/*********************************************************************/
/* 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 DOUBLE
#define PREFETCH_SIZE (8 * 16)
#else
#define PREFETCH_SIZE (1 * 64)
#endif
#define ALPHA f8
#define N r32
#define X1 r36
#define INCX r37
#define X2 r14
#define Y1 r15
#define Y2 r16
#define PRE1 r17
#define I r18
#define NAND15 r19
#define INCX5 r20
#define INCX16 r21
#define XX r22
#define PR r30
#define ARLC r31
PROLOGUE
.prologue
PROFCODE
{ .mfi
shladd INCX = INCX, BASE_SHIFT, r0
fcmp.eq p0, p6 = ALPHA, f0
.save ar.lc, ARLC
mov ARLC = ar.lc
}
{ .mib
cmp.ge p7, p0 = 0, N
tbit.z p0, p10 = X1, BASE_SHIFT
(p7) br.ret.sptk.many b0
}
.body
;;
{ .mmi
mov XX = X1
(p10) LDFD f32 = [X1], INCX
mov PR = pr
}
{ .mmi
shladd INCX5 = INCX, 2, INCX
shladd INCX16 = INCX, 4, r0
(p10) adds N = -1, N
}
;;
{ .mmi
shladd X2 = INCX, 2, X1
nop __LINE__
mov ar.ec = 5
}
{ .mmi
and NAND15 = 15, N
nop __LINE__
shr I = N, 4
}
;;
{ .mmi
adds I = -1, I
nop __LINE__
tbit.z p0, p12 = N, 3
}
{ .mmb
cmp.ge p9, p0 = 0, NAND15
adds PRE1 = PREFETCH_SIZE * SIZE + 192, XX
(p6) br.cond.dptk .L100 // if (alpha != 0) goto L3
}
;;
{ .mmi
(p10) STFD [XX] = f0
nop __LINE__
mov ar.lc = I
}
{ .mmb
cmp.gt p8, p0 = 0, I
(p8) br.cond.dpnt .L30
}
;;
.align 32
.L20:
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
lfetch.excl.nt1 [PRE1], INCX16
add X1 = INCX, X1
add X2 = INCX, X2
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX, X1
add X2 = INCX, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX, X1
add X2 = INCX, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX5, X1
add X2 = INCX5, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX, X1
add X2 = INCX, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX, X1
add X2 = INCX, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmi
add X1 = INCX, X1
add X2 = INCX, X2
nop __LINE__
}
;;
{.mmi
STFD [X1] = f0
STFD [X2] = f0
nop __LINE__
}
{.mmb
add X1 = INCX5, X1
add X2 = INCX5, X2
br.cloop.sptk.few .L20
}
;;
.align 16
.L30:
{ .mmi
(p12) STFD [X1] = f0
(p12) STFD [X2] = f0
mov ar.lc = ARLC
}
{ .mmb
(p12) add X1 = INCX, X1
(p12) add X2 = INCX, X2
(p9) br.ret.sptk.many b0
}
;;
{ .mmi
(p12) STFD [X1] = f0
(p12) add X1 = INCX, X1
tbit.z p0, p13 = N, 2
}
{ .mmi
(p12) STFD [X2] = f0
(p12) add X2 = INCX, X2
tbit.z p0, p14 = N, 1
}
;;
{ .mmi
(p12) STFD [X1] = f0
(p12) add X1 = INCX, X1
tbit.z p0, p15 = N, 0
}
{ .mmb
(p12) STFD [X2] = f0
(p12) add X2 = INCX, X2
nop __LINE__
}
;;
{ .mmb
(p12) STFD [X1] = f0
(p12) add X1 = INCX5, X1
nop __LINE__
}
{ .mmb
(p12) STFD [X2] = f0
(p12) add X2 = INCX5, X2
nop __LINE__
}
;;
{ .mmb
(p13) STFD [X1] = f0
(p13) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p13) STFD [X1] = f0
(p13) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p13) STFD [X1] = f0
(p13) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p13) STFD [X1] = f0
(p13) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p14) STFD [X1] = f0
(p14) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p14) STFD [X1] = f0
(p14) add X1 = INCX, X1
nop __LINE__
}
;;
{ .mmb
(p15) STFD [X1] = f0
nop __LINE__
br.ret.sptk.many b0
}
;;
.align 32
.L100:
{ .mmi
mov Y1 = X1
shladd Y2 = INCX, 2, X1
mov pr.rot = 0
}
{ .mmf
cmp.gt p8, p0 = 0, I
shladd X2 = INCX, 2, X1
(p10) FMPY f32 = ALPHA, f32
}
;;
{ .mmi
(p10) STFD [XX] = f32
cmp.eq p0, p7 = SIZE, INCX
mov ar.lc = I
}
{ .mbb
cmp.eq p16, p0 = r0, r0
(p7) br.cond.dpnt .L300
(p8) br.cond.dpnt .L120
}
;;
.align 32
.L110:
{ .mmf
(p21) STFD [Y1] = f6, 1 * SIZE
(p21) STFD [Y2] = f7, 1 * SIZE
(p20) FMPY f112 = ALPHA, f36
}
{ .mmf
(p16) lfetch.excl.nt1 [PRE1], 16 * SIZE
(p16) LDFPD f32, f37 = [X1], 2 * SIZE
(p20) FMPY f113 = ALPHA, f56
}
;;
{ .mmf
(p21) STFD [Y1] = f10, 1 * SIZE
(p21) STFD [Y2] = f11, 1 * SIZE
(p20) FMPY f114 = ALPHA, f41
}
{ .mfi
(p16) LDFPD f42, f47 = [X1], 2 * SIZE
(p20) FMPY f115 = ALPHA, f61
nop __LINE__
}
;;
{ .mmf
(p21) STFD [Y1] = f12, 1 * SIZE
(p21) STFD [Y2] = f13, 1 * SIZE
(p20) FMPY f116 = ALPHA, f46
}
{ .mfi
(p16) LDFPD f52, f57 = [X1], 2 * SIZE
(p20) FMPY f117 = ALPHA, f66
nop __LINE__
}
;;
{ .mmf
(p21) STFD [Y1] = f14, 5 * SIZE
(p21) STFD [Y2] = f15, 5 * SIZE
(p20) FMPY f118 = ALPHA, f51
}
{ .mfi
(p16) LDFPD f62, f67 = [X1], 2 * SIZE
(p20) FMPY f119 = ALPHA, f71
nop __LINE__
}
;;
{ .mmf
(p20) STFD [Y1] = f112, 1 * SIZE
(p20) STFD [Y2] = f113, 1 * SIZE
(p20) FMPY f6 = ALPHA, f76
}
{ .mfi
(p16) LDFPD f72, f77 = [X1], 2 * SIZE
(p20) FMPY f7 = ALPHA, f96
nop __LINE__
}
;;
{ .mmf
(p20) STFD [Y1] = f114, 1 * SIZE
(p20) STFD [Y2] = f115, 1 * SIZE
(p20) FMPY f10 = ALPHA, f81
}
{ .mfi
(p16) LDFPD f82, f87 = [X1], 2 * SIZE
(p20) FMPY f11 = ALPHA, f101
nop __LINE__
}
;;
{ .mmf
(p20) STFD [Y1] = f116, 1 * SIZE
(p20) STFD [Y2] = f117, 1 * SIZE
(p20) FMPY f12 = ALPHA, f86
}
{ .mfi
(p16) LDFPD f92, f97 = [X1], 2 * SIZE
(p20) FMPY f13 = ALPHA, f106
(p20) shladd X2 = INCX, 2, X1
}
;;
{ .mmf
(p20) STFD [Y1] = f118, 5 * SIZE
(p20) STFD [Y2] = f119, 5 * SIZE
(p20) FMPY f14 = ALPHA, f91
}
{ .mfb
(p16) LDFPD f102, f107 = [X1], 2 * SIZE
(p20) FMPY f15 = ALPHA, f111
br.ctop.sptk.few .L110
}
;;
.align 32
.L120:
{ .mmi
(p21) STFD [Y1] = f6, 1 * SIZE
(p21) STFD [Y2] = f7, 1 * SIZE
tbit.z p0, p13 = N, 2
}
{ .mmi
(p12) LDFPD f32, f33 = [X1], 2 * SIZE
(p12) LDFPD f36, f37 = [X2], 2 * SIZE
nop __LINE__
}
;;
{ .mmi
(p21) STFD [Y1] = f10, 1 * SIZE
(p21) STFD [Y2] = f11, 1 * SIZE
mov ar.lc = ARLC
}
{ .mmi
(p12) LDFPD f34, f35 = [X1]
(p12) LDFPD f38, f39 = [X2]
(p12) adds X1 = 6 * SIZE,X1
}
;;
{ .mmi
(p21) STFD [Y1] = f12, 1 * SIZE
(p21) STFD [Y2] = f13, 1 * SIZE
tbit.z p0, p14 = N, 1
}
{ .mmi
(p13) LDFPD f40, f41 = [X1], 2 * SIZE
nop __LINE__
nop __LINE__
}
;;
{ .mmi
(p21) STFD [Y1] = f14, 5 * SIZE
(p21) STFD [Y2] = f15, 5 * SIZE
mov pr = PR, -65474
}
{ .mib
(p13) LDFPD f42, f43 = [X1], 2 * SIZE
nop __LINE__
(p9) br.ret.sptk.many b0
}
;;
{ .mmi
(p14) LDFPD f44, f45 = [X1], 2 * SIZE
nop __LINE__
tbit.z p0, p15 = N, 0
}
;;
{ .mmi
(p15) LDFD f46 = [X1]
nop __LINE__
nop __LINE__
}
;;
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f32 = ALPHA, f32
}
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f36 = ALPHA, f36
}
;;
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f33 = ALPHA, f33
}
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f37 = ALPHA, f37
}
;;
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f34 = ALPHA, f34
}
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f38 = ALPHA, f38
}
;;
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f35 = ALPHA, f35
}
{ .mmf
nop __LINE__
nop __LINE__
(p12) FMPY f39 = ALPHA, f39
}
;;
{ .mmf
(p12) STFD [Y1] = f32, 1 * SIZE
nop __LINE__
(p13) FMPY f40 = ALPHA, f40
}
{ .mmf
(p12) STFD [Y2] = f36, 1 * SIZE
nop __LINE__
(p13) FMPY f41 = ALPHA, f41
}
;;
{ .mmf
(p12) STFD [Y1] = f33, 1 * SIZE
nop __LINE__
(p13) FMPY f42 = ALPHA, f42
}
{ .mmf
(p12) STFD [Y2] = f37, 1 * SIZE
nop __LINE__
(p13) FMPY f43 = ALPHA, f43
}
;;
{ .mmf
(p12) STFD [Y1] = f34, 1 * SIZE
nop __LINE__
(p14) FMPY f44 = ALPHA, f44
}
{ .mmf
(p12) STFD [Y2] = f38, 1 * SIZE
nop __LINE__
(p14) FMPY f45 = ALPHA, f45
}
;;
{ .mmf
(p12) STFD [Y1] = f35, 5 * SIZE
(p12) STFD [Y2] = f39, 5 * SIZE
(p15) FMPY f46 = ALPHA, f46
}
;;
{ .mmi
(p13) STFD [Y1] = f40, 1 * SIZE
;;
(p13) STFD [Y1] = f41, 1 * SIZE
nop __LINE__
}
;;
{ .mmi
(p13) STFD [Y1] = f42, 1 * SIZE
;;
(p13) STFD [Y1] = f43, 1 * SIZE
nop __LINE__
}
;;
{ .mmi
(p14) STFD [Y1] = f44, 1 * SIZE
;;
(p14) STFD [Y1] = f45, 1 * SIZE
nop __LINE__
}
;;
{ .mmb
(p15) STFD [Y1] = f46
nop __LINE__
br.ret.sptk.many b0
}
;;
.align 32
.L300:
{ .mmi
adds PRE1 = PREFETCH_SIZE * SIZE + 64, X1
nop __LINE__
mov.i ar.ec = 6
}
{ .mmb
cmp.gt p8, p0 = 0, I
nop __LINE__
(p8) br.cond.dpnt .L320
}
;;
.align 32
.L310:
{ .mmf
(p16) lfetch.excl.nt1 [PRE1], INCX16
(p16) LDFD f32 = [X1], INCX
(p21) FMPY f6 = ALPHA, f37
}
{ .mmb
(p22) STFD [Y1] = f12
(p22) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f38 = [X1], INCX
(p21) FMPY f7 = ALPHA, f43
nop __LINE__
}
{ .mmb
(p22) STFD [Y1] = f13
(p22) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f44 = [X1], INCX
(p21) FMPY f10 = ALPHA, f49
nop __LINE__
}
{ .mmb
(p22) STFD [Y1] = f14
(p22) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f50 = [X1], INCX
(p21) FMPY f11 = ALPHA, f55
nop __LINE__
}
{ .mmb
(p22) STFD [Y1] = f15
(p22) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f56 = [X1], INCX
(p21) FMPY f12 = ALPHA, f61
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f6
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f62 = [X1], INCX
(p21) FMPY f13 = ALPHA, f67
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f7
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f68 = [X1], INCX
(p21) FMPY f14 = ALPHA, f73
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f10
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f74 = [X1], INCX
(p21) FMPY f15 = ALPHA, f79
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f11
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f80 = [X1], INCX
(p21) FMPY f6 = ALPHA, f85
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f12
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f86 = [X1], INCX
(p21) FMPY f7 = ALPHA, f91
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f13
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f92 = [X1], INCX
(p21) FMPY f10 = ALPHA, f97
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f14
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f98 = [X1], INCX
(p21) FMPY f11 = ALPHA, f103
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f15
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f104 = [X1], INCX
(p21) FMPY f12 = ALPHA, f109
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f6
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f110 = [X1], INCX
(p21) FMPY f13 = ALPHA, f115
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f7
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f116 = [X1], INCX
(p21) FMPY f14 = ALPHA, f121
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f10
(p21) add Y1 = INCX, Y1
nop __LINE__
}
;;
{ .mfb
(p16) LDFD f122 = [X1], INCX
(p21) FMPY f15 = ALPHA, f127
nop __LINE__
}
{ .mmb
(p21) STFD [Y1] = f11
(p21) add Y1 = INCX, Y1
br.ctop.sptk.few .L310
}
;;
STFD [Y1] = f12
add Y1 = INCX, Y1
shladd Y2 = INCX, 2, X1
;;
STFD [Y1] = f13
add Y1 = INCX, Y1
shladd X2 = INCX, 2, X1
;;
STFD [Y1] = f14
add Y1 = INCX, Y1
;;
STFD [Y1] = f15
add Y1 = INCX, Y1
;;
.align 16
.L320:
{ .mmi
(p12) LDFD f48 = [X1], INCX
(p12) LDFD f52 = [X2], INCX
mov ar.lc = ARLC
}
;;
{ .mmi
(p12) LDFD f49 = [X1], INCX
(p12) LDFD f53 = [X2], INCX
mov pr = PR, -65474
}
{ .mmb
nop.m 0
nop.m 0
(p9) br.ret.sptk.many b0
}
;;
{ .mmi
(p12) LDFD f50 = [X1], INCX
(p12) LDFD f54 = [X2], INCX
tbit.z p0, p13 = N, 2
}
;;
{ .mmi
(p12) LDFD f51 = [X1], INCX5
(p12) LDFD f55 = [X2], INCX5
tbit.z p0, p14 = N, 1
}
;;
(p13) LDFD f56 = [X1], INCX
tbit.z p0, p15 = N, 0
;;
(p13) LDFD f57 = [X1], INCX
;;
{ .mfi
(p13) LDFD f58 = [X1], INCX
(p12) FMPY f48 = ALPHA, f48
}
{ .mfi
(p12) FMPY f52 = ALPHA, f52
}
;;
{ .mfi
(p13) LDFD f59 = [X1], INCX
(p12) FMPY f49 = ALPHA, f49
}
{ .mfi
(p12) FMPY f53 = ALPHA, f53
}
;;
{ .mfi
(p14) LDFD f60 = [X1], INCX
(p12) FMPY f50 = ALPHA, f50
}
{ .mfi
(p12) FMPY f54 = ALPHA, f54
}
;;
{ .mfi
(p14) LDFD f61 = [X1], INCX
(p12) FMPY f51 = ALPHA, f51
}
{ .mfi
(p12) FMPY f55 = ALPHA, f55
}
;;
{ .mmf
(p12) STFD [Y1] = f48
(p12) STFD [Y2] = f52
(p13) FMPY f56 = ALPHA, f56
}
{ .mmi
(p15) LDFD f62 = [X1]
(p12) add Y1 = INCX, Y1
(p12) add Y2 = INCX, Y2
}
;;
{ .mmf
(p12) STFD [Y1] = f49
(p12) STFD [Y2] = f53
(p13) FMPY f57 = ALPHA, f57
}
{ .mmi
(p12) add Y1 = INCX, Y1
(p12) add Y2 = INCX, Y2
nop __LINE__
}
;;
{ .mmf
(p12) STFD [Y1] = f50
(p12) STFD [Y2] = f54
(p13) FMPY f58 = ALPHA, f58
}
{ .mmi
(p12) add Y1 = INCX, Y1
(p12) add Y2 = INCX, Y2
nop __LINE__
}
;;
{ .mmf
(p12) STFD [Y1] = f51
(p12) STFD [Y2] = f55
(p13) FMPY f59 = ALPHA, f59
}
{ .mmi
(p12) add Y1 = INCX5, Y1
(p12) add Y2 = INCX5, Y2
nop __LINE__
}
;;
{ .mfi
(p13) STFD [Y1] = f56
(p14) FMPY f60 = ALPHA, f60
(p13) add Y1 = INCX, Y1
}
;;
{ .mfi
(p13) STFD [Y1] = f57
(p14) FMPY f61 = ALPHA, f61
(p13) add Y1 = INCX, Y1
}
;;
{ .mfi
(p13) STFD [Y1] = f58
(p15) FMPY f62 = ALPHA, f62
(p13) add Y1 = INCX, Y1
}
;;
{ .mmi
(p13) STFD [Y1] = f59
(p13) add Y1 = INCX, Y1
}
;;
{ .mmi
(p14) STFD [Y1] = f60
(p14) add Y1 = INCX, Y1
}
;;
{ .mmi
(p14) STFD [Y1] = f61
(p14) add Y1 = INCX, Y1
}
;;
{ .mib
(p15) STFD [Y1] = f62
mov pr = PR, -65474
br.ret.sptk.many b0
}
EPILOGUE