townforge/odeint
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
1,384 Commits 1 Branch 0 Tags 4.1 MiB
cc
Commit Graph

1384 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Pascal Germroth
92c0a40f04 MPI resize/copy: different inner state types for both sides. #10 2013-08-06 02:25:58 +02:00
Pascal Germroth
875a13dfa4 Guard OpenMP against Clang. 
(Or better rewrite the Jamfile to only build OpenMP targets for GCC & Intel toolsets?)
 2013-08-06 02:24:52 +02:00
mariomulansky
b0a35c06eb Use up to 64 threads for the benchmark. 2013-08-06 02:24:52 +02:00
mariomulansky
c9b1140b67 Enable binding threads to processors. 2013-08-06 02:24:52 +02:00
mariomulansky
54fd67d60f Use boost::ref 2013-08-06 02:24:52 +02:00
mariomulansky
d209d7cf2f Enable whole program optimisation for ICC 2013-08-06 02:24:26 +02:00
Karsten Ahnert
79c9b6c7ff adding iterator docs 2013-08-05 09:08:12 +02:00
Karsten Ahnert
c7d814ac7f adding time to system function call in velocity verlet 2013-08-05 00:56:19 +02:00
Karsten Ahnert
0ae9080585 adjusting template arguments in velocity verlet 2013-08-05 00:52:39 +02:00
Karsten Ahnert
fd568c6968 velocity verlet test with boost units 2013-08-04 22:38:15 +02:00
Karsten Ahnert
8333b2899a adding test for velocity verlet 2013-08-04 14:58:41 +02:00
Karsten Ahnert
caacc11927 Merge branch 'velocity_verlet' of github.com:headmyshoulder/odeint-v2 into velocity_verlet 2013-08-04 13:56:46 +02:00
Karsten Ahnert
d5fbd21356 correct indent 2013-08-04 13:56:40 +02:00
mariomulansky
8c4e77c06c Merge remote-tracking branch 'origin' 2013-07-28 20:00:39 -05:00
mariomulansky
303c15454f fixed #94 + test case (now in master) 2013-07-28 19:59:57 -05:00
mariomulansky
7f28ba75e8 fixed problem with icc 2013-07-28 19:26:32 -05:00
Karsten Ahnert
c25cfa86ab bug fix in velocity verlet test, but velocity verlet is still not finished 2013-07-25 22:17:36 +02:00
Karsten Ahnert
4688e3a0f0 continue with velocity verlet 2013-07-25 17:16:13 +02:00
Karsten Ahnert
4f9284367d cont velocity verlet 2013-07-24 10:21:16 +02:00
Karsten Ahnert
61bfb1bf26 adding velocity verlet 2013-07-23 14:09:26 +02:00
Pascal Germroth
4a6db9a52c MPI osc_chain_1d benchmark #4 2013-07-21 04:50:52 +02:00
Pascal Germroth
1ffb155345 Unit tests for MPI state #9, resize #11 and reduction #12 2013-07-21 04:45:59 +02:00
Pascal Germroth
7a709f7dc1 MPI state and algebra #9 #10 2013-07-21 04:45:59 +02:00
Pascal Germroth
744e108fc2 Show perfect speedup in OpenMP plot 2013-07-21 04:45:54 +02:00
Pascal Germroth
d76b77083d Fix wrong index in OpenMP osc_chain_1d 2013-07-21 04:45:46 +02:00
Pascal Germroth
66e32e3e67 Reuse value from previous iteration in simple case too. #3 
From <9792ca4f33/osc_chain_1d/openmp/system.hpp>.
Makes simple/split times comparable (change is minimal though).
 2013-07-19 04:15:52 +02:00
Pascal Germroth
fd5a419d51 1D-chain performance test. #3 
Based on <9792ca4f33/osc_chain_1d/openmp2>

Use osc_chain_speedup.{sh,gnu} to compute and plot speedup.
"split" uses openmp_state/openmp_algebra;
"simple" uses vector/openmp_range_algebra
 2013-07-19 02:33:58 +02:00
Pascal Germroth
daa54972d0 Update Lorenz ensemble sample to new OpenMP algebra 2013-07-19 02:26:35 +02:00
Pascal Germroth
de83cdcec9 Reorganize OpenMP algebra and state. #5 #6 #7 
openmp_range_algebra: parallel for over a random access container.
openmp_nested_algebra: processs parts of a split container in parallel.
openmp_state: a split container based on vector<vector<>>.
openmp_algebra: use a range_algebra on each part of that container.
 2013-07-19 02:23:12 +02:00
Pascal Germroth
53fa4494fd Remove OpenMP system function wrapper. 
Not very useful and doesn't cover all forms.
 2013-07-19 02:22:31 +02:00
Pascal Germroth
758d5b4299 ICC has problems with these tests. 
MPL errors (?!) and numerical errors (probably due to optimization)
 2013-07-19 02:21:28 +02:00
Pascal Germroth
5f25fb50e6 Add all external test cases to Jamroot 2013-07-19 02:21:07 +02:00
Pascal Germroth
797847039c Adjust gcc/clang warnings 2013-07-19 02:20:47 +02:00
Pascal Germroth
bd667cd2a8 Allow any kind of functor in openmp_wrapper #6 2013-07-07 22:53:00 +02:00
Karsten Ahnert
243b570dd3 added bjam config vexcl tests 2013-07-05 11:57:46 +02:00
headmyshoulder
a4f7f904f4 Merge pull request #93 from ddemidov/vexcl_norm_inf 
Introducing vector_space_norm_inf specification for VexCL
 2013-07-05 02:07:18 -07:00
Denis Demidov
a870c4557b Introducing vector_space_norm_inf specification for VexCL 
see #92
 2013-07-05 09:54:18 +04:00
mariomulansky
4871c2049c Merge branch 'master' of github.com:headmyshoulder/odeint-v2 2013-06-24 11:35:53 -05:00
mariomulansky
3332ab3879 clean up eigen algebra 2013-06-24 11:35:36 -05:00
mariomulansky
6b90748f83 fixed eigen algebra 2013-06-24 11:34:51 -05:00
Pascal Germroth
fe64b0811a OpenMP state #5, algebra #6 and resizer #7 
State splits a given Range into an InnerState, one for each thread. The algebra's for_eachN calls for_eachN in parallel on each part, using the InnerState's algebra. There's an openmp_wrapper to parallelize the system function; this needs a way to pass on the offset.

The idea was that this design should allow using OpenMP on each MPI node with a single-threaded inner_state: mpi_state< openmp_state< inner_state > > with mpi_wrapper(openmp_wrapper(system_function))
 2013-06-18 20:07:09 +02:00
Pascal Germroth
2277080c9f Trivial algebra for OpenMP #6 2013-06-15 23:34:26 +02:00
Pascal Germroth
d13e4388b3 Preprocessor macros to generate scale_sumN/for_eachN functors. 2013-06-15 23:32:29 +02:00
Pascal Germroth
ce30759861 Clang complains about unnecessary typenames 2013-06-15 23:32:29 +02:00
Pascal Germroth
cca844e5e7 Fix warnings 2013-06-15 23:32:29 +02:00
Pascal Germroth
f66e0440d7 More warnings 2013-06-15 23:32:29 +02:00
Mario Mulansky
1af6977c58 Merge pull request #86 from kylelutz/master 
fix bug/typo in thrust_algebra

Thanks!
 2013-05-29 19:46:05 -07:00
Kyle Lutz
31ab395e3b fix bug/typo in thrust_algebra 2013-05-29 22:17:47 -04:00
mariomulansky
66e5d66d5f some cleanup 2013-05-23 11:22:58 -05:00
mariomulansky
097ac25e03 even more elegant norm type deduction (recursive value_type analysis) 2013-05-23 11:07:56 -05:00