1
0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-10 03:17:07 +01:00
Commit Graph

183 Commits

Author SHA1 Message Date
e55397bc13 Staggerd cg 2017-11-24 14:18:30 +00:00
1d7ccc6b2c Declaring virtual functions as pure virtual functions. 2017-11-09 19:46:57 +00:00
9b8d1cc3da Staggered Schur decomposed matrix norm changed to not be the Schur anymore :(
Carleton wanted this for multimass / multishift
2017-11-07 14:48:45 +00:00
360efd0088 Improved treatment of reverse asked for by chris.
Truncate the basis.
Power method renormalises
2017-11-02 22:05:31 +00:00
fe4d9b003c More digits 2017-10-30 00:04:47 +00:00
84b441800f Merge branch 'develop' into feature/lanczos-reorg 2017-10-27 14:21:38 +01:00
f96c800d25 Passes reload of coarse basis 2017-10-27 09:43:22 +01:00
32a52d7583 Move the local coherence lanczos into algorithms.
Keep the I/O in the tester. Other people can copy this method to write other I/O formats.
2017-10-27 09:04:31 +01:00
0c4ddaea0b Cleaning up 2017-10-26 23:31:46 +01:00
034de160bf Staggered updates : Schur fixed and added a unit test for Test_staggered_cg_schur.cc giving stronger check 2017-10-26 20:58:46 +01:00
31f99574fa Moving these out of algorithms 2017-10-26 07:47:42 +01:00
a34c8a2961 Update to IRL; getting close to the structure I would like. 2017-10-26 07:45:56 +01:00
f6c3f6bf2d XML serialisation of parms and initialise from parms object 2017-10-25 23:47:59 +01:00
d83868fdbb Identity linear op added -- useful in circumstances where a linear op may or may not be needed.
Supply a trivial one if not needed
2017-10-25 23:47:10 +01:00
303e0b927d Improvements for coarse grid compressed lanczos 2017-10-25 23:46:33 +01:00
e325929851 ALl codes compile against the new Lanczos call signature 2017-10-13 14:02:43 +01:00
47af3565f4 Logging improvement; reunified the Lanczos codes 2017-10-13 13:23:07 +01:00
4b4d187935 Reunified the Lanczos implementations 2017-10-13 13:22:44 +01:00
9aff354ab5 Final version prior to reunification 2017-10-13 13:22:26 +01:00
cb9ff20249 Approx tests and lanczos improvement 2017-10-13 11:30:50 +01:00
9fe6ac71ea Starting reorg of Blocked lanczos 2017-10-11 10:12:07 +01:00
bf58557fb1 Block compressed Lanczos 2017-10-10 14:15:11 +01:00
a1d80282ec cb factorise 2017-10-10 13:49:31 +01:00
4eb8bbbebe Christop mods 2017-10-10 13:48:51 +01:00
bb7378cfc3 Schur for staggered 2017-10-10 12:02:18 +01:00
f0e084a88c Schur staggered 2017-10-10 10:00:43 +01:00
4f8b6f26b4 Merge branch 'develop' into feature/dwf-multirhs 2017-10-02 11:41:49 +01:00
fddeb29d6b Bug fix with spreadout FFT 2017-09-21 11:10:08 +01:00
946a8671b9 Merge pull request #129 from djm2131/feature/eofa
Add support for DWF with the exact one flavor algorithm
2017-09-21 10:15:21 +01:00
771a1b8e79 Merge pull request #128 from paboyle/feature/CG-reliable-update
Feature/cg reliable update
2017-09-21 10:12:03 +01:00
b83b2b1415 Stability improvement to BCG. Force m_rr hermitian beyond rounding. 2017-09-04 14:09:47 +01:00
d9cd4f0273 Staggered multinode block cg debugged. Missing global sum.
Code stalls and resumes on KNL at cambridge. Curious.

CG iterations 23ms each, then 3200 ms pauses. Mean bandwidth reports
as 200MB/s. Comms dominant in the report. However, the time behaviour suggests it
is *bursty*.... Could be swap to disk?
2017-08-23 15:07:18 +01:00
41f73ec083 Add ChronoForecast class for forecasting solutions across poles in the EOFA heatbath 2017-08-16 12:37:38 -04:00
9939b267d2 Added switching to fallback linear operator in reliable update CG, and added recalculation of b parameter on update. 2017-07-31 13:39:44 -04:00
9f280b82c4 Added mixed-precision CG with reliable updates 2017-07-25 11:30:41 -04:00
e504260f3d Able to run a test job splitting into multiple MPI subdomains. 2017-06-22 18:53:11 +01:00
b9104f3072 Block CG 2017-06-21 21:08:03 +01:00
e8b95bd35b Clean up finished. Could shrink Lanczos to around 400 lines at a push 2017-06-21 02:50:09 +01:00
7e35286860 Simplified lanczos, added Eigen diagonalisation.
Curious if we can deprecate dependencly on BLAS.
Will see when we get 48^3 running on our BG/Q port
2017-06-21 02:26:03 +01:00
0486ff8e79 Improved the lancos 2017-06-20 18:46:01 +01:00
e9cc21900f Block solver complete for staggered. Now stable on mass 0.003 and
gives 8x (!) speed up on Haswell laptop vs. standard CG for 8 RHS solves.

166 iterations vs. 537 iterations so algorithmic gain + 2x in flop rate gain.

Better than a slap in the face with a wet kipper.
2017-06-20 12:37:41 +01:00
cfe3cd76d1 Block solver improvements 2017-06-19 14:04:21 +01:00
2439999ec8 Warning elimination; drop to -O2 on G++ bad versions 2017-05-06 14:44:49 +01:00
20999c1370 Merge branch 'develop' into feature/hmc_generalise 2017-05-05 12:47:17 +01:00
ea9aef7baa New header for standard headers (was an issue with Remez.h and external compilation) 2017-05-02 18:26:11 +01:00
3344788fa1 Merge branch 'develop' into feature/hmc_generalise 2017-05-01 12:13:56 +01:00
8e161152e4 MultiRHS solver improvements with slice operations moved into lattice and sped up.
Block solver requires a lot of performance work.
2017-04-18 10:51:55 +01:00
3141ebac10 MultiRHS working, starting to optimise. Block doesn't and I thought it already was; puzzled. 2017-04-17 10:50:19 +01:00
7ede696126 Non compile of tests fixed 2017-04-16 23:40:00 +01:00
a8db024c92 Cleaning up the dense matrix and lanczos sector 2017-04-15 08:54:11 +01:00