle	.L120	testl	$15, X	jne	.L120		# Unaligned Mode/* Aligned Mode */.L110:	movl	M,  I		# rcx = n	sarl	$2, I	jle	.L112	ALIGN_4.L111:#ifdef HAVE_3DNOW	prefetchw	64 * SIZE(X)#endif	movapd	 0 * SIZE(X), %xmm2	movapd	 2 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  0 * SIZE(X)	movapd	%xmm4,  2 * SIZE(X)	movapd	 4 * SIZE(X), %xmm2	movapd	 6 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  4 * SIZE(X)	movapd	%xmm4,  6 * SIZE(X)	addl	$8 * SIZE, X	decl	I	jg	.L111	ALIGN_4.L112:	testl	$3, M	je	.L999	testl	$2, M	je	.L113	movapd	 0 * SIZE(X), %xmm2	movapd	 2 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  0 * SIZE(X)	movapd	%xmm4,  2 * SIZE(X)	addl	$4 * SIZE, X	ALIGN_3.L113:	testl	$1, M	je	.L999	movapd	 0 * SIZE(X), %xmm2	movapd	 %xmm2, %xmm3	SHUFPD_1 %xmm2, %xmm3	mulpd	 %xmm0, %xmm2	mulpd	 %xmm1, %xmm3	addpd	 %xmm3, %xmm2	movapd	%xmm2,  0 * SIZE(X)	jmp	.L999	ALIGN_3/* Unaligned Mode */.L120:	movl	M,  I		# rcx = n	sarl	$2, I	jle	.L122	ALIGN_4.L121:#ifdef HAVE_3DNOW	prefetchw	80 * SIZE(X)#endif	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	movsd	 2 * SIZE(X), %xmm4	movhpd	 3 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  0 * SIZE(X)	movhpd	%xmm2,  1 * SIZE(X)	movsd	%xmm4,  2 * SIZE(X)	movhpd	%xmm4,  3 * SIZE(X)	movsd	 4 * SIZE(X), %xmm2	movhpd	 5 * SIZE(X), %xmm2	movsd	 6 * SIZE(X), %xmm4	movhpd	 7 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  4 * SIZE(X)	movhpd	%xmm2,  5 * SIZE(X)	movsd	%xmm4,  6 * SIZE(X)	movhpd	%xmm4,  7 * SIZE(X)	addl	$8 * SIZE, X	decl	I	jg	.L121	ALIGN_4.L122:	testl	$3, M	je	.L999	testl	$2, M	je	.L123	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	movsd	 2 * SIZE(X), %xmm4	movhpd	 3 * SIZE(X), %xmm4	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  0 * SIZE(X)	movhpd	%xmm2,  1 * SIZE(X)	movsd	%xmm4,  2 * SIZE(X)	movhpd	%xmm4,  3 * SIZE(X)	addl	$4 * SIZE, X	ALIGN_3.L123:	testl	$1, M	je	.L999	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	movapd	 %xmm2, %xmm3	SHUFPD_1 %xmm2, %xmm3	mulpd	 %xmm0, %xmm2	mulpd	 %xmm1, %xmm3	addpd	 %xmm3, %xmm2	movsd	%xmm2,  0 * SIZE(X)	movhpd	%xmm2,  1 * SIZE(X)	ALIGN_3	jmp	.L999	ALIGN_4/* incx != 1 */.L150:	movl	X, XX	testl	$15, X	jne	.L160	movl	M,  I		# rcx = n	sarl	$2, I		# (n >> 3)	jle	.L152	ALIGN_4.L151:#ifdef HAVE_3DNOW	prefetchw	40 * SIZE(X)#endif	movapd	 0 * SIZE(X), %xmm2	addl	INCX, X	movapd	 0 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  0 * SIZE(XX)	addl	INCX, XX	movapd	%xmm4,  0 * SIZE(XX)	addl	INCX, XX	movapd	 0 * SIZE(X), %xmm2	addl	INCX, X	movapd	 0 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  0 * SIZE(XX)	addl	INCX, XX	movapd	%xmm4,  0 * SIZE(XX)	addl	INCX, XX	decl	I	jg	.L151	ALIGN_4.L152:	testl	$3, M	je	.L999	testl	$2, M	je	.L153	movapd	 0 * SIZE(X), %xmm2	addl	INCX, X	movapd	 0 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movapd	%xmm2,  0 * SIZE(XX)	addl	INCX, XX	movapd	%xmm4,  0 * SIZE(XX)	addl	INCX, XX	ALIGN_3.L153:	testl	$1, M	je	.L999	movapd	 0 * SIZE(X), %xmm2	movapd	 %xmm2, %xmm3	SHUFPD_1 %xmm2, %xmm3	mulpd	 %xmm0, %xmm2	mulpd	 %xmm1, %xmm3	addpd	 %xmm3, %xmm2	movapd	%xmm2,  0 * SIZE(X)	jmp	.L999	ALIGN_4.L160:	movl	M,  I		# rcx = n	sarl	$2, I		# (n >> 3)	jle	.L162	ALIGN_4.L161:#ifdef HAVE_3DNOW	prefetchw	40 * SIZE(X)#endif	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	addl	INCX, X	movsd	 0 * SIZE(X), %xmm4	movhpd	 1 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  0 * SIZE(XX)	movhpd	%xmm2,  1 * SIZE(XX)	addl	INCX, XX	movsd	%xmm4,  0 * SIZE(XX)	movhpd	%xmm4,  1 * SIZE(XX)	addl	INCX, XX	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	addl	INCX, X	movsd	 0 * SIZE(X), %xmm4	movhpd	 1 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  0 * SIZE(XX)	movhpd	%xmm2,  1 * SIZE(XX)	addl	INCX, XX	movsd	%xmm4,  0 * SIZE(XX)	movhpd	%xmm4,  1 * SIZE(XX)	addl	INCX, XX	decl	I	jg	.L161	ALIGN_4.L162:	testl	$3, M	je	.L999	testl	$2, M	je	.L163	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	addl	INCX, X	movsd	 0 * SIZE(X), %xmm4	movhpd	 1 * SIZE(X), %xmm4	addl	INCX, X	movapd	 %xmm2, %xmm3	movapd	 %xmm4, %xmm5	SHUFPD_1 %xmm2, %xmm3	SHUFPD_1 %xmm4, %xmm5	mulpd	 %xmm0, %xmm2	mulpd	 %xmm0, %xmm4	mulpd	 %xmm1, %xmm3	mulpd	 %xmm1, %xmm5	addpd	 %xmm3, %xmm2	addpd	 %xmm5, %xmm4	movsd	%xmm2,  0 * SIZE(XX)	movhpd	%xmm2,  1 * SIZE(XX)	addl	INCX, XX	movsd	%xmm4,  0 * SIZE(XX)	movhpd	%xmm4,  1 * SIZE(XX)	addl	INCX, XX	ALIGN_3.L163:	testl	$1, M	je	.L999	movsd	 0 * SIZE(X), %xmm2	movhpd	 1 * SIZE(X), %xmm2	movapd	 %xmm2, %xmm3	SHUFPD_1 %xmm2, %xmm3	mulpd	 %xmm0, %xmm2	mulpd	 %xmm1, %xmm3	addpd	 %xmm3, %xmm2	movsd	%xmm2,  0 * SIZE(X)	movhpd	%xmm2,  1 * SIZE(X)	ALIGN_4.L999:	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 + -