successive over-relaxation factor changed from 0.70 to 0.55 (latest TINKER valueO

fixed problem w/ storage of PME induced dipoles

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

@@ -341,7 +341,7 @@ void kSorUpdateMutualInducedAndGkField_kernel(
                    float* matrixProduct, float* matrixProductP )
 {
 
-    float polarSOR = 0.70f;
+    float polarSOR = 0.55f;
     int pos        = blockIdx.x*blockDim.x + threadIdx.x;
     while( pos < 3*cSim.atoms )
     {
@@ -378,7 +378,7 @@ void kSorUpdateMutualInducedAndGkFieldS_kernel(
                    float* matrixProduct, float* matrixProductP )
 {
 
-    float polarSOR = 0.70f;
+    float polarSOR = 0.55f;
     int pos        = blockIdx.x*blockDim.x + threadIdx.x;
     while( pos < 3*cSim.atoms )
     {

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

@@ -206,7 +206,7 @@ void kSorUpdateMutualInducedField_kernel(
                    float* matrixProduct, float* matrixProductP )
 {
 
-    float polarSOR = 0.70f;
+    float polarSOR = 0.55f;
     int pos        = blockIdx.x*blockDim.x + threadIdx.x;
     while( pos < 3*cSim.atoms )
     {

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

@@ -376,7 +376,7 @@ static void kSorUpdateMutualInducedField_kernel(
 
     int pos                        = blockIdx.x*blockDim.x + threadIdx.x;
     const float term               = (4.0f/3.0f)*(cSim.alphaEwald*cSim.alphaEwald*cSim.alphaEwald)/cAmoebaSim.sqrtPi;
-    const float polarSOR           = 0.70f;
+    const float polarSOR           = 0.55f;
 
     while( pos < 3*cSim.atoms )
     {   
@@ -521,7 +521,8 @@ static void cudaComputeAmoebaPmeMutualInducedFieldBySOR( amoebaGpuContext amoeba
 
     while( !done ){
 
-        // matrix multiply
+        //  apply SOR
+
         cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply( amoebaGpu, amoebaGpu->psWorkVector[0],  amoebaGpu->psWorkVector[1] );
         kCalculateAmoebaPMEInducedDipoleField( amoebaGpu );
 

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

@@ -83,7 +83,7 @@ void METHOD_NAME(kCalculateAmoebaPmeMutualInducedField, _kernel)(
         fieldPolarSum[1]                 = 0.0f;
         fieldPolarSum[2]                 = 0.0f;
 
-        if (x == y ){
+        if ( x == y ){
 
             // load shared data
 
@@ -104,11 +104,13 @@ void METHOD_NAME(kCalculateAmoebaPmeMutualInducedField, _kernel)(
 
 #ifdef USE_OUTPUT_BUFFER_PER_WARP
             unsigned int offset            = 3*(x + tgx + warp*cSim.paddedNumberOfAtoms);
+            load3dArrayBufferPerWarp( offset, fieldSum,       outputField );
+            load3dArrayBufferPerWarp( offset, fieldPolarSum,  outputFieldPolar );
 #else
             unsigned int offset            = 3*(x + tgx + (x >> GRIDBITS) * cSim.paddedNumberOfAtoms);
-#endif
             load3dArray( offset, fieldSum,      outputField );
             load3dArray( offset, fieldPolarSum, outputFieldPolar);
+#endif
 
         } else {