Re: [eigen] beta1 tagged (NOT RELEASED), first try

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]


2010/6/30 Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>:
>
> In data mercoledì 30 giugno 2010 16:29:00, Benoit Jacob ha scritto:
>> Please give this a good testing. Beta1 will only be released once it's
>> received enough testing.
>
> Oops, the last mail was sent too early due to a key mismatch of the user....
>
>
> Only the part about gcc 4.1.2 was unfinished. "make check" finishes with
> ------------------------------------------------------------------------------------------
> 99% tests passed, 2 tests failed out of 453
>
> Total Test time (real) = 202.76 sec
>
> The following tests FAILED:
>        451 - prec_inverse_4x4_1 (Failed)
>        452 - prec_inverse_4x4_2 (Failed)
> ------------------------------------------------------------------------------------------
>
> This is reproduceable. The output of those tests is :
>
> ------------------------------------------------------------------------------------------
> orzel@berlioz hg/eigen/build.gcc4.1.2% ./test/prec_inverse_4x4_1
> Initializing random number generator with seed 1277914963
> Repeating each test 10 times
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
>
> inverse_general_4x4, Scalar = float
> error_avg = 1.20849
> error_max = 13.5126
> Test ( inverse_general_4x4<Matrix4f>(200000 * g_repeat) ) failed in "/home/orzel/hg/eigen/test/prec_inverse_4x4.cpp" (67)
>    error_avg < (NumTraits<Scalar>::IsComplex ? 8.0 : 1.2)
>
> orzel@berlioz hg/eigen/build.gcc4.1.2%
> orzel@berlioz hg/eigen/build.gcc4.1.2% ./test/prec_inverse_4x4_2
> Initializing random number generator with seed 1277914971
> Repeating each test 10 times
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> error = 0
> inverse_general_4x4, Scalar = double
> error_avg = 1.20949
> error_max = 12.1809
> Test ( inverse_general_4x4<Matrix<double,4,4,RowMajor> >(200000 * g_repeat) ) failed in "/home/orzel/hg/eigen/test/prec_inverse_4x4.cpp" (67)
>    error_avg < (NumTraits<Scalar>::IsComplex ? 8.0 : 1.2)

OK, so here you get an average error of roughly 1.21, and the
precision test requires it to be no bigger than 1.2, so it fails.

Of course this 1.2 value comes from past experiment, so now that we
see a case where it should be increased, let's increase it. Can you
set it to 1.3 and add a comment that it used to be 1.2 and you raised
it because gcc 4.1 was giving 1.21.

Benoit


>
> ------------------------------------------------------------------------------------------
>
>
> Thomas
> --
> Thomas Capricelli <orzel@xxxxxxxxxxxxxxx>
> http://www.freehackers.org/thomas
>
>
>



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/