ECE498ALProgrammingMassivelyParallel.ppt

ECE 498ALProgramming Massively Parallel ProcessorsLecture 6: CUDA MemoriesPart 2;Md;Pd1,0;Every Md and Nd Element is used exactly twice in generating a 2X2 tile of P;Pd1,0;Each phase of a Thread Block uses one tile from Md and one from Nd;Tiled Matrix Multiplication Kernel;CUDA Code – Kernel Execution Configuration;First-order Size Considerations in G80;Md;G80 Shared Memory and Threading;Tiled Matrix Multiplication Kernel;Tiling Size Effects;Global variables declaration __host__ __device__... __global__, __constant__, __texture__ Function prototypes __global__ void kernelOne(…) float handyFunction(…) Main () allocate memory space on the device – cudaMalloc(d_GlblVarPtr, bytes ) transfer data from host to device – cudaMemCpy(d_GlblVarPtr, h_Gl…) execution configuration setup kernel call – kernelOneexecution configuration( args… ); transfer results from device to host – cudaMemCpy(h_GlblVarPtr,…) optional: compare against golden (host computed) solution Kernel – void kernelOne(type args,…) variables declaration - __local__, __shared__ automatic variables transparently assigned to registers or local memory syncthreads()… Other functions float handyFunction(int inVar…);