Commit Graph

120 Commits

Author SHA1 Message Date
Paolo Giannozzi 7c83eb4ab5 [skip-CI] Configure updated to work with new name of pgf90 (nvfortran) 2020-08-25 21:37:14 +02:00
Paolo Giannozzi 16ad4fb98a Workaround for gcc 10 and later does not work around due to bad syntax 2020-08-14 10:44:16 +02:00
fabrizio22 cb55b2ddf0 set configuration with libxc 2020-07-22 15:52:50 +02:00
Ye Luo 3340d739b9 Make PPC64LE to work with XL. 2020-07-05 17:33:41 -04:00
Paolo Giannozzi b9996e6c3b With gcc 10, add currently needed option -fallow-argument-mismatch 2020-06-29 15:42:39 +02:00
Paolo Giannozzi ef4297751c It seems reasonable to try mpiifort before mpif90 on Intel machines 2020-06-06 21:36:58 +02:00
Sasha Fonari dd04c7fb7e Build internal libbeef 2020-04-14 11:19:43 -04:00
giannozz f58b89c061 Merge branch 'develop' into 'develop'
Added libbeef autoconf .m4 file.

See merge request QEF/q-e!869
2020-04-08 20:19:29 +00:00
gusmaogabriels 22d799ede3 Added beef autoconf .m4 file. 2020-04-08 10:51:18 -04:00
Pietro Bonfa 17e71640f5 Select appropriate openmp implementation for MKL when building with PGI. This depends on the version 2020-04-05 17:19:07 +02:00
Paolo Giannozzi c176f1d8dd Fixes for compilation with OpenMP 2020-03-03 10:45:48 +00:00
Paolo Giannozzi f048394930 Merge branch 'develop' of gitlab.com:giannozz/q-e into ph_and_conf_fix 2020-02-29 09:30:48 +01:00
giannozz 20c9141a10 Merge branch 'fix-qe-configure-m4_elpa' into 'develop'
accept ELPA>2016 by configure

See merge request QEF/q-e!808
2020-02-29 08:02:07 +00:00
giannozz 0db1ca8624 Workarounds and fixes for yesterday's commits 2020-02-29 08:59:35 +01:00
Thomas Zeiser 708a6763ad accept ELPA>2016 by configure
Mimic the behavior of version 2015 and 2016 also for 2017-2019.

install/configure has to be updated, too. But that can easily be
done by running 'autoconf' and is therefore not included in the
commit.

To be able to actually use ELPA version > 2016, the fix described
in iissue #144 is required in addition.
2020-02-28 19:31:56 +01:00
Thomas Zeiser dd5787b49a fix detection of FFTW3 include file by configure
- If FFT_LIBS is set, it may nevertheless be necessary to use a
  special include path; thus, check FFT_INCLUDE and FFTW_INCLUDE.
  If they are set, add them to IFLAGS.
- If FFTW3 has to be autodetected, fix the test for fftw3.f03:
  + just 'include "fftw3.f03"' is not a valid Fortran program
    => 'use iso_c_binding // end' are missing
  + moreover, the file is in free form
    => '-ffree-form' as required as compile argument
  + the original 'for inc in $try_incdir' did not have any effect
    as the loop body did not make any use of $inc
    => only FFLAGS will be evaluated by AC_COMPILE_IFELSE

install/configure has to be updated, too. But that can easily be
done by running 'autoconf' and is therefore not included in the
commit.
2020-02-28 19:01:54 +01:00
giannozz ae33473560 More cleanup, documentation updated 2020-02-28 10:53:11 +01:00
giannozz 871f1d7f67 More configure simplfication 2020-02-27 17:59:04 +01:00
giannozz 2d4b020298 More cleanup of autoconf 2020-02-27 16:11:37 +01:00
giannozz 38f9231628 More cleanup of obsolete cases; no search for lapack if mkl is available 2020-02-27 14:42:27 +01:00
giannozz c13727f54a Removed useless and undocumented configure option --use-netlib. Now:
- if BLAS_LIBS and/or LAPACK_LIBS are present, configure use them with no check
- otherwise, configure searches suitable blas and lapack;
  If nothing is found, configure selects internal blas and/or lapack
2020-02-27 11:01:20 +01:00
giannozz 7020ecf7fc Removal of obsolete stuff 2020-02-26 22:05:54 +01:00
giannozz 717fba1eb2 Replace old gfortran option with newer one 2020-02-26 15:51:24 +01:00
Carlo Cavazzoni 2af0e629ab Revert "Merge branch 'revert-c25e9f31' into 'develop'"
This reverts merge request !775
2020-02-01 09:14:20 +00:00
Carlo Cavazzoni a91d32bc5f Revert "Merge branch 'develop' of gitlab.com:QEF/q-e into develop"
This reverts commit c25e9f3126
2020-02-01 08:58:44 +00:00
Pietro Delugas a7df32754a armpl detection for blas lapack and fft also for gfortran 2020-01-27 15:15:53 +01:00
Pietro Delugas b815a4e90c completing and fixing case arm+armflag+armpl
still to do case arm:gfortran+armpl
2020-01-27 10:58:11 +01:00
Pietro Delugas 4668ec6165 recognition of armflang+armpl case for blas and lapack
still to do for fftw
2020-01-27 10:58:11 +01:00
Pietro Delugas 3bd7bae5ba autoconf for aarch64
* added aarch64 as an arm architecture ( is it right ?)
* detection of armflang compiler when wrapped in mpif90

yet to do:
   * recognition of armpl library
   * export of __PGI to FOX_FLAGS
2020-01-27 10:58:11 +01:00
giannozz e8f6a2baf9 Debug flag for PGI compiler as well -Z may or may not work. 2020-01-22 13:57:28 +01:00
Pietro Delugas 5fb1195b08 adding options for linking and include hdf5 lib
options --with-hdf5-lib and --with-hdf5-include for configure
allow to specify linker and include paths when the authomatic detection
fails.

 Changes to be committed:
	modified:   install/configure
	modified:   install/m4/x_ac_qe_hdf5.m4
2019-12-18 15:58:27 +01:00
Pietro Delugas 16c6c2e2f3 updating configure for HDF5 usage
* HDF5_LIB variable in make.inc.in renamed HDF5_LIBS ( see issue #149)
* simplified syntax for HDF5 version number check in configure

 Changes to be committed:
	modified:   install/configure
	modified:   install/m4/x_ac_qe_hdf5.m4
	modified:   install/make.inc.in
2019-10-24 11:19:27 +02:00
giannozz 16804bed49 DFTI is always selected if MKL are available, also with OpenMP.
Configure simplified: here is no need to check for DFTI if MKL is available.
2019-10-04 15:05:10 +02:00
Paolo Giannozzi 8e7fd37fe0 Configure for PPC64 machines with Linux and new xlf compiler should not set
-D__XLF unless xlf is actually used
2019-08-23 21:01:15 +02:00
giannozz 009e67c5ca xlf porting
Misc changes to make QE run with ppc64 machines, linux and xlf:
call flush(*) replaced by standard (?) command flush(*)
getpid #define-d as integer for xlf
configure (partially) updated, sets -d__XLF
2019-08-22 15:42:49 +02:00
Ye Luo 4fab5fba59 Add pgi support for non-X86 systems. 2019-07-25 13:24:56 -04:00
Pietro Delugas ef3579c992 fixing configure for hdf5 2019-07-16 08:28:34 +00:00
Pietro d2631d096f Remove any definition of the macro 'vector' to circumvent bug in some preprocessors. 2019-06-13 17:22:10 +00:00
Pietro Delugas 4d12deefef adapt conf-scripts for hdf5 linked with zlib
previous script was adding  zlib path HDF5_LIBS with a wrong syntax
2019-04-05 09:56:34 +02:00
Davide Ceresoli d84cd35e47 1) Added ./configure options to compile and link with libxc:
--with-libxc            (yes|no) Use libXC for some XC functionals (default:
                          no)
  --with-libxc-prefix=DIR Directory where libxc was installed.
  --with-libxc-include=DIR
                          Directory where libxc Fortran headers were
                          installed.

2) If you set input_dft='functional+META', (i.e. 'pbe+meta'), you can now
   perform a nscf/bands calculation using a potential-only MGGA functional
   (such as TB09).
2019-02-12 10:10:11 +01:00
Pietro Delugas 526ab6181e Hdf5 config 2018-12-11 09:11:52 +00:00
Paolo Giannozzi d87b145b4b [skip-CI] Minor changes to configure and documentation, suggested by Nichols Romero 2018-12-10 10:29:05 +01:00
Paolo Giannozzi 78c5ef7ad9 [skip-CI] Minor updates 2018-11-28 20:04:39 +01:00
Paolo Giannozzi 701c7eaf6f More configure cleanup: removal of mysterious MKL_FLAG="-static-intel" option
that has no effect on MKL (it's a linker option); removal of MKL_FLAG from the
test for DFTI (was hindering detection); removal of unused or commented out
stuff.
2018-11-22 14:17:07 +01:00
Paolo Giannozzi e8b1eca456 [skip-CI] configure simplified: no need to separately test serial and parallel
compiler. There is still a problem with detection of DFTI, though.
2018-11-22 11:19:14 +01:00
Paolo Giannozzi cbad87c52e [skip-CI] configure wasn't working as expected in the serial case: gfortran
was invariably chosen, due to some obscure interaction with x_ac_qe_cuda.
Unsetting some obscure variables seems to work.
2018-11-21 19:01:22 +01:00
Paolo Giannozzi 3b4bbde193 Minor updates to build mechanism, mostly for MKL on Mac: is there anybody
with a Mac and MKL available for testing?
2018-10-23 13:28:14 +02:00
Paolo Giannozzi c290deba93 Detection of MKL libraries modified upon suggestion by Nichols Romero:
use MKLROOT instead of trying to guess all possible paths. In case MKLROOT
is not set, use MKLROOT=/opt/intel/mkl. Support for very old MKL removed.
User guide updated.
2018-10-20 11:18:14 +02:00
Pietro Delugas 4fa547e5ec update configure script management of hdf5 library 2018-09-18 13:23:51 +00:00
Paolo Giannozzi 4277567973 Obsolete configure option removed 2018-09-07 17:14:10 +02:00