[Bug target/106723] Unreplaced mode/code attributes in aarch64/arm backends

2022-08-24 Thread rearnsha at gcc dot gnu.org via Gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106723 --- Comment #3 from Richard Earnshaw --- I think we should error if the name matches some iteration values, but not others. If is valid in the assembler output 'as-is' then it's bad form to have an attribute that overloads this - and the fix

[Bug target/106723] Unreplaced mode/code attributes in aarch64/arm backends

2022-08-24 Thread jakub at gcc dot gnu.org via Gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106723 --- Comment #2 from Jakub Jelinek --- I think better would be to error out (or just warn) if we see a valid mode/code attribute which doesn't handle the corresponding mode/code. For the PR106721, that would diagnose all but the last issue where

[Bug target/106723] Unreplaced mode/code attributes in aarch64/arm backends

2022-08-24 Thread rguenth at gcc dot gnu.org via Gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106723 --- Comment #1 from Richard Biener --- I wonder if the generator should simply skip mode combinations when a mode attribute fails to expand? That could be also used as "conditional iteration" using a mode attribute expanding to "".