[llvm-dev] Publication & Project: Verificarlo: checking floating point accuracy through Monte Carlo Arithmetic

Pablo Oliveira via llvm-dev llvm-dev at lists.llvm.org
Mon Sep 26 09:31:48 PDT 2016


Thanks !

Pablo

Le lun. 26 sept. 2016 18:30, Renato Golin <renato.golin at linaro.org> a
écrit :

> Added on r282416.
>
> http://llvm.org/pubs/
>
> cheers,
> --renato
>
> On 26 September 2016 at 08:13, Pablo Oliveira via llvm-dev
> <llvm-dev at lists.llvm.org> wrote:
> > Hi,
> >
> > We have recently published a paper on floating point accuracy analysis
> > through Monte Carlo Arithmetic. We also released the open-source tool
> > Verificarlo (https://github.com/verificarlo/verificarlo) that relies on
> > LLVM for instrumenting floating point operations.
> >
> > Could you please add our paper to http://llvm.org/pubs/ ?
> >
> > Verificarlo: checking floating point accuracy through Monte Carlo
> > Arithmetic. Christophe Denis, Pablo de Oliveira Castro, and Eric Petit.
> > 23rd IEEE International Symposium on Computer Arithmetic (ARITH), July
> > 2016.
> >
> > The preprint is available online at
> > <https://hal.archives-ouvertes.fr/hal-01192668>.
> >
> >
> > Could you please also add our tool to http://llvm.org/ProjectsWithLLVM ?
> >
> > Verificarlo: A tool for automatic Montecarlo Arithmetic analysis.
> >
> > https://github.com/verificarlo/verificarlo
> >
> > Verificarlo is an LLVM based tool to automatically use the Monte Carlo
> > arithmetic in place of IEEE-754 floating point arithmetic. Verificarlo
> > instruments floating operations point at the LLVM Intermediate
> > Representation level replacing all floating point operations by their
> > Monte Carlo Arithmetic counterpart. It forces the results of floating
> > point operations to behave like random variables turning executions of a
> > program into trials of a Monte Carlo simulation and allowing statistics
> > on the effects of rounding error to be obtained over a number of
> > executions. In particular, Verificarlo tracks rounding and catastrophic
> > cancellation errors at a given virtual precision.
> >
> > Thanks,
> >
> > Pablo
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160926/f30a2cd6/attachment.html>


More information about the llvm-dev mailing list