📄 zdot_sse_opteron.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"#if defined(F_INTERFACE) && defined(F_INTERFACE_F2C)#define RESULT	ARG1	/* rdi */#define N	ARG2	/* rsi */#define X	ARG3	/* rdx */#define INCX	ARG4	/* rcx */#ifndef WINDOWS_ABI#define Y	ARG5	/* r8  */#define INCY	ARG6	/* r9  */#else#define Y	%r10#define INCY	%r11#endif#else#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#endif#define PREFETCHSIZE 224#ifdef BARCELONA#define MOVDDUP(a, b, c)	movddup	a(b), c#else#define MOVDDUP(a, b, c)	movlps	a(b), c;unpcklps c, c#endif	PROLOGUE	PROFCODE#ifdef WINDOWS_ABI#if defined(F_INTERFACE) && defined(F_INTERFACE_F2C)		movq	40(%rsp), Y		movq	48(%rsp), INCY#else		movq	40(%rsp), INCY#endif#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	salq	$ZBASE_SHIFT, INCX	salq	$ZBASE_SHIFT, INCY	pxor	%xmm0, %xmm0	pxor	%xmm1, %xmm1	pxor	%xmm2, %xmm2	pxor	%xmm3, %xmm3	cmpq	$0, N	jle	.L999	cmpq	$2 * SIZE, INCX	jne	.L50	cmpq	$2 * SIZE, INCY	jne	.L50	movq	N,  %rax	sarq	$3, %rax	jle	.L25	ALIGN_3.L21:	PREFETCH	(PREFETCHSIZE + 0) * SIZE(X)	movlpd	 0 * SIZE(X), %xmm4	movhpd	 2 * SIZE(X), %xmm4	movlpd	 4 * SIZE(X), %xmm6	movhpd	 6 * SIZE(X), %xmm6	movlpd	 8 * SIZE(X), %xmm8	movhpd	10 * SIZE(X), %xmm8	movlpd	12 * SIZE(X), %xmm10	movhpd	14 * SIZE(X), %xmm10	pshufd	 $0xb1, %xmm4,  %xmm5	pshufd	 $0xb1, %xmm6,  %xmm7	pshufd	 $0xb1, %xmm8,  %xmm9	pshufd	 $0xb1, %xmm10, %xmm11	movlpd	 0 * SIZE(Y), %xmm12	movhpd	 2 * SIZE(Y), %xmm12	movlpd	 4 * SIZE(Y), %xmm13	movhpd	 6 * SIZE(Y), %xmm13	movlpd	 8 * SIZE(Y), %xmm14	movhpd	10 * SIZE(Y), %xmm14	movlpd	12 * SIZE(Y), %xmm15	movhpd	14 * SIZE(Y), %xmm15	PREFETCH	(PREFETCHSIZE + 0) * SIZE(Y)	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	mulps	 %xmm13, %xmm6	mulps	 %xmm13, %xmm7	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	addps	 %xmm6, %xmm2	addps	 %xmm7, %xmm3	mulps	 %xmm14, %xmm8	mulps	 %xmm14, %xmm9	mulps	 %xmm15, %xmm10	mulps	 %xmm15, %xmm11	addps	 %xmm8,  %xmm0	addps	 %xmm9,  %xmm1	addps	 %xmm10, %xmm2	addps	 %xmm11, %xmm3	addq	$16 * SIZE, X	addq	$16 * SIZE, Y	decq	%rax	jg	.L21	ALIGN_3.L25:	testq	$4, N	jle	.L27	movlpd	 0 * SIZE(X), %xmm4	movhpd	 2 * SIZE(X), %xmm4	movlpd	 4 * SIZE(X), %xmm6	movhpd	 6 * SIZE(X), %xmm6	movlpd	 0 * SIZE(Y), %xmm12	movhpd	 2 * SIZE(Y), %xmm12	movlpd	 4 * SIZE(Y), %xmm13	movhpd	 6 * SIZE(Y), %xmm13	pshufd	 $0xb1, %xmm4, %xmm5	pshufd	 $0xb1, %xmm6, %xmm7	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	mulps	 %xmm13, %xmm6	mulps	 %xmm13, %xmm7	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	addps	 %xmm6, %xmm2	addps	 %xmm7, %xmm3	addq	$8 * SIZE, X	addq	$8 * SIZE, Y	ALIGN_3.L27:	testq	$2, N	jle	.L28	movlpd	 0 * SIZE(X), %xmm4	movhpd	 2 * SIZE(X), %xmm4	movlpd	 0 * SIZE(Y), %xmm12	movhpd	 2 * SIZE(Y), %xmm12	pshufd	 $0xb1, %xmm4, %xmm5	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	addq	 $4 * SIZE, X	addq	 $4 * SIZE, Y	ALIGN_3.L28:	testq	$1, N	jle	.L999	movsd	 (X), %xmm4	movsd	 (Y), %xmm12	pshufd	 $0xb1, %xmm4, %xmm5	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	jmp	 .L999	ALIGN_3.L50:#ifdef F_INTERFACE	testq	INCX, INCX		# if (incx < 0)	jge	.L51	movq	N, %rax		# n	decq	%rax			# n - 1	imulq	INCX, %rax		# (n - 1) * incx	subq	%rax, X	ALIGN_3.L51:	testq	INCY, INCY		# if (incy < 0)	jge	.L52	movq	N, %rax	decq	%rax			# (n - 1)	imulq	INCY, %rax		# (n - 1) * incy	subq	%rax, Y	ALIGN_3.L52:#endif	movq	N,  %rax	sarq	$3, %rax	jle	.L55	ALIGN_3.L53:	PREFETCH	(PREFETCHSIZE + 0) * SIZE(X)	movlpd	(X), %xmm4	addq	 INCX, X	movhpd	(X), %xmm4	addq	 INCX, X	movlpd	(X), %xmm6	addq	 INCX, X	movhpd	(X), %xmm6	addq	 INCX, X	movlpd	(X), %xmm8	addq	 INCX, X	movhpd	(X), %xmm8	addq	 INCX, X	movlpd	(X), %xmm10	addq	 INCX, X	movhpd	(X), %xmm10	addq	 INCX, X	pshufd	 $0xb1, %xmm4,  %xmm5	pshufd	 $0xb1, %xmm6,  %xmm7	pshufd	 $0xb1, %xmm8,  %xmm9	pshufd	 $0xb1, %xmm10, %xmm11	movlpd	(Y), %xmm12	addq	 INCY, Y	movhpd	(Y), %xmm12	addq	 INCY, Y	movlpd	(Y), %xmm13	addq	 INCY, Y	movhpd	(Y), %xmm13	addq	 INCY, Y	movlpd	(Y), %xmm14	addq	 INCY, Y	movhpd	(Y), %xmm14	addq	 INCY, Y	movlpd	(Y), %xmm15	addq	 INCY, Y	movhpd	(Y), %xmm15	addq	 INCY, Y	PREFETCH	(PREFETCHSIZE + 0) * SIZE(Y)	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	mulps	 %xmm13, %xmm6	mulps	 %xmm13, %xmm7	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	addps	 %xmm6, %xmm2	addps	 %xmm7, %xmm3	mulps	 %xmm14, %xmm8	mulps	 %xmm14, %xmm9	mulps	 %xmm15, %xmm10	mulps	 %xmm15, %xmm11	addps	 %xmm8,  %xmm0	addps	 %xmm9,  %xmm1	addps	 %xmm10, %xmm2	addps	 %xmm11, %xmm3	decq	%rax	jg	.L53	ALIGN_3.L55:	testq	$4, N	jle	.L57	movlpd	 (X), %xmm4	addq	 INCX, X	movhpd	 (X), %xmm4	addq	 INCX, X	movlpd	 (X), %xmm6	addq	 INCX, X	movhpd	 (X), %xmm6	addq	 INCX, X	movlpd	 (Y), %xmm12	addq	 INCY, Y	movhpd	 (Y), %xmm12	addq	 INCY, Y	movlpd	 (Y), %xmm13	addq	 INCY, Y	movhpd	 (Y), %xmm13	addq	 INCY, Y	pshufd	 $0xb1, %xmm4, %xmm5	pshufd	 $0xb1, %xmm6, %xmm7	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	mulps	 %xmm13, %xmm6	mulps	 %xmm13, %xmm7	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	addps	 %xmm6, %xmm2	addps	 %xmm7, %xmm3	ALIGN_3.L57:	testq	$2, N	jle	.L58	movlpd	 (X), %xmm4	addq	 INCX, X	movhpd	 (X), %xmm4	addq	 INCX, X	movlpd	 (Y), %xmm12	addq	 INCY, Y	movhpd	 (Y), %xmm12	addq	 INCY, Y	pshufd	 $0xb1, %xmm4, %xmm5	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	ALIGN_3.L58:	testq	$1, N	jle	.L999	movsd	 (X), %xmm4	movsd	 (Y), %xmm12	pshufd	 $0xb1, %xmm4, %xmm5	mulps	 %xmm12, %xmm4	mulps	 %xmm12, %xmm5	addps	 %xmm4, %xmm0	addps	 %xmm5, %xmm1	ALIGN_3.L999:	addps	%xmm2, %xmm0	addps	%xmm3, %xmm1		movhlps	%xmm0, %xmm2	movhlps	%xmm1, %xmm3#ifndef CONJ	addps	%xmm2, %xmm0	addps	%xmm3, %xmm1	pshufd	$1, %xmm0, %xmm2	pshufd	$1, %xmm1, %xmm3	subps	 %xmm2, %xmm0	addps	 %xmm3, %xmm1#else	addps	%xmm2, %xmm0	addps	%xmm1, %xmm3	pshufd	$1, %xmm0, %xmm2	pshufd	$1, %xmm3, %xmm1	addps	 %xmm2, %xmm0	subps	 %xmm3, %xmm1#endif#if defined(F_INTERFACE) && defined(F_INTERFACE_F2C)	movss	%xmm0, 0 * SIZE(RESULT)	movss	%xmm1, 1 * SIZE(RESULT)#elif defined(PACKED_RETURN) || defined(F_SUN)	unpcklps       %xmm1, %xmm0	#endif	RESTOREREGISTERS	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 + -