dksbase use Algoruthms base class for fft, colimator physics and greens function

CMakeLists.txt

@@ -39,22 +39,27 @@ IF (Boost_FOUND)
 ENDIF (Boost_FOUND)
 
 #find clFFT
-SET (clFFT_USE_STATIC_LIBS OFF)
+OPTION (ENABLE_AMD "Enable AMD libraries" OFF)
-FIND_PACKAGE(clFFT REQUIRED HINTS $ENV{CLFFT_PREFIX} $ENV{CLFFT_DIR} $ENV{CLFFT})
+IF (ENABLE_AMD)
-MESSAGE (STATUS "Found clFFT library: ${CLFFT_LIBRARIES}")
+  SET (clFFT_USE_STATIC_LIBS OFF)
-MESSAGE (STATUS "Found clFFT include dir: ${CLFFT_INCLUDE_DIRS}")
+  FIND_PACKAGE(clFFT REQUIRED HINTS $ENV{CLFFT_PREFIX} $ENV{CLFFT_DIR} $ENV{CLFFT})
-INCLUDE_DIRECTORIES (${CLFFT_INCLUDE_DIRS})
+  MESSAGE (STATUS "Found clFFT library: ${CLFFT_LIBRARIES}")
-LINK_DIRECTORIES (${CLFFT_LIBRARIES})
+  MESSAGE (STATUS "Found clFFT include dir: ${CLFFT_INCLUDE_DIRS}")
+  INCLUDE_DIRECTORIES (${CLFFT_INCLUDE_DIRS})
+  LINK_DIRECTORIES (${CLFFT_LIBRARIES})
 
-#find clRNG
+  #find clRNG
-#SET (clRNG_USE_STATIC_LIBS OFF)
+  #SET (clRNG_USE_STATIC_LIBS OFF)
-#FIND_PACKAGE(clRng REQUIRED HINTS &ENV{CLRNG_PREFIX} $ENV{CLRNG_DIR} $ENV{CLRNG})
+  #FIND_PACKAGE(clRng REQUIRED HINTS &ENV{CLRNG_PREFIX} $ENV{CLRNG_DIR} $ENV{CLRNG})
-#MESSAGE (STATUS "Found clRNG library: ${CLRNG_LIBRARIES}")
+  #MESSAGE (STATUS "Found clRNG library: ${CLRNG_LIBRARIES}")
-#MESSAGE (STATUS "Found clRNG include dir: ${CLRNG_INCLUDE_DIRS}")
+  #MESSAGE (STATUS "Found clRNG include dir: ${CLRNG_INCLUDE_DIRS}")
-#INCLUDE_DIRECTORIES (${CLFFT_INCLUDE_DIRS})
+  #INCLUDE_DIRECTORIES (${CLFFT_INCLUDE_DIRS})
-#LINK_DIRECTORIES (${CLRNG_LIBRARIES})
+  #LINK_DIRECTORIES (${CLRNG_LIBRARIES})
-#find_package(PkgConfig)
+  #find_package(PkgConfig)
-#pkg_check_modules(clRng REQUIRED)
+  #pkg_check_modules(clRng REQUIRED)
+
+  SET (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DDKS_AMD")
+ENDIF (ENABLE_AMD)
 
 #enable UQTK
 OPTION (USE_UQTK "Use UQTK" OFF)

src/Algorithms/CollimatorPhysics.h

@@ -5,10 +5,7 @@
 #include <string>
 #include "../DKSDefinitions.h"
 
-class DKSBaseMuSR;
-
 class DKSCollimatorPhysics {
-  friend class DKSBaseMuSR;
 
 protected:
 

src/DKSBase.cpp

@@ -103,25 +103,16 @@ DKSBase::DKSBase() {
 
 #ifdef DKS_CUDA
   cbase = new CudaBase();
-  cfft = new CudaFFT(cbase);
-  cgreens = new CudaGreensFunction(cbase);
   cchi = new CudaChiSquare(cbase);
-  ccol = new CudaCollimatorPhysics(cbase);
 #endif
 
-#ifdef DKS_OPENCL
+ls#ifdef DKS_OPENCL
   oclbase = new OpenCLBase();
-  oclfft = new OpenCLFFT(oclbase);
   oclchi = new OpenCLChiSquare(oclbase);
-  oclcol = new OpenCLCollimatorPhysics(oclbase);
-  oclgreens = new OpenCLGreensFunction(oclbase);
 #endif
 
 #ifdef DKS_MIC
   micbase = new MICBase();
-  micfft = new MICFFT(micbase);
-  miccol = new MICCollimatorPhysics(micbase);
-  micgreens = new MICGreensFunction(micbase);
   micchi = new MICChiSquare(micbase);
 #endif
 
@@ -139,26 +130,17 @@ DKSBase::DKSBase(const char* api_name, const char* device_name) {
 
 #ifdef DKS_CUDA
   cbase = new CudaBase();
-  cfft = new CudaFFT(cbase);
-  cgreens = new CudaGreensFunction(cbase);
   cchi = new CudaChiSquare(cbase);
-  ccol = new CudaCollimatorPhysics(cbase);
 #endif
 
 #ifdef DKS_OPENCL
   oclbase = new OpenCLBase();
-  oclfft = new OpenCLFFT(oclbase);
   oclchi = new OpenCLChiSquare(oclbase);
-  oclcol = new OpenCLCollimatorPhysics(oclbase);
-  oclgreens = new OpenCLGreensFunction(oclbase);
 #endif
 
 #ifdef DKS_MIC
   micbase = new MICBase();
-  micfft = new MICFFT(micbase);
   miccol = new MICCollimatorPhysics(micbase);
-  micgreens = new MICGreensFunction(micbase);
-  micchi = new MICChiSquare(micbase);
 #endif
 
 }
@@ -175,28 +157,22 @@ DKSBase::~DKSBase() {
   if (m_function_name != NULL)
     delete[] m_function_name;
 
+  delete dksfft;
+  delete dkscol;
+  delete dksgreens;
  
 #ifdef DKS_CUDA
-  delete cfft;
-  delete cgreens;
   delete cchi;
-  delete ccol;
   delete cbase;
 #endif
 
 #ifdef DKS_OPENCL
-  delete oclfft;
   delete oclchi;
-  delete oclcol;
   delete oclbase;
-  delete oclgreens;
 #endif
 
 
 #ifdef DKS_MIC
-  delete micfft;
-  delete miccol;
-  delete micgreens;
   delete micchi;
   delete micbase;
 #endif
@@ -311,38 +287,70 @@ int DKSBase::getDeviceList(std::vector<int> &devices) {
     return DKS_ERROR;
 }
 
+int DKSBase::setup() {
+
+  int ierr = DKS_ERROR;
+
+  if (apiOpenCL()) {
+    ierr = OPENCL_SAFECALL( DKS_SUCCESS );
+    //TODO: only enable if AMD libraries are available
+    dksfft = OPENCL_SAFEINIT_AMD( new OpenCLFFT(oclbase) );
+    dkscol = OPENCL_SAFEINIT_AMD( new OpenCLCollimatorPhysics(oclbase) );
+    dksgreens = OPENCL_SAFEINIT_AMD( new OpenCLGreensFunction(oclbase) );
+  } else if (apiCuda()) {
+    ierr = CUDA_SAFECALL( DKS_SUCCESS );
+    dksfft = CUDA_SAFEINIT( new CudaFFT(cbase) );
+    dkscol = CUDA_SAFEINIT( new CudaCollimatorPhysics(cbase) );
+    dksgreens = CUDA_SAFEINIT( new CudaGreensFunction(cbase) );
+  } else if (apiOpenMP()) {
+    ierr = MIC_SAFECALL( DKS_SUCCESS );
+    dksfft = MIC_SAFEINIT( new MICFFT(micbase) );
+    dkscol = MIC_SAFEINIT( new MICCollimatorPhysics(micbase) );
+    dksgreens = MIC_SAFEINIT( new MICGreensFunction(micbase) );
+  } else {
+    ierr = DKS_ERROR;
+  }
+
+  return ierr;
+}
+
 /*
   init device
 */
 int DKSBase::initDevice() {
 
+  int ierr = DKS_ERROR;
+
   //if api is not set default is OpenCL
   if (!m_api_set) {
     setDevice("-gpu", 4);
     setAPI(API_OPENCL, 6);
-    return OPENCL_SAFECALL( oclbase->ocl_setUp("-gpu") );
+    ierr = OPENCL_SAFECALL( oclbase->ocl_setUp("-gpu") );
   } else {
     if (apiOpenCL()) {
       if (!m_device_set) {
 	setDevice("-gpu", 4);
 	setAPI(API_OPENCL, 6);
-	return OPENCL_SAFECALL( oclbase->ocl_setUp("-gpu") );
+	ierr = OPENCL_SAFECALL( oclbase->ocl_setUp("-gpu") );
       } else {
 	setAPI(API_OPENCL, 6);
-	return OPENCL_SAFECALL( oclbase->ocl_setUp(m_device_name) );
+	ierr = OPENCL_SAFECALL( oclbase->ocl_setUp(m_device_name) );
       }
     } else if (apiCuda()) {
       setDevice("-gpu", 4);
       setAPI(API_CUDA, 4);			
-      return CUDA_SAFECALL(DKS_SUCCESS);
+      ierr = CUDA_SAFECALL(DKS_SUCCESS);
     } else if (apiOpenMP()) {
       setDevice("-mic", 4);
       setAPI(API_OPENMP, 6);
-      return MIC_SAFECALL(DKS_SUCCESS);
+      ierr = MIC_SAFECALL(DKS_SUCCESS);
     }
   }
 
-  return DKS_ERROR;
+  if (ierr == DKS_SUCCESS)
+    ierr = setup();
+
+  return ierr;
 }
 
 /* 
@@ -464,11 +472,11 @@ int DKSBase::syncDevice() {
 int DKSBase::setupFFT(int ndim, int N[3]) {
 
   if (apiCuda()) {
-    return CUDA_SAFECALL( cfft->setupFFT(ndim, N) );
+    return dksfft->setupFFT(ndim, N);
   } else if (apiOpenCL()) {
-    int ierr1 = OPENCL_SAFECALL( oclfft->setupFFT(ndim, N) );
+    int ierr1 = dksfft->setupFFT(ndim, N);
-    int ierr2 = OPENCL_SAFECALL( oclfft->setupFFTRC(ndim, N) );
+    int ierr2 = dksfft->setupFFTRC(ndim, N);
-    int ierr3 = OPENCL_SAFECALL( oclfft->setupFFTCR(ndim, N) );
+    int ierr3 = dksfft->setupFFTCR(ndim, N);
     if (ierr1 != DKS_SUCCESS || ierr2 != DKS_SUCCESS || ierr3 != DKS_SUCCESS)
       return DKS_ERROR;
 
@@ -476,8 +484,8 @@ int DKSBase::setupFFT(int ndim, int N[3]) {
   } else if (apiOpenMP()) {
     //micbase.mic_setupFFT(ndim, N);
     //BENI: setting up RC and CR transformations on MIC
-    int ierr1 = MIC_SAFECALL( micfft->setupFFTRC(ndim, N, 1.) );
+    int ierr1 = dksfft->setupFFTRC(ndim, N, 1.);
-    int ierr2 = MIC_SAFECALL( micfft->setupFFTCR(ndim, N, 1./(N[0]*N[1]*N[2])) );
+    int ierr2 = dksfft->setupFFTCR(ndim, N, 1./(N[0]*N[1]*N[2]));
     if (ierr1 != DKS_SUCCESS)
       return ierr1;
     if (ierr2 != DKS_SUCCESS)
@@ -492,11 +500,11 @@ int DKSBase::setupFFT(int ndim, int N[3]) {
 int DKSBase::setupFFTRC(int ndim, int N[3], double scale) {
 
   if (apiCuda())
-    return CUDA_SAFECALL(cfft->setupFFT(ndim, N));
+    return dksfft->setupFFT(ndim, N);
   if (apiOpenCL())
-    return OPENCL_SAFECALL(oclfft->setupFFTRC(ndim, N));
+    return dksfft->setupFFTRC(ndim, N);
   else if (apiOpenMP())
-    return MIC_SAFECALL(micfft->setupFFTRC(ndim, N, scale));
+    return dksfft->setupFFTRC(ndim, N, scale);
 
   return DKS_ERROR;
 
@@ -506,11 +514,11 @@ int DKSBase::setupFFTRC(int ndim, int N[3], double scale) {
 int DKSBase::setupFFTCR(int ndim, int N[3], double scale) {
 
   if (apiCuda())
-    return CUDA_SAFECALL(cfft->setupFFT(ndim, N));
+    return dksfft->setupFFT(ndim, N);
   if (apiOpenCL())
-    return OPENCL_SAFECALL(oclfft->setupFFTCR(ndim, N));
+    return dksfft->setupFFTCR(ndim, N);
   else if (apiOpenMP())
-    return MIC_SAFECALL(micfft->setupFFTCR(ndim, N, scale));
+    return dksfft->setupFFTCR(ndim, N, scale);
 
   return DKS_ERROR;
 
@@ -519,17 +527,10 @@ int DKSBase::setupFFTCR(int ndim, int N[3], double scale) {
 /* call OpenCL FFT function for selected platform */
 int DKSBase::callFFT(void * data_ptr, int ndim, int dimsize[3], int streamId) {
 
-  if (apiOpenCL()) {
+  if (apiOpenCL() || apiOpenMP()) 
-    //load kernel and execute
+    return dksfft->executeFFT(data_ptr, ndim, dimsize);
-    if ( loadOpenCLKernel("OpenCL/OpenCLKernels/OpenCLFFT.cl") == DKS_SUCCESS )
+  else if (apiCuda())
-      return OPENCL_SAFECALL( oclfft->executeFFT(data_ptr, ndim, dimsize) );
+    return dksfft->executeFFT(data_ptr, ndim, dimsize, streamId);
-    else
-      return DKS_ERROR;
-  } else if (apiCuda()) {
-    return CUDA_SAFECALL(cfft->executeFFT(data_ptr, ndim, dimsize, streamId));
-  } else if (apiOpenMP()) {
-    return MIC_SAFECALL(micfft->executeFFT(data_ptr, ndim, dimsize));
-  }
 
   DEBUG_MSG("No implementation for selected platform");
   return DKS_ERROR;
@@ -537,16 +538,10 @@ int DKSBase::callFFT(void * data_ptr, int ndim, int dimsize[3], int streamId) {
 
 /* call OpenCL IFFT function for selected platform */
 int DKSBase::callIFFT(void * data_ptr, int ndim, int dimsize[3], int streamId) {
-  if (apiOpenCL()) {
+  if (apiOpenCL() || apiOpenMP())
-    if ( loadOpenCLKernel("OpenCL/OpenCLKernels/OpenCLFFT.cl") == DKS_SUCCESS )
+      return dksfft->executeIFFT(data_ptr, ndim, dimsize);
-      return OPENCL_SAFECALL( oclfft->executeIFFT(data_ptr, ndim, dimsize) );
+  else if (apiCuda()) 
-    else
+    return dksfft->executeIFFT(data_ptr, ndim, dimsize, streamId);
-      return DKS_ERROR;
-  } else if (apiCuda()) {
-    return CUDA_SAFECALL( cfft->executeIFFT(data_ptr, ndim, dimsize, streamId) );
-  } else if (apiOpenMP()) {
-    return MIC_SAFECALL( micfft->executeIFFT(data_ptr, ndim, dimsize) );
-  }
 
   DEBUG_MSG("No implementation for selected platform");
   return DKS_ERROR;
@@ -557,13 +552,13 @@ int DKSBase::callNormalizeFFT(void * data_ptr, int ndim, int dimsize[3], int str
 
   if (apiOpenCL()) {
     if ( loadOpenCLKernel("OpenCL/OpenCLKernels/OpenCLFFT.cl") == DKS_SUCCESS )
-      return OPENCL_SAFECALL( oclfft->normalizeFFT(data_ptr, ndim, dimsize) );
+      return dksfft->normalizeFFT(data_ptr, ndim, dimsize);
     else 
       return DKS_ERROR;
   } else if (apiCuda()) {
-    return CUDA_SAFECALL( cfft->normalizeFFT(data_ptr, ndim, dimsize, streamId) );
+    return dksfft->normalizeFFT(data_ptr, ndim, dimsize, streamId);
   } else if (apiOpenMP()) {
-    return MIC_SAFECALL( micfft->normalizeFFT(data_ptr, ndim, dimsize) );
+    return dksfft->normalizeFFT(data_ptr, ndim, dimsize);
   }
 
   DEBUG_MSG("No implementation for selected platform");
@@ -574,11 +569,9 @@ int DKSBase::callNormalizeFFT(void * data_ptr, int ndim, int dimsize[3], int str
 int DKSBase::callR2CFFT(void * real_ptr, void * comp_ptr, int ndim, int dimsize[3], int streamId) {
 
   if (apiCuda())
-    return CUDA_SAFECALL( cfft->executeRCFFT(real_ptr, comp_ptr, ndim, dimsize, streamId) );
+    return dksfft->executeRCFFT(real_ptr, comp_ptr, ndim, dimsize, streamId);
-  else if (apiOpenCL())
+  else if (apiOpenCL() || apiOpenMP())
-    return OPENCL_SAFECALL( oclfft->executeRCFFT(real_ptr, comp_ptr, ndim, dimsize) );
+    return dksfft->executeRCFFT(real_ptr, comp_ptr, ndim, dimsize);
-  else if (apiOpenMP())
-    return MIC_SAFECALL( micfft->executeRCFFT(real_ptr,comp_ptr, ndim, dimsize) );
 
   DEBUG_MSG("No implementation for selected platform");
   return DKS_ERROR;
@@ -587,11 +580,9 @@ int DKSBase::callR2CFFT(void * real_ptr, void * comp_ptr, int ndim, int dimsize[
 /* call complex to real FFT */
 int DKSBase::callC2RFFT(void * real_ptr, void * comp_ptr, int ndim, int dimsize[3], int streamId) {
   if (apiCuda())
-    return CUDA_SAFECALL( cfft->executeCRFFT(real_ptr, comp_ptr, ndim, dimsize, streamId) );
+    return dksfft->executeCRFFT(real_ptr, comp_ptr, ndim, dimsize, streamId);
-  else if (apiOpenCL())
+  else if (apiOpenCL() || apiOpenMP())
-    return OPENCL_SAFECALL( oclfft->executeCRFFT(real_ptr, comp_ptr, ndim, dimsize) );
+    return dksfft->executeCRFFT(real_ptr, comp_ptr, ndim, dimsize);
-  else if (apiOpenMP())
-    return MIC_SAFECALL( micfft->executeCRFFT(comp_ptr,real_ptr, ndim, dimsize) );
 
   DEBUG_MSG("No implementation for selected platform");
   return DKS_ERROR;
@@ -600,72 +591,38 @@ int DKSBase::callC2RFFT(void * real_ptr, void * comp_ptr, int ndim, int dimsize[
 /* normalize complex to real iFFT */
 int DKSBase::callNormalizeC2RFFT(void * real_ptr, int ndim, int dimsize[3], int streamId) {
   if (apiCuda())
-    return CUDA_SAFECALL( cfft->normalizeCRFFT(real_ptr, ndim, dimsize, streamId) );
+    return dksfft->normalizeCRFFT(real_ptr, ndim, dimsize, streamId);
   else if (apiOpenCL())
-    return DKS_SUCCESS;
+    return DKS_ERROR;
   else if (apiOpenMP())
-    return DKS_SUCCESS;
+    return DKS_ERROR;
 
   DEBUG_MSG("No implementation for selected platform");
-  return DKS_SUCCESS;
+  return DKS_ERROR;
 }
 
 int DKSBase::callGreensIntegral(void *tmp_ptr, int I, int J, int K, int NI, int NJ, 
 				double hz_m0, double hz_m1, double hz_m2, int streamId) {
 
-  if (apiCuda()) {
+    return dksgreens->greensIntegral(tmp_ptr, I, J, K, NI, NJ, 
-    return CUDA_SAFECALL(cgreens->greensIntegral(tmp_ptr, I, J, K, NI, NJ, 
+				     hz_m0, hz_m1, hz_m2, streamId);
-						 hz_m0, hz_m1, hz_m2, streamId) );
-  } else if (apiOpenCL()) {
-    return OPENCL_SAFECALL(oclgreens->greensIntegral(tmp_ptr, I, J, K, NI, NJ, 
-						     hz_m0, hz_m1, hz_m2) );
-  } else if (apiOpenMP()) {
-    //BENI:
-    return MIC_SAFECALL(micgreens->greensIntegral(tmp_ptr, I, J, K, hz_m0, hz_m1, hz_m2));
-  } 
 
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 int DKSBase::callGreensIntegration(void *mem_ptr, void *tmp_ptr, 
 				   int I, int J, int K, int streamId) {
 
-  if (apiCuda())
+  return dksgreens->integrationGreensFunction(mem_ptr, tmp_ptr, I, J, K, streamId);
-    return CUDA_SAFECALL(cgreens->integrationGreensFunction(mem_ptr, tmp_ptr, I, J, K, streamId));
-  else if (apiOpenCL())
-    return OPENCL_SAFECALL(oclgreens->integrationGreensFunction(mem_ptr, tmp_ptr, I, J, K));
-  else if (apiOpenMP())
-    return MIC_SAFECALL(micgreens->integrationGreensFunction(mem_ptr, tmp_ptr, I, J, K));
-  
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 int DKSBase::callMirrorRhoField(void *mem_ptr, int I, int J, int K, int streamId) {
 
-  if (apiCuda()) 
+  return dksgreens->mirrorRhoField(mem_ptr, I, J, K, streamId);  
-    return CUDA_SAFECALL(cgreens->mirrorRhoField(mem_ptr, I, J, K, streamId));
-  else if (apiOpenCL())
-    return OPENCL_SAFECALL(oclgreens->mirrorRhoField(mem_ptr, I, J, K, streamId));
-  else if (apiOpenMP())
-    return MIC_SAFECALL(micgreens->mirrorRhoField(mem_ptr, I, J, K));
-  
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 int DKSBase::callMultiplyComplexFields(void *mem_ptr1, void *mem_ptr2, int size, int streamId) {
   
-  if (apiCuda())
+  return dksgreens->multiplyCompelxFields(mem_ptr1, mem_ptr2, size, streamId);
-    return CUDA_SAFECALL(cgreens->multiplyCompelxFields(mem_ptr1, mem_ptr2, size, streamId));
-  else if (apiOpenCL())
-    return OPENCL_SAFECALL(oclgreens->multiplyCompelxFields(mem_ptr1, mem_ptr2, size));
-  else if (apiOpenMP())
-    return MIC_SAFECALL(micgreens->multiplyCompelxFields(mem_ptr1, mem_ptr2, size));
-
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 
@@ -748,19 +705,7 @@ int DKSBase::callCollimatorPhysics(void *mem_ptr, void *par_ptr,
 				   int &numaddback, int &numdead) 
 {
 
-  if (apiCuda()) {
+  return dkscol->CollimatorPhysics(mem_ptr, par_ptr, numparticles);
-    return CUDA_SAFECALL(ccol->CollimatorPhysics(mem_ptr, par_ptr, numparticles));
-  } else if (apiOpenCL()) {
-    if (loadOpenCLKernel("OpenCL/OpenCLKernels/OpenCLCollimatorPhysics.cl") == DKS_SUCCESS)
-      return OPENCL_SAFECALL(oclcol->CollimatorPhysics(mem_ptr, par_ptr, numparticles));
-    else
-      return DKS_ERROR;
-
-  } else if (apiOpenMP()) {
-    return MIC_SAFECALL(miccol->CollimatorPhysics(mem_ptr, par_ptr, numparticles));
-  } 
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 
 }
 
@@ -768,13 +713,8 @@ int DKSBase::callCollimatorPhysics(void *mem_ptr, void *par_ptr,
 int DKSBase::callCollimatorPhysics2(void *mem_ptr, void *par_ptr, int numparticles) 
 {
 
-  if (apiCuda())
+  return dkscol->CollimatorPhysics(mem_ptr, par_ptr, numparticles);
-    return CUDA_SAFECALL( ccol->CollimatorPhysics(mem_ptr, par_ptr, numparticles) );
-  else if (apiOpenMP())
-    return MIC_SAFECALL( miccol->CollimatorPhysics(mem_ptr, par_ptr, numparticles) );
   
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 int DKSBase::callCollimatorPhysicsSoA(void *label_ptr, void *localID_ptr, 
@@ -783,28 +723,21 @@ int DKSBase::callCollimatorPhysicsSoA(void *label_ptr, void *localID_ptr,
 				      void *par_ptr, int numparticles)
 {
 
-  if (apiOpenMP()) {
+  
-    return MIC_SAFECALL( miccol->CollimatorPhysicsSoA(label_ptr, localID_ptr, 
+    return dkscol->CollimatorPhysicsSoA(label_ptr, localID_ptr, 
 					rx_ptr, ry_ptr, rz_ptr, 
 					px_ptr, py_ptr, pz_ptr,
-						      par_ptr,  numparticles) );
+					par_ptr,  numparticles);
-  }
 
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 }
 
 
 int DKSBase::callCollimatorPhysicsSort(void *mem_ptr, int numparticles, int &numaddback) 
 {
 
-  if (apiCuda())
-    return CUDA_SAFECALL(ccol->CollimatorPhysicsSort(mem_ptr, numparticles, numaddback));
-  else if (apiOpenMP())
-    return MIC_SAFECALL(miccol->CollimatorPhysicsSort(mem_ptr, numparticles, numaddback));
 
-  DEBUG_MSG("No implementation for selceted platform");
+  return dkscol->CollimatorPhysicsSort(mem_ptr, numparticles, numaddback);
-  return DKS_ERROR;
+
 }
 
 int DKSBase::callCollimatorPhysicsSortSoA(void *label_ptr, void *localID_ptr, 
@@ -813,15 +746,10 @@ int DKSBase::callCollimatorPhysicsSortSoA(void *label_ptr, void *localID_ptr,
 					  void *par_ptr, int numparticles, int &numaddback) 
 {
 
-  if (apiOpenMP()) {
+  return MIC_SAFECALL(dkscol->CollimatorPhysicsSortSoA(label_ptr, localID_ptr, 
-    return MIC_SAFECALL(miccol->CollimatorPhysicsSortSoA(label_ptr, localID_ptr, 
 						       rx_ptr, ry_ptr, rz_ptr, 
 						       px_ptr, py_ptr, pz_ptr,
 						       par_ptr,  numparticles, numaddback));
-  }
-  
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 
 }
 
@@ -844,15 +772,7 @@ int DKSBase::callParallelTTrackerPush(void *r_ptr, void *p_ptr, int npart,
 				      bool usedt, int streamId) 
 {
 
-  if (apiCuda()) 
+  return dkscol->ParallelTTrackerPush(r_ptr, p_ptr, npart, dt_ptr, dt, c, usedt, streamId);
-    return CUDA_SAFECALL(ccol->ParallelTTrackerPush(r_ptr, p_ptr, npart, dt_ptr, dt, c, 
-						    usedt, streamId));
-  else if (apiOpenMP())
-    return MIC_SAFECALL(miccol->ParallelTTrackerPush(r_ptr, p_ptr, npart, dt_ptr, dt, 
-						     c, usedt, streamId));
-   
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
 
 }
 
@@ -862,20 +782,8 @@ int DKSBase::callParallelTTrackerPushTransform(void *x_ptr, void *p_ptr,
 					       double c, bool usedt, int streamId)
 {
 
-  if (apiCuda()) {
+  return dkscol->ParallelTTrackerPushTransform(x_ptr, p_ptr, lastSec_ptr, orient_ptr,
-    return CUDA_SAFECALL(ccol->ParallelTTrackerPushTransform(x_ptr, p_ptr, 
+					       npart, nsec, dt_ptr, dt, c, usedt, streamId);
-							     lastSec_ptr, orient_ptr,
-							     npart, nsec, dt_ptr, dt, 
-							     c, usedt, streamId));
-  } else if (apiOpenMP()) {
-    return MIC_SAFECALL(miccol->ParallelTTrackerPushTransform(x_ptr, p_ptr, 
-							      lastSec_ptr, orient_ptr,
-							      npart, nsec, dt_ptr, dt, 
-							      c, usedt, streamId));
-  } 
-    
-  DEBUG_MSG("No implementation for selceted platform");
-  return DKS_ERROR;
   
 }
 

src/DKSBase.h

@@ -29,8 +29,11 @@
 #endif
 
 #include "OpenCL/OpenCLBase.h"
-#include "OpenCL/OpenCLFFT.h"
 #include "OpenCL/OpenCLChiSquare.h"
+#endif
+
+#ifdef DKS_AMD
+#include "OpenCL/OpenCLFFT.h"
 #include "OpenCL/OpenCLCollimatorPhysics.h"
 #include "OpenCL/OpenCLGreensFunction.h"
 #endif
@@ -52,6 +55,7 @@
 #include "MIC/MICGreensFunction.hpp"
 #endif
 
+#include "Algorithms/GreensFunction.h"
 #include "Algorithms/CollimatorPhysics.h"
 #include "Algorithms/FFT.h"
 
@@ -72,27 +76,22 @@ private:
   bool m_auto_tuning;
   bool m_use_config;
 
+  DKSFFT *dksfft;
+  DKSCollimatorPhysics *dkscol;
+  GreensFunction *dksgreens;
+
 #ifdef DKS_OPENCL	
   OpenCLBase *oclbase;
-  OpenCLFFT *oclfft;
   OpenCLChiSquare *oclchi;
-  OpenCLCollimatorPhysics *oclcol;
-  OpenCLGreensFunction *oclgreens;
 #endif
 
 #ifdef DKS_CUDA
   CudaBase *cbase;
-  CudaFFT *cfft;
-  CudaGreensFunction *cgreens;
   CudaChiSquare *cchi;
-  CudaCollimatorPhysics *ccol;
 #endif
 
 #ifdef DKS_MIC
   MICBase *micbase;
-  MICFFT *micfft;
-  MICCollimatorPhysics *miccol;
-  MICGreensFunction *micgreens;
   MICChiSquare *micchi;
 #endif
 
@@ -156,6 +155,11 @@ protected:
     return device_name;
   }
 
+  /** Private function to initialize objects based on the device used.
+   *  
+   */
+  int setup();
+
 public:
 
   /** 

src/DKSDefinitions.h

@@ -62,6 +62,12 @@
 #define OPENCL_SAFEINIT(x) ( NULL )
 #endif
 
+#ifdef DKS_AMD
+#define OPENCL_SAFEINIT_AMD(x) ( x )
+#else
+#define OPENCL_SAFEINIT_AMD(x) ( NULL )
+#endif
+
 #ifdef DKS_MIC
 #define MIC_SAFEINIT(x) ( x )
 #else

src/MIC/MICGreensFunction.hpp

@@ -29,8 +29,8 @@ public:
   ~MICGreensFunction();
 
   /* compute greens integral analytically */
-  int greensIntegral(void * tmpgreen_, int I, int J, int K, double hr_m0, double hr_m1, double hr_m2,
+  int greensIntegral(void * tmpgreen_, int I, int J, int K, int NI, int NJ,
-		     int streamId = -1);
+		     double hr_m0, double hr_m1, double hr_m2, int streamId = -1);
 
   /* perform the actual integration */
   int integrationGreensFunction(void * rho2_m, void * tmpgreen,int I,int J, int K, 

src/OpenCL/CMakeLists.txt

@@ -1,4 +1,7 @@
-SET (_SRCS
+#dont include FFT, GreensFunction and CollimatorPhysics if clFFT and clRNG not found
+
+IF (ENABLE_AMD)
+  SET (_SRCS
     OpenCLBase.cpp
     OpenCLFFT.cpp
     OpenCLChiSquare.cpp
@@ -7,7 +10,7 @@ SET (_SRCS
     OpenCLGreensFunction.cpp
     )
 
-SET (_HDRS
+  SET (_HDRS
     OpenCLBase.h
     OpenCLFFT.h
     OpenCLChiSquare.h
@@ -15,10 +18,19 @@ SET (_HDRS
     OpenCLChiSquareRuntime.h
     OpenCLGreensFunction.h
     )
+ELSE (ENABLE_AMD)
+  SET (_SRCS
+    OpenCLBase.cpp
+    OpenCLChiSquare.cpp
+    OpenCLChiSquareRuntime.cpp
+    )
 
-#INCLUDE_DIRECTORIES (
+  SET (_HDRS
-#  ${CMAKE_CURRENT_SOURCE_DIR}
+    OpenCLBase.h
-#)
+    OpenCLChiSquare.h
+    OpenCLChiSquareRuntime.h
+    )
+ENDIF (ENABLE_AMD)
 
 SET (_KERNELS
   OpenCLKernels/OpenCLChiSquare.cl