[PATCH] [opaque pointer type] Bitcode support for explicit type parameter on GEP.

Duncan P. N. Exon Smith dexonsmith at apple.com
Tue Feb 24 17:04:54 PST 2015


> On 2015 Feb 24, at 17:03, David Blaikie <dblaikie at gmail.com> wrote:
> 
> 
> 
> On Tue, Feb 24, 2015 at 4:27 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> 
> > On 2015-Feb-24, at 16:27, David Blaikie <dblaikie at gmail.com> wrote:
> >
> > Change to the more common VBR default of 6 bits rather than 8.
> >
> >
> > http://reviews.llvm.org/D7736
> >
> > Files:
> >  include/llvm/Bitcode/LLVMBitCodes.h
> >  include/llvm/IR/Instructions.h
> >  lib/Bitcode/Reader/BitcodeReader.cpp
> >  lib/Bitcode/Writer/BitcodeWriter.cpp
> >  test/Bitcode/function-encoding-rel-operands.ll
> >  tools/llvm-bcanalyzer/llvm-bcanalyzer.cpp
> >
> > EMAIL PREFERENCES
> >  http://reviews.llvm.org/settings/panel/emailpreferences/
> > <D7736.20637.patch>
> 
> Afterthought -- can you confirm whether GEPs already covered well
> for bitcode upgrading (both inbounds and not)?  If not, you should
> probably add testcases that 3.5 GEPs can be read going forward.
> 
> Yep, adding "llvm_unreachable" after 
>      case bitc::FUNC_CODE_INST_INBOUNDS_GEP_OLD:
>     case bitc::FUNC_CODE_INST_GEP_OLD:
> 
> (before the main block under INST_GEP) causes failures in test/Bitcode/{constantTest,memInstructions}.3.2.ll, looking at those ll files they seem to cover a myriad of geps, so I'm OK using that as coverage here.
> 

SGTM.



More information about the llvm-commits mailing list