[llvm] r183343 - R600: Add a pass that merge Vector Register
Vincent Lejeune
vljn at ovi.com
Wed Jun 5 23:02:31 PDT 2013
Thank for the fix.
I think it is false alarm, because RemapChan should always contains a value equals to Swizzle.
However I realize now that this loop is rather awkward there and should be encapsulated in a function,
and add an llvm_unreachable if the loop is never exited.
Btw was this compiled with clang 3.3 or 3.4 svn ? I didn't get the warning when building with clang 3.3rc
Vincent
----- Mail original -----
> De : NAKAMURA Takumi <geek4civic at gmail.com>
> À : Sean Silva <silvas at purdue.edu>
> Cc : Vincent Lejeune <vljn at ovi.com>; "llvm-commits at cs.uiuc.edu" <llvm-commits at cs.uiuc.edu>
> Envoyé le : Jeudi 6 juin 2013 4h18
> Objet : Re: [llvm] r183343 - R600: Add a pass that merge Vector Register
>
> Suppressed in r183371 with arbitrary value.
>
> Vincent, do you think it would be false alarm?
>
> 2013/6/6 Sean Silva <silvas at purdue.edu>:
>> I'm getting the following warning:
>>
>>
> /Users/sean/pg/llvm/llvm/lib/Target/R600/R600OptimizeVectorRegisters.cpp:183:49:
>> warning: variable 'Chan' is used uninitialized whenever
> 'for' loop exits
>> because its condition is false [-Wsometimes-uninitialized]
>> for (unsigned j = 0, je = RemapChan.size(); j < je; j++) {
>> ^~~~~~
>>
> /Users/sean/pg/llvm/llvm/lib/Target/R600/R600OptimizeVectorRegisters.cpp:193:17:
>> note: uninitialized use occurs here
>> .addImm(Chan);
>> ^~~~
>>
> /Users/sean/pg/llvm/llvm/lib/Target/R600/R600OptimizeVectorRegisters.cpp:183:49:
>> note: remove the condition if it is always true
>> for (unsigned j = 0, je = RemapChan.size(); j < je; j++) {
>> ^~~~~~
>>
> /Users/sean/pg/llvm/llvm/lib/Target/R600/R600OptimizeVectorRegisters.cpp:182:18:
>> note: initialize the variable 'Chan' to silence this warning
>> unsigned Chan;
>> ^
>> = 0
>> 1 warning generated.
>>
>> -- Sean Silva
>>
>> _______________________________________________
>> 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