|
|
Ublas : |
Subject: Re: [ublas] Floating point exceptions and unexpected call to sqrt with sparse integer vectors
From: Paul C. Leopardi (paul.leopardi_at_[hidden])
Date: 2009-08-11 06:03:33
On Tue, 11 Aug 2009, Mark Johnson wrote:
> Gdb shows that the exception is generated in a call to sqrt(-2147483648)
>
> [mj_at_lugha tmp]$ gdb ublas-int
> ...(gdb) run
> Starting program: /home/mj/tmp/ublas-int
>
> Program received signal SIGFPE, Arithmetic exception.
> 0x0000000000405468 in std::sqrt<int> (__x=-2147483648) at
>
[...]
> #3 0x0000000000403fbe in
>
> boost::numeric::ublas::detail::expression_type_check<boost::numeric::ublas:
>:mapped_vector<int, boost::numeric::ublas::map_std<unsigned long, int,
> std::allocator<std::pair<unsigned long const, int> > > >,
> boost::numeric::ublas::vector<int,
> boost::numeric::ublas::unbounded_array<int, std::allocator<int> > >
>
> > (e1=@0x7fffffffdee0, e2=@0x7fffffffde00) at
>
> /usr/include/boost/numeric/ublas/detail/vector_assign.hpp:42
> #4 0x0000000000403466 in
>
> boost::numeric::ublas::vector_assign<boost::numeric::ublas::scalar_assign,
> boost::numeric::ublas::mapped_vector<int,
> boost::numeric::ublas::map_std<unsigned long, int,
> std::allocator<std::pair<unsigned long const, int> > > >,
> boost::numeric::ublas::unit_vector<int, std::allocator<int> > >
> (v=@0x7fffffffdee0, e=@0x7fffffffdf20) at
> /usr/include/boost/numeric/ublas/detail/vector_assign.hpp:369
[...]
> Any ideas as to what's happening? I must admit I don't understand why
> the program is calling sqrt()!
Hi Mark,
Seems to have something to do with BOOST_UBLAS_TYPE_CHECK being defined
because NDEBUG is not defined. Try compiling again with -DNDEBUG to check
this. You could also try mapping your integer vectors to double vectors for
the purposes of uBLAS operations.
Sorry I can't explain why sqrt is being called. AFAICT the check should use
norm_inf which shouldn't use sqrt. You'll have to call on a uBLAS author to
dig deeper.
Best, Paul