Remove nonbondElectrostaticThreadsPerBlock, energyOutputBuffers

Set nonbondThreadsPerBlock to gpu->sim.nonbond_threads_per_block

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

@@ -195,13 +195,16 @@ void gpuPrintCudaAmoebaGmxSimulation(amoebaGpuContext amoebaGpu, FILE* log )
     (void) fprintf( log, "     sm_version                         %u\n",      gpu->sm_version );
     (void) fprintf( log, "     device                             %u\n",      gpu->device );
     (void) fprintf( log, "     sharedMemoryPerBlock               %u\n",      gpu->sharedMemoryPerBlock );
-    (void) fprintf( log, "     bOutputBufferPerWarp               %d\n",      amoebaGpu->bOutputBufferPerWarp );
+    (void) fprintf( log, "     amoebaBOutputBufferPerWarp         %d\n",      amoebaGpu->bOutputBufferPerWarp );
+    (void) fprintf( log, "     bOutputBufferPerWarp               %d\n",      gpu->bOutputBufferPerWarp );
+    //`amoebaGpu->nonbondOutputBuffers           = (amoebaGpu->nonbondBlocks*amoebaGpu->nonbondThreadsPerBlock)/GRID;
     (void) fprintf( log, "     paddedNumberOfAtoms                %u\n",      amoebaGpu->paddedNumberOfAtoms );
-    (void) fprintf( log, "     nonbondBlocks                      %u\n",      amoebaGpu->nonbondBlocks );
-    (void) fprintf( log, "     nonbondThreadsPerBlock             %u\n",      amoebaGpu->nonbondThreadsPerBlock );
-    (void) fprintf( log, "     nonbondElectrostaticThreadsPerBlock%u\n",      amoebaGpu->nonbondElectrostaticThreadsPerBlock );
-    (void) fprintf( log, "     nonbondOutputBuffers               %u\n",      amoebaGpu->nonbondOutputBuffers );
-    (void) fprintf( log, "     energyOutputBuffers                %u\n",      amoebaGpu->energyOutputBuffers );
+    (void) fprintf( log, "     amoebaNonbondBlocks                %u\n",      amoebaGpu->nonbondBlocks );
+    (void) fprintf( log, "     nonbondBlocks                      %u\n",      gpu->sim.nonbond_blocks );
+    (void) fprintf( log, "     amoebaNonbondThreadsPerBlock       %u\n",      amoebaGpu->nonbondThreadsPerBlock );
+    (void) fprintf( log, "     nonbondThreadsPerBlock             %u\n",      gpu->sim.nonbond_threads_per_block);
+    (void) fprintf( log, "     amoebaNonbondOutputBuffers         %u\n",      amoebaGpu->nonbondOutputBuffers );
+    (void) fprintf( log, "     nonbondOutputBuffers               %u\n",      gpu->sim.nonbondOutputBuffers );
     (void) fprintf( log, "     threadsPerBlock                    %u\n",      amoebaGpu->threadsPerBlock );
     (void) fprintf( log, "     fieldReduceThreadsPerBlock         %u\n",      amoebaGpu->fieldReduceThreadsPerBlock );
     (void) fprintf( log, "     outputBuffers                      %u\n",      amoebaGpu->outputBuffers );
@@ -2902,11 +2905,7 @@ void amoebaGpuBuildOutputBuffers( amoebaGpuContext amoebaGpu )
     
     amoebaGpu->nonbondBlocks                        = amoebaGpu->gpuContext->sim.blocks;
     amoebaGpu->threadsPerBlock                      = amoebaGpu->gpuContext->sim.threads_per_block;
-
-    // nonbondThreadsPerBlock & nonbondElectrostaticThreadsPerBlock need to be multiples of 32
-
-    amoebaGpu->nonbondThreadsPerBlock               = 192;
-    amoebaGpu->nonbondElectrostaticThreadsPerBlock  = 128;
+    amoebaGpu->nonbondThreadsPerBlock               = amoebaGpu->gpuContext->sim.nonbond_threads_per_block;
 
     amoebaGpu->fieldReduceThreadsPerBlock           = (amoebaGpu->paddedNumberOfAtoms*3 + amoebaGpu->gpuContext->natoms + amoebaGpu->nonbondBlocks - 1) / amoebaGpu->nonbondBlocks;
     amoebaGpu->fieldReduceThreadsPerBlock           = ((amoebaGpu->fieldReduceThreadsPerBlock + (amoebaGpu->gpuContext->grid - 1)) /  amoebaGpu->gpuContext->grid) *  amoebaGpu->gpuContext->grid;
@@ -2927,16 +2926,13 @@ void amoebaGpuBuildOutputBuffers( amoebaGpuContext amoebaGpu )
         amoebaGpu->nonbondOutputBuffers           = paddedNumberOfAtoms/GRID;
     }
     amoebaGpu->outputBuffers           = amoebaGpu->nonbondOutputBuffers;
-    //amoebaGpu->energyOutputBuffers     = max(amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->sim.localForces_threads_per_block)*amoebaGpu->sim.blocks;
-    amoebaGpu->energyOutputBuffers     = amoebaGpu->nonbondThreadsPerBlock*amoebaGpu->nonbondBlocks;
  
     if( amoebaGpu->log ){
         (void) fprintf(  amoebaGpu->log, "amoebaGpuBuildOutputBuffers: bOutputBufferPerWarp=%u nonbondBuffers=%u "
-                         "outputBuffers=%u energyBuffers=%d nonbondBlocks=%u fieldReduceThreadsPerBlock=%u\n",
+                         "outputBuffers=%u onbondBlocks=%u fieldReduceThreadsPerBlock=%u\n",
                          amoebaGpu->bOutputBufferPerWarp,
                          amoebaGpu->nonbondOutputBuffers,
                          amoebaGpu->outputBuffers,
-                         amoebaGpu->energyOutputBuffers,
                          amoebaGpu->nonbondBlocks,
                          amoebaGpu->fieldReduceThreadsPerBlock ); 
         (void) fflush( amoebaGpu->log );

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

@@ -56,9 +56,7 @@ struct _amoebaGpuContext {
     unsigned int paddedNumberOfAtoms;
     unsigned int nonbondBlocks;
     unsigned int nonbondThreadsPerBlock;
-    unsigned int nonbondElectrostaticThreadsPerBlock;
     unsigned int nonbondOutputBuffers;
-    unsigned int energyOutputBuffers;
     unsigned int threadsPerBlock;
     unsigned int fieldReduceThreadsPerBlock;
     unsigned int outputBuffers; 

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

@@ -777,9 +777,9 @@ void cudaComputeAmoebaElectrostatic( amoebaGpuContext amoebaGpu )
 
     if (gpu->bOutputBufferPerWarp){
 
-        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaElectrostaticN2Forces warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%lu shrd=%lu Ebuf=%u ixnCt=%lu workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaElectrostaticN2Forces warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%lu shrd=%lu ixnCt=%lu workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(ElectrostaticParticle), sizeof(ElectrostaticParticle)*threadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits ); (void) fflush( amoebaGpu->log );
+                        sizeof(ElectrostaticParticle), sizeof(ElectrostaticParticle)*threadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits ); (void) fflush( amoebaGpu->log );
 
 
         kCalculateAmoebaCudaElectrostaticN2ByWarpForces_kernel<<<amoebaGpu->nonbondBlocks, threadsPerBlock, sizeof(ElectrostaticParticle)*threadsPerBlock>>>(
@@ -800,9 +800,9 @@ void cudaComputeAmoebaElectrostatic( amoebaGpuContext amoebaGpu )
     } else {
 
 #ifdef AMOEBA_DEBUG
-        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaElectrostaticN2Forces no warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaElectrostaticN2Forces no warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u xnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(ElectrostaticParticle), sizeof(ElectrostaticParticle)*threadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        sizeof(ElectrostaticParticle), sizeof(ElectrostaticParticle)*threadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
 

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

@@ -430,9 +430,9 @@ void cudaComputeAmoebaFixedEAndGkFields( amoebaGpuContext amoebaGpu )
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "N2 no warp\n" );
-        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*threadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*threadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
 
@@ -472,9 +472,9 @@ void cudaComputeAmoebaFixedEAndGkFields( amoebaGpuContext amoebaGpu )
 #ifdef AMOEBA_DEBUG
     if( amoebaGpu->log ){
         gpu->psInteractionCount->Download();
-        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 
         amoebaGpu->psWorkArray_3_1->Download();

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

@@ -116,9 +116,9 @@ void cudaComputeAmoebaFixedEField( amoebaGpuContext amoebaGpu )
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "N2 no warp\n" );
-        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
 
@@ -154,9 +154,9 @@ void cudaComputeAmoebaFixedEField( amoebaGpuContext amoebaGpu )
 #ifdef AMOEBA_DEBUG
     if( amoebaGpu->log ){
         gpu->psInteractionCount->Download();
-        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
         amoebaGpu->psWorkArray_3_1->Download();
         amoebaGpu->psWorkArray_3_2->Download();

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

@@ -1895,10 +1895,10 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
 
         if( amoebaGpu->log ){
 
-            (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaKirkwood: blcks=%u tds=%u %u bPrWrp=%u atm=%lu shrd=%lu Ebuf=%u ixnCt=%lu workUnits=%u\n",
+            (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaKirkwood: blcks=%u tds=%u %u bPrWrp=%u atm=%lu shrd=%lu ixnCt=%lu workUnits=%u\n",
                             amoebaGpu->nonbondBlocks, threadsPerBlock, maxThreads, amoebaGpu->bOutputBufferPerWarp,
                             sizeof(KirkwoodParticle), sizeof(KirkwoodParticle)*threadsPerBlock,
-                            amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                            (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
             (void) fflush( amoebaGpu->log );
         }
     }
@@ -1917,10 +1917,10 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
     } else {
 
 #ifdef AMOEBA_DEBUG
-        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaKirkwoodN2Forces no warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaKirkwoodN2Forces no warp:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(KirkwoodParticle), sizeof(KirkwoodParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
 

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

@@ -1064,10 +1064,10 @@ void kCalculateAmoebaKirkwoodEDiff( amoebaGpuContext amoebaGpu )
 #ifdef AMOEBA_DEBUG
     if( amoebaGpu->log && timestep == 1 ){
         (void) fprintf( amoebaGpu->log, "kCalculateAmoebaCudaKirkwoodEDiffN2Forces: blocks=%u threads=%u bffr/Warp=%u atm=%lu shrd=%lu"
-                                        " Ebuf=%u ixnCt=%lu workUnits=%u sm=%d device=%d sharedMemoryPerBlock=%u\n",
+                                        " ixnCt=%lu workUnits=%u sm=%d device=%d sharedMemoryPerBlock=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(KirkwoodEDiffParticle), sizeof(KirkwoodEDiffParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits, gpu->sm_version, gpu->device, gpu->sharedMemoryPerBlock );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits, gpu->sm_version, gpu->device, gpu->sharedMemoryPerBlock );
         (void) fflush( amoebaGpu->log );
     }   
 #endif

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

@@ -544,10 +544,10 @@ static void cudaComputeAmoebaMutualInducedAndGkFieldMatrixMultiply( amoebaGpuCon
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "N2 no warp\n" );
-        (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaMutualInducedAndGkFieldMatrixMultiply numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaMutualInducedAndGkFieldMatrixMultiply numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(MutualInducedParticle), sizeof(MutualInducedParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
 

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

@@ -273,10 +273,10 @@ static void cudaComputeAmoebaMutualInducedFieldMatrixMultiply( amoebaGpuContext
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "N2 no warp\n" );
-        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaN2Forces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         amoebaGpu->nonbondBlocks, amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(MutualInducedParticle), sizeof(MutualInducedParticle)*amoebaGpu->nonbondThreadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
         kCalculateAmoebaMutualInducedFieldN2_kernel<<<amoebaGpu->nonbondBlocks, amoebaGpu->nonbondThreadsPerBlock, sizeof(MutualInducedParticle)*amoebaGpu->nonbondThreadsPerBlock>>>(

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

@@ -1238,10 +1238,10 @@ void cudaComputeAmoebaPmeDirectElectrostatic( amoebaGpuContext amoebaGpu )
                     threadsPerBlock, getThreadsPerBlock(amoebaGpu, sizeof(PmeDirectElectrostaticParticle)),
                     sizeof(PmeDirectElectrostaticParticle) );
 
-      (void) fprintf( amoebaGpu->log, "kCalculateAmoebaPmeDirectElectrostaticCutoffForces:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Obuf=%u ixnCt=%u workUnits=%u gpu->nonbond_threads_per_block=%u\n",
+      (void) fprintf( amoebaGpu->log, "kCalculateAmoebaPmeDirectElectrostaticCutoffForces:  numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u gpu->nonbond_threads_per_block=%u\n",
                       amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                       sizeof(PmeDirectElectrostaticParticle), (sizeof(PmeDirectElectrostaticParticle))*threadsPerBlock,
-                      amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits, gpu->sim.nonbond_threads_per_block );
+                      (*gpu->psInteractionCount)[0], gpu->sim.workUnits, gpu->sim.nonbond_threads_per_block );
       (void) fflush( amoebaGpu->log );
 #endif
 
@@ -1314,18 +1314,6 @@ void cudaComputeAmoebaPmeDirectElectrostatic( amoebaGpuContext amoebaGpu )
             }
         }
         (void) fflush( amoebaGpu->log );
-/*
-        gpu->psEnergy->Download();
-        double energy = 0.0;
-        for( unsigned int ii = 0; ii < gpu->sim.energyOutputBuffers; ii++ ){
-            if( (*gpu->psEnergy)[ii] !=  (*gpu->psEnergy)[ii] || (*gpu->psEnergy)[ii] == std::numeric_limits<double>::infinity() || (*gpu->psEnergy)[ii] == -std::numeric_limits<double>::infinity() ){
-                (void) fprintf( amoebaGpu->log,"Energy nan at index=%d\n", ii );
-            } else {
-               energy += (*gpu->psEnergy)[ii];
-            }
-        }
-        (void) fprintf( amoebaGpu->log,"Force sums: [%16.9e %16.9e %16.9e] Energy=%16.9e\n", forceSum[0], forceSum[1], forceSum[2], energy );
-*/
 
         if( 0 ){
             (void) fprintf( amoebaGpu->log,"DebugElec\n" );

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

@@ -468,9 +468,9 @@ static void cudaComputeAmoebaPmeDirectFixedEField( amoebaGpuContext amoebaGpu )
         (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaPmeDirectFixedEField:  threadsPerBlock=%u getThreadsPerBlock=%d sizeof=%u shrd=%u\n", 
                         threadsPerBlock, getThreadsPerBlock(amoebaGpu, sizeof(FixedFieldParticle)+sizeof(float3)),
                         (sizeof(FixedFieldParticle)+sizeof(float3)), (sizeof(FixedFieldParticle)+sizeof(float3))*threadsPerBlock );
-        (void) fprintf( amoebaGpu->log, "AmoebaCutoffForces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u warp=%d\n",
+        (void) fprintf( amoebaGpu->log, "AmoebaCutoffForces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u warp=%d\n",
                         amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
-                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock, amoebaGpu->energyOutputBuffers, 
+                        sizeof(FixedFieldParticle), sizeof(FixedFieldParticle)*amoebaGpu->nonbondThreadsPerBlock,
                         (*gpu->psInteractionCount)[0], gpu->sim.workUnits, gpu->bOutputBufferPerWarp );
         (void) fflush( amoebaGpu->log );
 /*

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

@@ -389,7 +389,7 @@ static void cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply( amoebaGpuConte
         (void) fflush( amoebaGpu->log );
     }
     int paddedNumberOfAtoms                    = amoebaGpu->gpuContext->sim.paddedNumberOfAtoms;
-    int maxSlots                               = amoebaGpu->gpuContext->sim.paddedNumberOfAtoms;
+    int maxSlots                               = 10;
     CUDAStream<float4>* debugArray             = new CUDAStream<float4>(maxSlots*paddedNumberOfAtoms, 1, "DebugArray");
     memset( debugArray->_pSysData,      0, sizeof( float )*4*maxSlots*paddedNumberOfAtoms);
     debugArray->Upload();
@@ -414,10 +414,10 @@ static void cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply( amoebaGpuConte
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "Cutoff -- use warp\n" );
-        (void) fprintf( amoebaGpu->log, "AmoebaCutoffForces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
-                        amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
+        (void) fprintf( amoebaGpu->log, "%s numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
+                        methodName, amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(MutualInducedParticle), sizeof(MutualInducedParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
                                                                  //gpu->sim.pInteractingWorkUnit,
@@ -436,10 +436,10 @@ static void cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply( amoebaGpuConte
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "Cutoff no warp\n" );
-        (void) fprintf( amoebaGpu->log, "AmoebaCutoffForces_kernel numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
-                        amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
+        (void) fprintf( amoebaGpu->log, "%s numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
+                        methodName, amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(MutualInducedParticle), sizeof(MutualInducedParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif
         kCalculateAmoebaPmeMutualInducedFieldCutoff_kernel<<<amoebaGpu->nonbondBlocks, threadsPerBlock, sizeof(MutualInducedParticle)*threadsPerBlock>>>(

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

@@ -123,6 +123,7 @@ void METHOD_NAME(kCalculateAmoebaPmeMutualInducedField, _kernel)(
                 fieldPolarSum[2]       += mask ? ijField[2].z : 0.0f;
 
 #ifdef AMOEBA_DEBUG
+/*
 if( atomI == targetAtom || (y+j) == targetAtom ){
             unsigned int index                 = atomI == targetAtom ? (y+j) : atomI;
             unsigned int pullBackIndex         = 0;
@@ -190,6 +191,7 @@ if( atomI == targetAtom || (y+j) == targetAtom ){
             debugArray[index].w                = + 10.0f;
 
 }
+*/
 #endif
 
             }

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

@@ -434,6 +434,7 @@ void cudaComputeAmoebaLabFrameMoments( amoebaGpuContext amoebaGpu )
                              amoebaGpu->psMolecularDipole->_pSysData[mIndex],
                              amoebaGpu->psMolecularDipole->_pSysData[mIndex+1],
                              amoebaGpu->psMolecularDipole->_pSysData[mIndex+2] );
+              if( ii == 30 )ii = gpu->natoms - 30;
         }
     }
 //    int64 kernelTime = AmoebaTiming::getTimeOfDay();
@@ -459,7 +460,7 @@ void cudaComputeAmoebaLabFrameMoments( amoebaGpuContext amoebaGpu )
         amoebaGpu->psLabFrameQuadrupole->Download();
         (void) fprintf( amoebaGpu->log, "psLabFrameQpole completed\n" );  (void) fflush( amoebaGpu->log );
 
-        int maxPrint = 10;
+        int maxPrint = 20;
         for( int ii = 0; ii < amoebaGpu->paddedNumberOfAtoms; ii++ ){
 
              int dipoleOffset     = 3*ii;

plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaVdw14_7.cu

@@ -525,10 +525,10 @@ void kCalculateAmoebaVdw14_7Forces( amoebaGpuContext amoebaGpu, int applyCutoff
 
 #ifdef AMOEBA_DEBUG_PRINT
     (void) fprintf( amoebaGpu->log, "Apply cutoff=%d warp=%d\n", applyCutoff, gpu->bOutputBufferPerWarp );
-    (void) fprintf( amoebaGpu->log, "numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+    (void) fprintf( amoebaGpu->log, "numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                     amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                     sizeof(Vdw14_7Particle), sizeof(Vdw14_7Particle)*threadsPerBlock,
-                    amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                    (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
     if( 0 ){  
         gpu->psInteractionCount->Download();
         amoebaGpu->psVdwWorkUnit->Download();

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

@@ -434,10 +434,10 @@ void kCalculateAmoebaWcaDispersionForces( amoebaGpuContext amoebaGpu )
 
 #ifdef AMOEBA_DEBUG
         (void) fprintf( amoebaGpu->log, "N2 no warp\n" );
-        (void) fprintf( amoebaGpu->log, "%s numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u Ebuf=%u ixnCt=%u workUnits=%u\n",
+        (void) fprintf( amoebaGpu->log, "%s numBlocks=%u numThreads=%u bufferPerWarp=%u atm=%u shrd=%u ixnCt=%u workUnits=%u\n",
                         methodName, amoebaGpu->nonbondBlocks, threadsPerBlock, amoebaGpu->bOutputBufferPerWarp,
                         sizeof(WcaDispersionParticle), sizeof(WcaDispersionParticle)*threadsPerBlock,
-                        amoebaGpu->energyOutputBuffers, (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
+                        (*gpu->psInteractionCount)[0], gpu->sim.workUnits );
         (void) fflush( amoebaGpu->log );
 #endif