[llvm] r176649 - [fast-isel] Add support for the expect intrinsic.

Andrew Trick atrick at apple.com
Thu Mar 7 21:16:45 PST 2013


On Mar 7, 2013, at 1:44 PM, Chad Rosier <mcrosier at apple.com> wrote:

> 
> On Mar 7, 2013, at 1:36 PM, Bill Wendling <wendling at apple.com> wrote:
> 
>> On Mar 7, 2013, at 12:42 PM, Chad Rosier <mcrosier at apple.com> wrote:
>> 
>>> Author: mcrosier
>>> Date: Thu Mar  7 14:42:17 2013
>>> New Revision: 176649
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=176649&view=rev
>>> Log:
>>> [fast-isel] Add support for the expect intrinsic.
>>> rdar://13370942
>>> 
>>> Modified:
>>>   llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
>>>   llvm/trunk/test/CodeGen/ARM/fast-isel-intrinsic.ll
>>> 
>>> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=176649&r1=176648&r2=176649&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
>>> +++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Thu Mar  7 14:42:17 2013
>>> @@ -698,6 +698,11 @@ bool FastISel::SelectCall(const User *I)
>>>    UpdateValueMap(Call, ResultReg);
>>>    return true;
>>>  }
>>> +  case Intrinsic::expect: {
>>> +//    unsigned ResultReg = getRegForValue(Call->getArgOperand(0));
>>> +//    UpdateValueMap(Call, ResultReg);
>>> +//    return true;
>>> +  }
>>>  }
>>> 
>> ?? Why is this all commented out?
> 
> Opps… I commented out the code to make sure the test failed prior to my change, but then forgot to uncomment them out.. :o/

Not your fault. This is 100% llvm-lit's fault. I've never figured out how to use it in a normal workflow:

Clean build in /clean
Fix in /fix
Implement test case in /fix
Verify that test fails with /clean

I end up copying test files around and have to remember to clean them up. It's super irritating, but I don't see myself diving into lit any time soon, and I know it's easier to complain than fix it because the tool and test configurations are interrelated. Hence:

PR15218: llvm-lit should use its current location as the tools-path
PR15473: llvm-lit should respect the test file's path from the command line.

-Andy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130307/90fc44fc/attachment.html>


More information about the llvm-commits mailing list