From d2514e28c2e2c7ddc00bfbdf0a6df9468c21631e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jakub=20Klinkovsk=C3=BD?= <klinkjak@fjfi.cvut.cz>
Date: Wed, 9 Dec 2015 23:27:46 +0100
Subject: [PATCH] Fixed order of fields printed to stdout in the CUDA benchmark
---
tests/benchmarks/tnl-cuda-benchmarks.h | 28 +++++++++++++++-----------
1 file changed, 16 insertions(+), 12 deletions(-)
diff --git a/tests/benchmarks/tnl-cuda-benchmarks.h b/tests/benchmarks/tnl-cuda-benchmarks.h
index 38c108b5d9..919b4ebe82 100644
--- a/tests/benchmarks/tnl-cuda-benchmarks.h
+++ b/tests/benchmarks/tnl-cuda-benchmarks.h
@@ -100,25 +100,28 @@ int main( int argc, char* argv[] )
cout << bandwidth << " GB/sec." << endl;
*/
- Real resultHost, resultDevice;
+ Real resultHost, resultDevice, timeHost, timeDevice;
cout << "Benchmarking scalar product on CPU: ";
timer.reset();
timer.start();
for( int i = 0; i < loops; i++ )
resultHost = hostVector.scalarProduct( hostVector2 );
timer.stop();
+ timeHost = timer.getTime();
bandwidth = 2 * datasetSize / timer.getTime();
- cout << bandwidth << " GB/sec." << endl;
+ cout << "bandwidth: " << bandwidth << " GB/sec, time: " << timer.getTime() << " sec." << endl;
- cout << "Benchmarking scalar product on GPU: " << endl;
+ cout << "Benchmarking scalar product on GPU: ";
timer.reset();
timer.start();
for( int i = 0; i < loops; i++ )
- resultDevice = deviceVector.scalarProduct( deviceVector );
- cout << "Time: " << timer.getTime() << endl;
+ resultDevice = deviceVector.scalarProduct( deviceVector2 );
timer.stop();
+ timeDevice = timer.getTime();
bandwidth = 2 * datasetSize / timer.getTime();
- cout << "Time: " << timer.getTime() << " bandwidth: " << bandwidth << " GB/sec." << endl;
+ cout << "bandwidth: " << bandwidth << " GB/sec, time: " << timer.getTime() << " sec." << endl;
+ cout << "CPU/GPU speedup: " << timeHost / timeDevice << endl;
+
if( resultHost != resultDevice )
{
cerr << "Error. " << resultHost << " != " << resultDevice << endl;
@@ -144,23 +147,24 @@ int main( int argc, char* argv[] )
#endif
#endif
- cout << "Benchmarking prefix-sum on CPU ..." << endl;
+ cout << "Benchmarking prefix-sum on CPU: ";
timer.reset();
timer.start();
hostVector.computePrefixSum();
- cout << "Time: " << timer.getTime() << endl;
timer.stop();
+ timeHost = timer.getTime();
bandwidth = 2 * datasetSize / loops / timer.getTime();
- cout << "Time: " << timer.getTime() << " bandwidth: " << bandwidth << " GB/sec." << endl;
+ cout << "bandwidth: " << bandwidth << " GB/sec, time: " << timer.getTime() << " sec." << endl;
- cout << "Benchmarking prefix-sum on GPU ..." << endl;
+ cout << "Benchmarking prefix-sum on GPU: ";
timer.reset();
timer.start();
deviceVector.computePrefixSum();
- cout << "Time: " << timer.getTime() << endl;
timer.stop();
+ timeDevice = timer.getTime();
bandwidth = 2 * datasetSize / loops / timer.getTime();
- cout << "Time: " << timer.getTime() << " bandwidth: " << bandwidth << " GB/sec." << endl;
+ cout << "bandwidth: " << bandwidth << " GB/sec, time: " << timer.getTime() << " sec." << endl;
+ cout << "CPU/GPU speedup: " << timeHost / timeDevice << endl;
for( int i = 0; i < size; i++ )
if( hostVector.getElement( i ) != deviceVector.getElement( i ) )
--
GitLab