[llvm-commits] [llvm] r166747 - /llvm/trunk/test/CodeGen/Mips/setgek.ll

[Reed Kotler]

I thought it was safe to match two registers but it seems that sometimes 
the compiler is flipping the operands to XOR  so it could mismatch and 
then fail the make check. It happened to me on one of my internal 
working branches so I changed it back to more general register matching.

I try and make the tests as good as possible as far as matching, but not 
doing things that could change and make the test fail (unless there was 
a real regression).

On 10/25/2012 10:50 PM, David Blaikie wrote:
> On Thu, Oct 25, 2012 at 6:29 PM, Reed Kotler <rkotler at mips.com> wrote:
>> Author: rkotler
>> Date: Thu Oct 25 20:29:42 2012
>> New Revision: 166747
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=166747&view=rev
>> Log:
>> fix test setgek.ll so that it will not give false "make check"
>> failure in some cases
> What exactly do you mean by some cases? Compiler output should be
> entirely stable. What about this test is varying?
>
>> Modified:
>>      llvm/trunk/test/CodeGen/Mips/setgek.ll
>>
>> Modified: llvm/trunk/test/CodeGen/Mips/setgek.ll
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Mips/setgek.ll?rev=166747&r1=166746&r2=166747&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/Mips/setgek.ll (original)
>> +++ llvm/trunk/test/CodeGen/Mips/setgek.ll Thu Oct 25 20:29:42 2012
>> @@ -12,7 +12,7 @@
>>     %conv = zext i1 %cmp to i32
>>     store i32 %conv, i32* @r1, align 4
>>   ; 16:  slti    ${{[0-9]+}}, -32768
>> -; 16:  move    $[[REGISTER:[0-9]+]], $t8
>> -; 16:  xor     ${{[0-9]+}}, $[[REGISTER]]
>> +; 16:  move    ${{[0-9]+}}, $t8
>> +; 16:  xor     ${{[0-9]+}}, ${{[0-9]+}}
>>     ret void
>>   }
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits