5.5. High-speed quadruple precision basic arithmetic library

The high-speed quad-precision basic arithmetic library (fast_dd) is a library that performs arithmetic operations by expressing quad-precision values in double-double format.
The double-double format is a format that represents a quadruple precision variable with two double precision real variables.
Since the calculation is performed using the double precision arithmetic instructions of the processor, it can be performed faster than the quadruple precision real type of the language processor.

The high-speed quad-precision basic operation library can be used from Fortran or C ++ programs.

5.5.1. How to combine Fortran versions

fast_dd can be used in conjunction with user programs written in Fortran or C ++. Use the Fortran / C ++ compiler made by Fujitsu for translation and combination.

Compile example is as following:

  1. Conbine SVE fast_dd archives.

[_LNlogin]$ frtpx -Kfast,optmsg=2,SVE a.f90 -SSL2

  1. Combine generic fast_dd archives.

[_LNlogin]$ frtpx -Kfast,optmsg=2,NOSVE a.f90 -SSL2

  1. Translate user program using thread parallel routine, conbine fast_dd archive.

[_LNlogin]$ frtpx -Kfast,openmp,optmsg=2,SVE b.f90 -SSL2

  1. Translate user program by native compiler, conbine SVE version fast_dd archives.

[_CNlogin]$ frt -Kfast,optmsg=2,SVE a.f90 -SSL2

See also

  • To translate a user program with a cross compiler and combine fast_dd, specify -SSL2 to frtpx command line.

  • If you are using the BLAS / LAPACK thread parallel version in your user program, specify -SSL2BLAMP .

  • To use fast_dd thread parallelized routines, add the -Kopenmp or -Kparallel option.

  • When joining with a login node, if -KSVE is enabled, combine libraries that are using SVE.

  • When joining with a login node, if -KNOSVE is enabled, combine general purpose libraries.

  • If combine with computing node, specifying -KSVE or -KNOSVE combine library using SVE.

  • -KSVE option’s default (when omitted) will be -KSVE.

  • To translate a user program written in MPI using a native compiler and combine fast_dd, use mpifrtpx, mpifrt command.

5.5.2. How to conbine C ++ version

Compile example is as following:

  1. Translate user program a.cpp,conbine SVE fast_dd archives.

[_LNlogin]$ FCCpx -Kfast,openmp,optmsg=2,SVE a.cpp -SSL2

  1. Translate user program a.cpp, combine generic fast_dd archives.

[_LNlogin]$ FCCpx -Kfast,openmp,optmsg=2,NOSVE a.cpp -SSL2

  1. Translate user program by native compiler , conbine SVE fast_dd archives.

[_CNlogin]$ FCC -Kfast,openmp,optmsg=2,SVE a.cpp -SSL2

See also

  • To translate a user program with a cross compiler and combine fast_dd, specify -SSL2 to FCCpx command line.

  • If you are using the BLAS / LAPACK thread parallel version in your user program, specify -SSL2BLAMP.

  • fast_dd is a thread-safe object, specify -Kopenmp, -Kparallel or -mt.

  • When joining with a login node, if -KSVE is enabled, combine libraries that are using SVE.

  • When joining with a login node, if -KNOSVE is enabled, combine general purpose libraries.

  • If combine with computing node, specifying -KSVE or -KNOSVE combine library using SVE.

  • -KSVE option’s default (when omitted) will be -KSVE.

  • To translate a user program written in MPI using a native compiler and combine fast_dd, use mpiFCCpx, mpiFCC command.