https://github.com/N-BodyShop/changa

sort by:
Revision Author Date Message Commit Date
877f7c1 SMP mode with GPU Manager Change-Id: I727c2613ac88e8ee372db1d90d870788505c26fc 08 October 2015, 04:14:33 UTC
666a056 A little GPU code documentation and a little cleanup. 05 October 2015, 23:02:07 UTC
0817252 A couple of GPU comments. Change-Id: Idd18eea64bc03feb7284bb4f9aad1b65d8a391ea 30 September 2015, 15:25:33 UTC
765eb1a Comment kernels, and correct call to Ewald*Kernel(). Change-Id: I4a58ab197339960f15730e83095538dced6d5305 25 September 2015, 23:35:34 UTC
81106e3 Merge branch 'master' into trq/cudadev 22 September 2015, 22:16:12 UTC
6f90836 Rewrite of the "newOrder" function in addDelParticles() to avoid non-scaling messaging. Change-Id: I09f5c698c31172e8c6f4f1c13d98e53cfa2bb26a 18 September 2015, 19:06:55 UTC
925a663 Merge branch 'master' into trq/cudadev 11 September 2015, 17:50:39 UTC
e85359d Enable multistepping for Ewald. 11 September 2015, 17:36:27 UTC
716d7db Small change to timings table title. Change-Id: I9b01ff54e01666598f7d133116aa68c22d7e8082 26 August 2015, 16:13:53 UTC
d4b07bc Enable restart using Tipsy binary array files. Also refactored the tipsy array reading code. Change-Id: Icb3c055203f01c47dcb70a2aedfc50dd0c08206e 25 August 2015, 18:20:35 UTC
94c4aff Implemented bGravStep on the GPU. 25 August 2015, 16:00:32 UTC
72310f4 cuda_typedef.h: eliminate dead classes. 25 August 2015, 00:12:35 UTC
895035d Make GPU routines less verbose. 24 August 2015, 22:40:30 UTC
3e715df Temporary fix for Ewald to avoid bad buffers. 24 August 2015, 21:49:50 UTC
2f4bf0b More timing info: Drift, Kick, Adjust and Finish Cache. Change-Id: I6b5b3925472f993061e41e660b367f26d6effc36 22 August 2015, 21:51:13 UTC
c6335b7 Change the tree build to use splitter keys given by DD instead sending the key of the first and the last particle. This is to make the tree build process scalable. Change-Id: I874b30b16cea36d0be1de29506c1145b9c5cba85 13 August 2015, 16:41:15 UTC
06ff01f Reset the obj time in the load balancing database if the load balancer is not being called for that iteration. During the shuffle stage, if the load for a rung does not exist and load for rung 0 does not exist then set it to 0.0 Also fix load bookkeeping in I/O, and some accounting for particles changing rungs. Change-Id: Id09f50ca9f787db113521ff9adac48397e91a3e6 12 August 2015, 19:13:33 UTC
06531b0 Merge branch 'lukasz/cudadev' into trq/cudadev 10 July 2015, 03:48:53 UTC
47f6ada taManager::resumeRemoteChunk(): commenceCalculateGravityLocal() can only be called once per iteration. 10 July 2015, 03:44:09 UTC
d1ad921 ckloop and CUDA defines. 09 July 2015, 21:52:00 UTC
9c943a9 Merge branch 'master' into trq/cudadev This is a large merge of recent work into Lukasz' cudadev branch. Conflicts: Compute.C Makefile.dep Makefile.in ParallelGravity.ci ParallelGravity.h TreePiece.cpp configure 08 July 2015, 19:34:02 UTC
5a28e16 Perform LB only when there is sufficient active TPs Change-Id: Ic202e142a906c7bcdbfdedcc8da96a2f09ebd052 06 July 2015, 19:03:20 UTC
a45b824 Bug fixes 18 June 2015, 08:11:11 UTC
3ffe4a6 Implementing fix for synchronizing GPU Manager device buffer table after transfers of remote chunk data 18 June 2015, 07:04:34 UTC
39951f1 TreePiece::sendParticlesDuringDD(): account for PHASE_FEEDBACK loads. Change-Id: Ia4df437f8070b4c2bcbd629a20ec5b3f3a0a86ac 18 April 2015, 00:13:50 UTC
62e7a95 Smooth Cache writeback: increase priority for better performance on small steps. Change-Id: I20869cc3e366776b66b9666b87018a7104cbcd18 16 April 2015, 05:01:08 UTC
de0a35c Merge branch 'master' into trq/smoothcomp 15 April 2015, 01:39:03 UTC
2ba10fc Change type of thread indices on the GPU from char to int to support large bucket sizes 03 April 2015, 09:24:34 UTC
21ee488 Ewald CUDA kernels: pass particleTable length instead of getting it from cachedData. On SMP machines the number of particles could come from another core. This would also break multiple treepieces per core. 02 April 2015, 04:23:58 UTC
340a2e6 In the load balancers check whether the object is migratable before accessing its obj userdata Change-Id: I6f7773757bc5f4a04e216620d3e4807e0729c9fa 21 March 2015, 22:34:15 UTC
11f8384 Make bUseCkLoopPar readonly variable and remove registerTP from ResumeFromSync Change-Id: Idcf14338a777961318a15de7397a18ae3e227c98 21 March 2015, 02:35:54 UTC
fcb8d28 Fixed bugs in smooth unpack and flush. Change-Id: I4d79c80d6ff8d4618e180ef9f7ac554a81709157 19 March 2015, 19:34:48 UTC
92a985a Fixed bug in smoothcompressed writeback. It now passes testcollapse. 19 March 2015, 15:25:32 UTC
5735cdc Implement smooth cache squeezing. Runs, but has bugs. 19 March 2015, 14:56:44 UTC
bfcc22c Adapt load balancers to modern style inheriting from CBase classes Change-Id: Iafc57f430e3ca43541a4a81cecc1b33576054508 14 March 2015, 17:08:25 UTC
64fc96f EntryTypeSmoothParticle::unpack(): add paranoid check because of extraData kludge. 11 March 2015, 19:47:14 UTC
6b14784 Consistently output tipsy binary integer arrays as 32 bit ints. 10 March 2015, 21:28:05 UTC
4c3c8f8 Further bug fixes 09 March 2015, 11:47:53 UTC
c2c4f39 Purge shared buffers from GPU Manager table before each step 09 March 2015, 11:27:22 UTC
25d08c6 Further fixes. Will need cleanup later. 04 March 2015, 11:08:45 UTC
c2957a2 Further bug fixes 03 March 2015, 10:28:05 UTC
0f3125e Bug fixes to previous commit 03 March 2015, 08:45:36 UTC
636048a Merge branch 'master' of charmgit:cosmo/changa Conflicts: TreePiece.cpp 03 March 2015, 03:45:24 UTC
d20fb56 More Doxygen comments. 03 March 2015, 03:42:48 UTC
731b61c Merge branch 'public' into ppl_master 02 March 2015, 16:39:38 UTC
4bc2ff9 Fix for the bug when running CUDA version of ChaNGa in smp mode. Work in progress. 02 March 2015, 11:18:03 UTC
5755ec0 Fix the crash caused by DD without Sorter. This crash happens because some TreePiece ends up with no particles. This fix calls the Sorter if any TreePiece ends with no particles. Change-Id: Ifdca1cb982d8f3dd7be36fb1c15d8c2232827595 28 February 2015, 23:10:58 UTC
06592b3 Merge branch 'master' of charmgit:cosmo/changa 24 February 2015, 19:17:37 UTC
76c652c load_NC_gas(): ignore missing FeMassFrac and OxMassFrac files. 18 February 2015, 02:55:01 UTC
b5f51ce Merge branch 'public' of github.com:N-BodyShop/changa_uw into public 17 February 2015, 23:59:09 UTC
94d0d28 assignKeys(): be sure boundary particle keys are set. This commit also has some timestep debugging. 15 February 2015, 01:48:16 UTC
de06b14 Be safe with particle pointers if calling newParticle(). 15 February 2015, 00:41:24 UTC
36d703b initSmoothPrioQueue(): correct logic for initial particle searching. 15 February 2015, 00:39:23 UTC
88e9da9 Allow change of dEtaCourant and dEtauDot on restart. 15 February 2015, 00:37:05 UTC
c832276 StellarFeedback(): don't use iLowhFix. 10 February 2015, 06:11:12 UTC
db9270f Merge branch 'public' of github.com:N-BodyShop/changa_uw into public 10 February 2015, 06:03:47 UTC
bea57df TreePiece::readFloatBinary(): copy DataManager pointer for Cooling. 07 February 2015, 03:36:04 UTC
e1aa69a Rung sanity check. 06 February 2015, 23:48:56 UTC
106ee94 Delete vestigial references to comlib/commlib that were never live in committed code Change-Id: I49c6d16862f72edbb92906c0e5b2d119e77e42a7 06 February 2015, 22:54:31 UTC
6fb3630 Adjust local Gravity and Remote Gravity priorities below SPH priorities. Change-Id: I9d3404b728439fd5815e42f8a1384c9b2c91b0f0 06 February 2015, 02:39:08 UTC
9fefc82 Handle ReSmooth case that finds no neighbors. 06 February 2015, 02:06:21 UTC
e91c2da Merge branch 'master' of http://charm.cs.uiuc.edu/gerrit/cosmo/changa 02 February 2015, 16:52:36 UTC
1037306 Allow verbosity to change on restart. Change-Id: I57882039bd5d5264ef9fecd25902167cff663a26 28 January 2015, 23:52:34 UTC
aafcffb Couple more Doxygen comments. 23 January 2015, 20:13:31 UTC
244f660 Delete unused Compare class. 23 January 2015, 20:12:13 UTC
9d052bf Don't assume a particular message type for responses to CkIO callbacks Change-Id: I7e09573cb6024e3a252609e281cee2e230472b30 21 January 2015, 05:00:15 UTC
5b18338 Be sure the load balancer has data for hydro-only simulations. 19 January 2015, 23:05:25 UTC
bb1eaab TreePiece::kick(): do an exponential decay in the COOLING_NONE case to avoid negative thermal energies. 18 December 2014, 18:47:48 UTC
f6f4066 Correct logic for outputing tipsy binary array files in parallel. 10 December 2014, 01:21:04 UTC
94ebc75 openCriterion*: comment out non-SMP safe markUsedBy(). 26 November 2014, 18:51:28 UTC
5eac65b PETreeMerger::mergeWalk(): cleanup and document. 26 November 2014, 18:47:23 UTC
cda2e27 Merge branch 'master' of charmgit:cosmo/changa Conflicts: configure 21 November 2014, 01:50:05 UTC
a608fb4 Update README and CHANGES for v3.1 21 November 2014, 01:40:25 UTC
83c3fc9 Fix nIOProc logic. 05 November 2014, 17:20:25 UTC
0b2e69a Merge branch 'ppl_master' into public 16 October 2014, 16:59:42 UTC
f22cbf6 Intra node load balancing using CkLoop Change-Id: Id1b94c1aef9257161f65ccddd27b1c756dc9d323 16 October 2014, 03:35:59 UTC
c34e9f7 Merge pull request #79 from N-BodyShop/starlog_fix Fix iOrder in starlog table. 09 October 2014, 03:25:37 UTC
86bdbc9 Fix iOrder in starlog table. 08 October 2014, 03:52:08 UTC
2e13680 Fix compilation error in ChaNGa due to the previous change related to HierarchOrbLB Change-Id: I6acfeea8a431159152220f939432fc2643360fd6 06 October 2014, 18:27:42 UTC
878a622 Use lbuserdata to send centroid data to the load balancers. Remove the logic to store the previous phase load since now the TreePiece itself stores the load per phase and sets its load accordingly before calling the load balancer. Change-Id: I4e89628a1796ffb5e4bce9952bb4b7022c4cfa59 06 October 2014, 17:51:51 UTC
26f952e Add HierarchOrbLB which is a hierarchical lb Change-Id: I9f017d59dc3b74df8d094515218894cdc8a5585f 03 October 2014, 20:01:09 UTC
01b44a6 Merge pull request #76 from N-BodyShop/extrastore_fix Adjust particlePointers in the tree nodes if we realloc myParticles. Tested on h229H2 23 September 2014, 16:01:19 UTC
fb7770c Adjust particlePointers in the tree nodes if we realloc myParticles. 23 September 2014, 03:53:33 UTC
373a34c Set cudatype to double when running in hexadecapole mode on the GPU 18 September 2014, 02:17:45 UTC
a3737b1 changed buffer size to 4 to prevent flaky seg faults 18 September 2014, 02:12:15 UTC
d6afe38 Merge pull request #72 from N-BodyShop/extrastore Allocate more memory instead of crashing when hitting the extraStore lim... Basic testing done. Note that this changes the semantics of "dExtraStore": it now specifies how much extra memory is allocated when we run out while creating new particles. 15 September 2014, 20:52:21 UTC
9a5479c TreePiece::newParticle(): get rid of abort statement. 15 September 2014, 02:41:50 UTC
f942742 Allocate more memory instead of crashing when hitting the extraStore limit. 14 September 2014, 02:58:35 UTC
399bf8c Make CHANGESOFT a configure option. (On by default.) Also fix thinko in other flags so that --disable-xxx works. 11 September 2014, 03:16:56 UTC
317ae49 Merge pull request #70 from N-BodyShop/trq/read_coolarray Trq/read coolarray 10 September 2014, 16:50:20 UTC
8f66316 Merge branch 'public' into trq/read_coolarray Conflicts: (adjacent changes) ParallelGravity.ci ParallelGravity.h 09 September 2014, 23:22:16 UTC
3b48843 Merge branch 'ppl_master' into public 09 September 2014, 22:17:37 UTC
da49072 Main::restartGas(): when restarting from NChilada files, find MaxIOrders. 09 September 2014, 22:14:55 UTC
f46dea7 Merge branch 'public' into trq/read_coolarray Conflicts: Sph.C: adjacent changes. 09 September 2014, 04:21:52 UTC
76b16bd Fix reading of coolontime when restarting from output files. 09 September 2014, 04:18:24 UTC
fcbd5e7 Overwrite existing directory (with warning) if producing tipsy files. 08 September 2014, 22:41:56 UTC
52ef60c Merge branch 'ppl_master' into public 04 September 2014, 23:41:56 UTC
138eb59 Merge branch 'master' into trq/ckio 04 September 2014, 23:36:18 UTC
6c6ddc1 Use COOL_SET_ARRAY? functions for reading ionization arrays. This provides infrastructure for restarting metal cooling and H2 runs. 04 September 2014, 23:18:12 UTC
c47ed7a Update HostCUDA.cu as per new GPU Manager 03 September 2014, 02:17:26 UTC
back to top