/*********************************************************************/
/* 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 OLD_M %rdi
#define OLD_N %rsi
#define OLD_K %rdx
#define M %r13
#define N %r14
#define K %r15
#define A %rcx
#define B %r8
#define C %r9
#define LDC %r10
#define I %r11
#define AO %rdi
#define BO %rsi
#define CO1 %rbx
#define CO2 %rbp
#define KK %rdx
#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 48 + STACKSIZE(%rsp)
#define OLD_B 56 + STACKSIZE(%rsp)
#define OLD_C 64 + STACKSIZE(%rsp)
#define OLD_LDC 72 + STACKSIZE(%rsp)
#define OLD_OFFSET 80 + STACKSIZE(%rsp)
#define OFFSET 224(%rsp)
#define J 232(%rsp)
#define KKK 240(%rsp)
#define AORIG 248(%rsp)
#endif
#define PREFETCHSIZE (8 * 1 + 2)
#define PREFETCH prefetcht0
#define ADD1 addpd
#define ADD2 addpd
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, OLD_M
movq ARG2, OLD_N
movq ARG3, OLD_K
movq OLD_A, A
movq OLD_B, B
movq OLD_C, C
#endif
subq $-16 * SIZE, A
subq $-16 * SIZE, B
movq OLD_M, M
movq OLD_N, N
movq OLD_K, K
movq OLD_LDC, LDC
movq OLD_OFFSET, KK
salq $ZBASE_SHIFT, LDC
movq KK, OFFSET
negq KK
#ifdef LN
movq M, %rax
salq $ZBASE_SHIFT, %rax
addq %rax, C
imulq K, %rax
addq %rax, A
#endif
#ifdef RT
movq N, %rax
salq $ZBASE_SHIFT, %rax
imulq K, %rax
addq %rax, B
movq N, %rax
imulq LDC, %rax
addq %rax, C
#endif
#ifdef RT
movq N, KK
subq OFFSET, KK
#endif
testq M, M
jle .L999
movq N, J
sarq $2, J
NOBRANCH
jle .L20
ALIGN_4
.L01:
#if defined(LT) || defined(RN)
movq A, AO
#else
movq A, AORIG
#endif
#ifdef RT
movq K, %rax
salq $2 + ZBASE_SHIFT, %rax
subq %rax, B
leaq (, LDC, 4), %rax
subq %rax, C
#endif
movq C, CO1
leaq (C, LDC, 2), CO2
#ifndef RT
leaq (C, LDC, 4), C
#endif
#ifdef LN
movq OFFSET, KK
addq M, KK
#endif
movq K, %rax
salq $ZBASE_SHIFT + 2, %rax
leaq (B, %rax), BB
#ifdef LT
movq OFFSET, KK
#endif
movq M, I
ALIGN_4
.L11:
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
subq %rax, AORIG
#endif
#if defined(LN) || defined(RT)
movq AORIG, AO
movq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (B, %rax, 4), BO
#else
movq B, BO
#endif
prefetchnta -16 * SIZE(BB)
subq $-8 * SIZE, BB
xorps %xmm1, %xmm1
movaps -16 * SIZE(AO), %xmm0
xorps %xmm2, %xmm2
xorps %xmm3, %xmm3
xorps %xmm4, %xmm4
xorps %xmm8, %xmm8
prefetcht0 2 * SIZE(CO1)
xorps %xmm9, %xmm9
prefetcht0 2 * SIZE(CO1, LDC)
xorps %xmm10, %xmm10
xorps %xmm11, %xmm11
xorps %xmm12, %xmm12
prefetcht0 2 * SIZE(CO2)
xorps %xmm13, %xmm13
prefetcht0 2 * SIZE(CO2, LDC)
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
NOBRANCH
jle .L15
ALIGN_3
.L12:
PREFETCH (PREFETCHSIZE + 0) * SIZE(AO)
ADD1 %xmm1, %xmm12
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm13
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm14
movaps -14 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm15
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm8
movaps -12 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -10 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm12
movaps -8 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm13
movaps -14 * SIZE(AO), %xmm0
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm14
movaps -6 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm15
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm8
movaps -4 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -2 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm12
movaps 0 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm13
movaps -12 * SIZE(AO), %xmm0
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm14
movaps 2 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm15
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm8
movaps 4 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps 6 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -10 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm12
movaps 8 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm13
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm14
movaps 10 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm15
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm8
movaps 12 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps 14 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -8 * SIZE(AO), %xmm0
subq $-8 * SIZE, AO
subq $-32 * SIZE, BO
subq $1, %rax
BRANCH
jg .L12
ALIGN_3
.L15:
#if defined(LT) || defined(RN)
movq KK, %rax
#else
movq K, %rax
subq KK, %rax
#endif
andq $3, %rax # if (k & 1)
BRANCH
je .L18
ALIGN_3
.L16:
ADD1 %xmm1, %xmm12
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm13
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm14
movaps -14 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm15
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
ADD1 %xmm1, %xmm8
movaps -12 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -10 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -14 * SIZE(AO), %xmm0
addq $2 * SIZE, AO
addq $8 * SIZE, BO
subq $1, %rax
BRANCH
jg .L16
ALIGN_3
.L18:
#if defined(LN) || defined(RT)
movq KK, %rax
#ifdef LN
subq $1, %rax
#else
subq $4, %rax
#endif
salq $ZBASE_SHIFT, %rax
movq AORIG, AO
leaq (AO, %rax, 1), AO
leaq (B, %rax, 4), BO
#endif
ADD1 %xmm1, %xmm12
ADD2 %xmm2, %xmm13
ADD1 %xmm3, %xmm14
ADD2 %xmm4, %xmm15
pcmpeqb %xmm0, %xmm0
psllq $63, %xmm0
#if defined(LN) || defined(LT)
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
xorps %xmm0, %xmm10
xorps %xmm0, %xmm12
xorps %xmm0, %xmm14
#else
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm9
xorps %xmm0, %xmm11
xorps %xmm0, %xmm13
xorps %xmm0, %xmm15
#endif
#else
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
xorps %xmm0, %xmm10
xorps %xmm0, %xmm12
xorps %xmm0, %xmm14
#else
shufps $0x04, %xmm0, %xmm0
xorps %xmm0, %xmm9
xorps %xmm0, %xmm11
xorps %xmm0, %xmm13
xorps %xmm0, %xmm15
#endif
#endif
haddpd %xmm9, %xmm8
haddpd %xmm11, %xmm10
haddpd %xmm13, %xmm12
haddpd %xmm15, %xmm14
#if defined(LN) || defined(LT)
movapd -16 * SIZE(BO), %xmm9
movapd -14 * SIZE(BO), %xmm11
movapd -12 * SIZE(BO), %xmm13
movapd -10 * SIZE(BO), %xmm15
subpd %xmm8, %xmm9
subpd %xmm10, %xmm11
subpd %xmm12, %xmm13
subpd %xmm14, %xmm15
#else
movapd -16 * SIZE(AO), %xmm9
movapd -14 * SIZE(AO), %xmm11
movapd -12 * SIZE(AO), %xmm13
movapd -10 * SIZE(AO), %xmm15
subpd %xmm8, %xmm9
subpd %xmm10, %xmm11
subpd %xmm12, %xmm13
subpd %xmm14, %xmm15
#endif
pcmpeqb %xmm7, %xmm7
psllq $63, %xmm7
#ifndef CONJ
shufps $0x04, %xmm7, %xmm7
#else
shufps $0x40, %xmm7, %xmm7
#endif
#if defined(LN) || defined(LT)
movddup -16 * SIZE(AO), %xmm0
movddup -15 * SIZE(AO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
pshufd $0x4e, %xmm11, %xmm10
pshufd $0x4e, %xmm13, %xmm12
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm8
xorpd %xmm7, %xmm10
xorpd %xmm7, %xmm12
xorpd %xmm7, %xmm14
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
mulpd %xmm0, %xmm13
mulpd %xmm1, %xmm12
mulpd %xmm0, %xmm15
mulpd %xmm1, %xmm14
addpd %xmm8, %xmm9
addpd %xmm10, %xmm11
addpd %xmm12, %xmm13
addpd %xmm14, %xmm15
#endif
#ifdef RN
movddup -16 * SIZE(BO), %xmm0
movddup -15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
movddup -14 * SIZE(BO), %xmm2
movddup -13 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm9, %xmm2
mulpd %xmm8, %xmm3
subpd %xmm2, %xmm11
subpd %xmm3, %xmm11
movddup -12 * SIZE(BO), %xmm0
movddup -11 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm9, %xmm0
mulpd %xmm8, %xmm1
subpd %xmm0, %xmm13
subpd %xmm1, %xmm13
movddup -10 * SIZE(BO), %xmm2
movddup -9 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm9, %xmm2
mulpd %xmm8, %xmm3
subpd %xmm2, %xmm15
subpd %xmm3, %xmm15
movddup -6 * SIZE(BO), %xmm0
movddup -5 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
addpd %xmm10, %xmm11
movddup -4 * SIZE(BO), %xmm0
movddup -3 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm11, %xmm0
mulpd %xmm10, %xmm1
subpd %xmm0, %xmm13
subpd %xmm1, %xmm13
movddup -2 * SIZE(BO), %xmm2
movddup -1 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm11, %xmm2
mulpd %xmm10, %xmm3
subpd %xmm2, %xmm15
subpd %xmm3, %xmm15
movddup 4 * SIZE(BO), %xmm0
movddup 5 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm13, %xmm12
xorpd %xmm7, %xmm12
mulpd %xmm0, %xmm13
mulpd %xmm1, %xmm12
addpd %xmm12, %xmm13
movddup 6 * SIZE(BO), %xmm2
movddup 7 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm13, %xmm12
xorpd %xmm7, %xmm12
mulpd %xmm13, %xmm2
mulpd %xmm12, %xmm3
subpd %xmm2, %xmm15
subpd %xmm3, %xmm15
movddup 14 * SIZE(BO), %xmm0
movddup 15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm14
mulpd %xmm0, %xmm15
mulpd %xmm1, %xmm14
addpd %xmm14, %xmm15
#endif
#ifdef RT
movddup 14 * SIZE(BO), %xmm0
movddup 15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm14
mulpd %xmm0, %xmm15
mulpd %xmm1, %xmm14
addpd %xmm14, %xmm15
movddup 12 * SIZE(BO), %xmm2
movddup 13 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm14
mulpd %xmm15, %xmm2
mulpd %xmm14, %xmm3
subpd %xmm2, %xmm13
subpd %xmm3, %xmm13
movddup 10 * SIZE(BO), %xmm0
movddup 11 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm14
mulpd %xmm15, %xmm0
mulpd %xmm14, %xmm1
subpd %xmm0, %xmm11
subpd %xmm1, %xmm11
movddup 8 * SIZE(BO), %xmm2
movddup 9 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm15, %xmm14
xorpd %xmm7, %xmm14
mulpd %xmm15, %xmm2
mulpd %xmm14, %xmm3
subpd %xmm2, %xmm9
subpd %xmm3, %xmm9
movddup 4 * SIZE(BO), %xmm0
movddup 5 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm13, %xmm12
xorpd %xmm7, %xmm12
mulpd %xmm0, %xmm13
mulpd %xmm1, %xmm12
addpd %xmm12, %xmm13
movddup 2 * SIZE(BO), %xmm0
movddup 3 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm13, %xmm12
xorpd %xmm7, %xmm12
mulpd %xmm13, %xmm0
mulpd %xmm12, %xmm1
subpd %xmm0, %xmm11
subpd %xmm1, %xmm11
movddup 0 * SIZE(BO), %xmm2
movddup 1 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm13, %xmm12
xorpd %xmm7, %xmm12
mulpd %xmm13, %xmm2
mulpd %xmm12, %xmm3
subpd %xmm2, %xmm9
subpd %xmm3, %xmm9
movddup -6 * SIZE(BO), %xmm0
movddup -5 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
addpd %xmm10, %xmm11
movddup -8 * SIZE(BO), %xmm2
movddup -7 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm11, %xmm2
mulpd %xmm10, %xmm3
subpd %xmm2, %xmm9
subpd %xmm3, %xmm9
movddup -16 * SIZE(BO), %xmm0
movddup -15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
#endif
#ifdef LN
subq $2 * SIZE, CO1
subq $2 * SIZE, CO2
#endif
movsd %xmm9, 0 * SIZE(CO1)
movhpd %xmm9, 1 * SIZE(CO1)
movsd %xmm11, 0 * SIZE(CO1, LDC)
movhpd %xmm11, 1 * SIZE(CO1, LDC)
movsd %xmm13, 0 * SIZE(CO2)
movhpd %xmm13, 1 * SIZE(CO2)
movsd %xmm15, 0 * SIZE(CO2, LDC)
movhpd %xmm15, 1 * SIZE(CO2, LDC)
#if defined(LN) || defined(LT)
movapd %xmm9, -16 * SIZE(BO)
movapd %xmm11, -14 * SIZE(BO)
movapd %xmm13, -12 * SIZE(BO)
movapd %xmm15, -10 * SIZE(BO)
#else
movapd %xmm9, -16 * SIZE(AO)
movapd %xmm11, -14 * SIZE(AO)
movapd %xmm13, -12 * SIZE(AO)
movapd %xmm15, -10 * SIZE(AO)
#endif
#ifndef LN
addq $2 * SIZE, CO1
addq $2 * SIZE, CO2
#endif
#if defined(LT) || defined(RN)
movq K, %rax
subq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (BO, %rax, 4), BO
#endif
#ifdef LN
subq $1, KK
#endif
#ifdef LT
addq $1, KK
#endif
#ifdef RT
movq K, %rax
salq $ZBASE_SHIFT, %rax
addq %rax, AORIG
#endif
decq I
BRANCH
jg .L11
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
leaq (B, %rax, 4), B
#endif
#if defined(LT) || defined(RN)
movq BO, B
#endif
#ifdef RN
addq $4, KK
#endif
#ifdef RT
subq $4, KK
#endif
subq $1, J
BRANCH
jg .L01
ALIGN_4
.L20:
testq $2, N
BRANCH
jle .L30
#if defined(LT) || defined(RN)
movq A, AO
#else
movq A, AORIG
#endif
#ifdef RT
movq K, %rax
salq $1 + ZBASE_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, KK
addq M, KK
#endif
#ifdef LT
movq OFFSET, KK
#endif
movq M, I
ALIGN_4
.L21:
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
subq %rax, AORIG
#endif
#if defined(LN) || defined(RT)
movq AORIG, AO
movq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (B, %rax, 2), BO
#else
movq B, BO
#endif
xorps %xmm1, %xmm1
movaps -16 * SIZE(AO), %xmm0
xorps %xmm2, %xmm2
xorps %xmm3, %xmm3
xorps %xmm4, %xmm4
xorps %xmm8, %xmm8
prefetcht0 2 * SIZE(CO1)
xorps %xmm9, %xmm9
prefetcht0 2 * SIZE(CO2)
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
NOBRANCH
jle .L25
ALIGN_3
.L22:
PREFETCH (PREFETCHSIZE + 0) * SIZE(AO)
ADD1 %xmm1, %xmm8
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -14 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -14 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm8
movaps -12 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -10 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -12 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm8
movaps -8 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -6 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -10 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm8
movaps -4 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -2 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -8 * SIZE(AO), %xmm0
subq $-8 * SIZE, AO
subq $-16 * SIZE, BO
subq $1, %rax
BRANCH
jg .L22
ALIGN_3
.L25:
#if defined(LT) || defined(RN)
movq KK, %rax
#else
movq K, %rax
subq KK, %rax
#endif
andq $3, %rax # if (k & 1)
BRANCH
je .L28
ALIGN_3
.L26:
ADD1 %xmm1, %xmm8
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
ADD1 %xmm3, %xmm10
movaps -14 * SIZE(BO), %xmm3
ADD2 %xmm4, %xmm11
pshufd $0x4e, %xmm3, %xmm4
mulpd %xmm0, %xmm3
mulpd %xmm0, %xmm4
movaps -14 * SIZE(AO), %xmm0
addq $2 * SIZE, AO
addq $4 * SIZE, BO
subq $1, %rax
BRANCH
jg .L26
ALIGN_3
.L28:
#if defined(LN) || defined(RT)
movq KK, %rax
#ifdef LN
subq $1, %rax
#else
subq $2, %rax
#endif
salq $ZBASE_SHIFT, %rax
movq AORIG, AO
leaq (AO, %rax, 1), AO
leaq (B, %rax, 2), BO
#endif
ADD1 %xmm1, %xmm8
ADD2 %xmm2, %xmm9
ADD1 %xmm3, %xmm10
ADD2 %xmm4, %xmm11
pcmpeqb %xmm0, %xmm0
psllq $63, %xmm0
#if defined(LN) || defined(LT)
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
xorps %xmm0, %xmm10
#else
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm9
xorps %xmm0, %xmm11
#endif
#else
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
xorps %xmm0, %xmm10
#else
shufps $0x04, %xmm0, %xmm0
xorps %xmm0, %xmm9
xorps %xmm0, %xmm11
#endif
#endif
haddpd %xmm9, %xmm8
haddpd %xmm11, %xmm10
#if defined(LN) || defined(LT)
movapd -16 * SIZE(BO), %xmm9
movapd -14 * SIZE(BO), %xmm11
subpd %xmm8, %xmm9
subpd %xmm10, %xmm11
#else
movapd -16 * SIZE(AO), %xmm9
movapd -14 * SIZE(AO), %xmm11
subpd %xmm8, %xmm9
subpd %xmm10, %xmm11
#endif
pcmpeqb %xmm7, %xmm7
psllq $63, %xmm7
#ifndef CONJ
shufps $0x04, %xmm7, %xmm7
#else
shufps $0x40, %xmm7, %xmm7
#endif
#if defined(LN) || defined(LT)
movddup -16 * SIZE(AO), %xmm0
movddup -15 * SIZE(AO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm8
xorpd %xmm7, %xmm10
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
addpd %xmm8, %xmm9
addpd %xmm10, %xmm11
#endif
#ifdef RN
movddup -16 * SIZE(BO), %xmm0
movddup -15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
movddup -14 * SIZE(BO), %xmm2
movddup -13 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm9, %xmm2
mulpd %xmm8, %xmm3
subpd %xmm2, %xmm11
subpd %xmm3, %xmm11
movddup -10 * SIZE(BO), %xmm0
movddup -9 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
addpd %xmm10, %xmm11
#endif
#ifdef RT
movddup -10 * SIZE(BO), %xmm0
movddup -9 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm0, %xmm11
mulpd %xmm1, %xmm10
addpd %xmm10, %xmm11
movddup -12 * SIZE(BO), %xmm2
movddup -11 * SIZE(BO), %xmm3
pshufd $0x4e, %xmm11, %xmm10
xorpd %xmm7, %xmm10
mulpd %xmm11, %xmm2
mulpd %xmm10, %xmm3
subpd %xmm2, %xmm9
subpd %xmm3, %xmm9
movddup -16 * SIZE(BO), %xmm0
movddup -15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
#endif
#ifdef LN
subq $2 * SIZE, CO1
subq $2 * SIZE, CO2
#endif
movsd %xmm9, 0 * SIZE(CO1)
movhpd %xmm9, 1 * SIZE(CO1)
movsd %xmm11, 0 * SIZE(CO2)
movhpd %xmm11, 1 * SIZE(CO2)
#if defined(LN) || defined(LT)
movapd %xmm9, -16 * SIZE(BO)
movapd %xmm11, -14 * SIZE(BO)
#else
movapd %xmm9, -16 * SIZE(AO)
movapd %xmm11, -14 * SIZE(AO)
#endif
#ifndef LN
addq $2 * SIZE, CO1
addq $2 * SIZE, CO2
#endif
#if defined(LT) || defined(RN)
movq K, %rax
subq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (BO, %rax, 2), BO
#endif
#ifdef LN
subq $1, KK
#endif
#ifdef LT
addq $1, KK
#endif
#ifdef RT
movq K, %rax
salq $ZBASE_SHIFT, %rax
addq %rax, AORIG
#endif
decq I
BRANCH
jg .L21
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
leaq (B, %rax, 2), B
#endif
#if defined(LT) || defined(RN)
movq BO, B
#endif
#ifdef RN
addq $2, KK
#endif
#ifdef RT
subq $2, KK
#endif
ALIGN_4
.L30:
testq $1, N
BRANCH
jle .L999
#if defined(LT) || defined(RN)
movq A, AO
#else
movq A, AORIG
#endif
#ifdef RT
movq K, %rax
salq $ZBASE_SHIFT, %rax
subq %rax, B
subq LDC, C
#endif
movq C, CO1
#ifndef RT
addq LDC, C
#endif
#ifdef LN
movq OFFSET, KK
addq M, KK
#endif
#ifdef LT
movq OFFSET, KK
#endif
movq M, I
ALIGN_4
.L31:
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
subq %rax, AORIG
#endif
#if defined(LN) || defined(RT)
movq AORIG, AO
movq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (B, %rax, 1), BO
#else
movq B, BO
#endif
xorps %xmm1, %xmm1
movaps -16 * SIZE(AO), %xmm0
xorps %xmm2, %xmm2
xorps %xmm8, %xmm8
prefetcht0 2 * SIZE(CO1)
xorps %xmm9, %xmm9
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
NOBRANCH
jle .L35
ALIGN_3
.L32:
PREFETCH (PREFETCHSIZE + 0) * SIZE(AO)
ADD1 %xmm1, %xmm8
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
movaps -14 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm10
movaps -14 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm11
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
movaps -12 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm8
movaps -12 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
movaps -10 * SIZE(AO), %xmm0
ADD1 %xmm1, %xmm10
movaps -10 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm11
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
movaps -8 * SIZE(AO), %xmm0
subq $-8 * SIZE, AO
subq $-8 * SIZE, BO
subq $1, %rax
BRANCH
jg .L32
addpd %xmm10, %xmm8
addpd %xmm11, %xmm9
ALIGN_3
.L35:
#if defined(LT) || defined(RN)
movq KK, %rax
#else
movq K, %rax
subq KK, %rax
#endif
andq $3, %rax
BRANCH
je .L38
ALIGN_3
.L36:
ADD1 %xmm1, %xmm8
movaps -16 * SIZE(BO), %xmm1
ADD2 %xmm2, %xmm9
pshufd $0x4e, %xmm1, %xmm2
mulpd %xmm0, %xmm1
mulpd %xmm0, %xmm2
movaps -14 * SIZE(AO), %xmm0
addq $2 * SIZE, AO
addq $2 * SIZE, BO
subq $1, %rax
BRANCH
jg .L36
ALIGN_3
.L38:
#if defined(LN) || defined(RT)
movq KK, %rax
subq $1, %rax
salq $ZBASE_SHIFT, %rax
movq AORIG, AO
leaq (AO, %rax, 1), AO
leaq (B, %rax, 1), BO
#endif
ADD1 %xmm1, %xmm8
ADD2 %xmm2, %xmm9
pcmpeqb %xmm0, %xmm0
psllq $63, %xmm0
#if defined(LN) || defined(LT)
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
#else
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm9
#endif
#else
#ifndef CONJ
shufps $0x40, %xmm0, %xmm0
xorps %xmm0, %xmm8
#else
shufps $0x04, %xmm0, %xmm0
xorps %xmm0, %xmm9
#endif
#endif
haddpd %xmm9, %xmm8
#if defined(LN) || defined(LT)
movapd -16 * SIZE(BO), %xmm9
movapd -14 * SIZE(BO), %xmm11
subpd %xmm8, %xmm9
#else
movapd -16 * SIZE(AO), %xmm9
movapd -14 * SIZE(AO), %xmm11
subpd %xmm8, %xmm9
#endif
pcmpeqb %xmm7, %xmm7
psllq $63, %xmm7
#ifndef CONJ
shufps $0x04, %xmm7, %xmm7
#else
shufps $0x40, %xmm7, %xmm7
#endif
#if defined(LN) || defined(LT)
movddup -16 * SIZE(AO), %xmm0
movddup -15 * SIZE(AO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
#endif
#if defined(RN) || defined(RT)
movddup -16 * SIZE(BO), %xmm0
movddup -15 * SIZE(BO), %xmm1
pshufd $0x4e, %xmm9, %xmm8
xorpd %xmm7, %xmm8
mulpd %xmm0, %xmm9
mulpd %xmm1, %xmm8
addpd %xmm8, %xmm9
#endif
#ifdef LN
subq $2 * SIZE, CO1
#endif
movsd %xmm9, 0 * SIZE(CO1)
movhpd %xmm9, 1 * SIZE(CO1)
#if defined(LN) || defined(LT)
movapd %xmm9, -16 * SIZE(BO)
#else
movapd %xmm9, -16 * SIZE(AO)
#endif
#ifndef LN
addq $2 * SIZE, CO1
#endif
#if defined(LT) || defined(RN)
movq K, %rax
subq KK, %rax
salq $ZBASE_SHIFT, %rax
leaq (AO, %rax, 1), AO
leaq (BO, %rax, 1), BO
#endif
#ifdef LN
subq $1, KK
#endif
#ifdef LT
addq $1, KK
#endif
#ifdef RT
movq K, %rax
salq $ZBASE_SHIFT, %rax
addq %rax, AORIG
#endif
decq I
BRANCH
jg .L31
#ifdef LN
movq K, %rax
salq $ZBASE_SHIFT, %rax
leaq (B, %rax, 1), B
#endif
#if defined(LT) || defined(RN)
movq BO, B
#endif
#ifdef RN
addq $1, KK
#endif
#ifdef RT
subq $1, KK
#endif
ALIGN_4
.L999:
movq 0(%rsp), %rbx
movq 8(%rsp), %rbp
movq 16(%rsp), %r12
movq 24(%rsp), %r13
movq 32(%rsp), %r14
movq 40(%rsp), %r15
#ifdef WINDOWS_ABI
movq 48(%rsp), %rdi
movq 56(%rsp), %rsi
movups 64(%rsp), %xmm6
movups 80(%rsp), %xmm7
movups 96(%rsp), %xmm8
movups 112(%rsp), %xmm9
movups 128(%rsp), %xmm10
movups 144(%rsp), %xmm11
movups 160(%rsp), %xmm12
movups 176(%rsp), %xmm13
movups 192(%rsp), %xmm14
movups 208(%rsp), %xmm15
#endif
addq $STACKSIZE, %rsp
ret
EPILOGUE