https://sankethbk.github.io/blog/tags/profiling/Recent content in Profiling on Sanketh's BlogHugo -- 0.166.0en-usFri, 04 Sep 2026 00:00:00 +0000https://sankethbk.github.io/blog/posts/gpu/2026-09-04-2-cuda-profiling/Fri, 04 Sep 2026 00:00:00 +0000https://sankethbk.github.io/blog/posts/gpu/2026-09-04-2-cuda-profiling/<h1 id="cuda-profiling-and-performance-measurement">CUDA Profiling and Performance Measurement</h1> <p>The notes up to this point covered what makes CUDA kernels fast in theory: coalescing, tiling, occupancy, and asynchronous execution. This note covers how to <strong>measure</strong> whether a kernel is actually fast and how to find the bottleneck when it is not.</p> <p>Profiling is what separates guessing from optimization. Without it, you can apply every best practice and still be slower than a simpler implementation.</p>