https://akeelmedina22.github.io/tags/scc/Recent content in SCC on Akeel Ather MedinaHugoen-usThu, 10 Sep 2026 00:00:00 +0500https://akeelmedina22.github.io/posts/scc-2026-inference/Thu, 10 Sep 2026 00:00:00 +0500https://akeelmedina22.github.io/posts/scc-2026-inference/<p>As a participant representing TeamEPCC in the ISC Student Cluster Competition 2026, I&rsquo;ve recently had the chance to work on a custom-built HPC cluster that we shipped to Hamburg for the in-person competition. I wanted to give my overall reflections on the competition and my experience.</p> <p>The SCC has never included llama.cpp as a micro-benchmark, usually focusing on HPC-specific benchmarks like HPL and HPCG. This year introduced llama.cpp, which allows LLM inference with minimal setup on a wide range of hardware. It is a plain C/C++ implementation with no dependencies. It runs surprisingly well on Apple silicon! Understanding how to use llama.cpp is fairly simple, but maximizing performance for specific hardware could be a whole MSc thesis. The source code is very well written, and it is easy to find and read specific implementations of custom kernels. But the majority of performance comes from tuning the runtime parameters, so understanding the workload, and how those parameters change the character of your problem on your hardware, is what actually matters. That makes it quite different from the other benchmarks we were given, like HPL.</p>