farzonl wrote: > Usually when new ISD nodes are added they are expanded for all types, so that > every backend will get at least working code even if it is not optimal. The > targets can then come along and override the defaults for the types they are > interested in, to get better results. > > For tan I would expect most vector types would want to scalarize, so marking > them as expand would make sense. If more types than are necessary get marked > as Expand that shouldn't be an issue, it looks like we already do that for a > number of other nodes.
Should this be considered a termporary change until their is parity across alll backends? I'm trying to understand why no on else has done this for sin\cos\log\exp\etc. https://github.com/llvm/llvm-project/pull/94559 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
