Sumários
Memory bank conflicts + Advanced GPU algorithms
30 Novembro 2023, 16:30 • Alcides Fonseca
- Memory Bank Conflicts
- Thread Divergence (review)
- Shared memory usage (review)
- Advanced Algorithms: Reduction, Prefix Sum, Filter, QuickSort
Test + Guest Lecture
23 Novembro 2023, 16:30 • Alcides Fonseca
- Test #3
- Guest Lecture by João Reis, from Cloudflare
GPGPU
16 Novembro 2023, 16:30 • Alcides Fonseca
- The GPU hardware model (Multiprocessors and Warps)
- The GPU Memory model (Global memory, Shared Memory and Local/Private memory)
- Threads and Blocks, how to choose the 3D grid and block size. Dynamically loading the GPU properties.
- Block-wide barriers with __syncthreads;
- Atomics on the GPU
- How to make use of Shared Memory to reduce conflicts and have faster access times.
- OpenCL
- Write GPGPU code from Java (JavaCL) and Python (PyCuda and Numba)