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

/* 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"

#if !defined(HAVE_SSE) || !defined(HAVE_MMX)

#error  You have to check your configuration.

#endif

#define STACK	16

#define ARGS	 0

#define STACK_M	 4 + STACK + ARGS(%esi)

#define STACK_N	 8 + STACK + ARGS(%esi)

#define STACK_K	12 + STACK + ARGS(%esi)

#define STACK_A	20 + STACK + ARGS(%esi)

#define STACK_B	24 + STACK + ARGS(%esi)

#define STACK_C	28 + STACK + ARGS(%esi)

#define STACK_LDC	32 + STACK + ARGS(%esi)

#define STACK_OFFT	36 + STACK + ARGS(%esi)

#define TRMASK	 0(%esp)

#define K	16(%esp)

#define N	20(%esp)

#define M	24(%esp)

#define A	28(%esp)

#define C	32(%esp)

#define J	36(%esp)

#define OLD_STACK 40(%esp)

#define OFFSET  44(%esp)

#define KK	48(%esp)

#define KKK	52(%esp)

#define AORIG	56(%esp)

#define BORIG	60(%esp)

#define BUFFER 128(%esp)

#ifdef HAVE_3DNOW

#define PREFETCH     prefetch

#define PREFETCHW    prefetchw

#define PREFETCHSIZE (16 * 10 + 8)

#else

#define PREFETCH     prefetcht0

#define PREFETCHW    prefetcht0

#define PREFETCHSIZE   96

#endif

#define B	%edi

#define AA	%edx

#define	BB	%ecx

#define LDC	%ebp

#define CO1	%esi

#define STACK_ALIGN	4096

#define STACK_OFFSET	1024

#if !defined(HAVE_SSE2) || defined(OPTERON)

#define movsd	movlps

#endif

#ifdef HAVE_SSE2

#define xorps	pxor

#endif

	PROLOGUE

	pushl	%ebp

	pushl	%edi

	pushl	%esi

	pushl	%ebx

	PROFCODE

	movl	%esp, %esi	# save old stack

	subl	$128 + LOCAL_BUFFER_SIZE, %esp

	andl	$-STACK_ALIGN, %esp

	STACK_TOUCHING

	movss	STACK_M, %xmm0

	movl	STACK_N, %eax

	movss	STACK_K, %xmm1

	movss	STACK_A, %xmm2

	movl	STACK_B, B

	movss	STACK_C, %xmm3

	movl	STACK_LDC, LDC

	movss	STACK_OFFT, %xmm4

	movss	%xmm1, K

	movl	%eax,  N

	movss	%xmm0, M

	movss	%xmm2, A

	movss	%xmm3, C

	movl	%esi,  OLD_STACK

	movss	%xmm4, OFFSET

	movss	%xmm4, KK

	leal	(, LDC, SIZE), LDC

#ifdef LN

       movl	M, %eax

       leal	(, %eax, SIZE), %eax

       addl	%eax, C

       imull	K, %eax

       addl	%eax, A

#endif

#ifdef RT

       movl	N, %eax

       leal	(, %eax, SIZE), %eax

       imull	K, %eax

       addl	%eax, B

       movl	N, %eax

       imull	LDC, %eax

       addl	%eax, C

#endif

#ifdef RN

	negl	KK

#endif	

#ifdef RT

       movl	N, %eax

       subl	OFFSET, %eax

       movl	%eax, KK

#endif

#if defined(LN) || defined(LT)

	movl	  $0x3f800000,   0 + TRMASK	#  1.0

	movl	  $0x00000000,   4 + TRMASK	#  0.0

	movl	  $0x3f800000,   8 + TRMASK	#  1.0

	movl	  $0x00000000,  12 + TRMASK	#  0.0

#endif

	testl	$1, N

	jle	.L100

#ifdef LN

	movl	OFFSET, %eax

	addl	M, %eax

	movl	%eax, KK

#endif	

	leal	BUFFER, BB

#ifdef RT

       movl	K, %eax

       sall	$BASE_SHIFT, %eax

       subl	%eax, B

#endif

#if defined(LN) || defined(RT)

	movl	KK, %eax

	movl	B, BORIG

        sall	$BASE_SHIFT, %eax

	leal	(B,  %eax, 1), B

	leal	(BB, %eax, 4), BB

#endif	

#ifdef LT

	movl	OFFSET, %eax

	movl	%eax, KK

#endif

#if defined(LT) || defined(RN)

	movl	KK, %eax

#else

	movl	K,  %eax

	subl	KK, %eax

#endif

	sarl	$3, %eax

	jle	.L103

	ALIGN_4

.L102:

	movsd	 0 * SIZE(B), %xmm3

	movhps	 2 * SIZE(B), %xmm3

	movsd	 4 * SIZE(B), %xmm7

	movhps	 6 * SIZE(B), %xmm7

#ifdef HAVE_SSE2

	pshufd	 $0x00, %xmm3, %xmm0

	pshufd	 $0x55, %xmm3, %xmm1

	pshufd	 $0xaa, %xmm3, %xmm2

	pshufd	 $0xff, %xmm3, %xmm3

	pshufd	 $0x00, %xmm7, %xmm4

	pshufd	 $0x55, %xmm7, %xmm5

	pshufd	 $0xaa, %xmm7, %xmm6

	pshufd	 $0xff, %xmm7, %xmm7

#else

	movaps	%xmm3, %xmm0

	shufps	 $0x00, %xmm0, %xmm0

	movaps	%xmm3, %xmm1

	shufps	 $0x55, %xmm1, %xmm1

	movaps	%xmm3, %xmm2

	shufps	 $0xaa, %xmm2, %xmm2

	shufps	 $0xff, %xmm3, %xmm3

	movaps	%xmm7, %xmm4

	shufps	 $0x00, %xmm4, %xmm4

	movaps	%xmm7, %xmm5

	shufps	 $0x55, %xmm5, %xmm5

	movaps	%xmm7, %xmm6

	shufps	 $0xaa, %xmm6, %xmm6

	shufps	 $0xff, %xmm7, %xmm7

#endif

	movaps	%xmm0,  0 * SIZE(BB)

	movaps	%xmm1,  4 * SIZE(BB)

	movaps	%xmm2,  8 * SIZE(BB)

	movaps	%xmm3, 12 * SIZE(BB)

	movaps	%xmm4, 16 * SIZE(BB)

	movaps	%xmm5, 20 * SIZE(BB)

	movaps	%xmm6, 24 * SIZE(BB)

	movaps	%xmm7, 28 * SIZE(BB)

	addl	$ 8 * SIZE, B

	addl	$32 * SIZE, BB

	decl	%eax

	BRANCH

	jne	.L102

	ALIGN_2

.L103:

#if defined(LT) || defined(RN)

	movl	KK, %eax

#else

	movl	K,  %eax

	subl	KK, %eax

#endif

	andl	$7, %eax

	BRANCH

	jle	.L105

	ALIGN_2

.L104:

	movss	 0 * SIZE(B), %xmm0

	shufps	 $0x00, %xmm0, %xmm0

	movaps	%xmm0,  0 * SIZE(BB)

	addl	$1 * SIZE, B

	addl	$4 * SIZE, BB

	decl	%eax

	jne	.L104

	ALIGN_4

.L105:

#if defined(LT) || defined(RN)

	movl	A, AA

#else

	movl	A, %eax

	movl	%eax, AORIG

#endif

#ifdef RT

	subl	LDC, C

#endif

	movl	C, CO1

#ifndef RT

	addl	LDC, C

#endif

	movl	M,  %ebx

	sarl	$3, %ebx	# i = (m >> 2)

	jle	.L130

	ALIGN_4

.L110:

#ifdef LN

       movl	K, %eax

       sall	$3 + BASE_SHIFT, %eax

       subl	%eax, AORIG

#endif

#if defined(LN) || defined(RT)

	movl	KK, %eax

	movl	AORIG, AA

	sall	$3 + BASE_SHIFT, %eax

	addl	%eax, AA

#endif

	leal	BUFFER, BB

#if defined(LN) || defined(RT)

	movl	KK, %eax

	sall	$BASE_SHIFT, %eax

	leal	(BB, %eax, 4), BB

#endif	

	movaps	 0 * SIZE(BB), %xmm2

	xorps	%xmm4, %xmm4

	movaps	 0 * SIZE(AA), %xmm0

	xorps	%xmm5, %xmm5

	movaps	 16 * SIZE(BB), %xmm3

	xorps	%xmm6, %xmm6

	movaps	 16 * SIZE(AA), %xmm1

	xorps	%xmm7, %xmm7

	PREFETCHW      7 * SIZE(CO1)

#if defined(LT) || defined(RN)

	movl	KK, %eax

#else

	movl	K, %eax

	subl	KK, %eax

#endif

	sarl	$3, %eax

	je	.L112

	ALIGN_2

.L111:

	mulps	%xmm2, %xmm0

	mulps	 4 * SIZE(AA), %xmm2

	addps	%xmm0, %xmm4

	movaps	 8 * SIZE(AA), %xmm0

	addps	%xmm2, %xmm6

	movaps	 4 * SIZE(BB), %xmm2

	mulps	%xmm2, %xmm0

	mulps	12 * SIZE(AA), %xmm2

	addps	%xmm0, %xmm5

	movaps	32 * SIZE(AA), %xmm0

	addps	%xmm2, %xmm7

	movaps	 8 * SIZE(BB), %xmm2

	mulps	%xmm2, %xmm1

	mulps	20 * SIZE(AA), %xmm2

	addps	%xmm1, %xmm4

	movaps	24 * SIZE(AA), %xmm1

	addps	%xmm2, %xmm6

	movaps	12 * SIZE(BB), %xmm2

	mulps	%xmm2, %xmm1

	mulps	28 * SIZE(AA), %xmm2

	addps	%xmm1, %xmm5

	movaps	48 * SIZE(AA), %xmm1

	addps	%xmm2, %xmm7

	movaps	32 * SIZE(BB), %xmm2

	mulps	%xmm3, %xmm0

	mulps	36 * SIZE(AA), %xmm3

	addps	%xmm0, %xmm4

	movaps	40 * SIZE(AA), %xmm0

	addps	%xmm3, %xmm6

	movaps	20 * SIZE(BB), %xmm3

	mulps	%xmm3, %xmm0

	mulps	44 * SIZE(AA), %xmm3

	addps	%xmm0, %xmm5

	movaps	64 * SIZE(AA), %xmm0

	addps	%xmm3, %xmm7

	movaps	24 * SIZE(BB), %xmm3

	mulps	%xmm3, %xmm1

	mulps	52 * SIZE(AA), %xmm3

	addps	%xmm1, %xmm4

	movaps	56 * SIZE(AA), %xmm1

	addps	%xmm3, %xmm6

	movaps	28 * SIZE(BB), %xmm3

	mulps	%xmm3, %xmm1

	mulps	60 * SIZE(AA), %xmm3

	addps	%xmm1, %xmm5

	movaps	80 * SIZE(AA), %xmm1

	addps	%xmm3, %xmm7

	movaps	48 * SIZE(BB), %xmm3

	addl   $64 * SIZE, AA

	addl   $32 * SIZE, BB

	decl   %eax

	jne    .L111

	ALIGN_2

.L112:

#if defined(LT) || defined(RN)

	movl	KK, %eax

#else

	movl	K, %eax

	subl	KK, %eax

#endif

	andl	$7, %eax		# if (k & 1)

	BRANCH

	je .L114

.L113:

	movaps	 0 * SIZE(BB), %xmm2

	movaps	 0 * SIZE(AA), %xmm0

	mulps	%xmm2, %xmm0

	addps	%xmm0, %xmm4

	mulps	 4 * SIZE(AA), %xmm2

	addps	%xmm2, %xmm6

	addl	$8 * SIZE, AA

	addl	$4 * SIZE, BB

	subl	$1, %eax

	jg	.L113

	ALIGN_4

.L114:

	addps	%xmm5, %xmm4

	addps	%xmm7, %xmm6

#if defined(LN) || defined(RT)

	movl	KK, %eax

#ifdef LN

	subl	$8, %eax

#else

	subl	$1, %eax

#endif

	movl	AORIG, AA

	movl	BORIG, B

	leal	BUFFER, BB

	sall	$BASE_SHIFT, %eax

	leal	(AA, %eax, 8), AA

	leal	(B,  %eax, 1), B

	leal	(BB, %eax, 4), BB

#endif

#if defined(LN) || defined(LT)

	movsd	 0 * SIZE(B), %xmm2

	movhps	 2 * SIZE(B), %xmm2

	movsd	 4 * SIZE(B), %xmm5

	movhps	 6 * SIZE(B), %xmm5

	subps	%xmm4,  %xmm2

	subps	%xmm6,  %xmm5

	xorps	%xmm0,  %xmm0

	movaps	 %xmm2, %xmm3

	unpcklps %xmm0, %xmm2

	unpckhps %xmm0, %xmm3

	movaps	 %xmm5, %xmm7

	unpcklps %xmm0, %xmm5

	unpckhps %xmm0, %xmm7

#else

	movaps	 0 * SIZE(AA), %xmm0

	movaps	 4 * SIZE(AA), %xmm1

	subps	%xmm4, %xmm0

	subps	%xmm6, %xmm1

#endif

#if defined(LN) || defined(LT)

	movaps	TRMASK, %xmm6

#endif

#ifdef LN

	movss	63 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm7

	movaps	%xmm7, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movss	62 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movsd	60 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	58 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	56 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	54 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm7

	movaps	%xmm7, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movsd	52 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	50 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	48 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	45 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm5

	movaps	%xmm5, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movss	44 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	42 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	40 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	36 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm5

	movaps	%xmm5, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movsd	34 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	32 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	27 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm3

	movaps	%xmm3, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movss	26 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	24 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	18 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm3

	movaps	%xmm3, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movsd	16 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	 9 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm2

	movaps	%xmm2, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movss	 8 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movss	 0 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm2

#endif

#ifdef LT

	movss	 0 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm2

	movaps	%xmm2, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movss	 1 * SIZE(AA), %xmm0

	shufps	$0x05, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm2

	movsd	 2 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	 4 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	 6 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	 9 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm2

	movaps	%xmm2, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movsd	10 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	12 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	14 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	18 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm3

	movaps	%xmm3, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movss	19 * SIZE(AA), %xmm0

	shufps	$0x05, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm3

	movsd	20 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	22 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	27 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm3

	movaps	%xmm3, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movsd	28 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	30 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	36 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm5

	movaps	%xmm5, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movss	37 * SIZE(AA), %xmm0

	shufps	$0x05, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm5

	movsd	38 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	45 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm5

	movaps	%xmm5, %xmm1

	shufps	$0xee, %xmm1, %xmm1

	movsd	46 * SIZE(AA), %xmm0

	shufps	$0x50, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	54 * SIZE(AA), %xmm0

	shufps	$0x00, %xmm6,  %xmm0

	mulps	%xmm0, %xmm7

	movaps	%xmm7, %xmm1

	shufps	$0x44, %xmm1, %xmm1

	movss	55 * SIZE(AA), %xmm0

	shufps	$0x05, %xmm0, %xmm0

	mulps	%xmm1, %xmm0

	subps	%xmm0, %xmm7

	movss	63 * SIZE(AA), %xmm0

	movaps	 %xmm6, %xmm1

	shufps	$0x00, %xmm0,  %xmm1

	mulps	%xmm1, %xmm7

#endif

#if defined(RN) || defined(RT)

	movss	 0 * SIZE(B), %xmm6

	shufps	$0x00, %xmm6, %xmm6

	mulps	%xmm6, %xmm0

	mulps	%xmm6, %xmm1

#endif

#if defined(LN) || defined(LT)

	shufps	$0x88, %xmm3, %xmm2

	shufps	$0x88, %xmm7, %xmm5

	movlps	%xmm2,   0 * SIZE(B)

	movhps	%xmm2,   2 * SIZE(B)

	movlps	%xmm5,   4 * SIZE(B)

	movhps	%xmm5,   6 * SIZE(B)

#ifdef HAVE_SSE2

	pshufd	$0x00, %xmm2, %xmm0

	pshufd	$0x55, %xmm2, %xmm1

	pshufd	$0xaa, %xmm2, %xmm4

	pshufd	$0xff, %xmm2, %xmm6

#else

	movaps	%xmm2, %xmm0

	shufps	$0x00, %xmm0, %xmm0

	movaps	%xmm2, %xmm1

	shufps	$0x55, %xmm1, %xmm1

	movaps	%xmm2, %xmm4

	shufps	$0xaa, %xmm4, %xmm4

	movaps	%xmm2, %xmm6

	shufps	$0xff, %xmm6, %xmm6

#endif

	movaps	%xmm0,   0 * SIZE(BB)

	movaps	%xmm1,   4 * SIZE(BB)

	movaps	%xmm4,   8 * SIZE(BB)

	movaps	%xmm6,  12 * SIZE(BB)

#ifdef HAVE_SSE2

	pshufd	$0x00, %xmm5, %xmm0

	pshufd	$0x55, %xmm5, %xmm1

	pshufd	$0xaa, %xmm5, %xmm4

	pshufd	$0xff, %xmm5, %xmm6

#else

	movaps	%xmm5, %xmm0

	shufps	$0x00, %xmm0, %xmm0

	movaps	%xmm5, %xmm1

	shufps	$0x55, %xmm1, %xmm1

	movaps	%xmm5, %xmm4

	shufps	$0xaa, %xmm4, %xmm4

	movaps	%xmm5, %xmm6

	shufps	$0xff, %xmm6, %xmm6

#endif

	movaps	%xmm0,  16 * SIZE(BB)

	movaps	%xmm1,  20 * SIZE(BB)

	movaps	%xmm4,  24 * SIZE(BB)

	movaps	%xmm6,  28 * SIZE(BB)

#else

	movaps	%xmm0,   0 * SIZE(AA)

	movaps	%xmm1,   4 * SIZE(AA)

#endif

#ifdef LN

	subl	$8 * SIZE, CO1

#endif

#if defined(LN) || defined(LT)

	movlps	%xmm2, 0 * SIZE(CO1)

	movhps	%xmm2, 2 * SIZE(CO1)

	movlps	%xmm5, 4 * SIZE(CO1)

	movhps	%xmm5, 6 * SIZE(CO1)

#else

	movlps	%xmm0, 0 * SIZE(CO1)

	movhps	%xmm0, 2 * SIZE(CO1)

	movlps	%xmm1, 4 * SIZE(CO1)

	movhps	%xmm1, 6 * SIZE(CO1)

#endif

#ifndef LN

	addl	$8 * SIZE, CO1

#endif

#if defined(LT) || defined(RN)

	movl	K,  %eax

	subl	KK, %eax

	leal	(,%eax, SIZE), %eax

	leal	(AA, %eax, 8), AA

#ifdef LT

	addl	$8 * SIZE, B

#endif

#endif

#ifdef LN

	subl	$8, KK

	movl	BORIG, B

#endif

#ifdef LT

	addl	$8, KK

#endif

#ifdef RT

	movl	K, %eax

	movl	BORIG, B

	sall	$3 + BASE_SHIFT, %eax

	addl	%eax, AORIG

#endif

	decl	%ebx			# i --

	jg	.L110

	ALIGN_2	

.L130:

	testl	$4, M