/*********************************************************************/

/* Copyright 2009, 2010 The University of Texas at Austin.           */

/* All rights reserved.                                              */

/*                                                                   */

/* Redistribution and use in source and binary forms, with or        */

/* without modification, are permitted provided that the following   */

/* conditions are met:                                               */

/*                                                                   */

/*   1. Redistributions of source code must retain the above         */

/*      copyright notice, this list of conditions and the following  */

/*      disclaimer.                                                  */

/*                                                                   */

/*   2. Redistributions in binary form must reproduce the above      */

/*      copyright notice, this list of conditions and the following  */

/*      disclaimer in the documentation and/or other materials       */

/*      provided with the distribution.                              */

/*                                                                   */

/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */

/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */

/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */

/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */

/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */

/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */

/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */

/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */

/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */

/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */

/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */

/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */

/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */

/*    POSSIBILITY OF SUCH DAMAGE.                                    */

/*                                                                   */

/* The views and conclusions contained in the software and           */

/* documentation are those of the authors and should not be          */

/* interpreted as representing official policies, either expressed   */

/* or implied, of The University of Texas at Austin.                 */

/*********************************************************************/

#define ASSEMBLER

#include "common.h"

#ifdef movsd

#undef movsd

#endif

#ifdef PENTIUM3

#ifdef HAVE_SSE

#define PREFETCH	prefetcht0

#define PREFETCHW	prefetcht0

#define PREFETCHSIZE	(16 * 2)

#endif

#define movsd		movlps

#endif

#ifdef PENTIUM4

#define PREFETCH	prefetcht0

#define PREFETCHW	prefetcht0

#define PREFETCHSIZE	(16 * 4)

#endif

#if defined(CORE2) || defined(PENRYN) || defined(DUNNINGTON) || defined(NEHALEM)

#define PREFETCH	prefetcht0

#define PREFETCHW	prefetcht0

#define PREFETCHSIZE	(16 * 7)

#endif

#ifdef OPTERON

#define PREFETCH	prefetchnta

#define PREFETCHW	prefetchw

#define PREFETCHSIZE	(16 * 3)

#define movsd		movlps

#endif

#ifdef BARCELONA

#define PREFETCH	prefetchnta

#define PREFETCHW	prefetchw

#define PREFETCHSIZE	(16 * 5)

#endif

#ifdef ATOM

#define PREFETCH	prefetchnta

#define PREFETCHW	prefetcht0

#define PREFETCHSIZE	(16 * 6)

#endif

#ifdef NANO

#define PREFETCH	prefetcht0

#define PREFETCHSIZE	(16 * 4)

#endif

#define STACKSIZE	16

#define M		 4 + STACKSIZE(%esp)

#define N		 8 + STACKSIZE(%esp)

#define ALPHA		16 + STACKSIZE(%esp)

#define A		20 + STACKSIZE(%esp)

#define STACK_LDA	24 + STACKSIZE(%esp)

#define STACK_X		28 + STACKSIZE(%esp)

#define STACK_INCX	32 + STACKSIZE(%esp)

#define Y		36 + STACKSIZE(%esp)

#define STACK_INCY	40 + STACKSIZE(%esp)

#define BUFFER		44 + STACKSIZE(%esp)

#define I	%eax

#define J	%ebx

#define INCX	J

#define INCY	%ecx

#define A1	%esi

#define X	%edx

#define Y1	%edi

#define LDA	%ebp

	PROLOGUE

	pushl	%ebp

	pushl	%edi

	pushl	%esi

	pushl	%ebx

	PROFCODE

	movl	STACK_LDA,  LDA

	movl	STACK_X,    X

	movl	STACK_INCX, INCX

	movl	STACK_INCY, INCY

	leal	(,INCX, SIZE), INCX

	leal	(,INCY, SIZE), INCY

	leal	(,LDA,  SIZE), LDA

	subl	$-32 * SIZE, A

	cmpl	$0, N

	jle	.L999

	cmpl	$0, M

	jle	.L999

	movl	BUFFER, Y1

	movl	M,  I

	sarl	$3, I

	jle	.L05

	ALIGN_4

.L02:

	movss	(X), %xmm0

	addl	INCX, X

	movss	(X), %xmm1

	addl	INCX, X

	unpcklps %xmm1, %xmm0

	movss	(X), %xmm2

	addl	INCX, X

	movss	(X), %xmm3

	addl	INCX, X

	unpcklps %xmm3, %xmm2

	movss	(X), %xmm4

	addl	INCX, X

	movss	(X), %xmm5

	addl	INCX, X

	unpcklps %xmm5, %xmm4

	movss	(X), %xmm6

	addl	INCX, X

	movss	(X), %xmm7

	addl	INCX, X

	unpcklps %xmm7, %xmm6

	movlps	%xmm0, 0 * SIZE(Y1)

	movlps	%xmm2, 2 * SIZE(Y1)

	movlps	%xmm4, 4 * SIZE(Y1)

	movlps	%xmm6, 6 * SIZE(Y1)

	addl	$8 * SIZE, Y1

	decl	I

	jg	.L02

	ALIGN_4

.L05:

	movl	M,  I

	andl	$7, I

	jle	.L10

	ALIGN_2

.L06:

	movss	(X), %xmm0

	addl	INCX, X

	movss	%xmm0, 0 * SIZE(Y1)

	addl	$SIZE, Y1

	decl	I

	jg	.L06

	ALIGN_4

.L10:

	movl	Y, Y1

	movl	N,  J

	sarl	$1, J

	jle	.L20

	ALIGN_3

.L11:

	movl	BUFFER, X

	addl	$32 * SIZE, X

	movl	A, A1

	leal	(A1, LDA, 2), %eax

	movl	%eax, A

	xorps	%xmm0, %xmm0

	xorps	%xmm1, %xmm1

	movaps	-32 * SIZE(X), %xmm2

	movaps	-28 * SIZE(X), %xmm3

	movl	M,   I

	sarl	$4,  I

	jle	.L15

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	movsd	-32 * SIZE(A1, LDA), %xmm5

	movhps	-30 * SIZE(A1, LDA), %xmm5

	movsd	-28 * SIZE(A1), %xmm6

	movhps	-26 * SIZE(A1), %xmm6

	movsd	-28 * SIZE(A1, LDA), %xmm7

	movhps	-26 * SIZE(A1, LDA), %xmm7

	decl	I

	jle	.L13

	ALIGN_4

.L12:

#ifdef PREFETCH

	PREFETCH PREFETCHSIZE * SIZE(A1)

#endif

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	movsd	-24 * SIZE(A1), %xmm4

	movhps	-22 * SIZE(A1), %xmm4

	mulps	%xmm2, %xmm5

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm5, %xmm1

	movsd	-24 * SIZE(A1, LDA), %xmm5

	movhps	-22 * SIZE(A1, LDA), %xmm5

	mulps	%xmm3, %xmm6

	addps	%xmm6, %xmm0

	movsd	-20 * SIZE(A1), %xmm6

	movhps	-18 * SIZE(A1), %xmm6

	mulps	%xmm3, %xmm7

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm7, %xmm1

	movsd	-20 * SIZE(A1, LDA), %xmm7

	movhps	-18 * SIZE(A1, LDA), %xmm7

#ifdef PREFETCH

	PREFETCH PREFETCHSIZE * SIZE(A1, LDA)

#endif

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	movsd	-16 * SIZE(A1), %xmm4

	movhps	-14 * SIZE(A1), %xmm4

	mulps	%xmm2, %xmm5

	movaps	-16 * SIZE(X), %xmm2

	addps	%xmm5, %xmm1

	movsd	-16 * SIZE(A1, LDA), %xmm5

	movhps	-14 * SIZE(A1, LDA), %xmm5

	mulps	%xmm3, %xmm6

	addps	%xmm6, %xmm0

	movsd	-12 * SIZE(A1), %xmm6

	movhps	-10 * SIZE(A1), %xmm6

	mulps	%xmm3, %xmm7

	movaps	-12 * SIZE(X), %xmm3

	addps	%xmm7, %xmm1

	movsd	-12 * SIZE(A1, LDA), %xmm7

	movhps	-10 * SIZE(A1, LDA), %xmm7

	addl	$16 * SIZE, A1

	addl	$16 * SIZE, X

	decl	I

	jg	.L12

	ALIGN_4

.L13:

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	movsd	-24 * SIZE(A1), %xmm4

	movhps	-22 * SIZE(A1), %xmm4

	mulps	%xmm2, %xmm5

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm5, %xmm1

	movsd	-24 * SIZE(A1, LDA), %xmm5

	movhps	-22 * SIZE(A1, LDA), %xmm5

	mulps	%xmm3, %xmm6

	addps	%xmm6, %xmm0

	movsd	-20 * SIZE(A1), %xmm6

	movhps	-18 * SIZE(A1), %xmm6

	mulps	%xmm3, %xmm7

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm7, %xmm1

	movsd	-20 * SIZE(A1, LDA), %xmm7

	movhps	-18 * SIZE(A1, LDA), %xmm7

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	mulps	%xmm2, %xmm5

	movaps	-16 * SIZE(X), %xmm2

	addps	%xmm5, %xmm1

	mulps	%xmm3, %xmm6

	addps	%xmm6, %xmm0

	mulps	%xmm3, %xmm7

	movaps	-12 * SIZE(X), %xmm3

	addps	%xmm7, %xmm1

	addl	$16 * SIZE, A1

	addl	$16 * SIZE, X

	ALIGN_4

.L15:

	testl	$8, M

	jle	.L16

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	movsd	-32 * SIZE(A1, LDA), %xmm5

	movhps	-30 * SIZE(A1, LDA), %xmm5

	movsd	-28 * SIZE(A1), %xmm6

	movhps	-26 * SIZE(A1), %xmm6

	movsd	-28 * SIZE(A1, LDA), %xmm7

	movhps	-26 * SIZE(A1, LDA), %xmm7

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	mulps	%xmm2, %xmm5

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm5, %xmm1

	mulps	%xmm3, %xmm6

	addps	%xmm6, %xmm0

	mulps	%xmm3, %xmm7

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm7, %xmm1

	addl	$8 * SIZE, A1

	addl	$8 * SIZE, X

	ALIGN_4

.L16:

	testl	$4, M

	jle	.L17

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	movsd	-32 * SIZE(A1, LDA), %xmm5

	movhps	-30 * SIZE(A1, LDA), %xmm5

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	mulps	%xmm2, %xmm5

	addps	%xmm5, %xmm1

	movaps	%xmm3, %xmm2

	addl	$4 * SIZE, A1

	ALIGN_4

.L17:

	testl	$2, M

	jle	.L18

#ifdef movsd

	xorps	%xmm4, %xmm4

#endif

	movsd	-32 * SIZE(A1), %xmm4

#ifdef movsd

	xorps	%xmm5, %xmm5

#endif

	movsd	-32 * SIZE(A1, LDA), %xmm5

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	mulps	%xmm2, %xmm5

	addps	%xmm5, %xmm1

	movhlps	%xmm2, %xmm2

	addl	$2 * SIZE, A1

	ALIGN_4

.L18:

	testl	$1, M

	jle	.L19

	movss	-32 * SIZE(A1), %xmm4

	mulss	%xmm2, %xmm4

	addss	%xmm4, %xmm0

	movss	-32 * SIZE(A1, LDA), %xmm5

	mulss	%xmm2, %xmm5

	addss	%xmm5, %xmm1

	ALIGN_4

.L19:

#ifdef HAVE_SSE3

	haddps	%xmm0, %xmm0

	haddps	%xmm1, %xmm1

	haddps	%xmm0, %xmm0

	haddps	%xmm1, %xmm1

#else

	movhlps	%xmm0, %xmm2

	movhlps	%xmm1, %xmm3

	addps	%xmm2, %xmm0

	addps	%xmm3, %xmm1

	movaps	%xmm0, %xmm2

	shufps	$1, %xmm0, %xmm0

	movaps	%xmm1, %xmm3

	shufps	$1, %xmm1, %xmm1

	addss	%xmm2, %xmm0

	addss	%xmm3, %xmm1

#endif

	movss	ALPHA, %xmm7

	mulss	%xmm7, %xmm0

	mulss	%xmm7, %xmm1

	addss	(Y1), %xmm0

	addss	(Y1, INCY), %xmm1

	movss	%xmm0, (Y1)

	movss	%xmm1, (Y1, INCY)

	leal	(Y1, INCY, 2), Y1

	decl	J

	jg	.L11

	ALIGN_4

.L20:

	testl	$1, N

	jle	.L999

	movl	BUFFER, X

	addl	$32 * SIZE, X

	movl	A, A1

	xorps	%xmm0, %xmm0

	xorps	%xmm1, %xmm1

	movaps	-32 * SIZE(X), %xmm2

	movaps	-28 * SIZE(X), %xmm3

	movl	M,   I

	sarl	$4,  I

	jle	.L25

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	movsd	-28 * SIZE(A1), %xmm6

	movhps	-26 * SIZE(A1), %xmm6

	decl	I

	jle	.L23

	ALIGN_4

.L22:

#ifdef PREFETCH

	PREFETCH PREFETCHSIZE * SIZE(A1)

#endif

	mulps	%xmm2, %xmm4

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm4, %xmm0

	movsd	-24 * SIZE(A1), %xmm4

	movhps	-22 * SIZE(A1), %xmm4

	mulps	%xmm3, %xmm6

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm6, %xmm0

	movsd	-20 * SIZE(A1), %xmm6

	movhps	-18 * SIZE(A1), %xmm6

	mulps	%xmm2, %xmm4

	movaps	-16 * SIZE(X), %xmm2

	addps	%xmm4, %xmm0

	movsd	-16 * SIZE(A1), %xmm4

	movhps	-14 * SIZE(A1), %xmm4

	mulps	%xmm3, %xmm6

	movaps	-12 * SIZE(X), %xmm3

	addps	%xmm6, %xmm0

	movsd	-12 * SIZE(A1), %xmm6

	movhps	-10 * SIZE(A1), %xmm6

	addl	$16 * SIZE, A1

	addl	$16 * SIZE, X

	decl	I

	jg	.L22

	ALIGN_4

.L23:

	mulps	%xmm2, %xmm4

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm4, %xmm0

	movsd	-24 * SIZE(A1), %xmm4

	movhps	-22 * SIZE(A1), %xmm4

	mulps	%xmm3, %xmm6

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm6, %xmm0

	movsd	-20 * SIZE(A1), %xmm6

	movhps	-18 * SIZE(A1), %xmm6

	mulps	%xmm2, %xmm4

	movaps	-16 * SIZE(X), %xmm2

	addps	%xmm4, %xmm0

	mulps	%xmm3, %xmm6

	movaps	-12 * SIZE(X), %xmm3

	addps	%xmm6, %xmm0

	addl	$16 * SIZE, A1

	addl	$16 * SIZE, X

	ALIGN_4

.L25:

	testl	$8, M

	jle	.L26

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	movsd	-28 * SIZE(A1), %xmm6

	movhps	-26 * SIZE(A1), %xmm6

	mulps	%xmm2, %xmm4

	movaps	-24 * SIZE(X), %xmm2

	addps	%xmm4, %xmm0

	mulps	%xmm3, %xmm6

	movaps	-20 * SIZE(X), %xmm3

	addps	%xmm6, %xmm0

	addl	$8 * SIZE, A1

	addl	$8 * SIZE, X

	ALIGN_4

.L26:

	testl	$4, M

	jle	.L27

	movsd	-32 * SIZE(A1), %xmm4

	movhps	-30 * SIZE(A1), %xmm4

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	movaps	%xmm3, %xmm2

	addl	$4 * SIZE, A1

	ALIGN_4

.L27:

	testl	$2, M

	jle	.L28

#ifdef movsd

	xorps	%xmm4, %xmm4

#endif

	movsd	-32 * SIZE(A1), %xmm4

	mulps	%xmm2, %xmm4

	addps	%xmm4, %xmm0

	movhlps	%xmm2, %xmm2

	addl	$2 * SIZE, A1

	ALIGN_4

.L28:

	testl	$1, M

	jle	.L29

	movss	-32 * SIZE(A1), %xmm4

	mulss	%xmm2, %xmm4

	addss	%xmm4, %xmm0

	ALIGN_4

.L29:

#ifdef HAVE_SSE3

	haddps	%xmm0, %xmm0

	haddps	%xmm0, %xmm0

#else

	movhlps	%xmm0, %xmm2

	addps	%xmm2, %xmm0

	movaps	%xmm0, %xmm2

	shufps	$1, %xmm0, %xmm0

	addss	%xmm2, %xmm0

#endif

	movss	ALPHA, %xmm7

	mulss	%xmm7, %xmm0

	addss	(Y1), %xmm0

	movss	%xmm0, (Y1)

	ALIGN_4

.L999:

	popl	%ebx

	popl	%esi

	popl	%edi	

	popl	%ebp

	ret

	EPILOGUE