Is this work around good for gcc9 too ?
cheers
Henri
Le 13/03/2021 à 21:06, Charo del Genio a écrit :
> Dear all,
> as per the subject line, AmberTools20 fails to compile with gcc10. This is due to numerous parts of the code that are not exactly compliant with the C and FORTRAN standards. I think it may be a
> really good idea to fix them all. For the time being, there is a workaround, which consists of two steps.
>
>
> First, 2 lines in AmberTools/src/sander/mt19937.F90 need to be modified. These are line 187, which needs to read
> mult_a = ieor(ishft(int(z'6C07'), 16), z'8965') ! but this is okay
> and line 229, which needs to read
> mult_b = ieor(ishft(int(z'5D58'), 16), z'8B65')
>
>
> Then, in the configuration step, the following options must be passed to cmake
> -DCMAKE_C_FLAGS=-fcommon -DCMAKE_CXX_FLAGS=-fcommon -DCMAKE_Fortran_FLAGS='-fallow-argument-mismatch -fallow-invalid-boz'
>
>
> Please note that this is a workaround, and I think that the best thing would be to actually fix the invalid uses of BOZ literals and the argument mismatches in the code.
>
> Oh, by the way, are the RXSGLD fixes going to be in an upcoming update patch, or are they going directly in the AmberTools21 code?
>
>
> Cheers,
>
> Charo
>
>
>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Mar 13 2021 - 14:00:02 PST
