Commit Graph

88 Commits

Author SHA1 Message Date
Pietro Bonfa c7f5a80894 Always check GPU allocations to avoid misleading error messages in accelerated subroutines 2020-08-03 13:21:30 +02:00
Paolo Giannozzi f8c6b655d8 Duplicated SAVE 2020-04-29 08:45:15 +02:00
Pietro Bonfa 46d2a63bb9 Fixed and increased error reporting in LAXlib 2020-04-28 12:24:06 +02:00
Pietro Bonfa c735d989ef Implementation of serial eigensolver with CU_BLAS, from C. Cavazzoni 2020-04-22 13:15:41 +02:00
Pietro Bonfa f451772e75 Fix GPU compilation with __USE_CUSOLVER 2020-04-11 11:27:54 +02:00
Paolo Giannozzi 439e62af37 Complex BLAS functions ZLADIV and ZDOTC in pzhptrd replaced (by a simple
division the former, by dot_product the latter). Fixes issue #165 and likely
other obscure segfaults in some machines (notably, Macintosh but not only).
In general, all ZDOTC should be replaced either by DDOT when possible, or
by dot_product otherwise. There are many left in linear-response codes. This
one was likely the last instance in PP (via projwfc.x that does not actually
need to use pzhptrd, but it still does)
2020-04-08 16:21:44 +02:00
Pietro Bonfa ce9a2e9552 Updated LAXlib timers 2020-03-28 12:08:39 +01:00
carcava f514bf7def bug fix in GPU LA driver, for spin polarized simulations 2020-03-06 19:28:44 +01:00
giannozz bcb35c86b9 Merge branch 'scalapack' into 'develop'
Default ndiag/northo

See merge request QEF/q-e!818
2020-03-06 12:43:19 +00:00
Fabio Affinito 56181b3a60 fix to adapt test.f90 to the new laxlib apis 2020-03-06 11:21:55 +01:00
giannozz 72f2cb7952 This small change in the default number of processors for linear algebra
(use all, not half) was present in v.6.5 but it has disappeared since.
If there are good reasons not to revert to 6.5 behavior, please close it
2020-03-04 17:01:15 +01:00
Thomas Zeiser 918a61c276 Updated ELPA API to support ELPA 2017-2019
based on issue #144 and the commit
  From: Ifeanyi <ifeanyi@nb-20-96.ictp.it>
  Date: Fri, 29 Nov 2019 10:48:42 +0100
  Subject: [PATCH] Updated ELPA API
  fcfd0874c5

with one change in PW/src/setup.f90 removed which has already been
applied by 61c416dc43
2020-02-28 19:54:12 +01:00
carcava 6dad052c07 adding GPU version of some basic routines 2020-02-24 13:06:31 +01:00
Pietro 9373181df4 Adding GPU subroutines to LAXlib 2020-02-05 15:57:05 +00:00
Paolo Giannozzi 462d8d63fe It is safe to check that arrays allocate into desc_init are not already
allocated
2020-02-04 10:35:48 +00:00
Paolo Giannozzi b89ca39069 Allocations moved inside desc_init 2020-02-03 22:06:52 +00:00
Paolo Giannozzi e265446d5d More desc_init harmonization: second version of desc_init moved to laxlib as
well. Not sure what is the difference between the two versions, though.
2020-02-03 20:21:20 +00:00
Paolo Giannozzi dee8f970d2 desc_init moved into LAXlib 2020-02-03 11:02:41 +00:00
carcava fbd3123d4f adding single precision version of few subroutines, and making LAXlib
the same as for the mixed-precision branch
2020-02-02 11:25:06 +01:00
Carlo Cavazzoni 62d106d25c - fix errors in the interface 2020-02-01 22:48:13 +01:00
Carlo Cavazzoni 1664399a74 - adding subroutine to initialize multiple descriptors 2020-02-01 22:37:14 +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
Paolo Giannozzi df3cbf7dd4 Unused (#ifdef'ed) code in new laxlib contains references to two routines
that are present only iwth an appended _x and no interface. Most linkers
don't complain but old PGI did. For the time being: added _x to calls.
2020-01-30 22:22:03 +01:00
giannozz 2d7b169cef globla => global 2020-01-29 09:42:52 +01:00
Paolo Giannozzi ffd53eb4da Dependencies updated
Two routines in KS_Solvers/PPCG aligned to latest LAXlib changes
2020-01-28 14:58:32 +00:00
fabioaffinito 14d1d4f270 Deleted LAXlib/a 2019-11-26 16:13:11 +00:00
Fabio Affinito d3affeae5d test 2019-11-26 17:05:47 +01:00
Carlo Cavazzoni aa10dc7765 - bug fix: inconsistent interface block
- clean up
2019-09-22 08:28:12 +02:00
Carlo Cavazzoni 2ab74effc0 Merge branch 'develop' into laxlib 2019-09-20 15:08:36 +02:00
Carlo Cavazzoni 2274ab71d8 Merge branch 'develop' of gitlab.com:QEF/q-e into develop 2019-09-20 15:06:47 +02:00
Lorenzo Paulatto d4055fd986 New UPF tool and a few minor fixes 2019-09-20 12:57:10 +00:00
giannozz fad4581a03 qe_pztrtri and qe_pdtrtri: split communicators were created on all processors
but freed only on active processors. Should fix the crashes occurring in long
runs, reported in issue #140. Found by Carlo Cavazzoni.
2019-09-10 10:34:12 +02:00
Carlo Cavazzoni b1d9885cb4 - eliminating useless module la_param
- renaming staff to reduce possibility of conflict
- adding dependency from .fh files
2019-08-18 17:10:53 +02:00
Carlo Cavazzoni 188e125b66 - adding missing interfaces for non scalapack build 2019-08-13 02:57:08 +02:00
Carlo Cavazzoni 004301add1 - re-factoring of LAXlib
now QE do not "use" modules of LAXlib any longer, but it just include interface blocks.
  In principle they can now be compiled independently.
  All this beside possible errors.
  Further clean-ups are now possible, within LAXlib and in QE source codes
2019-08-13 01:16:24 +02:00
Carlo Cavazzoni e358b952bd - blacs related variable always instantiated 2019-08-10 22:35:01 +02:00
Carlo Cavazzoni 04eb22ab8b - missing dependency 2019-08-10 22:11:25 +02:00
Carlo Cavazzoni 27adf6d690 - more disentanglement with LAXlib, quite some change inside LAXlib, still few outside.
Next we have to deal with the removal of the use descriptors stuff
2019-08-10 18:49:26 +02:00
Paolo Giannozzi ebbb788aa8 No need to add a new variable "isolveph" containing information on the
diagonalization for the phonon code: existing variable "isolve" does the job.
make.depend updated
2019-08-07 19:46:20 +02:00
Carlo Cavazzoni 2dc1e177d3 - variable info is now used also outside SCALAPACK macro 2019-08-07 15:08:15 +02:00
Carlo Cavazzoni 5fbc6ecc9c - LAXlib made independent from other module 2019-08-07 14:27:02 +02:00
Federico Ficarelli 0a8988251e Fix LAXLib test build without MPI 2019-05-23 11:04:18 +02:00
Paolo Giannozzi 10c77bf912 Upon suggestion by G. Cantele: add OpenMP check preventing a compilation error 2019-02-21 10:04:17 +01:00
Paolo Giannozzi e256a3e8dc [Skip-CI] Makefile and documentation cleanup 2019-02-14 09:45:46 +01:00
Paolo Giannozzi 5a6a4417dd Also LAXlib and UtilXlib lib should make better usage of parallel_include 2019-02-04 10:25:07 +01:00
Stefano De Gironcoli 75ea65a1a7 some more openMP do loop parallelization 2018-08-14 01:36:48 +02:00
Stefano de Gironcoli a241241d27 updated dependencies 2018-08-05 16:52:11 +02:00
Stefano de Gironcoli d936f16226 export_gstart_2_* and set_mpi_comm_4_* moved to LAXLIB
their call corrected in init_run and mp_global
a recently added bug in ppcg_k when npol=1 corrected
2018-08-05 16:52:11 +02:00
Paolo Giannozzi 7f32483fff IN all Makefiles, LIBS replaced by QELIBS, to prevent unlikely but not
impossible clashes with LIBS as defined by the OS or as used by the configure's
of external libraries (e.g. FoX, for which one such problem has been reported)
2018-07-11 19:58:41 +02:00