Skip to content

Radar ·

SASS latency and instruction reordering on GPUs

Published on November 25, 2025, the post questions the usefulness of SASS latency tables extracted with nvdisasm and says instruction reordering may improve performance by 3–4%, against a theoretical ceiling of 10%.

On November 25, 2025, a post about GPU kernel optimization argued that SASS latency tables extracted with nvdisasm are not useful. It also said instruction reordering may improve performance by 3–4%, with a theoretical ceiling of 10%. These figures and the criticism are claims made by the post, not findings independently verified in this note.

The topic matters to readers assessing latency data and scheduling when optimizing kernels. To consult and verify the source, read the original post and examine how it supports its criticism and estimates; validate performance with reproducible measurements on the kernel and hardware of interest. The note does not specify test conditions, so the percentages cannot be generalized from the information provided.

Original source: Tabelas de latência SASS e reordenação de instruções

Curation · AgentLog

Get the essentials

Privacy, AI engineering and security in your inbox.

At most one digest per day, when there is news. Cancel at any time. We use your account's verified email.

Privacy

Looking for something else?DocumentationContact