[llvm] r196288 - Reland 196270 "Generalize debug info / EH emission in AsmPrinter"

Tom Stellard tom at stellard.net
Fri Oct 10 13:27:18 PDT 2014


On Fri, Oct 10, 2014 at 12:49:08PM -0700, David Blaikie wrote:
> On Fri, Oct 10, 2014 at 12:44 PM, Tom Stellard <tom at stellard.net> wrote:
> 
> > On Fri, Oct 10, 2014 at 11:08:19AM -0700, David Blaikie wrote:
> > > On Tue, Dec 3, 2013 at 11:26 AM, Timur Iskhodzhanov <timurrrr at google.com
> > >
> > > wrote:
> > >
> > > > Can you please "svn up -r196270" and try running the R600 tests and add
> > > > printfs to DwarfDebug::{begin,end}Function?
> > > >
> > > > e.g. see
> > > > http://lab.llvm.org:8011/builders/clang-x86_64-ubuntu/builds/10693
> > > >
> > > > I can guess that these tests don't actually need debug info (as they
> > don't
> > > > have any metadata),
> > > > but it is obvious that DwarfDebug::endFunction() gets called without
> > > > calling beginFunction() first.
> > > >
> > >
> > > Do we have an answer to this? Or a test case? etc?
> > >
> > > Otherwise I'll be rather inclined to remove it as dead. Nothing exercises
> > > it.
> > >
> >
> > What are you planning to remove?
> >
> 
> This bit in particular:
> 
> +  // Every beginFunction(MF) call should be followed by an endFunction(MF)
> call,
> +  // though the beginFunction may not be called at all.
> +  // We should handle both cases.
> +  if (CurFn == 0)
> +    CurFn = MF;
> +  else
> +    assert(CurFn == MF);
> 

That's fine with me.  If there are bugs in the R600 backend that is
preventing some useful change let me know and I'll take a look at it.

-Tom

> 
> >
> > -Tom
> > >
> > >
> > > >
> > > >
> > > > 2013/12/3 Tom Stellard <tom at stellard.net>
> > > >
> > > >> On Tue, Dec 03, 2013 at 11:03:18PM +0400, Timur Iskhodzhanov wrote:
> > > >> > 2013/12/3 Eric Christopher <echristo at gmail.com>
> > > >> >
> > > >> > > > +  /// Please note that some AsmPrinter implementationss may not
> > > >> call
> > > >> > >
> > > >> > > "implementations"
> > > >> > >
> > > >> >
> > > >> > Oops! r196304
> > > >> >
> > > >> >  > +  /// beginFunction at all.
> > > >> > >
> > > >> > > Wat? They call endFunction but not beginFunction?
> > > >> >
> > > >> >
> > > >> > +Tom
> > > >> > You might want to revisit how AMDGPUAsmPrinter interacts with
> > DwarfDebug
> > > >> > through AsmPrinter.
> > > >> >
> > > >>
> > > >> Can you give me a little more background on the issue?  I don't know
> > > >> anything about DwarfDebug and if R600 is using it, then it is probably
> > > >> an accident.
> > > >>
> > > >> -Tom
> > > >>
> > > >> >
> > > >> > > -eric
> > > >> > >
> > > >>
> > > >> > _______________________________________________
> > > >> > llvm-commits mailing list
> > > >> > llvm-commits at cs.uiuc.edu
> > > >> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> > > >>
> > > >>
> > > >
> > > > _______________________________________________
> > > > llvm-commits mailing list
> > > > llvm-commits at cs.uiuc.edu
> > > > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> > > >
> > > >
> >



More information about the llvm-commits mailing list