From ecd167e8c45da5f158a86c8cd9a9de8353410450 Mon Sep 17 00:00:00 2001
From: Tomas Oberhuber <tomas.oberhuber@fjfi.cvut.cz>
Date: Thu, 4 Mar 2010 15:47:35 +0000
Subject: [PATCH] Fixing tnlLongVectorCUDA :: setNewSize.
---
src/Makefile.am | 6 ++-
src/core/tnl-cuda-kernels.h | 12 +++---
src/core/tnlCUDAKernelsTester.h | 2 +-
src/core/tnlFieldCUDA2D.h | 7 +++-
src/core/tnlLongVectorCUDA.h | 63 +++++++++++-------------------
src/core/tnlLongVectorCUDATester.h | 6 ++-
src/core/tnlObject.cpp | 10 +----
src/core/tnlObject.h | 7 ----
src/tnl-benchmarks.h | 14 +++----
9 files changed, 52 insertions(+), 75 deletions(-)
diff --git a/src/Makefile.am b/src/Makefile.am
index 004eeba807..31a09732ce 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -72,5 +72,7 @@ endif
-TESTS = tnl-unit-tests \
- tnl-benchmarks
\ No newline at end of file
+TESTS = tnl-unit-tests
+
+##tnl-unit-tests \
+## tnl-benchmarks
\ No newline at end of file
diff --git a/src/core/tnl-cuda-kernels.h b/src/core/tnl-cuda-kernels.h
index 9b5bbc2ed0..75b97d7507 100644
--- a/src/core/tnl-cuda-kernels.h
+++ b/src/core/tnl-cuda-kernels.h
@@ -403,7 +403,7 @@ bool tnlCUDAReduction( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
@@ -648,7 +648,7 @@ bool tnlCUDASimpleReduction5( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
@@ -874,7 +874,7 @@ bool tnlCUDASimpleReduction4( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
@@ -1085,7 +1085,7 @@ bool tnlCUDASimpleReduction3( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
@@ -1261,7 +1261,7 @@ bool tnlCUDASimpleReduction2( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
@@ -1420,7 +1420,7 @@ bool tnlCUDASimpleReduction1( const int size,
if( ! deviceAux )
{
int sizeAlloc = :: Max( 1, size / desBlockSize );
- if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
+ if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
return false;
deviceAux = deviceAuxVct. Data();
}
diff --git a/src/core/tnlCUDAKernelsTester.h b/src/core/tnlCUDAKernelsTester.h
index 10e871a698..b70f8598c0 100644
--- a/src/core/tnlCUDAKernelsTester.h
+++ b/src/core/tnlCUDAKernelsTester.h
@@ -106,7 +106,7 @@ template< class T > class tnlCUDAKernelsTester : public CppUnit :: TestCase
{
const int size = host_input. GetSize();
tnlLongVectorCUDA< T > device_input;
- if( ! device_input. SetNewSize( size ) )
+ if( ! device_input. setNewSize( size ) )
return false;
device_input. copyFrom( host_input );
diff --git a/src/core/tnlFieldCUDA2D.h b/src/core/tnlFieldCUDA2D.h
index c008a04987..16ce0b6780 100644
--- a/src/core/tnlFieldCUDA2D.h
+++ b/src/core/tnlFieldCUDA2D.h
@@ -70,7 +70,7 @@ template< class T > class tnlFieldCUDA2D : public tnlLongVectorCUDA< T >
{
x_size = new_x_size;
y_size = new_y_size;
- return tnlLongVectorCUDA< T > :: SetNewSize( x_size * y_size );
+ return tnlLongVectorCUDA< T > :: setNewSize( x_size * y_size );
};
bool SetNewDimensions( const tnlFieldCUDA2D< T >& f )
@@ -78,6 +78,11 @@ template< class T > class tnlFieldCUDA2D : public tnlLongVectorCUDA< T >
return SetNewDimensions( f. GetXSize(), f. GetYSize() );
};
+ bool SetNewDimensions( const tnlField2D< T >& f )
+ {
+ return SetNewDimensions( f. GetXSize(), f. GetYSize() );
+ };
+
void SetSharedData( T* _data, const int _x_size, const int _y_size )
{
tnlLongVectorCUDA< T > :: SetSharedData( _data, _x_size * _y_size );
diff --git a/src/core/tnlLongVectorCUDA.h b/src/core/tnlLongVectorCUDA.h
index bd1adcbeac..989c3264c1 100644
--- a/src/core/tnlLongVectorCUDA.h
+++ b/src/core/tnlLongVectorCUDA.h
@@ -44,6 +44,7 @@ using namespace std;
#include <core/tnlObject.h>
#include <core/param-types.h>
#include <core/tnlLongVector.h>
+#include <debug/tnlDebug.h>
template< class T > class tnlLongVectorCUDA : public tnlObject
{
@@ -51,17 +52,9 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
//! Constructor with given size
tnlLongVectorCUDA( const char* name = 0, int _size = 0 )
#ifdef HAVE_CUDA
- : tnlObject( name ), size( _size ), data( NULL ), shared_data( false )
+ : tnlObject( name ), size( 0 ), data( NULL ), shared_data( false )
{
- if( size && cudaMalloc( ( void** ) &data, ( size ) * sizeof( T ) ) != cudaSuccess )
- {
- cerr << "Unable to allocate new long vector '" << GetName() << "' with size "
- << size << " on CUDA device for "
- << GetName() << "." << endl;
- data = NULL;
- abort();
- }
- //data ++;
+ if( _size ) setNewSize( _size );
};
#else
{
@@ -72,16 +65,10 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
//! Constructor with another long vector as template
tnlLongVectorCUDA( const tnlLongVectorCUDA& v )
#ifdef HAVE_CUDA
- : tnlObject( v ), size( v. size ), data( NULL ), shared_data( false )
+ : tnlObject( v ), size( 0 ), data( NULL ), shared_data( false )
{
- if( size )
- if( cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
- {
- cerr << "Unable to allocate new long vector '" << GetName() << "' with size " << size << " on CUDA device." << endl;
- data = NULL;
- abort();
- }
- //data ++;
+ if( v. GetSize() )
+ setNewSize( v. size );
};
#else
{
@@ -92,15 +79,9 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
//! Constructor with another long vector as template
tnlLongVectorCUDA( const tnlLongVector< T >& v )
#ifdef HAVE_CUDA
- : tnlObject( v ), size( v. GetSize() ), data( NULL ), shared_data( false )
+ : tnlObject( v ), size( 0 ), data( NULL ), shared_data( false )
{
- if( size && cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
- {
- cerr << "Unable to allocate new long vector with size " << size << " on CUDA device." << endl;
- data = NULL;
- abort();
- }
- //data ++;
+ if( v. GetSize() ) setNewSize( v. GetSize() );
};
#else
{
@@ -116,19 +97,21 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
}
- bool SetNewSize( int _size )
+ bool setNewSize( int _size )
#ifdef HAVE_CUDA
{
- if( debug )
- cout << "Setting new size to " << _size << " for " << GetName() << endl;
+ dbgFunctionName( "tnlLongVectorCUDA", "setNewSize" );
+ dbgCout( "Setting new size to " << _size << " for " << GetName() );
if( size == _size ) return true;
if( data && ! shared_data )
{
+ dbgCout( "Freeing allocated memory on CUDA device of " << GetName() );
cudaFree( data );
data = NULL;
}
size = _size;
shared_data = false;
+ cerr << "###########" << endl;
if( cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
{
cerr << "Unable to allocate new long vector with size "
@@ -147,9 +130,14 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
};
#endif
- bool SetNewSize( const tnlLongVectorCUDA< T >& v )
+ bool setNewSize( const tnlLongVectorCUDA< T >& v )
{
- return SetNewSize( v. GetSize() );
+ return setNewSize( v. GetSize() );
+ };
+
+ bool setNewSize( const tnlLongVector< T >& v )
+ {
+ return setNewSize( v. GetSize() );
};
void SetSharedData( T* _data, const int _size )
@@ -239,11 +227,11 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
virtual
~tnlLongVectorCUDA()
{
+ dbgFunctionName( "tnlLongVectorCUDA", "~tnlLOngVectorCUDA" );
#ifdef HAVE_CUDA
if( data && ! shared_data )
{
- if( debug )
- cout << "Freeing allocated memory of " << GetName() << " on CUDA device." << endl;
+ dbgCout( "Freeing allocated memory of " << GetName() << " on CUDA device." );
if( cudaFree( data ) != cudaSuccess )
{
cerr << "Unable to free alocated memory on CUDA device of " << GetName() << "." << endl;
@@ -254,11 +242,6 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
#endif
};
- static void setDebug( bool _debug )
- {
- debug = _debug;
- };
-
private:
int size;
@@ -290,6 +273,4 @@ template< class T > bool tnlLongVector< T > :: copyFrom( const tnlLongVectorCUDA
}
-template< class T > bool tnlLongVectorCUDA< T > :: debug;
-
#endif /* TNLLONGVECTORCUDA_H_ */
diff --git a/src/core/tnlLongVectorCUDATester.h b/src/core/tnlLongVectorCUDATester.h
index 0f3e100d0b..885a53331e 100644
--- a/src/core/tnlLongVectorCUDATester.h
+++ b/src/core/tnlLongVectorCUDATester.h
@@ -25,6 +25,7 @@
#include <cppunit/TestResult.h>
#include <cppunit/TestCaller.h>
#include <cppunit/TestCase.h>
+#include <cppunit/Message.h>
#include <core/tnlLongVectorCUDA.h>
#include <core/tnlLongVector.h>
@@ -164,13 +165,14 @@ template< class T > class tnlLongVectorCUDATester : public CppUnit :: TestCase
tnlLongVectorCUDA< T > cuda_vector_1;
CPPUNIT_ASSERT( !cuda_vector_1 );
- cuda_vector_1. SetNewSize( 100 );
+ cuda_vector_1. setNewSize( 100 );
+
CPPUNIT_ASSERT( cuda_vector_1 );
CPPUNIT_ASSERT( cuda_vector_1. GetSize() == 100 );
tnlLongVectorCUDA< T > cuda_vector_2;
CPPUNIT_ASSERT( !cuda_vector_2 );
- cuda_vector_2. SetNewSize( cuda_vector_1 );
+ cuda_vector_2. setNewSize( cuda_vector_1 );
CPPUNIT_ASSERT( cuda_vector_2. GetSize() == 100 );
tnlLongVectorCUDA< T > cuda_vector_3( "cuda-vector", 100 );
diff --git a/src/core/tnlObject.cpp b/src/core/tnlObject.cpp
index ca49633088..af114326ea 100644
--- a/src/core/tnlObject.cpp
+++ b/src/core/tnlObject.cpp
@@ -25,7 +25,6 @@
const char magic_number[] = "SIM33";
-bool tnlObject :: debug = false;
//--------------------------------------------------------------------------
tnlObject :: tnlObject( )
{
@@ -33,9 +32,9 @@ tnlObject :: tnlObject( )
//--------------------------------------------------------------------------
tnlObject :: tnlObject( const char* name )
{
+ dbgFunctionName( "tnlObject", "tnlObject" );
+ dbgCout( "Initiating object " << GetName() );
SetName( name );
- if( debug )
- cout << "Initiating object " << GetName() << endl;
}
//--------------------------------------------------------------------------
tnlObject :: tnlObject( const tnlObject& object )
@@ -85,11 +84,6 @@ bool tnlObject :: Load( istream& file )
return true;
}
//--------------------------------------------------------------------------
-void tnlObject :: setDebug( bool _debug )
-{
- debug = _debug;
-}
-//--------------------------------------------------------------------------
bool GetObjectType( istream& file, tnlString& type )
{
dbgFunctionName( "", "GetObjectType" );
diff --git a/src/core/tnlObject.h b/src/core/tnlObject.h
index 6c91c5e6be..3378f613d3 100644
--- a/src/core/tnlObject.h
+++ b/src/core/tnlObject.h
@@ -59,9 +59,6 @@ class tnlObject
//! Method for restoring the object from a file
virtual bool Load( istream& file );
- //! Debug moder setter
- static void setDebug( bool );
-
//! Destructor
virtual ~tnlObject(){};
@@ -69,10 +66,6 @@ class tnlObject
//! Object name
tnlString name;
-
- //! Turns on debuging messages
- static bool debug;
-
};
bool GetObjectType( istream& file, tnlString& type );
diff --git a/src/tnl-benchmarks.h b/src/tnl-benchmarks.h
index 07213972ce..b9b4a467a7 100644
--- a/src/tnl-benchmarks.h
+++ b/src/tnl-benchmarks.h
@@ -28,10 +28,10 @@ bool transferBenchmark( const int size,
double& device_to_device_band_width )
{
- tnlLongVector< T > host_vector( size );
- tnlLongVector< T > host_vector2( size );
- tnlLongVectorCUDA< T > device_vector( size );
- tnlLongVectorCUDA< T > device_vector2( size );
+ tnlLongVector< T > host_vector( "transferBenchmark:host-vector", size );
+ tnlLongVector< T > host_vector2( "transferBenchmark:host-vector-2", size );
+ tnlLongVectorCUDA< T > device_vector( "transferBenchmark:device-vector", size );
+ tnlLongVectorCUDA< T > device_vector2( "transferBenchmark:device-vector-2", size );
for( int i = 0; i < size; i ++ )
host_vector[ i ] = i + 1;
@@ -122,9 +122,9 @@ template< class T >
void reductionBenchmark( const int size,
const int algorithm )
{
- tnlLongVector< T > host_vector( size );
- tnlLongVectorCUDA< T > device_vector( size );
- tnlLongVectorCUDA< T > device_aux( size / 2 );
+ tnlLongVector< T > host_vector( "reductionBenchmark:host-vector", size );
+ tnlLongVectorCUDA< T > device_vector( "reductionBenchmark:device-vector", size );
+ tnlLongVectorCUDA< T > device_aux( "reductionBenchmark:device-aux", size / 2 );
for( int i = 0; i < size; i ++ )
host_vector[ i ] = i + 1;
--
GitLab