https://akurniawan.github.io/categories/systems/Recent content in Systems on Aditya Kurniawan (Akur)Hugoen-usMon, 14 Sep 2026 00:00:00 +0700https://akurniawan.github.io/posts/return-std-move-costs-instructions-not-cycles/Mon, 14 Sep 2026 00:00:00 +0700https://akurniawan.github.io/posts/return-std-move-costs-instructions-not-cycles/I wanted to understand named return value optimisation, got curious what the pessimised version costs, and my first numbers were reproducible to 0.01% and wrong. Instructions and cycles instead of wall-clock, and the pointer question turned out to be an allocator question.https://akurniawan.github.io/posts/how-i-benchmark/Sun, 13 Sep 2026 00:00:00 +0700https://akurniawan.github.io/posts/how-i-benchmark/The same benchmark landed at 3.47 ms and 7.01 ms on different afternoons while reporting under 0.3% variation both times. These are the habits that survived that.https://akurniawan.github.io/posts/nvlink-c2c-what-changes-for-cpp/Sun, 06 Sep 2026 00:00:00 +0700https://akurniawan.github.io/posts/nvlink-c2c-what-changes-for-cpp/The CUDA Programming Guide assumes the host and the device keep separate memories with a copy between them. My DGX Spark doesn’t. So I went and learned what NVLink-C2C actually is compared to PCIe, how coherence behaves across it, and what that changes for someone writing C++ on this machine.https://akurniawan.github.io/posts/gpu-timeline-gaps/Sun, 16 Aug 2026 00:00:00 +0700https://akurniawan.github.io/posts/gpu-timeline-gaps/A program can report 98.8% device utilization while leaving its compute units idle 39% of the time. This is what I learned building a TUI for nsys timelines, and why the second number is the only one that matters.