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

/* 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"

#define M	%rdi

#define N	%rsi

#define K	%rdx

#define A	%rcx

#define B	%r8

#define C	%r9

#define LDC	%r10

#define I	%r11

#define AO	%r13

#define BO	%r14

#define	CO1	%r15

#define CO2	%rbx

#define KK	%rbp

#define BB	%r12

#ifndef WINDOWS_ABI

#define STACKSIZE 128

#define OLD_LDC		 8 + STACKSIZE(%rsp)

#define OLD_OFFSET	16 + STACKSIZE(%rsp)

#define OFFSET	 48(%rsp)

#define J	 56(%rsp)

#define KKK	 64(%rsp)

#define AORIG	 72(%rsp)

#else

#define STACKSIZE 256

#define OLD_A		40 + STACKSIZE(%rsp)

#define OLD_B		48 + STACKSIZE(%rsp)

#define OLD_C		56 + STACKSIZE(%rsp)

#define OLD_LDC		64 + STACKSIZE(%rsp)

#define OLD_OFFSET	72 + STACKSIZE(%rsp)

#define OFFSET	 224(%rsp)

#define J	 232(%rsp)

#define KKK	 240(%rsp)

#define AORIG	 248(%rsp)

#endif

#define PREFETCH     prefetcht0

#define PREFETCHSIZE (8 * 8 + 3)

	PROLOGUE

	PROFCODE

	subq	$STACKSIZE, %rsp

	movq	%rbx,  0(%rsp)

	movq	%rbp,  8(%rsp)

	movq	%r12, 16(%rsp)

	movq	%r13, 24(%rsp)

	movq	%r14, 32(%rsp)

	movq	%r15, 40(%rsp)

#ifdef WINDOWS_ABI

	movq	%rdi,    48(%rsp)

	movq	%rsi,    56(%rsp)

	movups	%xmm6,   64(%rsp)

	movups	%xmm7,   80(%rsp)

	movups	%xmm8,   96(%rsp)

	movups	%xmm9,  112(%rsp)

	movups	%xmm10, 128(%rsp)

	movups	%xmm11, 144(%rsp)

	movups	%xmm12, 160(%rsp)

	movups	%xmm13, 176(%rsp)

	movups	%xmm14, 192(%rsp)

	movups	%xmm15, 208(%rsp)

	movq	ARG1,      M

	movq	ARG2,      N

	movq	ARG3,      K

	movq	OLD_A,     A

	movq	OLD_B,     B

	movq	OLD_C,     C

#endif

	movq	OLD_LDC,   LDC

	movq	OLD_OFFSET, KK

	movq	KK, OFFSET

	leaq	(, LDC, SIZE), LDC

#ifdef LN

       leaq	(, M, SIZE), %rax

       addq	%rax, C

       imulq	K, %rax

       addq	%rax, A

#endif

#ifdef RT

       leaq	(, N, SIZE), %rax

       imulq	K, %rax

       addq	%rax, B

       movq	N, %rax

       imulq	LDC, %rax

       addq	%rax, C

#endif

#ifdef RN

	negq	KK

#endif	

#ifdef RT

       movq	N, %rax

       subq	OFFSET, %rax

       movq	%rax, KK

#endif

	movq	N,  J

	sarq	$1, J

	jle	.L40

	ALIGN_4

.L10:

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

	movq	A, AO

#else

	movq	A, AORIG

#endif

#ifdef RT

       movq	K, %rax

       salq	$1 + BASE_SHIFT, %rax

       subq	%rax, B

       leaq	(, LDC, 2), %rax

       subq	%rax, C

#endif

	movq	C, CO1

	leaq	(C, LDC, 1), CO2

#ifndef RT

	leaq	(C, LDC, 2), C

#endif

#ifdef LN

	movq	OFFSET, %rax

	addq	M, %rax

	movq	%rax, KK

#endif	

	movq	K, %rax

	salq	$BASE_SHIFT + 1, %rax

	leaq	(B, %rax), BB

#ifdef LT

	movq	OFFSET, %rax

	movq	%rax, KK

#endif

	movq	M,  I

	sarq	$2, I

	jle	.L20

	ALIGN_4

.L11:

#ifdef LN

       movq	K, %rax

       salq	$2 + BASE_SHIFT, %rax

       subq	%rax, AORIG

#endif

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

	movq	KK, %rax

	leaq	(, %rax, SIZE), %rax

	movq	AORIG, AO

	leaq	(AO, %rax, 4), AO

	leaq	(B,  %rax, 2), BO

#else

	movq	B, BO

#endif	

	prefetcht0	  0 * SIZE(BB)

	subq	   $-8 * SIZE, BB

	movsd	 0 * SIZE(AO), %xmm0

	xorps	%xmm2,  %xmm2

	movsd	 1 * SIZE(AO), %xmm4

	xorps	%xmm5,  %xmm5

	movsd	 2 * SIZE(AO), %xmm5

	xorps	%xmm6,  %xmm6

	xorps	%xmm7,  %xmm7

	movsd	 0 * SIZE(BO), %xmm1

	xorps	%xmm8,  %xmm8

	xorps	%xmm9,  %xmm9

	movsd	 1 * SIZE(BO), %xmm3

	xorps	%xmm10, %xmm10

	xorps	%xmm11, %xmm11

	prefetcht0     3 * SIZE(CO1)

	xorps	%xmm12, %xmm12

	xorps	%xmm13, %xmm13

	prefetcht0     3 * SIZE(CO2)

	xorps	%xmm14, %xmm14

	xorps	%xmm15, %xmm15

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

	movq	KK, %rax

#else

	movq	K, %rax

	subq	KK, %rax

#endif

	sarq	$2, %rax

	je	.L15

	ALIGN_4

.L12:

	addsd	 %xmm2, %xmm13

	PREFETCH (PREFETCHSIZE + 0) * SIZE(AO)

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm7, %xmm14

	movsd	 3 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm15

	PREFETCH ((PREFETCHSIZE) >> 1 + 0) * SIZE(BO)

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	addsd	 %xmm0, %xmm8

	movsd	 4 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm6

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm4, %xmm10

	movsd	 5 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 2 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm12

	movsd	 6 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm6

	movsd	 3 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm13

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm7, %xmm14

	movsd	 7 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm15

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	addsd	 %xmm0, %xmm8

	movsd	 8 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm6

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm4, %xmm10

	movsd	 9 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 4 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm12

	movsd	10 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm6

	movsd	 5 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm13

	PREFETCH (PREFETCHSIZE + 8) * SIZE(AO)

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm7, %xmm14

	movsd	11 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm15

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	addsd	 %xmm0, %xmm8

	movsd	12 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm6

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm4, %xmm10

	movsd	13 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 6 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm12

	movsd	14 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm6

	movsd	 7 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm13

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm7, %xmm14

	movsd	15 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm2

	subq   $-16 * SIZE, AO

	addsd	 %xmm6, %xmm15

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	addsd	 %xmm0, %xmm8

	movsd	 0 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm6

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addq   $  8 * SIZE, BO

	addsd	 %xmm4, %xmm10

	movsd	 1 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm2

	decq   %rax

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 0 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm12

	movsd	 2 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm6

	movsd	 1 * SIZE(BO), %xmm3

	jne    .L12

	ALIGN_4

.L15:

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

	movq	KK, %rax

#else

	movq	K, %rax

	subq	KK, %rax

#endif

	andq	$3, %rax

	BRANCH

	je .L19

	ALIGN_4

.L16:

	addsd	 %xmm2, %xmm13

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm7, %xmm14

	movsd	 3 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm15

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	addsd	 %xmm0, %xmm8

	movsd	 4 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm6

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm4, %xmm10

	movsd	 5 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 2 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm12

	movsd	 6 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm6

	movsd	 3 * SIZE(BO), %xmm3

	addq	$4 * SIZE, AO

	addq	$2 * SIZE, BO

	decq	%rax

	BRANCH

	jg	.L16

	ALIGN_4

.L19:

	addsd	 %xmm2, %xmm13

	addsd	 %xmm7, %xmm14

	addsd	 %xmm6, %xmm15

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

	movq	KK, %rax

#ifdef LN

	subq	$4, %rax

#else

	subq	$2, %rax

#endif

	leaq	(, %rax, SIZE), %rax

	movq	AORIG, AO

	leaq	(AO, %rax, 4), AO

	leaq	(B,  %rax, 2), BO

#endif

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

	movsd	 0 * SIZE(BO), %xmm0

	movsd	 1 * SIZE(BO), %xmm1

	movsd	 2 * SIZE(BO), %xmm2

	movsd	 3 * SIZE(BO), %xmm3

	movsd	 4 * SIZE(BO), %xmm4

	movsd	 5 * SIZE(BO), %xmm5

	movsd	 6 * SIZE(BO), %xmm6

	movsd	 7 * SIZE(BO), %xmm7

	subsd	%xmm8,  %xmm0

	subsd	%xmm9,  %xmm1

	subsd	%xmm10, %xmm2

	subsd	%xmm11, %xmm3

	subsd	%xmm12, %xmm4

	subsd	%xmm13, %xmm5

	subsd	%xmm14, %xmm6

	subsd	%xmm15, %xmm7

#else

	movsd	 0 * SIZE(AO), %xmm0

	movsd	 1 * SIZE(AO), %xmm2

	movsd	 2 * SIZE(AO), %xmm4

	movsd	 3 * SIZE(AO), %xmm6

	movsd	 4 * SIZE(AO), %xmm1

	movsd	 5 * SIZE(AO), %xmm3

	movsd	 6 * SIZE(AO), %xmm5

	movsd	 7 * SIZE(AO), %xmm7

	subsd	%xmm8,  %xmm0

	subsd	%xmm10, %xmm2

	subsd	%xmm12, %xmm4

	subsd	%xmm14, %xmm6

	subsd	%xmm9,  %xmm1

	subsd	%xmm11, %xmm3

	subsd	%xmm13, %xmm5

	subsd	%xmm15, %xmm7

#endif

#ifdef LN

	movsd	15 * SIZE(AO), %xmm8

	mulsd	 %xmm8, %xmm6

	movsd	14 * SIZE(AO), %xmm9

	mulsd	 %xmm8, %xmm7

	movsd	13 * SIZE(AO), %xmm11

	movaps	%xmm9,  %xmm10

	movsd	12 * SIZE(AO), %xmm13

	mulsd	%xmm6,  %xmm9

	movsd	10 * SIZE(AO), %xmm8

	mulsd	%xmm7,  %xmm10

	subsd	%xmm9,  %xmm4

	movsd	 9 * SIZE(AO), %xmm9

	subsd	%xmm10, %xmm5

	movaps	%xmm11,  %xmm12

	mulsd	%xmm6,  %xmm11

	mulsd	%xmm7,  %xmm12

	subsd	%xmm11, %xmm2

	movsd	 8 * SIZE(AO), %xmm11

	subsd	%xmm12, %xmm3

	movaps	%xmm13,  %xmm14

	mulsd	%xmm6,  %xmm13

	mulsd	%xmm7,  %xmm14

	subsd	%xmm13, %xmm0

	subsd	%xmm14, %xmm1

	mulsd	 %xmm8, %xmm4

	mulsd	 %xmm8, %xmm5

	movsd	 5 * SIZE(AO), %xmm8

	movaps	%xmm9,  %xmm10

	mulsd	%xmm4,  %xmm9

	mulsd	%xmm5,  %xmm10

	subsd	%xmm9,  %xmm2

	movsd	 4 * SIZE(AO), %xmm9

	subsd	%xmm10, %xmm3

	movaps	%xmm11,  %xmm12

	mulsd	%xmm4,  %xmm11

	mulsd	%xmm5,  %xmm12

	subsd	%xmm11, %xmm0

	movsd	 0 * SIZE(AO), %xmm11

	subsd	%xmm12, %xmm1

	mulsd	 %xmm8, %xmm2

	mulsd	 %xmm8, %xmm3

	movaps	%xmm9,  %xmm10

	mulsd	%xmm2,  %xmm9

	mulsd	%xmm3,  %xmm10

	subsd	%xmm9,  %xmm0

	subsd	%xmm10, %xmm1

	mulsd	 %xmm11, %xmm0

	mulsd	 %xmm11, %xmm1

#endif

#ifdef LT

	movsd	 0 * SIZE(AO), %xmm8

	mulsd	 %xmm8, %xmm0

	movsd	 1 * SIZE(AO), %xmm9

	mulsd	 %xmm8, %xmm1

	movsd	 2 * SIZE(AO), %xmm11

	movaps	%xmm9,  %xmm10

	movsd	 3 * SIZE(AO), %xmm13

	mulsd	%xmm0,  %xmm9

	movsd	 5 * SIZE(AO), %xmm8

	mulsd	%xmm1,  %xmm10

	subsd	%xmm9,  %xmm2

	movsd	 6 * SIZE(AO), %xmm9

	subsd	%xmm10, %xmm3

	movaps	%xmm11, %xmm12

	mulsd	%xmm0,  %xmm11

	mulsd	%xmm1,  %xmm12

	subsd	%xmm11, %xmm4

	movsd	 7 * SIZE(AO), %xmm11

	subsd	%xmm12, %xmm5

	movaps	%xmm13, %xmm14

	mulsd	%xmm0,  %xmm13

	mulsd	%xmm1,  %xmm14

	subsd	%xmm13, %xmm6

	subsd	%xmm14, %xmm7

	mulsd	 %xmm8, %xmm2

	mulsd	 %xmm8, %xmm3

	movsd	10 * SIZE(AO), %xmm8

	movaps	%xmm9,  %xmm10

	mulsd	%xmm2,  %xmm9

	mulsd	%xmm3,  %xmm10

	subsd	%xmm9,  %xmm4

	movsd	11 * SIZE(AO), %xmm9

	subsd	%xmm10, %xmm5

	movaps	%xmm11, %xmm12

	mulsd	%xmm2,  %xmm11

	mulsd	%xmm3,  %xmm12

	subsd	%xmm11, %xmm6

	subsd	%xmm12, %xmm7

	mulsd	 %xmm8, %xmm4

	mulsd	 %xmm8, %xmm5

	movsd	15 * SIZE(AO), %xmm8

	movaps	%xmm9,  %xmm10

	mulsd	%xmm4,  %xmm9

	mulsd	%xmm5,  %xmm10

	subsd	%xmm9,  %xmm6

	subsd	%xmm10, %xmm7

	mulsd	 %xmm8, %xmm6

	mulsd	 %xmm8, %xmm7

#endif

#ifdef RN

	movsd	 0 * SIZE(BO), %xmm8

	mulsd	 %xmm8, %xmm0

	movsd	 1 * SIZE(BO), %xmm9

	mulsd	 %xmm8, %xmm2

	movsd	 3 * SIZE(BO), %xmm13

	mulsd	 %xmm8, %xmm4

	mulsd	 %xmm8, %xmm6

	movaps	 %xmm9, %xmm10

	movaps	 %xmm9, %xmm11

	movaps	 %xmm9, %xmm12

	mulsd	 %xmm0, %xmm9

	mulsd	 %xmm2, %xmm10

	mulsd	 %xmm4, %xmm11

	mulsd	 %xmm6, %xmm12

	subsd	 %xmm9,  %xmm1

	subsd	 %xmm10, %xmm3

	subsd	 %xmm11, %xmm5

	subsd	 %xmm12, %xmm7

	mulsd	 %xmm13, %xmm1

	mulsd	 %xmm13, %xmm3

	mulsd	 %xmm13, %xmm5

	mulsd	 %xmm13, %xmm7

#endif

#ifdef RT

	movsd	 3 * SIZE(BO), %xmm8

	mulsd	 %xmm8, %xmm1

	movsd	 2 * SIZE(BO), %xmm9

	mulsd	 %xmm8, %xmm3

	movsd	 0 * SIZE(BO), %xmm13

	mulsd	 %xmm8, %xmm5

	mulsd	 %xmm8, %xmm7

	movaps	 %xmm9, %xmm10

	movaps	 %xmm9, %xmm11

	movaps	 %xmm9, %xmm12

	mulsd	 %xmm1, %xmm9

	mulsd	 %xmm3, %xmm10

	mulsd	 %xmm5, %xmm11

	mulsd	 %xmm7, %xmm12

	subsd	 %xmm9,  %xmm0

	subsd	 %xmm10, %xmm2

	subsd	 %xmm11, %xmm4

	subsd	 %xmm12, %xmm6

	mulsd	 %xmm13, %xmm0

	mulsd	 %xmm13, %xmm2

	mulsd	 %xmm13, %xmm4

	mulsd	 %xmm13, %xmm6

#endif

#ifdef LN

	subq	$4 * SIZE, CO1

	subq	$4 * SIZE, CO2

#endif

	movsd	%xmm0,  0 * SIZE(CO1)

	movsd	%xmm2,  1 * SIZE(CO1)

	movsd	%xmm4,  2 * SIZE(CO1)

	movsd	%xmm6,  3 * SIZE(CO1)

	movsd	%xmm1,  0 * SIZE(CO2)

	movsd	%xmm3,  1 * SIZE(CO2)

	movsd	%xmm5,  2 * SIZE(CO2)

	movsd	%xmm7,  3 * SIZE(CO2)

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

	movsd	%xmm0,   0 * SIZE(BO)

	movsd	%xmm1,   1 * SIZE(BO)

	movsd	%xmm2,   2 * SIZE(BO)

	movsd	%xmm3,   3 * SIZE(BO)

	movsd	%xmm4,   4 * SIZE(BO)

	movsd	%xmm5,   5 * SIZE(BO)

	movsd	%xmm6,   6 * SIZE(BO)

	movsd	%xmm7,   7 * SIZE(BO)

#else

	movsd	%xmm0,   0 * SIZE(AO)

	movsd	%xmm2,   1 * SIZE(AO)

	movsd	%xmm4,   2 * SIZE(AO)

	movsd	%xmm6,   3 * SIZE(AO)

	movsd	%xmm1,   4 * SIZE(AO)

	movsd	%xmm3,   5 * SIZE(AO)

	movsd	%xmm5,   6 * SIZE(AO)

	movsd	%xmm7,   7 * SIZE(AO)

#endif

#ifndef LN

	addq	$4 * SIZE, CO1

	addq	$4 * SIZE, CO2

#endif

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

	movq	K,  %rax

	subq	KK, %rax

	leaq	(,%rax, SIZE), %rax

	leaq	(AO, %rax, 4), AO

	leaq	(BO, %rax, 2), BO

#endif

#ifdef LN

	subq	$4, KK

#endif

#ifdef LT

	addq	$4, KK

#endif

#ifdef RT

       movq	K, %rax

       salq	$2 + BASE_SHIFT, %rax

       addq	%rax, AORIG

#endif

	decq	I			# i --

	jg	.L11

	ALIGN_4

.L20:

	testq	$2, M

	BRANCH

	je	.L30

#ifdef LN

       movq	K, %rax

       salq	$1 + BASE_SHIFT, %rax

       subq	%rax, AORIG

#endif

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

	movq	KK, %rax

	leaq	(, %rax, SIZE), %rax

	movq	AORIG, AO

	leaq	(AO, %rax, 2), AO

	leaq	(B,  %rax, 2), BO

#else

	movq	B, BO

#endif	

	movsd	 0 * SIZE(AO), %xmm0

	xorps	%xmm2,  %xmm2

	movsd	 1 * SIZE(AO), %xmm4

	xorps	%xmm5,  %xmm5

	movsd	 2 * SIZE(AO), %xmm5

	xorps	%xmm6,  %xmm6

	movsd	 3 * SIZE(AO), %xmm7

	movsd	 0 * SIZE(BO), %xmm1

	xorps	%xmm8,  %xmm8

	xorps	%xmm9,  %xmm9

	movsd	 1 * SIZE(BO), %xmm3

	xorps	%xmm10, %xmm10

	xorps	%xmm11, %xmm11

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

	movq	KK, %rax

#else

	movq	K, %rax

	subq	KK, %rax

#endif

	sarq	$2, %rax

	je	.L25

	ALIGN_4

.L22:

	PREFETCH (PREFETCHSIZE + 0) * SIZE(AO)

	addsd	 %xmm2, %xmm9

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm6, %xmm11

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	movsd	 2 * SIZE(BO), %xmm1

	addsd	 %xmm0, %xmm8

	movsd	 4 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm4, %xmm10

	movsd	 5 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm6

	movsd	 3 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 4 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm8

	movsd	 6 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm7, %xmm10

	movsd	 7 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm6

	movsd	 5 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm9

	movaps	 %xmm0, %xmm2

	mulsd	 %xmm1, %xmm0

	addsd	 %xmm6, %xmm11

	movaps	 %xmm4, %xmm6

	mulsd	 %xmm1, %xmm4

	movsd	 6 * SIZE(BO), %xmm1

	addsd	 %xmm0, %xmm8

	movsd	 8 * SIZE(AO), %xmm0

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm4, %xmm10

	movsd	 9 * SIZE(AO), %xmm4

	mulsd	 %xmm3, %xmm6

	movsd	 7 * SIZE(BO), %xmm3

	addsd	 %xmm2, %xmm9

	movaps	 %xmm5, %xmm2

	mulsd	 %xmm1, %xmm5

	addsd	 %xmm6, %xmm11

	movaps	 %xmm7, %xmm6

	mulsd	 %xmm1, %xmm7

	movsd	 8 * SIZE(BO), %xmm1

	addsd	 %xmm5, %xmm8

	movsd	10 * SIZE(AO), %xmm5

	mulsd	 %xmm3, %xmm2

	addsd	 %xmm7, %xmm10

	movsd	11 * SIZE(AO), %xmm7

	mulsd	 %xmm3, %xmm6

	movsd	 9 * SIZE(BO), %xmm3