[llvm-dev] [Xray] Help with Xray

[Dangeti Tharun kumar via llvm-dev]

On Mon, Jan 7, 2019 at 5:17 PM Dean Michael Berris <dean.berris at gmail.com>
wrote:

> On Mon, Jan 7, 2019 at 10:26 PM Dangeti Tharun kumar
> <cs15mtech11002 at iith.ac.in> wrote:
> >
> >
> >
> > On Mon, Jan 7, 2019 at 3:21 PM Dean Michael Berris <
> dean.berris at gmail.com> wrote:
> >>
> >> On Mon, Jan 7, 2019 at 8:43 PM Dangeti Tharun kumar
> >> <cs15mtech11002 at iith.ac.in> wrote:
> >> >
> >> > Hi Dean,
> >> >
> >> > I have tried with -instr-map-1 and -instr-map-2, it didn't work.
> >> >
> >>
> >> Yeah, I'm looking through the code and it looks like we're always just
> >> using the function id when we should be using the symbol name. Let me
> >> land a patch to fix this.
> >
> >
> > Thanks Dean.
> >
> >> > Is there a way to find the function name from the identifier?
> >>
> >> You can try this manually by using the `llvm-xray extract -symbolize`
> >> tool to get a YAML file that maps the function id's to symbols, but
> >> that's a roundabout way of doing it. Ideally the graph-diff tool
> >> should just do it.
> >>
> > I have tried this strangely I got this error
> >
> > $> llvm-xray extract -symbolize xray-log.clang-8.*
> >
> > YAML:1:1: error: Got empty plain scalar
> > llvm-xray: Cannot extract instrumentation map from 'xray-log.clang-8.*'
> > Failed loading YAML document from 'xray-log.clang-8.*
> >
>
> Sorry, that was unclear -- you can extract the instrumentation map
> from an XRay-instrumented binary, not the traces.
>
> > but this is working
> >
> > llvm-xray convert xray-log.clang-8.*
> -instr_map=../avx/llvm/xray-build/bin/clang -output-format=yaml 1>out.yaml
> >
> > Am I missing anything?
> >
>
> Nope, my instructions were unclear. :)
>
> I've uploaded a patch which should make this work -- if you're able to
> build the tool(s) with the patch applied, then it would be great to
> hear your feedback:
>
> https://reviews.llvm.org/D56383
>
> Thanks for your prompt response. I have applied the patch. Unfortunately,
it did not solve the problem, I am still seeing function ids.

I'm going to need to come up with test-cases locally to ensure that
> this doesn't regress (and that it works as expected) but it would have
> to wait 'til tomorrow.
>
> Cheers
>
> --
> Dean
>


-- 
Regards,
DTharun
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20190107/a9a454fa/attachment.html>