[GiNaC-list] multithreading

[Cristobal Navarro]

i understand the frustation :(
i once opened a topic too, asking if a particular case would be thread
safe,
ginac is one of the best symbolic math libraries around, and it has proveed
that it works great on large cluster problems, its just that the middle
multithreading escenario is missing support, lets hope that in the future
there could be plans for doing some work on the thread safety.

On Mon, Aug 23, 2010 at 12:13 PM, Kraus Philipp
<philipp.kraus at flashpixx.de>wrote:

>
> Am 23.08.2010 um 16:39 schrieb Cristobal Navarro:
>
>
>  in some cases,
>> parallel symbolic algorithms obtain great performance increase, it all
>> depends how you model the problem.
>> im using ginac with openMPI to paralelize huge tree generators, i had to
>> go the processes paralelism since multithread is not safe, but would be nice
>> to be able to do that on the future
>>
>
> I' working with MPI (on boost) and in this case it works on my test system.
> But I would like to portage my algorithm without MPI support (on a simple
> desktop system MPI support is not installed, but on these systems I can use
> multithreading).
> I have tried to compile ginac with -pthread on CXXFLAGS and the CLN
> library. I analyse the compiled code with valgrind, you can show the
> excerpt:
>
> ==51729== Possible data race during write of size 4 at 0x1ea4564 by thread
> #3
> ==51729==    at 0x3AB9: GiNaC::refcounted::add_reference() (in main)
> ==51729==    by 0xF7C49F:
> GiNaC::ptr<GiNaC::basic>::ptr(GiNaC::ptr<GiNaC::basic> const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C4B9:
> GiNaC::ptr<GiNaC::basic>::ptr(GiNaC::ptr<GiNaC::basic> const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C4D3: GiNaC::ex::ex(GiNaC::ex const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C4ED: GiNaC::ex::ex(GiNaC::ex const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C51F: GiNaC::expair::expair(GiNaC::expair const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C539: GiNaC::expair::expair(GiNaC::expair const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7E4F3: void std::_Construct<GiNaC::expair,
> GiNaC::expair>(GiNaC::expair*, GiNaC::expair const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
> ==51729==    by 0xF7F3E9: GiNaC::expair*
> std::__uninitialized_copy_aux<__gnu_cxx::__normal_iterator<GiNaC::expair
> const*, std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*>(__gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> __gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*, __false_type) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
> ==51729==    by 0xF7F43C: GiNaC::expair*
> std::uninitialized_copy<__gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*>(__gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> __gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*) (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
> ==51729==    by 0xF7F45E: GiNaC::expair*
> std::__uninitialized_copy_a<__gnu_cxx::__normal_iterator<GiNaC::expair
> const*, std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*, GiNaC::expair>(__gnu_cxx::__normal_iterator<GiNaC::expair
> const*, std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> __gnu_cxx::__normal_iterator<GiNaC::expair const*,
> std::vector<GiNaC::expair, std::allocator<GiNaC::expair> > >,
> GiNaC::expair*, std::allocator<GiNaC::expair>) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
> ==51729==    by 0xF7F518: std::vector<GiNaC::expair,
> std::allocator<GiNaC::expair> >::vector(std::vector<GiNaC::expair,
> std::allocator<GiNaC::expair> > const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
> in the same run:
>
> ==51729== Possible data race during read of size 4 at 0x1ea42f4 by thread
> #3
> ==51729==    at 0x3AB6: GiNaC::refcounted::add_reference() (in main)
> ==51729==    by 0xF7C5AB: GiNaC::ptr<GiNaC::basic>::ptr(GiNaC::basic&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C5C5: GiNaC::ptr<GiNaC::basic>::ptr(GiNaC::basic&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C5ED: GiNaC::ex::ex() (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF7C604: GiNaC::ex::ex() (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xFCBC50: GiNaC::expairseq::expairseq() (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0x10BD804: GiNaC::mul::mul(GiNaC::ex const&, GiNaC::ex
> const&) (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0x10BD88C: GiNaC::mul::mul(GiNaC::ex const&, GiNaC::ex
> const&) (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0x10FF9D0: GiNaC::exminus(GiNaC::ex const&) (in
> /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0x10FFDD0: GiNaC::operator-(GiNaC::ex const&, GiNaC::ex
> const&) (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0x10BDB0C: GiNaC::mul::derivative(GiNaC::symbol const&)
> const (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
> ==51729==    by 0xF965E3: GiNaC::basic::diff(GiNaC::symbol const&, unsigned
> int) const (in /Developer/opt/Ginac/1.5.8/lib/libginac-1.5.0.dylib)
>
>
> HTH the developers
>
>
> Phil
>
> _______________________________________________
> GiNaC-list mailing list
> GiNaC-list at ginac.de
> https://www.cebix.net/mailman/listinfo/ginac-list
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.cebix.net/pipermail/ginac-list/attachments/20100823/2e35cd26/attachment-0001.htm>