Re: [eigen] Re: Infinite loop in gebp

2013/2/26 Sameer Agarwal <sameeragarwal@xxxxxxxxxx>

Thanks Gael. Good to know that I did not introduce some craziness, since this code has been running fine before this. I am guessing Apple has no plans on upgrading g++ beyond 4.2.1 due to GPL issues.

Sameer

On Tue, Feb 26, 2013 at 2:58 PM, Gael Guennebaud <gael.guennebaud@xxxxxxxxx> wrote:

Hi Sameer,

indeed, I've already observed wrong code generation with this apple
compiler, and I know that other heavily templated libraries observed
similar problems. In my case it concerned our iterative solvers.So
I'd not spend too much time on it but just to be sure it might still
be interesting to check with other gcc versions and valgrind.

gael

On Tue, Feb 26, 2013 at 6:24 PM, Sameer Agarwal
<sameeragarwal@xxxxxxxxxx> wrote:
> Sorry, I reported the wrong version of gcc. Here is the version string.. It
> is 4.2.1 instead of 4.1.2
>
> 686-apple-darwin11-g++-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5666) (dot 3)
>
> Sameer
>
>
>
> On Tue, Feb 26, 2013 at 9:19 AM, Sameer Agarwal <sameeragarwal@xxxxxxxxxx>
> wrote:
>>
>> Hi Guys,
>>
>> I am using eigen 3.1.2 on a mac and I am observing an infinite loop in
>>
>> Eigen::internal::gebp_kernel<double, double, long, 4, 4, false,
>> false>::operator()
>>
>> This is with g++4.1.2 -O3 -DNDEBUG -fast -msse3
>>
>> the same code works just fine with clang 4.0.0
>>
>> I can provide instructions for reproducing this (its corrector_test.cc of
>> the Ceres Solver code base).
>>
>> But before I do that, is there anything know about this compiler/Eigen/OS
>> combination?
>>
>> Thank you,
>> Sameer
>>
>