[PATCH 02/11] Support the new DiagnosticRemarks

Tobias Grosser tobias at grosser.es
Tue Jun 10 04:57:10 PDT 2014

On 10/06/2014 13:52, Andreas Simbuerger wrote:
> Hash: SHA1
>>> Mhhh. Actually that is a bit short-sighted you're right. After
>>> giving it a second thought it shouldn't be restricted at all und
>>> just print something like: "The size of the detected Scop can be
>>> increased by fixing the following errors."
>> This makes more sense.
>> Btw, it seems we do not report errors for regions where there is
>> not even a child region detected as valid.
>> It may be useful to report in this case errors for the smallest
>> region that contains at least one loop.
> Yep, the main idea behind it is that it might get a bit overwhelming
> as soon as we process larger files and spam the user with errors about
> each leaf in the region tree.

That's why I proposed to only handle regions that have at least a single 

Anyway, let's leave this out for later as well.

>> (I know some people dislike if I propose to reduce the initial
>> patch, so please let me know if my 'leave out' comments are
>> discouraging. My understanding is that as soon as we got the basic
>> patch in, all these commits should be very simple and should not
>> require a lot of discussions).
> Nah no problem at all, I just split the patch. I don't have to do a
> lot of work there, git rebase does the hard work ;-).



More information about the llvm-commits mailing list