1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
|
#!/usr/bin/perl
#
#########################################################################
# #
# Open Dynamics Engine, Copyright (C) 2001,2002 Russell L. Smith. #
# All rights reserved. Email: russ@q12.org Web: www.q12.org #
# #
# This library is free software; you can redistribute it and/or #
# modify it under the terms of EITHER: #
# (1) The GNU Lesser General Public License as published by the Free #
# Software Foundation; either version 2.1 of the License, or (at #
# your option) any later version. The text of the GNU Lesser #
# General Public License is included with this library in the #
# file LICENSE.TXT. #
# (2) The BSD-style license that is included with this library in #
# the file LICENSE-BSD.TXT. #
# #
# This library is distributed in the hope that it will be useful, #
# but WITHOUT ANY WARRANTY; without even the implied warranty of #
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the files #
# LICENSE.TXT and LICENSE-BSD.TXT for more details. #
# #
#########################################################################
# multi-dot-product code generator. this code generator is based on the
# dot-product code generator.
#
# code generation parameters, set in a parameters file:
# FNAME : name of source file to generate - a .c file will be made
# N1 : block size (number of `a' vectors to dot)
# UNROLL1 : inner loop unrolling factor (1..)
# FETCH : max num of a[i]'s and b[i]'s to load ahead of muls
# LAT1 : load -> mul latency (>=1)
# LAT2 : mul -> add latency (>=1). if this is 1, use fused mul-add
#
#############################################################################
require ("BuildUtil");
# get and check code generation parameters
error ("Usage: BuildMultidot <parameters-file>") if $#ARGV != 0;
do $ARGV[0];
if (!defined($FNAME) || !defined($N1) || !defined($UNROLL1) ||
!defined($FETCH) || !defined($LAT1) || !defined($LAT2)) {
error ("code generation parameters not defined");
}
# check parameters
error ("bad N1") if $N1 < 2;
error ("bad UNROLL1") if $UNROLL1 < 1;
error ("bad FETCH") if $FETCH < 1;
error ("bad LAT1") if $LAT1 < 1;
error ("bad LAT2") if $LAT2 < 1;
#############################################################################
open (FOUT,">$FNAME.c") or die "can't open $FNAME.c for writing";
# file and function header
output (<<END);
/* generated code, do not edit. */
#include "ode/matrix.h"
END
output ("void dMultidot$N1 (");
for ($i=0; $i<$N1; $i++) {
output ("const dReal *a$i, ");
}
output ("const dReal *b, dReal *outsum, int n)\n{\n");
output ("dReal ");
for ($i=0; $i<$UNROLL1; $i++) {
for ($j=0; $j<$N1; $j++) {
output ("p$i$j,");
output ("m$i$j,") if $LAT2 > 1;
}
output ("q$i,");
}
for ($i=0; $i<$N1; $i++) {
output ("sum$i");
output (",") if $i < ($N1-1);
}
output (";\n");
for ($i=0; $i<$N1; $i++) {
output ("sum$i = 0;\n");
}
output (<<END);
n -= $UNROLL1;
while (n >= 0) {
END
@load = (); # slot where a[i]'s and b[i]'s loaded
@mul = (); # slot where multiply i happened
@add = (); # slow where add i happened
for ($i=0; $i<$N1; $i++) {
output ("p0$i = a$i [0];\n");
}
output ("q0 = b[0];\n");
push (@load,0);
$slot=0; # one slot for every load/mul/add/nop issued
for (;;) {
$startslot = $slot;
# do next load
if (($#load - $#mul) < $FETCH && ($#load+1) < $UNROLL1) {
push (@load,$slot);
for ($j=0; $j<$N1; $j++) {
output ("p$#load$j = a$j [$#load];\n");
}
output ("q$#load = b[$#load];\n");
$slot++;
}
# do next multiply
if ($#load > $#mul && $slot >= ($load[$#mul+1] + $LAT1) &&
($#mul+1) < $UNROLL1) {
push (@mul,$slot);
if ($LAT2 > 1) {
for ($j=0; $j<$N1; $j++) {
output ("m$#mul$j = p$#mul$j * q$#mul;\n");
}
}
else {
for ($j=0; $j<$N1; $j++) {
output ("sum$j += p$#mul$j * q$#mul;\n");
}
last if ($#mul+1) >= $UNROLL1;
}
$slot++;
}
# do next add
if ($LAT2 > 1) {
if ($#mul > $#add && $slot >= ($mul[$#add+1] + $LAT2)) {
push (@add,$slot);
for ($j=0; $j<$N1; $j++) {
output ("sum$j += m$#add$j;\n");
}
$slot++;
last if ($#add+1) >= $UNROLL1;
}
}
if ($slot == $startslot) {
# comment ("nop");
$slot++;
}
}
for ($j=0; $j<$N1; $j++) {
output ("a$j += $UNROLL1;\n");
}
output ("b += $UNROLL1;\n");
output ("n -= $UNROLL1;\n");
output ("}\n");
output ("n += $UNROLL1;\n");
output ("while (n > 0) {\n");
output ("q0 = *b;\n");
for ($j=0; $j<$N1; $j++) {
output ("sum$j += (*a$j) * q0;\n");
output ("a$j++;\n");
}
output ("b++;\n");
output ("n--;\n");
output ("}\n");
for ($j=0; $j<$N1; $j++) {
output ("outsum[$j] = sum$j;\n");
}
output ("}\n");
|
