Reduced memory use

plugins/amoeba/platforms/cuda/src/AmoebaCudaData.cpp

@@ -111,9 +111,6 @@ void AmoebaCudaData::initializeGpu( void ) {
         if( getHasAmoebaGeneralizedKirkwood() && !getHasAmoebaMultipole() ){
             throw OpenMMException("GK force requires Multipole force\n");
         }
-        amoebaGpuBuildOutputBuffers( amoebaGpu );
-        amoebaGpuBuildThreadBlockWorkList( amoebaGpu );
-        amoebaGpuBuildScalingList( amoebaGpu );
         amoebaGpuSetConstants( amoebaGpu );
         gpuInitialized = true;
         if( log ){

plugins/amoeba/platforms/cuda/src/kernels/AmoebaGpu.cpp

@@ -213,8 +213,6 @@ void gpuPrintCudaAmoebaGmxSimulation(amoebaGpuContext amoebaGpu, FILE* log )
     gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_3_2, log );
     gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_3_3, log );
     gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_3_4, log );
-    gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_3_5, log );
-    gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_3_6, log );
 
     gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_1_1, log );
     gpuPrintCudaStreamFloat(  amoebaGpu->psWorkArray_1_2, log );
@@ -2689,8 +2687,6 @@ void amoebaGpuShutDown(amoebaGpuContext gpu)
     delete gpu->psWorkArray_3_2; 
     delete gpu->psWorkArray_3_3; 
     delete gpu->psWorkArray_3_4; 
-    delete gpu->psWorkArray_3_5; 
-    delete gpu->psWorkArray_3_6; 
 
     delete gpu->psWorkArray_1_1; 
     delete gpu->psWorkArray_1_2; 
@@ -2824,8 +2820,6 @@ void amoebaGpuBuildOutputBuffers( amoebaGpuContext amoebaGpu )
     // used GK
     amoebaGpu->psWorkArray_3_3            = new CUDAStream<float>(3*paddedNumberOfAtoms, (amoebaGpu->outputBuffers), "AmoebaField_3_3");
     amoebaGpu->psWorkArray_3_4            = new CUDAStream<float>(3*paddedNumberOfAtoms, (amoebaGpu->outputBuffers), "AmoebaField_3_4");
-    amoebaGpu->psWorkArray_3_5            = new CUDAStream<float>(3*paddedNumberOfAtoms, (amoebaGpu->outputBuffers), "AmoebaField_3_5");
-    amoebaGpu->psWorkArray_3_6            = new CUDAStream<float>(3*paddedNumberOfAtoms, (amoebaGpu->outputBuffers), "AmoebaField_3_6");
 
     amoebaGpu->psWorkArray_1_1            = new CUDAStream<float>(  paddedNumberOfAtoms, (amoebaGpu->outputBuffers), "AmoebaField_1_1");
     amoebaGpu->amoebaSim.pWorkArray_1_1   = amoebaGpu->psWorkArray_1_1->_pDevStream[0];

plugins/amoeba/platforms/cuda/src/kernels/amoebaGpuTypes.h

@@ -70,8 +70,6 @@ struct _amoebaGpuContext {
     CUDAStream<float>*  psWorkArray_3_2; 
     CUDAStream<float>*  psWorkArray_3_3; 
     CUDAStream<float>*  psWorkArray_3_4; 
-    CUDAStream<float>*  psWorkArray_3_5; 
-    CUDAStream<float>*  psWorkArray_3_6; 
 
     CUDAStream<float>*  psWorkArray_1_1; 
     CUDAStream<float>*  psWorkArray_1_2; 

plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaUtilities.cu

@@ -100,16 +100,6 @@ void kClearFields_3( amoebaGpuContext amoebaGpu, unsigned int numberToClear )
                                                                 amoebaGpu->psWorkArray_3_4->_pDevStream[0] );
         LAUNCHERROR("kClearFields_3_4");
     }
-    if( numberToClear > 4 ){
-        kClearFields_kernel<<<amoebaGpu->nonbondBlocks, 384>>>( amoebaGpu->paddedNumberOfAtoms*3*amoebaGpu->outputBuffers,
-                                                                amoebaGpu->psWorkArray_3_5->_pDevStream[0] );
-        LAUNCHERROR("kClearFields_3_5");
-
-        kClearFields_kernel<<<amoebaGpu->nonbondBlocks, 384>>>( amoebaGpu->paddedNumberOfAtoms*3*amoebaGpu->outputBuffers,
-                                                                amoebaGpu->psWorkArray_3_6->_pDevStream[0] );
-        LAUNCHERROR("kClearFields_3_6");
-    }
-
 }
 
 // clear psWorkArray_1_1 & psWorkArray_1_2