You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Dec 22, 2021. It is now read-only.
Repository navigation
This repository was archived by the owner on Dec 22, 2021. It is now read-only.
any_true SIMD instruction returns 1 if any lane is non-zero and returns 0 otherwise. Notably, the result does not depend on whether we look at 8-bit, 16-bit, 32-bit, or 64-bit lanes: if the SIMD vector has any bit set, the result is 1; otherwise, the result is 0. Therefore, it would make sense to replace i8x16.any_true, i16x8.any_true, i32x4.any_true, and i64x2.any_true with a single v128.any_true instruction.
I more or less agree - in Firefox the any_true implementations boil away and uniformly become v128.any_bit_set. There's no clear reason to have multiple ops.
Activity
lars-t-hansen commentedon Jan 4, 2021
I more or less agree - in Firefox the any_true implementations boil away and uniformly become v128.any_bit_set. There's no clear reason to have multiple ops.
dtig commentedon Jan 5, 2021
No objections to v128.any_true operation, V8 generates the same code for all the variants as well.
Maratyszcza commentedon Jan 14, 2021
Was implemented in #423
Rename i8x16.any_true to v128.any_true and remove other variants
Rename i8x16.any_true to v128.any_true and remove other variants
[spectext] Rename i8x16.any_true to v128.any_true
[spectext] Rename i8x16.any_true to v128.any_true
[spectext] Rename i8x16.any_true to v128.any_true
Remove tests of {i16x8,i32x4}.any_true
Remove tests of {i16x8,i32x4}.any_true (#13400)