[llvm-commits] [llvm] r160796 - in /llvm/trunk: include/llvm/CodeGen/Passes.h include/llvm/InitializePasses.h lib/CodeGen/CMakeLists.txt lib/CodeGen/EarlyIfConversion.cpp lib/CodeGen/MachineTraceMetrics.cpp lib/CodeGen/MachineTraceMetrics.h

Chad Rosier mcrosier at apple.com
Thu Jul 26 14:18:28 PDT 2012





On Jul 26, 2012, at 2:16 PM, Hal Finkel <hfinkel at anl.gov> wrote:

> On Thu, 26 Jul 2012 13:21:52 -0700
> Chad Rosier <mcrosier at apple.com> wrote:
> 
>> 
>> On Jul 26, 2012, at 12:40 PM, Jakob Stoklund Olesen wrote:
>> 
>>> 
>>> On Jul 26, 2012, at 11:55 AM, Chad Rosier <mcrosier at apple.com>
>>> wrote:
>>> 
>>>>> //===----------------------------------------------------------------------===//
>>>>> 
>>>>> #define DEBUG_TYPE "early-ifcvt"
>>>> 
>>>> Seems reasonable that this debug info would be enabled with early
>>>> if-converstion, but was this intentional?  I could imagine this
>>>> passing being used by other passes as well (e.g., a trace
>>>> scheduler).
>>> 
>>> Yes, it was intentional, because for now early-ifcvt is the only
>>> client, and
>>> 
>>> --debug-only doesn't accept multiple domains.
> 
> What would be the preferred way to change this? Should we accept
> multiple instances of debug-only, or have debug-only take a
> comma-separated list?

Funny you should ask.  I was just working on a fix that implemented this as a comma separated list.

 Chad

> -Hal
> 
>> 
>> Ah, yes.  I've run into this problem on several occasions.
>> 
>> Chad
>> 
>>> 
>>> When things settle down, and we gain multiple clients, I can change
>>> it.
>>> 
>>> /jakob
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 
> 
> 
> -- 
> Hal Finkel
> Postdoctoral Appointee
> Leadership Computing Facility
> Argonne National Laboratory



More information about the llvm-commits mailing list