University:
Boston University
Course:
GRS CS 910 | Directed Study: Systems
Academic year:

2023
Views:

376

Pages:

13
Author:

rachedelgrecomhdd

Parallel thinking ‣ Latency? Throughput? Execution time - Latency: time to complete a task - Throughput: number of tasks in a fixed time Bandwidth ‣ Decisions!?! - Depends on the problem “If you were plowing a field, which would you rather use: two strong oxen or 1024 chicken?” Seymour Cray Parallel thinking ‣ 1024 chicken? We better have a strategy! - Rethink our algorithms to be more parallel friendly - Massively parallel: ‣ Data parallelism ‣ Load balancing ‣ Regular computations ‣ Data access ‣ Avoid conflicts ‣ and so on... Computer Unified Device Architecture (CUDA) ‣ Parallel computer architecture developed by NVIDIA ‣ General purpose programming model ‣ Specially designed for General Purpose GPU computing: - Offers a compute designed API - Explicit GPU memory managing CUDA enabled GPUs ‣ What is the difference between GPUs - CUDA enabled GPUs - GPUs classified according to compute capability CUDA Programming Guide Appendix A CUDA Programming Guide Appendix F CUDA - Main features ‣ C/C++ with extensions ‣ Heterogeneous programming model: - Operates in CPU (host) and GPU (device) Host CPU Device GPU CUDA - Threads ‣ In CUDA, a kernel is executed by many threads - A thread is a sequence of executions - Multi-thread: many threads will be running at the same time __global__ void vec_add (float *A, float *B, float *C, int N){ int i = threadIdx.x + blockDim.x*blockIdx.x; if (i>=N) {return;} C[i] = A[i] + B[i]; } void vec_add (float *A, float *B, float *C, int N){ for (int i=0; i>> (); // Launch kernel ‣ Others threadIdx.x blockIdx.x CUDA - Program execution Allocate and initialize data on CPU Allocate data on GPU Transfer data from CPU to GPU Run kernel Transfer data from GPU to CPU CUDA - Vector add example __global__ void vec_add (float *A, float *B, float *C, int N) { // Using 1D threadID and block ID int i = threadIdx.x + blockDim.x*blockIdx.x; if (i>=N) {return;} C[i] = A[i] + B[i]; } int main() { .... // Launch kernel with N/256 blocks of 256 threads int blocks = int(N‐0.5)/256 + 1; vec_add<<>> (A_d, B_d, C_d, N); }