https://pablodelgado.org/tags/performance/Recent content in Performance on Share what you knowHugoen-usSat, 25 Apr 2026 12:00:00 +0000https://pablodelgado.org/posts/2026/04/25/first-post/Sat, 25 Apr 2026 12:00:00 +0000https://pablodelgado.org/posts/2026/04/25/first-post/<p>In this post, I&rsquo;ll iteratively optimize an implementation of matrix multiplication written in CUDA.<label for="sn-44602fdfdbc819199a0a4d6fcb203a4d-0" class="margin-toggle sidenote-number"></label> <input type="checkbox" id="sn-44602fdfdbc819199a0a4d6fcb203a4d-0" class="margin-toggle"/> <span class="sidenote">This post is a recreation of <a href="https://siboehm.com/articles/22/CUDA-MMM">Si Boehm&rsquo;s excellent worklog</a>, used here as a template to demonstrate Tufte-CSS features: side notes, margin notes, side figures, full-width figures, code blocks, and references.</span> My goal is not to build a cuBLAS replacement, but to deeply understand the most important performance characteristics of the GPUs that are used for modern deep learning. This includes coalescing global memory accesses, shared memory caching and occupancy optimizations, among others.</p>