Bug#921193: libmkl-rt: Octave returns wrong results when large arrays are multiplied

[Mo Zhou]

control: tags -1 +wontfix

Hi Ido,

As discussed in [1], I think this issue is not fixable because there
is no bug in either MKL or Octave. The GEMM computation error is just
because the clash between libgomp and libiomp.

If you need to use Octave against MKL, please set the environment
variable to switch threading model to GNU OpenMP or TBB:

  MKL_THREADING_LAYER=gnu
  MKL_THREADING_LAYER=tbb

Please don't set it as MKL_THREADING_LAYER=intel, which could cause
the original problem you described.

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=921207