Optimize runtime `countl_zero` #1948

Open

Labels

opened

Branch miss predictions in countl_zero have significant performance impact now that the interpreter is using it for the CLZ instruction. Change the runtime countl_zero function to a branchless implementation.

to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

No one assigned

Labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests