Re: [Starpu-devel] segfault in StarPU memory allocator

[Alfredo Buttari <alfredo.buttari@enseeiht.fr>]

Hello,
I managed to make some progress but I'm still far from finding the
problem. I installed valgrind 3.13 (btw 3.12 fails on an instruction
generated by the intel compiler). Unfortunately the code does not run
to completion when using valgrind because it takes too long and the
job exceeds the imposed time limits.

I have compiled both StarPU and my code with the address sanitizer in
gcc but the code runs fine and no problem is reported (apart from some
minor leaks).

I'll try a few more things.





On Mon, Jul 2, 2018 at 11:45 PM Samuel Thibault
<samuel.thibault@inria.fr> wrote:
>
> Alfredo Buttari, le lun. 02 juil. 2018 16:57:27 +0200, a ecrit:
> > Unfortunately valgrind does not like CUDA code:
> >
> > ==1292== Memcheck, a memory error detector
> > ==1292== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
> > ==1292== Using Valgrind-3.10.0 and LibVEX; rerun with -h for copyright 
> > info
> > ==1292== Command: ./dqrm_test
> > ==1292==
> > vex amd64->IR: unhandled instruction bytes: 0xF 0xAE 0x64 0x24 0x40
> > 0x48 0x8B 0x73
> > vex amd64->IR:   REX=0 REX.W=0 REX.R=0 REX.X=0 REX.B=0
> > vex amd64->IR:   VEX=0 VEX.L=0 VEX.nVVVV=0x0 ESC=0F
> > vex amd64->IR:   PFX.66=0 PFX.F2=0 PFX.F3=0
> > ==1292== valgrind: Unrecognised instruction at address 0x4015f48.
> > ==1292==    at 0x4015F48: _dl_runtime_resolve_xsave (in 
> > /usr/lib64/ld-2.17.so)
> > [cut]
>
> Mmm, it's not CUDA that it doesn't like, but the glibc itself. Valgrind
> 3.10 is indeed quite old, you need to upgrade it to get it to work with
> recent glibc.
>
> Samuel



-------------------------------------------
Alfredo Buttari, PhD
CNRS-IRIT
2 rue Camichel, 31071 Toulouse, France
http://buttari.perso.enseeiht.fr