{

__device__  __constant__  void *devInFramePtrs      [20];
__device__  __constant__  void *devOutFramePtrs     [20];
__device__  __constant__  void *devProcFramePtrs    [60];
__device__  __constant__  void *devProcOutFramePtrs [60];
__device__  __constant__ AlgorithmParms *devAlgoParmsPtr;
__device__  __constant__ AlgorithmStats *devStatParmsPtr;
__device__   float diamondOffsetsGlobal[36];

}

.global .align 4 .b8 devInFramePtrs[80];
.global .align 4 .b8 devOutFramePtrs[80];
.global .align 4 .b8 devProcFramePtrs[240];
.global .align 4 .b8 devProcOutFramePtrs[240];
.global .align 4 .u32 devAlgoParmsPtr;
.global .align 4 .u32 devStatParmsPtr;
.global .align 4 .b8 diamondOffsetsGlobal[144];

err = cudaMemcpyToSymbol("devInFramePtrs",  gDevInFramePtrs, sizeof(void *) * 20, 0, cudaMemcpyHostToDevice);
err = cudaMemcpyToSymbol("devOutFramePtrs", gDevOutFramePtrs, sizeof(void *) * 20, 0, cudaMemcpyHostToDevice);
err = cudaMemcpyToSymbol("devProcFramePtrs", gDevProcFramePtrs, sizeof(FRAME_BASE_TYPE *) * numDevInProcFramePtrs3, 0, cudaMemcpyHostToDevice);
err = cudaMemcpyToSymbol("devProcOutFramePtrs", gDevProcOutFramePtrs, sizeof(FRAME_BASE_TYPE *) * numDevOutProcFramePtrs3, 0, cudaMemcpyHostToDevice);
err = cudaMemcpyToSymbol("diamondOffsetsGlobal", &(diamondOffset[0][0]), sizeof(float) * 36, 0, cudaMemcpyHostToDevice);