> cat ./posts/*.md
Parallel programming notes
Everything I'm learning while parallelizing algorithms for multi-GPU environments — thesis progress on heavy weight perfect matching for bipartite graphs, and write-ups from LeetGPU problems.
- [leetgpu]
Sample 1 - LeetGPU — reduction, from shared memory to warp shuffles
Four iterations of a sum reduction kernel and the bandwidth each one actually reaches.
- [thesis]
Sample 2 - Baseline auction bidding kernel — first numbers
A naive one-thread-per-vertex bidding kernel, where it stalls, and the coalescing fix that followed.
- [thesis]
Sample 3 - Thesis kickoff — heavy weight perfect matching on the GPU
Framing the problem, why the sequential auction/Hungarian family stalls, and the target for a GPU implementation.