[LLVMdev] quantitative comparison of correctness of llvm-gcc 2.x versions

[John Regehr]

Please link to the graphs if that seems appropriate-- I will leave them 
there and hopefully also update them as new versions of LLVM come out. 
It would be great if these graphs can serve as an advertisement for LLVM 
as well as an advertisement for my work.  I hope to create analogous 
graphs for gcc 4.x sometime, and also to stress-test the x64 ports of both 
compiler families.

There is a large amount of work remaining in our program generator, which 
at present does not even output structs or pointers.  The goal is for it 
to emit the full range of constructs found in normal C programming 
practice without compromising on the "almost strictly conforming" thing.

John Regehr



On Tue, 18 Nov 2008, Vikram S. Adve wrote:

> Unfortunately, the data in the paper doesn't show that, through no
> fault of the authors :-(.  It might be nice to add a qualification and
> a pointer to this graph along with the paper, if John doesn't object.
>
> --Vikram
> Associate Professor, Computer Science
> University of Illinois at Urbana-Champaign
> http://llvm.org/~vadve
>
>
>
> On Nov 18, 2008, at 12:33 PM, Chris Lattner wrote:
>
>>
>> On Nov 18, 2008, at 10:08 AM, John Regehr wrote:
>>
>>> http://www.cs.utah.edu/~regehr/compiler_correctness/llvm_gcc_x86/
>>>
>>> I think these graphs speak for themselves.  Feedback is welcome.
>>
>> Hey, it looks like we made some progress. ;-)
>>
>> Would you mind if I added your paper to llvm.org/pubs?
>>
>> -Chris
>> _______________________________________________
>> LLVM Developers mailing list
>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>