Re: [DynInst_API:] cuda not supported by dyninstAPI?

[Hui Zhang <hzhang86@xxxxxxxxxx>]

Oops....my fault. I need to submit it as a job to run, but I forget to do that with dyninst.

Btw, I can still parse the addresses if I treat it as a static binary instead of a process so that I can do it locally, right?

Thanks

On Thu, Feb 1, 2018 at 3:31 PM, Bill Williams <bill@xxxxxxxxxxx> wrote:

This looks to me like libcuda.so is not in the library search path when you launch the executable. Does vecAddSub run without Dyninst from this same environment?
________________________________________
From: Hui Zhang <hzhang86@xxxxxxxxxx>
Sent: Thursday, February 1, 2018 2:26 PM
To: Bill Williams
Cc: John Mellor-Crummey; dyninst-api@xxxxxxxxxxx
Subject: Re: [DynInst_API:] cuda not supported by dyninstAPI?

I can't figure out the problem from the log. Thanks!

On Thu, Feb 1, 2018 at 3:17 PM, Bill Williams <bill@xxxxxxxxxxx<mailto:bill@cs.wisc.edu>> wrote:
Run with DYNINST_DEBUG_STARTUP=1 in your environment and if the problem isn't immediately clear to you from the log, send me stderr from that run.

--bw
________________________________________
From: Hui Zhang <hzhang86@xxxxxxxxxx<mailto:hzhang86@xxxxxxxxxx>>
Sent: Thursday, February 1, 2018 2:13 PM
To: John Mellor-Crummey
Cc: Bill Williams; dyninst-api@xxxxxxxxxxx<mailto:dyninst-api@xxxxxxxxxxx>
Subject: Re: [DynInst_API:] cuda not supported by dyninstAPI?

Hello, John

You mean a full path to the executable? Yes, I tried both, it didn't make a difference.
I can use addr2line to get the source line info (for the frames in cuda runtime or the user code), and the stack trace I'm parsing is on the host side. Both the dyninst and the binary seem to be built properly.
I don't have a clue...

On Thu, Feb 1, 2018 at 3:01 PM, John Mellor-Crummey <johnmc@xxxxxxxx<mailto:johnmc@rice.edu><mailto:johnmc@rice.edu<mailto:johnmc@xxxxxxxx>>> wrote:
Bill,

The EM_CUDA support should have nothing to do with this. libcuda.so.1 is a host binary.

Hui,

Did you use a full path to the file?

FYI: libcuda.so.1 is typically stripped.

> On Feb 1, 2018, at 1:55 PM, Bill Williams <bill@xxxxxxxxxxx<mailto:bill@cs.wisc.edu><mailto:bill@cs.wisc.edu<mailto:bill@xxxxxxxxedu>>> wrote:
>
> Does this persist with the latest commit on master? I just merged support for EM_CUDA within the last two weeks.
>
> --bw
>
> ________________________________________
> From: Dyninst-api <dyninst-api-bounces@xxxxxxxxedu<mailto:dyninst-api-bounces@xxxxxxxxxxx><mailto:dyninst-api-bounces@xxxxxxxxxxx<mailto:dyninst-api-bounces@cs.wisc.edu>>> on behalf of Hui Zhang <hzhang86@xxxxxxxxxx<mailto:hzhang86@xxxxxxxxxx><mailto:hzhang86@xxxxxxxxxx<mailto:hzhang86@xxxxxxxxxx>>>
> Sent: Thursday, February 1, 2018 1:37 PM
> To: dyninst-api@xxxxxxxxxxx<mailto:dyninst-api@xxxxxxxxxxx><mailto:dyninst-api@xxxxxxxxedu<mailto:dyninst-api@cs.wisc.edu>>
> Subject: [DynInst_API:] cuda not supported by dyninstAPI?
>
> Hello,
>
> I'm using DyninstAPI to retrieve source line information from a binary that's built with libcuda.so, but it reports this error:
>
> <program name unknown>: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory
> --SERIOUS-- #0: The process exited during startup. This is likely due to one of two reasons:
> A). The application is mis-built and unable to load. Try running the application outside of Dyninst and see if it loads properly.
> B). libdyninstAPI_RT is mis-built. Try loadin
> --FATAL-- #68: Dyninst was unable to create the specified process
> --FATAL-- #68: create process failed bootstrap
> error in bpatch.createProcess
>
> Is it because dyninstAPI doesn't support cuda?
>
> Thanks
>
> --
> Best regards
>
> Hui Zhang
> Ph.D. candidate in Computer Engineering
> University of Maryland at College Park
>
> _______________________________________________
> Dyninst-api mailing list
> Dyninst-api@xxxxxxxxxxx<mailto:Dyninst-api@xxxxxxxxxxx><mailto:Dyninst-api@xxxxxxxxedu<mailto:Dyninst-api@cs.wisc.edu>>

> https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api




--
Best regards

Hui Zhang
Ph.D. candidate in Computer Engineering
University of Maryland at College Park



--
Best regards

Hui Zhang
Ph.D. candidate in Computer Engineering
University of Maryland at College Park

--

Best regards

Hui Zhang

Ph.D. candidate in Computer Engineering

University of Maryland at College Park