@nicbarkeragain Clang failed to recognize it should use a select op in the IR and using a ternary helped.

Source

• URL: https://x.com/jakubtomsu_/status/1947602738478452869
• Author: Jakub Tomšů (@jakubtomsu_)
• Posted: 2025-07-22 10:20:52

Branch

1/ @jakubtomsu_

@nicbarkeragain

Clang failed to recognize it should use a select op in the IR and using a ternary helped. This gets also unrolled and vectorized and it ends up being faster.

GCC seems to always catch this, no idea why is clang being dumb even in a trivial case like this

https://godbolt.org/z/jMqazcenq

Related

• Spine: 2025-07-22-branchless-programming-is-a-term-used-to-describe