zara
/
cudaSamples
mirror of https://github.com/ArchaeaSoftware/cudahandbook.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
Browse Source

Port reduction sample code to new error handling

divergence
Nicholas Wilt 9 years ago
parent
4bfea533a8
commit
096e5506f0
2 changed files with 32 additions and 32 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 32
      reduction/reduction.cu
  2. 32
      reduction/reductionTemplated.cu

32
reduction/reduction.cu
Unescape View File

@ -81,13 +81,13 @@ TimedReduction(
cudaEvent_t stop = 0; cudaEvent_t stop = 0;
cudaError_t status; cudaError_t status;
CUDART_CHECK( cudaMalloc( &deviceAnswer, sizeof(int) ) ); cuda(Malloc( &deviceAnswer, sizeof(int) ) );
CUDART_CHECK( cudaMalloc( &partialSums, cBlocks*sizeof(int) ) ); cuda(Malloc( &partialSums, cBlocks*sizeof(int) ) );
CUDART_CHECK( cudaEventCreate( &start ) ); cuda(EventCreate( &start ) );
CUDART_CHECK( cudaEventCreate( &stop ) ); cuda(EventCreate( &stop ) );
CUDART_CHECK( cudaThreadSynchronize() ); cuda(ThreadSynchronize() );
CUDART_CHECK( cudaEventRecord( start, 0 ) ); cuda(EventRecord( start, 0 ) );
hostReduction( hostReduction(
deviceAnswer, deviceAnswer,
partialSums, partialSums,
@ -95,8 +95,8 @@ TimedReduction(
N, N,
cBlocks, cBlocks,
cThreads ); cThreads );
CUDART_CHECK( cudaEventRecord( stop, 0 ) ); cuda(EventRecord( stop, 0 ) );
CUDART_CHECK( cudaMemcpy( cuda(Memcpy(
answer, answer,
deviceAnswer, deviceAnswer,
sizeof(int), sizeof(int),
@ -159,13 +159,13 @@ usPerInvocation( int cIterations, size_t N,
double ret = 0.0f; double ret = 0.0f;
chTimerTimestamp start, stop; chTimerTimestamp start, stop;
CUDART_CHECK( cudaMalloc( &smallArray, N*sizeof(int) ) ); cuda(Malloc( &smallArray, N*sizeof(int) ) );
CUDART_CHECK( cudaMalloc( &partialSums, 1*sizeof(int) ) ); cuda(Malloc( &partialSums, 1*sizeof(int) ) );
chTimerGetTime( &start ); chTimerGetTime( &start );
for ( int i = 0; i < cIterations; i++ ) { for ( int i = 0; i < cIterations; i++ ) {
pfnReduction( partialSums, partialSums, smallArray, N, 1, 256 ); pfnReduction( partialSums, partialSums, smallArray, N, 1, 256 );
} }
CUDART_CHECK( cudaThreadSynchronize() ); cuda(ThreadSynchronize() );
chTimerGetTime( &stop ); chTimerGetTime( &stop );
ret = chTimerElapsedTime( &start, &stop ); ret = chTimerElapsedTime( &start, &stop );
ret = (ret / (double) cIterations) * 1e6; ret = (ret / (double) cIterations) * 1e6;
@ -209,10 +209,10 @@ main( int argc, char *argv[] )
hostData = (int *) malloc( cInts*sizeof(int) ); hostData = (int *) malloc( cInts*sizeof(int) );
if ( ! hostData ) if ( ! hostData )
goto Error; goto Error;
CUDART_CHECK( cudaSetDevice( device ) ); cuda(SetDevice( device ) );
CUDART_CHECK( cudaSetDeviceFlags( cudaDeviceMapHost ) ); cuda(SetDeviceFlags( cudaDeviceMapHost ) );
CUDART_CHECK( cudaMalloc( &deviceData, cInts*sizeof(int) ) ); cuda(Malloc( &deviceData, cInts*sizeof(int) ) );
CUDART_CHECK( cudaGetDeviceProperties( &props, 0 ) ); cuda(GetDeviceProperties( &props, 0 ) );
sum = 0; sum = 0;
for ( size_t i = 0; i < cInts; i++ ) { for ( size_t i = 0; i < cInts; i++ ) {
@ -220,7 +220,7 @@ main( int argc, char *argv[] )
sum += value; sum += value;
hostData[i] = value; hostData[i] = value;
} }
CUDART_CHECK( cudaMemcpy( deviceData, hostData, cInts*sizeof(int), cuda(Memcpy( deviceData, hostData, cInts*sizeof(int),
cudaMemcpyHostToDevice ) ); cudaMemcpyHostToDevice ) );
{ {

32
reduction/reductionTemplated.cu
Unescape View File

@ -125,18 +125,18 @@ TimedReduction(
cudaEvent_t stop = 0; cudaEvent_t stop = 0;
cudaError_t status; cudaError_t status;
CUDART_CHECK( cudaMalloc( &deviceAnswer, sizeof(ReductionType) ) ); cuda(Malloc( &deviceAnswer, sizeof(ReductionType) ) );
CUDART_CHECK( cudaMalloc( &partialSums, cBlocks*sizeof(ReductionType) ) ); cuda(Malloc( &partialSums, cBlocks*sizeof(ReductionType) ) );
CUDART_CHECK( cudaEventCreate( &start ) ); cuda(EventCreate( &start ) );
CUDART_CHECK( cudaEventCreate( &stop ) ); cuda(EventCreate( &stop ) );
CUDART_CHECK( cudaThreadSynchronize() ); cuda(ThreadSynchronize() );
CUDART_CHECK( cudaEventRecord( start, 0 ) ); cuda(EventRecord( start, 0 ) );
hostReduction( deviceAnswer, partialSums, deviceIn, N, cBlocks, cThreads ); hostReduction( deviceAnswer, partialSums, deviceIn, N, cBlocks, cThreads );
CUDART_CHECK( cudaEventRecord( stop, 0 ) ); cuda(EventRecord( stop, 0 ) );
CUDART_CHECK( cudaMemcpy( answer, deviceAnswer, sizeof(T), cudaMemcpyDeviceToHost ) ); cuda(Memcpy( answer, deviceAnswer, sizeof(T), cudaMemcpyDeviceToHost ) );
CUDART_CHECK( cudaEventElapsedTime( &ms, start, stop ) ) cuda(EventElapsedTime( &ms, start, stop ) )
ret = ms * 1000.0f; ret = ms * 1000.0f;
// fall through to free resources before returning // fall through to free resources before returning
@ -203,13 +203,13 @@ usPerInvocation( int cIterations, size_t N,
double ret = 0.0f; double ret = 0.0f;
chTimerTimestamp start, stop; chTimerTimestamp start, stop;
CUDART_CHECK( cudaMalloc( &smallArray, N*sizeof(T) ) ); cuda(Malloc( &smallArray, N*sizeof(T) ) );
CUDART_CHECK( cudaMalloc( &partialSums, 1*sizeof(ReductionType) ) ); cuda(Malloc( &partialSums, 1*sizeof(ReductionType) ) );
chTimerGetTime( &start ); chTimerGetTime( &start );
for ( int i = 0; i < cIterations; i++ ) { for ( int i = 0; i < cIterations; i++ ) {
pfnReduction( partialSums, partialSums, smallArray, N, 1, 256 ); pfnReduction( partialSums, partialSums, smallArray, N, 1, 256 );
} }
CUDART_CHECK( cudaThreadSynchronize() ); cuda(ThreadSynchronize() );
chTimerGetTime( &stop ); chTimerGetTime( &stop );
ret = chTimerElapsedTime( &start, &stop ); ret = chTimerElapsedTime( &start, &stop );
ret = (ret / (double) cIterations) * 1e6; ret = (ret / (double) cIterations) * 1e6;
@ -246,10 +246,10 @@ ShmooReport( size_t N, bool bFloat )
hostData = (T *) malloc( N*sizeof(T) ); hostData = (T *) malloc( N*sizeof(T) );
if ( ! hostData ) if ( ! hostData )
goto Error; goto Error;
CUDART_CHECK( cudaSetDeviceFlags( cudaDeviceMapHost ) ); cuda(SetDeviceFlags( cudaDeviceMapHost ) );
CUDART_CHECK( cudaMalloc( &deviceData, N*sizeof(T) ) ); cuda(Malloc( &deviceData, N*sizeof(T) ) );
CUDART_CHECK( cudaGetDeviceProperties( &props, 0 ) ); cuda(GetDeviceProperties( &props, 0 ) );
CUDART_CHECK( cudaMemcpy( deviceData, hostData, N*sizeof(T), cudaMemcpyHostToDevice ) ); cuda(Memcpy( deviceData, hostData, N*sizeof(T), cudaMemcpyHostToDevice ) );
forkPrint( g_fileShmoo, "N\tThreads(1)\tus(1)\t" forkPrint( g_fileShmoo, "N\tThreads(1)\tus(1)\t"
"Threads(2)\tus(2)\t" "Threads(2)\tus(2)\t"

Write Preview
Loading…
Cancel
Save