I'm getting the following build error with the Chemora code base on Supermike:
COMPILING /work/sbrandt/Cactus/arrangements/McLachlanCaKernel/ML_BSSN_Host/src/ML_BSSN_Host_convertFromADMBaseGamma.cc /work/sbrandt/Cactus/arrangements/McLachlanCaKernel/ML_BSSN_Host/src/ML_BSSN_Host_convertFromADMBaseGamma.cc(191): error: argument of type "__m256d" is incompatible with parameter of type "double" ptrdiff_t dir1 = Sign(beta1L); ^ Apparently it's unhappy with some of the vectorization. Any idea what should be done about it?
Cheers, Steve
On 26 Nov 2013, at 18:55, "Steven R. Brandt" sbrandt@cct.lsu.edu wrote:
I'm getting the following build error with the Chemora code base on Supermike:
COMPILING /work/sbrandt/Cactus/arrangements/McLachlanCaKernel/ML_BSSN_Host/src/ML_BSSN_Host_convertFromADMBaseGamma.cc /work/sbrandt/Cactus/arrangements/McLachlanCaKernel/ML_BSSN_Host/src/ML_BSSN_Host_convertFromADMBaseGamma.cc(191): error: argument of type "__m256d" is incompatible with parameter of type "double" ptrdiff_t dir1 = Sign(beta1L); ^ Apparently it's unhappy with some of the vectorization. Any idea what should be done about it?
Hi Steve,
See https://github.com/ianhinder/Kranc/issues/64. Erik says he committed a correction, presumably to Kranc or Vectors. If you are using the chemora versions of McLachlan, Kranc and Vectors, you might get away with backporting the relevant fix, assuming you don't want to merge everything to be up-to-date with the current master branches.
users@lists.einsteintoolkit.org