cuda_presentation.pptVIP

MY STEPS INTO CUDA;Hello World! with Device Code;Function Type Qualifiers;Heterogeneous;Kernel launch;Addition on the Device;Heterogeneous Computing;Simple Processing Flow;Simple Processing Flow;Simple Processing Flow;Addition on the Device: main();Addition on the Device: main();Moving to Parallel Blocks ;C(0,0);Vector Addition on the Device;Vector Addition on the Device;Vector Addition on the Device: main();Vector Addition on the Device: main();__global__ void add(int *a, int *b, int *c) { c[blockIdx.x] = a[blockIdx.x] + b[blockIdx.x]; };add 1, 1 (); add N, 1 (); add 1, N (); add M, N (); ;;Which thread will operate on the red element?;Use the built-in variable blockDim.x for threads per block int index = threadIdx.x + blockIdx.x * blockDim.x; Combined version of add() to use parallel threads and parallel blocks;Addition with Blocks and Threads: main();Addition with Blocks and Threads: main(); Avoid accessing beyond the end of the arrays:;Block (0,0);Memory hierarchy;Shared memory and local memory;Registers The fastest form of memory on the multi-processor. Is only accessible by the thread. Has the lifetime of the thread. Shared Memory Can be as fast as a register when there are no bank conflicts or when reading from the same address. Accessible by any thread of the block from which it was created. Has the lifetime of the block. Global memory Potentially 150x slower than register or shared memory Accessible from either the host or device. Has the lifetime of the application. Local memory Resides in global memory and can be 150x slower than register or shared memory. Is only accessible by the thread. Has the lifetime of the thread. ;Topics we skipped;CUDA programming guide /article/15/article-15244-1.html /Download/index.aspx?lang=en-us (notebook driver download) /cuda-downloads (toolkit、sdk download);Thank you!