On Jun 4, 2007, at 11:59 PM, Anton Korobeynikov wrote:

> Tanya,
>
>> The attached patch creates a global to track functions that should
>> not be inlined (noinline attribute). The llvm inliner will need to be
>> modified to make use of it.
> Why don't use the function attribute for "noinline"? This seems to be
> natural way to handle such things.

Attribute is natural way, if significant percentage of total  
functions is
annotated with "noinline", which is not the case here.

-
Devang
_______________________________________________
llvm-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

Reply via email to