3 * Copyright 2007,2008,2009,2010 Free Software Foundation, Inc.
5 * This file is part of GNU Radio
7 * GNU Radio is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 3, or (at your option)
12 * GNU Radio is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License along
18 * with this program; if not, write to the Free Software Foundation, Inc.,
19 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
22 #if defined(HAVE_CONFIG_H)
25 #include <gcell/gc_job_manager.h>
26 #include <boost/date_time/posix_time/posix_time_types.hpp>
30 #include <boost/scoped_array.hpp>
33 // handle to embedded SPU executable that contains benchmark routines
34 // (The name of the variable (benchmark_procs) is the name of the spu executable.)
35 extern spe_program_handle_t benchmark_procs;
37 static gc_proc_id_t gcp_benchmark_udelay = GCP_UNKNOWN_PROC;
39 #define BENCHMARK_PUT 0x1
40 #define BENCHMARK_GET 0x2
41 #define BENCHMARK_GET_PUT (BENCHMARK_PUT|BENCHMARK_GET)
46 power_of_2_p(unsigned long x)
48 int nbits = sizeof(x) * 8;
49 for (int i = 0; i < nbits; i++)
58 init_jd(gc_job_desc *jd, unsigned int usecs,
59 unsigned char *getbuf, unsigned char *putbuf, size_t buflen,
62 jd->proc_id = gcp_benchmark_udelay;
64 jd->input.arg[0].u32 = usecs;
67 switch(getput_mask & BENCHMARK_GET_PUT){
71 jd->eaa.arg[0].direction = GCJD_DMA_GET;
72 jd->eaa.arg[0].ea_addr = ptr_to_ea(getbuf);
73 jd->eaa.arg[0].get_size = buflen;
78 jd->eaa.arg[0].direction = GCJD_DMA_PUT;
79 jd->eaa.arg[0].ea_addr = ptr_to_ea(putbuf);
80 jd->eaa.arg[0].put_size = buflen;
83 case BENCHMARK_GET_PUT:
85 jd->eaa.arg[0].direction = GCJD_DMA_GET;
86 jd->eaa.arg[0].ea_addr = ptr_to_ea(getbuf);
87 jd->eaa.arg[0].get_size = buflen;
88 jd->eaa.arg[1].direction = GCJD_DMA_PUT;
89 jd->eaa.arg[1].ea_addr = ptr_to_ea(putbuf);
90 jd->eaa.arg[1].put_size = buflen;
96 run_test(unsigned int nspes, unsigned int usecs, unsigned int dma_size,
97 int getput_mask, int njobs_at_once)
99 using namespace boost::posix_time;
101 int NJDS = njobs_at_once;
102 gc_job_desc *all_jds[NJDS];
105 static const unsigned int BUFSIZE = (32 << 10) * NJDS;
106 unsigned char *getbuf = new unsigned char[BUFSIZE];
107 boost::scoped_array<unsigned char> _getbuf(getbuf);
108 unsigned char *putbuf = new unsigned char[BUFSIZE];
109 boost::scoped_array<unsigned char> _putbuf(putbuf);
112 // touch all pages to force allocation now
113 for (unsigned int i = 0; i < BUFSIZE; i += 4096){
119 opts.program_handle = gc_program_handle_from_address(&benchmark_procs);
121 //opts.enable_logging = true;
122 //opts.log2_nlog_entries = 13;
123 gc_job_manager_sptr mgr = gc_make_job_manager(&opts);
125 if ((gcp_benchmark_udelay = mgr->lookup_proc("benchmark_udelay")) == GCP_UNKNOWN_PROC){
126 fprintf(stderr, "lookup_proc: failed to find \"benchmark_udelay\"\n");
130 // allocate and init all job descriptors
131 for (int i = 0; i < NJDS; i++){
132 if (gbi + dma_size > BUFSIZE)
135 all_jds[i] = mgr->alloc_job_desc();
136 if (all_jds[i] == 0){
137 fprintf(stderr, "alloc_job_desc() returned 0\n");
140 init_jd(all_jds[i], usecs, &getbuf[gbi], &putbuf[gbi], dma_size, getput_mask);
145 ptime t_start(microsec_clock::universal_time());
147 for (int iter = 0; iter < niter; iter++){
150 for (int i = 0; i < NJDS; i++){
151 if (!mgr->submit_job(all_jds[i])){
152 printf("submit_job(jds[%d]) failed, status = %d\n",
153 i, all_jds[i]->status);
157 int n = mgr->wait_jobs(NJDS, all_jds, done, GC_WAIT_ALL);
159 fprintf(stderr, "mgr->wait_jobs failed\n");
163 fprintf(stderr, "mgr->wait_jobs returned short count. Expected %d, got %d\n",
169 ptime t_stop(microsec_clock::universal_time());
170 double delta = (t_stop - t_start).total_microseconds() * 1e-6;
171 printf("nspes: %2d udelay: %4d elapsed_time: %7.3f dma_size: %5d dma_throughput: %7.3e round_trip: %gus\n",
172 mgr->nspes(), usecs, delta, dma_size,
173 (double) NJDS * niter * dma_size / delta * (getput_mask == BENCHMARK_GET_PUT ? 2.0 : 1.0),
174 delta / niter * 1e6);
180 fprintf(stderr, "usage: benchmark_dma [-p] [-g] [-n <nspes>] [-u <udelay>] [-s <dma_size>] [-N <njobs_at_a_time>]\n");
181 fprintf(stderr, " you must specify one or both of -p (put) and -g (get)\n");
186 main(int argc, char **argv)
188 unsigned int nspes = 0;
189 unsigned int usecs = 0;
190 unsigned int dma_size = 32 << 10;
191 int njobs_at_once = -1;
195 while ((ch = getopt(argc, argv, "n:u:s:pgN:")) != EOF){
198 nspes = strtol(optarg, 0, 0);
202 usecs = strtol(optarg, 0, 0);
206 njobs_at_once = strtol(optarg, 0, 0);
210 dma_size = strtol(optarg, 0, 0);
212 fprintf(stderr, "-s <dma_size> must be > 0\n");
218 getput_mask |= BENCHMARK_PUT;
222 getput_mask |= BENCHMARK_GET;
232 if (njobs_at_once < 0)
233 njobs_at_once = nspes;
235 if (getput_mask == 0){
240 run_test(nspes, usecs, dma_size, getput_mask, njobs_at_once);