[llvm-commits] [PATCH] Add Pattern ID Information

[Bruno Cardoso Lopes]

Hi,

> You are right. I meant that the code which lowers generic shuffles (with cross-lane access) to AVX256 is missing. Due to the restrictive nature of pshufs on avx256, we need to manually lower these shuffles to pshufs + permute or even vinsertf128 + shuf + permute. The linux calling convention is implemented, not the win64 one.

Actually, a good part of the generic cross-lane shuffles are already
supported, that happens because the target independent legalizer will
break it into smaller (not optimal) legal shuffles when no legal
target node is selected. There's room for improvements indeed.
Regarding the win64 abi, just for the record, I'm not tackling it,
will leave that to someone who cares! :)

-- 
Bruno Cardoso Lopes
http://www.brunocardoso.cc