Back to Browse

32d Nvidia GPU ISA and DAXPY loop implementation

39 views
Feb 28, 2026
8:29

Nvidia GPU Instruction Set Architecture (ISA) is split into two layers: a virtual ISA (PTX) and a native hardware ISA (SASS). This dual-layered approach allows code to remain compatible across different GPU generations while still optimizing for specific hardware. In the context of high-performance computing and GPU programming, DAXPY is a standard linear algebra operation used as a benchmark for performance and memory bandwidth We cover how the ISA can be used to optimise DAXPY performance #gpu #engineering #placement Watch at 1.5x-2x speed

Download

1 formats

Video Formats

360pmp47.4 MB

Right-click 'Download' and select 'Save Link As' if the file opens in a new tab.

32d Nvidia GPU ISA and DAXPY loop implementation | NatokHD