forked from grame-cncm/faust
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbench.d
259 lines (209 loc) · 7.48 KB
/
bench.d
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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
/************************************************************************
IMPORTANT NOTE : this file contains two clearly delimited sections :
the ARCHITECTURE section (in two parts) and the USER section. Each section
is governed by its own copyright and license. Please check individually
each section for license and copyright information.
*************************************************************************/
/*******************BEGIN ARCHITECTURE SECTION (part 1/2)****************/
/************************************************************************
FAUST Architecture File
Copyright (C) 2003-2019 GRAME, Centre National de Creation Musicale
---------------------------------------------------------------------
This Architecture section is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 3 of
the License, or (at your option) any later version.
This program 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
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; If not, see <http://www.gnu.org/licenses/>.
EXCEPTION : As a special exception, you may create a larger work
that contains this FAUST architecture section and distribute
that work under terms of your choice, so long as this FAUST
architecture section is not modified.
************************************************************************
************************************************************************/
import std.datetime;
import core.cpuid;
import core.stdc.stdio;
import core.stdc.stdlib;
version (Posix)
{
import core.sys.posix.pwd;
import core.sys.posix.unistd;
import core.sys.posix.pthread;
}
import dplug.core.nogc;
@trusted:
@nogc:
nothrow:
float[] gBuffer = null; // a buffer of NV*VSize samples
__gshared ulong COUNT = 2000; // number of measures
__gshared ulong NV = 4096; // number of vectors in BIG buffer (should exceed cache)
__gshared ulong ITER = 10; // number of iterations per measure
__gshared ulong VSIZE = 4096; // size of a vector in samples
__gshared ulong IDX = 0; // current vector number (0 <= VIdx < NV)
bool setRealtimePriority()
{
version (Posix)
{
passwd* pw;
int err;
uid_t uid;
int policy;
sched_param param;
uid = getuid();
pw = getpwnam("root");
setuid(pw.pw_uid);
pthread_getschedparam(pthread_self(), &policy, ¶m);
policy = SCHED_RR;
param.sched_priority = 50;
err = pthread_setschedparam(pthread_self(), policy, ¶m);
setuid(uid);
return (err != -1);
}
else
{
return false;
}
}
double toSeconds(Duration dur)
{
return dur.total!"nsecs" / 1e9;
}
version (D_InlineAsm_X86) version = InlineAsm_X86_Any;
else version (D_InlineAsm_X86_64) version = InlineAsm_X86_Any;
void avoidDenormals()
{
version (InlineAsm_X86_Any)
{
uint mxcsr;
asm nothrow @nogc { stmxcsr mxcsr; }
if (sse2)
mxcsr |= 0x8040;
else
mxcsr |= 0x8000;
asm nothrow @nogc { ldmxcsr mxcsr; }
}
}
/******************************************************************************
*******************************************************************************
VECTOR INTRINSICS
*******************************************************************************
*******************************************************************************/
<<includeIntrinsic>>
/********************END ARCHITECTURE SECTION (part 1/2)****************/
class dsp {}
alias FAUSTFLOAT = float;
// dummy implementations
struct Meta { @nogc void opDispatch(string s, A...)(A) nothrow {} }
struct UI { @nogc void opDispatch(string s, A...)(A) nothrow {} }
/**************************BEGIN USER SECTION **************************/
<<includeclass>>
/***************************END USER SECTION ***************************/
/*******************BEGIN ARCHITECTURE SECTION (part 2/2)***************/
__gshared mydsp DSP;
shared static this() { DSP = mallocNew!mydsp; }
shared static ~this() { DSP.destroyFree; }
/**
* Bench by calling COUNT times the compute() method for
* the computation of vsize samples
*/
enum MEGABYTE = 1048576.0;
void statistic(string name, MonoTime[] timing)
{
double lo, hi, tot;
double mega = cast(double)(VSIZE*ITER)/MEGABYTE; // mega samples
// skip first 10 values to avoid cache bias ???
lo = hi = tot = mega/(timing[11] - timing[10]).toSeconds;
for (int i = 11; i < COUNT; i++) {
double delta = mega/(timing[i] - timing[i-1]).toSeconds;
if (delta < lo) {
lo = delta;
} else if (delta > hi) {
hi = delta;
}
tot += delta;
}
printf("\t" ~ "%f" ~
"\t" ~ "%f" ~ "\t" ~ "MB/s inputs" ~
"\t" ~ "%f" ~ "\t" ~ "MB/s outputs" ~
"\t" ~ "%f" ~
"\t" ~ "%f" ~ "\n",
hi, hi*4*DSP.getNumInputs(), hi*4*DSP.getNumOutputs(), tot/(COUNT-11), lo);
}
void allocBuffer()
{
ulong BSIZE = NV * VSIZE;
gBuffer = mallocSlice!float(BSIZE);
int R0_0 = 0;
for (int j = 0; j < BSIZE; j++) {
int R0temp0 = (12345 + (1103515245 * R0_0));
gBuffer[j] = 4.656613e-10f*R0temp0;
R0_0 = R0temp0;
}
}
void deallocBuffer()
{
freeSlice(gBuffer);
gBuffer = null;
}
float* nextVect()
{
IDX = (1+IDX)%NV;
return &gBuffer[IDX*VSIZE];
}
void bench(string name)
{
int numInChan = DSP.getNumInputs();
int numOutChan = DSP.getNumOutputs();
assert(numInChan < 256);
assert(numOutChan < 256);
float*[256] inChannel;
float*[256] outChannel;
// allocate input buffers (initialized with white noise)
allocBuffer();
scope(exit) deallocBuffer();
// allocate output channels (not initialized)
for (int i = 0; i < numOutChan; i++) outChannel[i] = mallocSlice!float(VSIZE).ptr;
scope(exit) for (int i = 0; i < numOutChan; i++) freeSlice(outChannel[i][0..VSIZE]);
// init the dsp with a resoneable sampling rate)
DSP.initialize(48000);
MonoTime[] timing = mallocSlice!MonoTime(COUNT);
scope(exit) freeSlice(timing);
for (int i = 0; i < COUNT; i++) {
timing[i] = MonoTime.currTime;
for (int k = 0; k < ITER; k++) {
// allocate new input buffers to avoid L2 cache
for (int c = 0; c < numInChan; c++) { inChannel[c] = nextVect(); }
DSP.compute(cast(uint)VSIZE,inChannel,outChannel);
}
}
statistic(name, timing);
}
//-------------------------------------------------------------------------
// MAIN
//-------------------------------------------------------------------------
// lopt : Scan Command Line long int Arguments
long lopt(string[] args, string longname, string shortname, long def)
{
for (ulong i = 2, n = args.length; i < n; i++) {
if (args[i-1] == shortname || args[i-1] == longname) {
return atoi(assumeZeroTerminated(args[i]));
}
}
return def;
}
void main(string[] args)
{
avoidDenormals();
VSIZE = lopt(args, "--vector-size", "-vec", 4096);
NV = lopt(args, "--num-vector", "-n", 20000);
COUNT = lopt(args, "--count", "-c", 1000);
ITER = lopt(args, "--iteration", "-i", 10);
//setRealtimePriority();
bench(args[0]);
}
/********************END ARCHITECTURE SECTION (part 2/2)****************/