quantum-espresso/UtilXlib
Ivan Carnimeo 671cf7d9be Merge branch 'develop_omp5' into develop 2024-07-22 15:50:54 +03:00
..
tests Make CCE fortran work with OpenMP offload. 2023-06-07 18:36:35 -04:00
CMakeLists.txt Merge branch 'develop_omp5' into develop 2024-07-22 15:50:54 +03:00
Makefile small fix 2023-04-27 10:15:55 +03:00
Makefile.test First draft of CUDA Fortran enabled UtilXlib 2018-02-15 18:08:38 +01:00
README.md [skip-CI] CP user guide updated 2021-09-02 18:44:12 +00:00
c_mkdir.c clib/ is no more 2021-08-29 18:08:27 +00:00
clib_wrappers.f90 Move Modules/wrappers.f90 to UtilXlib/clib_wrappers.f90 2021-06-19 19:44:23 -05:00
clocks_handler.f90 protect nvtx calls in roctx case and viceversa 2023-04-27 10:24:47 +03:00
copy.c clib/ is no more 2021-08-29 18:08:27 +00:00
cptimer.c clib/ is no more 2021-08-29 18:08:27 +00:00
data_buffer.f90 Remove or reduce MPI message splitting in CUDA Fortran implementation of UtilXlib 2019-02-11 09:49:52 +00:00
device_helper.f90 Merge branch 'develop_omp5' into 'develop_omp5' 2024-01-11 16:09:33 +00:00
divide.f90 Remove all unnecessary mem ops in cegterg. 2018-05-27 21:54:46 -05:00
error_handler.f90 Small things 2023-03-16 14:59:02 +00:00
eval_infix.c clib/ is no more 2021-08-29 18:08:27 +00:00
export_gstart_2_solvers.f90 - LAXlib made independent from other module 2019-08-07 14:27:02 +02:00
find_free_unit.f90 partially reversed previuos commit, in case of failure find_free_unit returns a negative value and prints an info message without stopping the program. 2017-07-30 17:28:48 +00:00
fletcher32.c clib/ is no more 2021-08-29 18:08:27 +00:00
fletcher32_mod.f90 Fletcher-32 checksum implemented in clib/fletcher32.c 2017-09-03 15:01:51 +00:00
hash.f90 simple example of use of the fletcher32 check sum functionality 2017-09-03 15:12:46 +00:00
md5.c clib/ is no more 2021-08-29 18:08:27 +00:00
md5.h clib/ is no more 2021-08-29 18:08:27 +00:00
md5_from_file.c clib/ is no more 2021-08-29 18:08:27 +00:00
mem_counter.f90 Update 3 polar EPW benchmark and the memcounter 2019-04-28 12:47:35 +01:00
memstat.c More obsolete stuff 2024-01-30 17:13:14 +01:00
memusage.c clib/ is no more 2021-08-29 18:08:27 +00:00
mp.f90 avoid exposure of parallel_include in mp module in Utilxlib 2024-06-25 12:00:14 +02:00
mp_bands_util.f90 Replicated routine "set_bgrp_index" replaced by "divide" 2017-12-23 22:00:32 +01:00
mp_base.f90 revert few unwanted changes 2023-09-11 18:39:33 +02:00
mp_base_gpu.f90 Introduce CUDA support in CMake with some refactoring. 2021-01-18 14:50:50 +00:00
nvtx_wrapper.f90 Move Modules/wrappers.f90 to UtilXlib/clib_wrappers.f90 2021-06-19 19:44:23 -05:00
onemkl_omp.f90 fix cpu compilation 2023-04-20 19:25:20 +03:00
parallel_include.f90 UtilXlib directory created to contain a library (libutil.a) for 2017-07-26 11:15:20 +00:00
print_mem.f90 Fix GNU warning. 2021-06-27 13:28:11 -05:00
ptrace.c clib/ is no more 2021-08-29 18:08:27 +00:00
rocblas.f90 Define rocblas API types based on the library. 2024-01-24 16:44:07 -05:00
roctx_wrapper.f90 uniform roctx wrappers with nvtx ones 2023-04-26 18:26:59 +03:00
set_mpi_comm_4_solvers.f90 revert few unwanted changes 2023-09-11 18:39:33 +02:00
thread_util.f90 Fixes for NAG compiler glitches, courtesy Themos Tsikas 2019-06-12 20:55:06 +02:00
util_param.f90 Adding RMM-DIIS cpu version 2021-08-12 20:08:31 +02:00

README.md

UtilXlib

This library implements various basic tasks such as timing, tracing, optimized memory accesses and an abstraction layer for the MPI subroutines.

The following pre-processor directives can be used to enable/disable some features:

  • __MPI : activates MPI support.
  • __TRACE : activates verbose output for debugging purposes
  • __CUDA : activates CUDA Fortran based interfaces.
  • __GPU_MPI : use CUDA aware MPI calls instead of standard sync-send-update method (experimental).

Usage of wrapper interfaces for MPI

This library offers a number of interfaces to abstract the MPI APIs and to optionally relax the dependency on a MPI library.

mp_* interfaces present in the library can only be called after the initialization performed by the subroutine mp_start and before the finalization done by mp_end. All rules have exceptions and indeed subroutines mp_count_nodes, mp_type_create_column_section and mp_type_free can also be called outside the aforementioned window.

If CUDA Fortran support is enabled, almost all interfaces accept input data declared with the device attribute. Note however that CUDA Fortran support should be considered experimental.

CUDA specific notes

All calls to message passing interfaces are synchronous with respect to both MPI and CUDA streams. The code will synchronize the device before starting the communication, also in those cases where communication may be avoided (for example in serial version). A different behaviour may be observed when the default stream synchronization behaviour is overridden by the user (see cudaStreamCreateWithFlags).

Be careful when using CUDA-aware MPI. Some implementations are not complete. The library will not check for the CUDA-aware MPI APIs during the initialization, but may report failure codes during the execution. If you encounter problems when adding the flag __GPU_MPI it might be that the MPI library does not support some CUDA-aware APIs.

Testing

Partial unit testing is available in the tests sub-directory. See the README.md file in that directory for further information.