GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-04-21 09:09:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated the benchmarking code to print the number of flops processed instead of the number of bytes.

Browse Source
This commit is contained in:
Benoit Steiner 2016-01-28 16:51:40 -08:00
parent 8217281ae4
commit a68864b6bc
3 changed files with 9 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
bench/tensors/benchmark.h
View File

@ -41,10 +41,9 @@ class Benchmark {
void RunWithArg(int arg);
};
} // namespace testing
void SetBenchmarkBytesProcessed(int64_t);
void SetBenchmarkFlopsProcessed(int64_t);
void StopBenchmarkTiming();
void StartBenchmarkTiming();
#define BENCHMARK(f) \
static ::testing::Benchmark* _benchmark_##f __attribute__((unused)) = \
(new ::testing::Benchmark(#f, f))

14
bench/tensors/benchmark_main.cc
View File

@ -23,7 +23,7 @@
#include <time.h>
#include <map>
static int64_t g_bytes_processed;
static int64_t g_flops_processed;
static int64_t g_benchmark_total_time_ns;
static int64_t g_benchmark_start_time_ns;
typedef std::map<std::string, ::testing::Benchmark*> BenchmarkMap;
@ -124,7 +124,7 @@ void Benchmark::Run() {
}
}
void Benchmark::RunRepeatedlyWithArg(int iterations, int arg) {
g_bytes_processed = 0;
g_flops_processed = 0;
g_benchmark_total_time_ns = 0;
g_benchmark_start_time_ns = NanoTime();
if (fn_ != NULL) {
@ -153,10 +153,10 @@ void Benchmark::RunWithArg(int arg) {
}
char throughput[100];
throughput[0] = '\0';
if (g_benchmark_total_time_ns > 0 && g_bytes_processed > 0) {
double mib_processed = static_cast<double>(g_bytes_processed)/1e6;
if (g_benchmark_total_time_ns > 0 && g_flops_processed > 0) {
double mflops_processed = static_cast<double>(g_flops_processed)/1e6;
double seconds = static_cast<double>(g_benchmark_total_time_ns)/1e9;
snprintf(throughput, sizeof(throughput), " %8.2f MiB/s", mib_processed/seconds);
snprintf(throughput, sizeof(throughput), " %8.2f MFlops/s", mflops_processed/seconds);
}
char full_name[100];
if (fn_range_ != NULL) {
@ -175,8 +175,8 @@ void Benchmark::RunWithArg(int arg) {
fflush(stdout);
}
} // namespace testing
void SetBenchmarkBytesProcessed(int64_t x) {
g_bytes_processed = x;
void SetBenchmarkFlopsProcessed(int64_t x) {
g_flops_processed = x;
}
void StopBenchmarkTiming() {
if (g_benchmark_start_time_ns != 0) {

2
bench/tensors/tensor_benchmarks.h
View File

@ -367,7 +367,7 @@ template <typename Device> class BenchmarkSuite {
}
#endif
StopBenchmarkTiming();
SetBenchmarkBytesProcessed(num_items);
SetBenchmarkFlopsProcessed(num_items);
}