Hello everyone, Thank you very much for comments. I agree, that we should prepare some design description document. I'll try to prepapre it as soon as possible. I nthe meantime, I've prepared a new patch which emits warning when -fopenmp option is found and changed doc file from html to rst. Best regards, Alexey Bataev
----- Original Message ----- From: Chandler Carruth Sent: 12/17/12 10:31 PM To: Dmitri Gribenko Subject: Re: [cfe-commits] [PATCH] First OpenMP patch On Mon, Dec 17, 2012 at 9:14 AM, Dmitri Gribenko < [email protected] > wrote: What is more important than documentation, is the change in behavior. Someone might have used '-fopenmp' with Clang before, and now they will get an incomplete implementation. We might also introduce a warning to Clang like "experimental OpenMP support enabled" for the time being. Just to be clear, this is what I was worried about. Previously, the '-fopenmp' flag was handled just like many GCC flags that LLVM will accept and ignore. Now there appears to be support for the feature, but it doesn't do what the user wants... Anyways, the real point is that we should have some write up (preferably committed) which describes the plan and rationale for OpenMP support in Clang. Otherwise we're likely te stay in the exact spot you point out Hal: FIXME's all over the place for unimplemented functionality. =/ That's actually what I'm hoping to avoid.
patch.p
Description: Attachment: patch.p
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
