[llvm-dev] A Study on "Dependence capturing strength" of four Dependence Analyzers in LLVM on SPEC 2017 benchmarks

[Michael Kruse via llvm-dev]

Hi Adil,

thank you for this analysis. For the Polly part, you seem to have used
"PolyhedralInfo", which has been the result of a GSoC project. In the
current implementation, it does not support runtime conditions. Since
checking assumptions at runtime (e.g. no aliasing, no overflow) is
required in most cases, I'd expect it to not be very successful. In
the data I looked at, there is only "0" for PolyhedralInfo.

Michael



2018-02-07 3:35 GMT-06:00 ADIL ARUN DANGUI via llvm-dev
<llvm-dev at lists.llvm.org>:
> Hi,
>
> I am Adil Arun Dangui, a 3rd year B.Tech student at IIT Hyderabad with a
> CGPA of 9.17. I am interested in Compilers, Compiler Optimizations and
> excited about LLVM and Polly. I have some familiarity of LLVM structure as
> we studied it in the Compiler Engineering course at IITH.
>
>
> Here is the link directing to our study on "Dependence capturing strength"
> of four Dependence Analyzers in LLVM on SPEC 2017 benchmarks. You might want
> to have a look at our work. Comments are welcome.
>
>
> Thank you.
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>