diff --git a/RTCP/Cobalt/GPUProc/src/CMakeLists.txt b/RTCP/Cobalt/GPUProc/src/CMakeLists.txt
index 03094df4b14299a6478b67672b10383c382b5557..33fb73b240310e3201e3d041b9eadf2b1c358c09 100644
--- a/RTCP/Cobalt/GPUProc/src/CMakeLists.txt
+++ b/RTCP/Cobalt/GPUProc/src/CMakeLists.txt
@@ -54,8 +54,8 @@ list(APPEND _gpuproc_sources
#Kernels/UHEP_InvFIR_Kernel.cc
#Kernels/UHEP_TransposeKernel.cc
#Kernels/UHEP_TriggerKernel.cc
- cuda/Pipelines/Pipeline.cc
-# cuda/Pipelines/UHEP_Pipeline.cc
+ Pipelines/Pipeline.cc
+# Pipelines/UHEP_Pipeline.cc
cuda/SubbandProcs/SubbandProc.cc
cuda/SubbandProcs/SubbandProcInputData.cc
cuda/SubbandProcs/SubbandProcOutputData.cc
diff --git a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/Pipeline.cc b/RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.cc
similarity index 100%
rename from RTCP/Cobalt/GPUProc/src/cuda/Pipelines/Pipeline.cc
rename to RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.cc
diff --git a/RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.h b/RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.h
index 0718dcdac67368e0034c519ca6aa9232ba7f4d7f..b09516353d02eed582d95e36bb683953b9fe1ba1 100644
--- a/RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.h
+++ b/RTCP/Cobalt/GPUProc/src/Pipelines/Pipeline.h
@@ -1,6 +1,5 @@
//# Pipeline.h
-//#
-//# Copyright (C) 2013 ASTRON (Netherlands Institute for Radio Astronomy)
+//# Copyright (C) 2012-2013 ASTRON (Netherlands Institute for Radio Astronomy)
//# P.O. Box 2, 7990 AA Dwingeloo, The Netherlands
//#
//# This file is part of the LOFAR software suite.
@@ -19,23 +18,130 @@
//#
//# $Id$
-// \file
-// Support for the beamformer pipeline.
+#ifndef LOFAR_GPUPROC_CUDA_PIPELINE_H
+#define LOFAR_GPUPROC_CUDA_PIPELINE_H
-#ifndef LOFAR_GPUPROC_PIPELINE_H
-#define LOFAR_GPUPROC_PIPELINE_H
+#include <string>
+#include <vector>
-#if defined (USE_CUDA) && defined (USE_OPENCL)
-# error "Either CUDA or OpenCL must be enabled, not both"
-#endif
+#include <Common/LofarTypes.h>
+#include <MACIO/RTmetadata.h>
+#include <CoInterface/Parset.h>
+#include <CoInterface/SmartPtr.h>
+#include <CoInterface/Pool.h>
+#include <CoInterface/OMPThread.h>
+#include <CoInterface/TABTranspose.h>
-#if defined (USE_CUDA)
-# include <GPUProc/cuda/Pipelines/Pipeline.h>
-#elif defined (USE_OPENCL)
-# include <GPUProc/opencl/Pipelines/Pipeline.h>
-#else
-# error "Either CUDA or OpenCL must be enabled, not neither"
-#endif
+#include <GPUProc/gpu_wrapper.h>
+#include <GPUProc/PerformanceCounter.h>
+#include <GPUProc/SubbandProcs/SubbandProc.h>
+#include <GPUProc/SubbandProcs/KernelFactories.h>
+
+#include <GPUProc/MPIReceiver.h>
+
+namespace LOFAR
+{
+ namespace Cobalt
+ {
+ using MACIO::RTmetadata;
+
+ class Pipeline
+ {
+ public:
+ Pipeline(const Parset &, const std::vector<size_t> &subbandIndices,
+ const std::vector<gpu::Device> &devices,
+ Pool<struct MPIRecvData> &pool,
+ RTmetadata &mdLogger, const std::string &mdKeyPrefix,
+ unsigned hostID = 0);
+
+ ~Pipeline();
+
+ // allocate resources, such as GPU buffers.
+ //
+ // Pipeline deploys delayed construction, because the resources may still
+ // be occupied by a previous observation at the time of construction.
+ //
+ // An alternative to delayed allocation could be to retry the GPU malloc
+ // with a timeout, but that could potentially dead-lock two concurrent
+ // observations.
+ void allocateResources();
+
+ // for each subband get data from input stream, sync, start the kernels to process all data, write output in parallel
+ void processObservation();
+
+ struct Output
+ {
+ // output data queue
+ SmartPtr< Queue< SmartPtr<SubbandProcOutputData> > > queue;
+ };
+
+ std::vector< SmartPtr<SubbandProc> > subbandProcs;
+
+ protected:
+ const Parset &ps;
+ const std::vector<gpu::Device> devices;
+
+ const std::vector<size_t> subbandIndices; // [localSubbandIdx]
+
+ // Whether we're the pipeline that processes the first subband.
+ // If true, we log our progress at INFO. Otherwise, at DEBUG.
+ const bool processingSubband0;
+
+ const size_t nrSubbandsPerSubbandProc;
+
+ RTmetadata &itsMdLogger; // non-const to be able to use its log()
+ const std::string itsMdKeyPrefix;
+
+ // Threads that write to outputProc, and need to
+ // be killed when they stall at observation end.
+ OMPThreadSet outputThreads;
+
+ Pool<struct MPIRecvData> &mpiPool;
+
+ std::vector<struct Output> writePool; // [localSubbandIdx]
+
+ KernelFactories factories;
+
+ // For each block, transpose all subbands from all stations, and divide the
+ // work over the subbandProcs
+ void transposeInput();
+ template<typename SampleT> void transposeInput();
+
+ // preprocess subbands on the CPU
+ void preprocessSubbands(SubbandProc &subbandProc);
+
+ // process subbands on the GPU
+ void processSubbands(SubbandProc &subbandProc);
+
+ // Post-process subbands on the CPU
+ void postprocessSubbands(SubbandProc &subbandProc);
+
+ void writeBeamformedOutput(
+ unsigned globalSubbandIdx,
+ Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
+ Queue< SmartPtr<SubbandProcOutputData> > &outputQueue,
+ Queue< SmartPtr<SubbandProcOutputData> > &spillQueue );
+
+ void writeCorrelatedOutput(
+ unsigned globalSubbandIdx,
+ Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
+ Queue< SmartPtr<SubbandProcOutputData> > &outputQueue );
+
+ public:
+ // Send subbands to Storage
+ void writeOutput(
+ unsigned globalSubbandIdx,
+ Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
+ Queue< SmartPtr<SubbandProcOutputData> > &outputQueue );
+
+ // Output send engine, takes care of the host connections and the multiplexing.
+ TABTranspose::MultiSender multiSender;
+
+ // MPI rank for this node
+ const unsigned hostID;
+ };
+ }
+}
#endif
diff --git a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/UHEP_Pipeline.cc b/RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.cc
similarity index 100%
rename from RTCP/Cobalt/GPUProc/src/cuda/Pipelines/UHEP_Pipeline.cc
rename to RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.cc
diff --git a/RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.h b/RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.h
index 8fdcc994b2974170d7418a7057e5702bbed0abbd..617d168b42be3e1324d90a882967b7e382608684 100644
--- a/RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.h
+++ b/RTCP/Cobalt/GPUProc/src/Pipelines/UHEP_Pipeline.h
@@ -1,6 +1,5 @@
//# UHEP_Pipeline.h
-//#
-//# Copyright (C) 2013 ASTRON (Netherlands Institute for Radio Astronomy)
+//# Copyright (C) 2012-2013 ASTRON (Netherlands Institute for Radio Astronomy)
//# P.O. Box 2, 7990 AA Dwingeloo, The Netherlands
//#
//# This file is part of the LOFAR software suite.
@@ -19,23 +18,35 @@
//#
//# $Id$
-// \file
-// Include the right GPU API include with our options.
+#ifndef LOFAR_GPUPROC_CUDA_UHEP_PIPELINE_H
+#define LOFAR_GPUPROC_CUDA_UHEP_PIPELINE_H
-#ifndef LOFAR_GPUPROC_UHEP_PIPELINE_H
-#define LOFAR_GPUPROC_UHEP_PIPELINE_H
+#include <CoInterface/Parset.h>
-#if defined (USE_CUDA) && defined (USE_OPENCL)
-# error "Either CUDA or OpenCL must be enabled, not both"
-#endif
+#include <GPUProc/gpu_wrapper.h>
+#include "Pipeline.h"
+#include <GPUProc/PerformanceCounter.h>
-#if defined (USE_CUDA)
-# include <GPUProc/cuda/Pipelines/UHEP_Pipeline.h>
-#elif defined (USE_OPENCL)
-# include <GPUProc/opencl/Pipelines/UHEP_Pipeline.h>
-#else
-# error "Either CUDA or OpenCL must be enabled, not neither"
-#endif
+namespace LOFAR
+{
+ namespace Cobalt
+ {
+
+ class UHEP_Pipeline : public Pipeline
+ {
+ public:
+ UHEP_Pipeline(const Parset &, Pool<struct MPIRecvData> &pool);
+
+ void doWork();
+
+ gpu::Module beamFormerProgram, transposeProgram, invFFTprogram, invFIRfilterProgram, triggerProgram;
+
+ PerformanceCounter beamFormerCounter, transposeCounter, invFFTcounter, invFIRfilterCounter, triggerCounter;
+ PerformanceCounter beamFormerWeightsCounter, samplesCounter;
+ };
+
+ }
+}
#endif
diff --git a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/Pipeline.h b/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/Pipeline.h
deleted file mode 100644
index b09516353d02eed582d95e36bb683953b9fe1ba1..0000000000000000000000000000000000000000
--- a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/Pipeline.h
+++ /dev/null
@@ -1,147 +0,0 @@
-//# Pipeline.h
-//# Copyright (C) 2012-2013 ASTRON (Netherlands Institute for Radio Astronomy)
-//# P.O. Box 2, 7990 AA Dwingeloo, The Netherlands
-//#
-//# This file is part of the LOFAR software suite.
-//# The LOFAR software suite is free software: you can redistribute it and/or
-//# modify it under the terms of the GNU General Public License as published
-//# by the Free Software Foundation, either version 3 of the License, or
-//# (at your option) any later version.
-//#
-//# The LOFAR software suite is distributed in the hope that it will be useful,
-//# but WITHOUT ANY WARRANTY; without even the implied warranty of
-//# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-//# GNU General Public License for more details.
-//#
-//# You should have received a copy of the GNU General Public License along
-//# with the LOFAR software suite. If not, see <http://www.gnu.org/licenses/>.
-//#
-//# $Id$
-
-#ifndef LOFAR_GPUPROC_CUDA_PIPELINE_H
-#define LOFAR_GPUPROC_CUDA_PIPELINE_H
-
-#include <string>
-#include <vector>
-
-#include <Common/LofarTypes.h>
-#include <MACIO/RTmetadata.h>
-#include <CoInterface/Parset.h>
-#include <CoInterface/SmartPtr.h>
-#include <CoInterface/Pool.h>
-#include <CoInterface/OMPThread.h>
-#include <CoInterface/TABTranspose.h>
-
-#include <GPUProc/gpu_wrapper.h>
-#include <GPUProc/PerformanceCounter.h>
-#include <GPUProc/SubbandProcs/SubbandProc.h>
-#include <GPUProc/SubbandProcs/KernelFactories.h>
-
-#include <GPUProc/MPIReceiver.h>
-
-namespace LOFAR
-{
- namespace Cobalt
- {
- using MACIO::RTmetadata;
-
- class Pipeline
- {
- public:
- Pipeline(const Parset &, const std::vector<size_t> &subbandIndices,
- const std::vector<gpu::Device> &devices,
- Pool<struct MPIRecvData> &pool,
- RTmetadata &mdLogger, const std::string &mdKeyPrefix,
- unsigned hostID = 0);
-
- ~Pipeline();
-
- // allocate resources, such as GPU buffers.
- //
- // Pipeline deploys delayed construction, because the resources may still
- // be occupied by a previous observation at the time of construction.
- //
- // An alternative to delayed allocation could be to retry the GPU malloc
- // with a timeout, but that could potentially dead-lock two concurrent
- // observations.
- void allocateResources();
-
- // for each subband get data from input stream, sync, start the kernels to process all data, write output in parallel
- void processObservation();
-
- struct Output
- {
- // output data queue
- SmartPtr< Queue< SmartPtr<SubbandProcOutputData> > > queue;
- };
-
- std::vector< SmartPtr<SubbandProc> > subbandProcs;
-
- protected:
- const Parset &ps;
- const std::vector<gpu::Device> devices;
-
- const std::vector<size_t> subbandIndices; // [localSubbandIdx]
-
- // Whether we're the pipeline that processes the first subband.
- // If true, we log our progress at INFO. Otherwise, at DEBUG.
- const bool processingSubband0;
-
- const size_t nrSubbandsPerSubbandProc;
-
- RTmetadata &itsMdLogger; // non-const to be able to use its log()
- const std::string itsMdKeyPrefix;
-
- // Threads that write to outputProc, and need to
- // be killed when they stall at observation end.
- OMPThreadSet outputThreads;
-
- Pool<struct MPIRecvData> &mpiPool;
-
- std::vector<struct Output> writePool; // [localSubbandIdx]
-
- KernelFactories factories;
-
- // For each block, transpose all subbands from all stations, and divide the
- // work over the subbandProcs
- void transposeInput();
- template<typename SampleT> void transposeInput();
-
- // preprocess subbands on the CPU
- void preprocessSubbands(SubbandProc &subbandProc);
-
- // process subbands on the GPU
- void processSubbands(SubbandProc &subbandProc);
-
- // Post-process subbands on the CPU
- void postprocessSubbands(SubbandProc &subbandProc);
-
- void writeBeamformedOutput(
- unsigned globalSubbandIdx,
- Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
- Queue< SmartPtr<SubbandProcOutputData> > &outputQueue,
- Queue< SmartPtr<SubbandProcOutputData> > &spillQueue );
-
- void writeCorrelatedOutput(
- unsigned globalSubbandIdx,
- Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
- Queue< SmartPtr<SubbandProcOutputData> > &outputQueue );
-
- public:
- // Send subbands to Storage
- void writeOutput(
- unsigned globalSubbandIdx,
- Queue< SmartPtr<SubbandProcOutputData> > &inputQueue,
- Queue< SmartPtr<SubbandProcOutputData> > &outputQueue );
-
- // Output send engine, takes care of the host connections and the multiplexing.
- TABTranspose::MultiSender multiSender;
-
- // MPI rank for this node
- const unsigned hostID;
- };
- }
-}
-
-#endif
-
diff --git a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/UHEP_Pipeline.h b/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/UHEP_Pipeline.h
deleted file mode 100644
index 617d168b42be3e1324d90a882967b7e382608684..0000000000000000000000000000000000000000
--- a/RTCP/Cobalt/GPUProc/src/cuda/Pipelines/UHEP_Pipeline.h
+++ /dev/null
@@ -1,52 +0,0 @@
-//# UHEP_Pipeline.h
-//# Copyright (C) 2012-2013 ASTRON (Netherlands Institute for Radio Astronomy)
-//# P.O. Box 2, 7990 AA Dwingeloo, The Netherlands
-//#
-//# This file is part of the LOFAR software suite.
-//# The LOFAR software suite is free software: you can redistribute it and/or
-//# modify it under the terms of the GNU General Public License as published
-//# by the Free Software Foundation, either version 3 of the License, or
-//# (at your option) any later version.
-//#
-//# The LOFAR software suite is distributed in the hope that it will be useful,
-//# but WITHOUT ANY WARRANTY; without even the implied warranty of
-//# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-//# GNU General Public License for more details.
-//#
-//# You should have received a copy of the GNU General Public License along
-//# with the LOFAR software suite. If not, see <http://www.gnu.org/licenses/>.
-//#
-//# $Id$
-
-#ifndef LOFAR_GPUPROC_CUDA_UHEP_PIPELINE_H
-#define LOFAR_GPUPROC_CUDA_UHEP_PIPELINE_H
-
-#include <CoInterface/Parset.h>
-
-#include <GPUProc/gpu_wrapper.h>
-#include "Pipeline.h"
-#include <GPUProc/PerformanceCounter.h>
-
-namespace LOFAR
-{
- namespace Cobalt
- {
-
- class UHEP_Pipeline : public Pipeline
- {
- public:
- UHEP_Pipeline(const Parset &, Pool<struct MPIRecvData> &pool);
-
- void doWork();
-
- gpu::Module beamFormerProgram, transposeProgram, invFFTprogram, invFIRfilterProgram, triggerProgram;
-
- PerformanceCounter beamFormerCounter, transposeCounter, invFFTcounter, invFIRfilterCounter, triggerCounter;
- PerformanceCounter beamFormerWeightsCounter, samplesCounter;
- };
-
- }
-}
-
-#endif
-