Notes on systems, speed, & the craft of software.
I'm Abdul Basith. I write long-form pieces about performance engineering, machine learning infrastructure, and whatever else I'm taking apart to understand.
Writing
- When Doubling the Math Costs Nothing: Overhead-Bound Inference in Robot ModelsI switched a robot model from 16-bit to 32-bit and it didn’t get slower. That one result explains why small robot models don’t speed up on bigger GPUs, why lower precision can hurt, and why the same GPU ran 1.65× slower on different machines.
- Thirty-two threads, one instructionA ground-up introduction to CUDA for programmers who have never written a kernel, where every number was measured on a real GPU.