Skip to content

Ensure constants for ISPC fast approximate exponential are double precision - #169

Merged
janbridley merged 4 commits into
mainfrom
fix/constants-in-fast-ispc-math
Nov 3, 2026
Merged

janbridley merged 4 commits into
mainfrom
fix/constants-in-fast-ispc-math

Conversation

@janbridley

@janbridley janbridley commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor

ISPC uses single precision by default, meaning uniform double X; gets truncated to a float before saving. Adding the d prefix resolves this, i.e. uniform double Xd;. The fast ISPC exponentials are now slightly more accurate and correctly handle overflows.

This also has a small CI fix to update a dependency.

Description

Motivation and Context

Resolves: #???

How Has This Been Tested?

Checklist:

@janbridley
janbridley requested a review from DomFijan October 30, 2026 20:19
@janbridley
janbridley merged commit dec0711 into main Nov 3, 2026
18 checks passed
@janbridley
janbridley deleted the fix/constants-in-fast-ispc-math branch November 3, 2026 15:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant