Skip to content

QuantizeAvx2: Manually hoist zero_point loads#1124

Closed
swolchok wants to merge 1 commit into
pytorch:mainfrom
swolchok:export-D36415089
Closed

QuantizeAvx2: Manually hoist zero_point loads#1124
swolchok wants to merge 1 commit into
pytorch:mainfrom
swolchok:export-D36415089

Conversation

@swolchok
Copy link
Copy Markdown
Contributor

Summary: When T is unsigned char, the compiler can't prove that the storages in this loop don't alias qparams, so it loads zero_point on each iteration. We manually hoist the loads to get around this.

Differential Revision: D36415089

Summary: When T is unsigned char, the compiler can't prove that the storages in this loop don't alias qparams, so it loads zero_point on each iteration. We manually hoist the loads to get around this.

Differential Revision: D36415089

fbshipit-source-id: 7a8d3e8a5ce7e5d39b207e605e3042be83904de8
@facebook-github-bot
Copy link
Copy Markdown
Contributor

This pull request was exported from Phabricator. Differential Revision: D36415089

pytorch-bot Bot pushed a commit that referenced this pull request Feb 26, 2026
Summary:
Pull Request resolved: #1124

When T is unsigned char, the compiler can't prove that the storages in this loop don't alias qparams, so it loads zero_point on each iteration. We manually hoist the loads to get around this.

Reviewed By: jianyuh

Differential Revision: D36415089

fbshipit-source-id: 74bedd92b0ea9139a29170966e93fd7cce5e8c99
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants