Commit Graph

316 Commits

Author SHA1 Message Date
Tong Dong Qiu
81c0a3d9f9 Simplified CPU fallback warnings 2021-03-12 11:15:02 +01:00
Giacomo Marchiori
b4aa28771f Added fpgaSolver 2021-03-12 11:14:59 +01:00
Markus Blatt
3ec181e231
Merge pull request #3089 from Tongdongq/exact-ilu-decomp-gpu
Exact ilu decomp gpu
2021-03-08 17:25:04 +01:00
Tong Dong Qiu
87e3f1d72d Added timers to ChowPatelIlu 2021-03-03 17:12:46 +01:00
Tong Dong Qiu
b87e9dad9a Updated comments 2021-03-03 16:36:27 +01:00
Tong Dong Qiu
0caae966b8 Add more waiting for opencl functions, using shortened opencl kernel params in WellContributions 2021-03-03 14:04:06 +01:00
Tong Dong Qiu
688d8ff627 Shorten opencl kernel params, add source file for openclKernels 2021-03-03 13:09:24 +01:00
Tong Dong Qiu
61bf85bf24 Add waiting for enqueueWriteBuffer 2021-03-03 10:37:01 +01:00
Tong Dong Qiu
18bf7c4b19 Replace diagfinding with std::find 2021-03-03 09:50:33 +01:00
Tong Dong Qiu
5b4deab7e4 Generating opencl kernels 2021-03-01 10:26:44 +01:00
Tong Dong Qiu
141af23db5 Exact ILU decomp is now performed on GPU 2021-03-01 10:26:35 +01:00
Tong Dong Qiu
748dc4ccb3 Reduced header inclusion, also moved WellContributions OpenCL initialization 2021-03-01 09:42:56 +01:00
Franz Georg Fuchs
17126ba229 throw specific error if json file does not exist 2021-02-24 11:31:36 +01:00
Markus Blatt
00df3d06c2 Allow usage of non-gpu devices with OpenCL.
This is needed to use the CPU backend of PoCL http://portablecl.org/
which seems handy for debugging.
2021-02-09 10:45:19 +01:00
tqiu
c8dca99fad Renamed fgpilu to ChowPatelIlu 2021-02-03 17:43:54 +01:00
tqiu
a64a342104 Added symmetry check in Debug mode 2021-02-01 11:19:38 +01:00
tqiu
363cc31316 Made waiting for a GPUcopy more explicit 2021-01-28 15:40:10 +01:00
tqiu
d222dffcfd Improved initial guess of L 2021-01-28 13:46:01 +01:00
tqiu
837a83fc16 Added error checking and std::call_once 2021-01-26 17:32:18 +01:00
tqiu
1e09b1f4d9 Added comments and rewrote if 2021-01-26 15:42:38 +01:00
tqiu
dba20adf04 Bugfix: extra swap was done on odd number of sweeps 2021-01-20 11:00:30 +01:00
tqiu
48ebef7808 Updated comments 2021-01-19 11:06:39 +01:00
tqiu
123e3fa89e Use std::find and added comments 2021-01-18 17:10:46 +01:00
tqiu
a8e524fc9d Minor C to C++ changes 2021-01-18 14:11:07 +01:00
tqiu
0f9ae3695d Made reordering optional for WellContributions
Removed unnecessary alloc
2021-01-13 12:34:56 +01:00
tqiu
f26e58c6ca Reordering is now actually skipped when NONE is chosen.
Also made reordering optional for openclSolver.
2021-01-12 15:25:47 +01:00
tqiu
71692ff52b Disabled FGPILU by default 2021-01-12 15:21:19 +01:00
tqiu
9f92a69037 Added CPU and GPU implementations of Fine-Grained Parallel ILU (FGPILU) 2021-01-12 15:21:19 +01:00
Markus Blatt
c037bd762d
Merge pull request #2990 from atgeirr/testing-precfactory-operators
Only add AMG preconditioners to factory if sensible.
2020-12-22 09:52:53 +01:00
Atgeirr Flø Rasmussen
d910d42f6e Only add AMG preconditioners to factory if sensible.
Also add test using a new operator class, that would not compile
without the change.
2020-12-17 14:54:59 +01:00
Markus Blatt
065e2f34f1 Do not include config.h in header (not even in *_impl.hpp)
I was getting warnings about redefined definitions (e.g. HAVE_PARMETIS).
2020-12-11 17:53:24 +01:00
Atgeirr Flø Rasmussen
374764fc6f
Merge pull request #2946 from atgeirr/add-more-amg-options
Add more AMG options to allow setting from JSON.
2020-11-27 08:57:04 +01:00
Atgeirr Flø Rasmussen
6e31c77234 Add more AMG options to allow setting from JSON. 2020-11-26 15:37:50 +01:00
Tor Harald Sandve
f924ac7ffe
Merge pull request #2933 from totto82/dunefem
WIP make Flow compile with dune-fem
2020-11-26 15:10:34 +01:00
Atgeirr Flø Rasmussen
a0ca8dd0b5 Use better-suited AMG criterion.
Use SymmetricDependency instead of SymmetricMatrixDependency.
Our matrices are structurally symmetric, but not numerically symmetric.
2020-11-25 16:40:03 +01:00
Markus Blatt
b21b6cf28c
Merge pull request #2914 from ducbueno/rm-oclcontainer
Removed WellContributionsOCLContainer class
2020-11-19 21:27:55 +01:00
Tor Harald Sandve
c4c0b7a756 Make it compile with dune-fem 2020-11-19 15:45:37 +01:00
Markus Blatt
ea3c899080
Merge pull request #2882 from atgeirr/msw-add-well-contrib
Implement addWellContributions() for multi-segment wells
2020-11-18 13:44:24 +01:00
Jose Eduardo Bueno
a9c2a684c6 Removed WellContributionsOCLContainer class (rebase) 2020-11-18 09:14:31 -03:00
Markus Blatt
64fa5a2b30
Merge pull request #2858 from Tongdongq/ilu-reorder-strategy
Added cmdline parameter --ilu-reorder-strategy
2020-11-17 09:35:12 +01:00
Atgeirr Flø Rasmussen
db7cd1053f Add the multMatrixImpl() utility. 2020-11-16 11:37:44 +01:00
Tor Harald Sandve
fbccdbf68e Make a flow version that uses the polyhedreal grid
Currently the simulator creats the polyhedreal grid from an eclGrid from opm-common

TODO
 - make it possible to create the grid directly from DGF or MRST format
 - fix issue on norne.
2020-11-10 15:49:59 +01:00
Atgeirr Flø Rasmussen
adfb8ef771
Merge pull request #2898 from blattms/always-accu-to-one-proc-as-default
Always accumulates data in AMG to 1 process on coarse levels.
2020-11-05 15:41:57 +01:00
Markus Blatt
d83ae17e5b Always accumulates data in AMG to 1 process on coarse levels.
In OPM the matrix graph might be unsymmetric as we do not store
the full sparsity pattern for copy rows but only the diagonal.
Unfortunately, DUNE assumes that matrices from finite elements and
finite volumes have a symmetric sparsity pattern for copy rows to
and uses this assumption to create the graphs for PTScotch/ParMETIS
more easily. But PTScotch/ParMetis assume a symmetric graph.
2020-11-03 22:13:27 +01:00
Arne Morten Kvarving
4121e0a3c4 fix typo 2020-11-03 13:48:09 +01:00
tqiu
d25405d39a Renamed parameter to OpenclIluReorder and moved error checking to BdaBridge constructor. 2020-10-19 11:36:25 +02:00
tqiu
a372a1e4bd Added cmdline parameter --ilu-reorder-strategy 2020-10-19 11:33:32 +02:00
Atgeirr Flø Rasmussen
e7d1d6603a Silence unused argument warning. 2020-10-19 09:13:37 +02:00
Bård Skaflestad
9a71bbe4b0 Move 'gridForConn' Into Cuda/OpenCL Block
It is only used within this context and produces a warning of the
form

  ISTLSolverEbos.hpp:128:25: warning: unused variable ‘gridForConn’

unless the build configures accelerator support.
2020-10-19 02:28:05 +02:00
Markus Blatt
44fadfb88a Fixes compilation of CUDA.
nvcc exits compilation if the header dune/istl/basearray.hh (form DUNE
2.6) is included as it does not seem to understand the friend declaration
there (friend class for a struct).

```
/usr/include/dune/istl/basearray.hh:101:49: error: ‘typename Dune::Imp::base_array_unmanaged<B, A>::RealIterator’ names ‘template<class B, class A> template<class T> struct Dune::Imp::base_array_unmanaged<B, A>::RealIterator’, which is not a type
       friend class RealIterator<const ValueType>;
                                                 ^
```
2020-10-14 19:23:57 +02:00