fc0b82e | Balint Joo | 07 January 2010, 17:53:05 UTC | Bumped version no to 3.37.3 | 07 January 2010, 17:53:05 UTC |
ebb3def | Balint Joo | 07 January 2010, 17:32:54 UTC | Merged. Resolved conflicts | 07 January 2010, 17:32:54 UTC |
5a7f317 | Balint Joo | 07 January 2010, 17:04:08 UTC | Checked and fixed regressions | 07 January 2010, 17:04:08 UTC |
9d50d55 | U-KZINTI\edwards | 01 January 2010, 19:41:52 UTC | Induced changes. | 01 January 2010, 19:41:52 UTC |
2bf65e7 | U-KZINTI\edwards | 01 January 2010, 18:53:36 UTC | Info on 3.37.2 | 01 January 2010, 18:53:36 UTC |
349ed97 | U-KZINTI\edwards | 01 January 2010, 18:49:58 UTC | Induced changes | 01 January 2010, 18:49:58 UTC |
7c2016d | U-KZINTI\edwards | 01 January 2010, 18:49:53 UTC | Induced changes | 01 January 2010, 18:49:53 UTC |
81e2bae | U-KZINTI\edwards | 01 January 2010, 18:49:48 UTC | Induced changes | 01 January 2010, 18:49:48 UTC |
5ae1b8a | U-KZINTI\edwards | 01 January 2010, 18:46:30 UTC | Changed to 3.37.2 | 01 January 2010, 18:46:30 UTC |
7e3ed59 | U-KZINTI\edwards | 01 January 2010, 18:42:20 UTC | Added PROP_AND_MATELEM_COLORVEC. Combines both the prop and the peram calculation. Saves on re-reading solution vectors. Also saves on memory - only need 1 solution vector at a time. | 01 January 2010, 18:42:20 UTC |
1b7183a | U-KZINTI\edwards | 01 January 2010, 18:41:11 UTC | Combines both the PROP_COLORVEC and PROP_MATELEM_COLORVEC into one task, in an effort to avoid reading the solutions from a map object. | 01 January 2010, 18:41:11 UTC |
e273267 | U-KZINTI\edwards | 01 January 2010, 18:40:44 UTC | Added inline_prop_and_matelem_colorvec. | 01 January 2010, 18:40:44 UTC |
b05c435 | Robert Edwards | 01 January 2010, 04:04:28 UTC | Use a temporary for vec_source instead of calling peekSpin repeatedly. | 01 January 2010, 04:04:28 UTC |
97b7fea | U-KZINTI\edwards | 31 December 2009, 05:41:56 UTC | Renamed prop_matelem_colorvec code to prop_matelem_lm_colorvec (a low memory version). The default prop_matelem_colorvec behavior is to lower the number of lookups from disk, at the expense of hold 4x (spin_sinks) the amount of temporaries. | 31 December 2009, 05:41:56 UTC |
b260029 | U-KZINTI\edwards | 30 December 2009, 17:41:21 UTC | Switched back to num_orthog=20. | 30 December 2009, 17:41:21 UTC |
371827e | U-KZINTI\edwards | 30 December 2009, 17:35:25 UTC | Induced changes | 30 December 2009, 17:35:25 UTC |
1d48881 | U-KZINTI\edwards | 30 December 2009, 17:35:09 UTC | Induced changes | 30 December 2009, 17:35:09 UTC |
84105af | U-KZINTI\edwards | 30 December 2009, 17:34:27 UTC | Induced changes. | 30 December 2009, 17:34:27 UTC |
f1c737d | U-KZINTI\edwards | 30 December 2009, 17:32:59 UTC | Changed to 3.37.1 | 30 December 2009, 17:32:59 UTC |
00bded3 | U-KZINTI\edwards | 30 December 2009, 17:32:20 UTC | Changed metric for new output. | 30 December 2009, 17:32:20 UTC |
114a2f2 | U-KZINTI\edwards | 28 December 2009, 23:03:24 UTC | New input/output. | 28 December 2009, 23:03:24 UTC |
1eea43b | U-KZINTI\edwards | 28 December 2009, 23:01:32 UTC | Switched to using a funcmap factory. Seems to fix some segfault. | 28 December 2009, 23:01:32 UTC |
b33a4bb | U-KZINTI\edwards | 27 December 2009, 20:12:15 UTC | Turned back on MapObject for colorvectors. | 27 December 2009, 20:12:15 UTC |
d99dfc0 | U-KZINTI\edwards | 27 December 2009, 19:54:19 UTC | Switched to MapObject. | 27 December 2009, 19:54:19 UTC |
013e963 | U-KZINTI\edwards | 27 December 2009, 19:54:06 UTC | Added include of xml_group_reader.h | 27 December 2009, 19:54:06 UTC |
101ce37 | U-KZINTI\edwards | 27 December 2009, 19:36:30 UTC | Fixed to compile with new registeries. | 27 December 2009, 19:36:30 UTC |
0239b42 | U-KZINTI\edwards | 27 December 2009, 18:49:54 UTC | Info on 3.37.0 | 27 December 2009, 18:49:54 UTC |
6b5ba84 | U-KZINTI\edwards | 27 December 2009, 18:47:54 UTC | Hopefully fixed up the metric paths. | 27 December 2009, 18:47:54 UTC |
f1c8cad | U-KZINTI\edwards | 27 December 2009, 18:42:25 UTC | Block code no longer exists. | 27 December 2009, 18:42:25 UTC |
93b2107 | U-KZINTI\edwards | 27 December 2009, 18:39:32 UTC | Now use to MapObject. | 27 December 2009, 18:39:32 UTC |
d425c03 | U-KZINTI\edwards | 27 December 2009, 18:37:21 UTC | Lots of fixes. Now seems to work. | 27 December 2009, 18:37:21 UTC |
99c43fd | U-KZINTI\edwards | 27 December 2009, 18:36:36 UTC | Put const char* error on a single line of output. | 27 December 2009, 18:36:36 UTC |
dcd8964 | U-KZINTI\edwards | 27 December 2009, 16:31:57 UTC | Turned off a debugging line. | 27 December 2009, 16:31:57 UTC |
b5a6da5 | U-KZINTI\edwards | 27 December 2009, 06:03:08 UTC | New version of copying object maps. This is like reading a disk map into memory, or copying a disk object onto another disk object. Probably the most useful is copying a memory object onto a disk object. | 27 December 2009, 06:03:08 UTC |
65b71a3 | U-KZINTI\edwards | 24 December 2009, 19:59:18 UTC | Switched to new MapObject format. | 24 December 2009, 19:59:18 UTC |
85c940c | U-KZINTI\edwards | 24 December 2009, 19:50:57 UTC | Converted the reset of the readers to new factory format. Removed MapObject readers. Those are superceded by the stand-alone versions. | 24 December 2009, 19:50:57 UTC |
652345e | U-KZINTI\edwards | 24 December 2009, 19:27:55 UTC | Fixed name of map object group. | 24 December 2009, 19:27:55 UTC |
f58e231 | U-KZINTI\edwards | 24 December 2009, 18:45:22 UTC | Merge branch 'branch-pre-chroma3-37-0' into devel | 24 December 2009, 18:45:22 UTC |
d4a2a0b | U-KZINTI\edwards | 24 December 2009, 18:40:08 UTC | Removed obsolete files. | 24 December 2009, 18:40:08 UTC |
8fe963b | U-KZINTI\edwards | 24 December 2009, 05:16:35 UTC | Info on 3.37.0 | 24 December 2009, 05:16:35 UTC |
556d495 | U-KZINTI\edwards | 24 December 2009, 05:11:35 UTC | Removed SubsetVectors. Overhauled calls to use factory for MapObject. Cleaned up uses of MapObjects. Removed some unused routines. | 24 December 2009, 05:11:35 UTC |
fd5dbc2 | U-KZINTI\edwards | 24 December 2009, 05:11:09 UTC | Removed all uses of SubsetVectors. | 24 December 2009, 05:11:09 UTC |
7f71e69 | Balint Joo | 17 December 2009, 16:49:00 UTC | Backpatched mods for QUDA v0.2 from devel | 17 December 2009, 16:49:00 UTC |
576bd2d | Balint Joo | 17 December 2009, 16:38:38 UTC | Patches for lib QUDA v0.2 | 17 December 2009, 16:38:38 UTC |
98b36a5 | U-KZINTI\edwards | 16 December 2009, 05:28:52 UTC | Induced changes. | 16 December 2009, 05:28:52 UTC |
dff69b6 | U-KZINTI\edwards | 16 December 2009, 05:27:19 UTC | Changed to 3.37.0 | 16 December 2009, 05:27:19 UTC |
29b8b73 | Balint Joo | 14 December 2009, 21:36:30 UTC | Merge branch 'devel' of git://git.jlab.org/pub/lattice/usqcd/chroma into devel | 14 December 2009, 21:36:30 UTC |
cd8cbfd | Balint Joo | 14 December 2009, 21:32:36 UTC | Added inline_read_subset_vectors_memory Works same as inline subset vectors | 14 December 2009, 21:32:36 UTC |
88da611 | U-KZINTI\edwards | 10 December 2009, 15:24:07 UTC | Added VECTOR_SMEAR to propagator smearing. | 10 December 2009, 15:24:07 UTC |
9931141 | U-KZINTI\edwards | 06 December 2009, 02:41:29 UTC | Merge branch 'devel' of ssh://git.jlab.org/pub/lattice/usqcd/chroma into devel | 06 December 2009, 02:41:29 UTC |
1863872 | Balint Joo | 04 December 2009, 17:40:55 UTC | Backpatched several bugfixes from devel - in qio_read_obj_funcmap record XML error - in prop colorvec added catch clause for char* type exceptions - in GPU solvers, zeroed out padding factors. | 04 December 2009, 17:40:55 UTC |
2278bf1 | Balint Joo | 03 December 2009, 18:53:31 UTC | Backpatched these, from devel | 03 December 2009, 18:53:31 UTC |
ccb6ca8 | Balint Joo | 03 December 2009, 18:52:21 UTC | Commented out padding factors in the QUDA interface (set to zero) They appear to have caused some solver instabilities. I will reinstate them when Mike tells me to | 03 December 2009, 18:52:21 UTC |
c5d06f1 | Balint Joo | 03 December 2009, 18:47:17 UTC | With the new GPU lib, padding caused problems. I have reset the pad factors to zero. This seems to cure the problems, at the cost of some performance. Will reset this when I hear from Mike. | 03 December 2009, 18:47:17 UTC |
5b9ac9b | U-KZINTI\edwards | 03 December 2009, 16:42:37 UTC | Changed to use new MapObject. | 03 December 2009, 16:42:37 UTC |
77bfe3b | U-KZINTI\edwards | 03 December 2009, 16:42:06 UTC | Induced changes | 03 December 2009, 16:42:06 UTC |
6f58050 | U-KZINTI\edwards | 03 December 2009, 16:41:54 UTC | Fixed filenames of pt thingies. | 03 December 2009, 16:41:54 UTC |
76feee9 | Balint Joo | 02 December 2009, 20:27:45 UTC | Merge remote branch 'origin/devel' into devel | 02 December 2009, 20:27:45 UTC |
f8079fe | Balint Joo | 01 December 2009, 20:47:48 UTC | Added code to monitor bandwidth | 01 December 2009, 20:47:48 UTC |
2d9c6d0 | Kostas Orginos | 25 November 2009, 18:26:50 UTC | added pt colorvec | 25 November 2009, 18:26:50 UTC |
dad13aa | Balint Joo | 25 November 2009, 00:52:34 UTC | *) Pulled in backpatch for new QUDA library organization. version as it was, did not compile. | 25 November 2009, 00:52:34 UTC |
3d70636 | Balint Joo | 24 November 2009, 14:19:09 UTC | Merged in master | 24 November 2009, 14:19:09 UTC |
daf21e8 | Balint Joo | 24 November 2009, 13:57:41 UTC | Added QUDA padding code | 24 November 2009, 13:57:41 UTC |
4c1d999 | Balint Joo | 23 November 2009, 20:34:29 UTC | *) Added a READ_SUBSET_VECTORS measurement to read new subset vectors | 23 November 2009, 20:34:29 UTC |
1c10391 | Balint Joo | 23 November 2009, 19:06:35 UTC | *) Added a type string instead of a type code NB: Binary Writer '\n' terminates std::strings (writeln) So always needed to add 1 to the string length *) Added Checksumming on Value Records. Added Checksumming on Map | 23 November 2009, 19:06:35 UTC |
e161592 | Balint Joo | 23 November 2009, 14:38:09 UTC | *) In the first release version invert_quda.h has been rolled into just quda.h | 23 November 2009, 14:38:09 UTC |
02aaeb4 | Balint Joo | 20 November 2009, 23:35:44 UTC | *) Ooopsie. I had a mistype here that broke regressions. Regressions now pass. But how strong are they? *) Still to do for completeness: -) Put Checksums onto the disk map records (and check these) -) Use free flow strings to encode Map Type? How to initialize these in the traits. | 20 November 2009, 23:35:44 UTC |
5f2d340 | Balint Joo | 20 November 2009, 22:47:59 UTC | *) Moved subset-vectors to be a map of pair structs *) Followed through and patched up places affected by this | 20 November 2009, 22:47:59 UTC |
3fe269d | Balint Joo | 20 November 2009, 16:46:38 UTC | Added back old style of linking flags for backward compatibility with older libQUDA-s | 20 November 2009, 16:46:38 UTC |
71d21a5 | Balint Joo | 20 November 2009, 15:26:52 UTC | Changed incude and libpath to cope with new library organization in libquda | 20 November 2009, 15:26:52 UTC |
0fa5a43 | Balint Joo | 19 November 2009, 19:36:05 UTC | *) Added READ_MAP_OBJECT_DISK task to read MapObjDisk thingies. This is based on a 'funcmap' file local to inline_read_map_obj_disk.cc - The reader functions are in an anonymous namespace - there is a registerAll function in this namespace which is called the first time the measurement is run. - New map types will need to be added here, tho the only thing that differentiates the reader functions are template types on the map... | 19 November 2009, 19:36:05 UTC |
d176494 | Balint Joo | 19 November 2009, 15:50:14 UTC | *) Hunted down crash bug in QIO_READ for subset vectors *) Added some more catch clauses to chroma - will be useful later *) This current setup appears to be able to cope with 24^3x128 lattices with 162 evecs in about 12GiB of front end. 48-62Gib Should be able to cope easily with 4 jobs. Apart from a few tidyups (MapObjReader) we're pretty much there. NB: It may be worth storing the SubsetVectors as a MapObjDisk for evecs and another binary file for evecs, because it takes some 5 minutes to read the current QIO format, and create the MapObjDisk for the evecs... | 19 November 2009, 15:50:14 UTC |
2dc7174 | Balint Joo | 18 November 2009, 21:08:53 UTC | *) Added type traits to deal with Map Key/Value information. These live in util/ferm/map_obj/map_obj_disk_traits.h Sadly, every time we use a new Kind of MapObjectDisk, we have to add traits classes here. *) Changed unit test: t_map_obj_disk *) Added a layer of indirection, so that named objects hold their data by Handle<>. This is hidden from the user. The plus is that now one can have NamedObjects with Constructor parameters by doing create< Type, Param1Type>( param 1 ); *) Use this to automagically open a MapObjectDisk backing store for evectors. | 18 November 2009, 21:08:53 UTC |
cf2827f | Balint Joo | 18 November 2009, 18:40:10 UTC | Compiles and xcheck passes | 18 November 2009, 18:40:10 UTC |
18dcd31 | Balint Joo | 18 November 2009, 16:55:28 UTC | I have no idea why this keeps being 'modified' | 18 November 2009, 16:55:28 UTC |
ef1e3df | Balint Joo | 18 November 2009, 16:50:35 UTC | *) Moved the SourceVectors construct to have map-like semantics. *) Currently these are implemented as a MemoryMap *) Added update possibility to the map. Currently this is used only in inline_block_colorvecs_w.cc which may well be dead code TODO: - add type information to DiskMaps and implement a DiskMapReader - allow the Memory Map in SubsetVectors to be swapped out with DiskMaps - add an inline measurement to convert old subset vectors to new DiskMap based ones | 18 November 2009, 16:50:35 UTC |
3fecc70 | Balint Joo | 17 November 2009, 22:50:22 UTC | Updated cpp_wilson_dslash. Why has it changed | 17 November 2009, 22:50:22 UTC |
326e2be | Balint Joo | 17 November 2009, 22:49:22 UTC | Merge branch 'mapObj' of ssh://qcd7ni01/home/bjoo/Devel/QCD/jlab-standard-chroma-build/chroma/chroma into fromqcd7ni01 | 17 November 2009, 22:49:22 UTC |
24fbf90 | Balint Joo | 17 November 2009, 22:48:18 UTC | Just a change in header comments | 17 November 2009, 22:48:18 UTC |
86b2063 | Balint Joo | 17 November 2009, 22:47:19 UTC | Committing to allow transfer to laptop. Doesnt compile currently | 17 November 2009, 22:47:19 UTC |
cdb340a | Balint Joo | 17 November 2009, 14:28:01 UTC | Updated submodule commit reference | 17 November 2009, 14:28:01 UTC |
6a641e3 | Balint Joo | 17 November 2009, 14:18:31 UTC | Induced changes: autoconf | 17 November 2009, 14:18:31 UTC |
f1ed977 | Balint Joo | 17 November 2009, 14:18:19 UTC | Induced changes: automake | 17 November 2009, 14:18:19 UTC |
87e80e5 | Balint Joo | 17 November 2009, 14:17:58 UTC | Induced changes: autoheader | 17 November 2009, 14:17:58 UTC |
aba79de | Balint Joo | 17 November 2009, 14:17:45 UTC | Induced changes: aclocal.m4 | 17 November 2009, 14:17:45 UTC |
5a503b3 | Balint Joo | 17 November 2009, 14:17:13 UTC | Fixed Makefile.in | 17 November 2009, 14:17:13 UTC |
39185c5 | Balint Joo | 17 November 2009, 02:26:38 UTC | Induced Change: autoconf | 17 November 2009, 02:26:38 UTC |
eb14df0 | Balint Joo | 17 November 2009, 02:26:24 UTC | Induced Change: automake | 17 November 2009, 02:26:24 UTC |
f86e10f | Balint Joo | 17 November 2009, 02:26:07 UTC | Induced Change: autoheader | 17 November 2009, 02:26:07 UTC |
d186ebf | Balint Joo | 17 November 2009, 02:25:56 UTC | Induced Change: aclocal | 17 November 2009, 02:25:56 UTC |
4cd61fe | Balint Joo | 17 November 2009, 02:25:06 UTC | Updated cpp_wilson_dslahs | 17 November 2009, 02:25:06 UTC |
de3d5cb | Balint Joo | 17 November 2009, 02:19:11 UTC | Backed out because, I need a read funcmap type stuff. Maybe I don't really want to faff with this inline. | 17 November 2009, 02:19:11 UTC |
62be519 | Balint Joo | 17 November 2009, 01:49:45 UTC | Oops | 17 November 2009, 01:49:45 UTC |
2bfa5ab | Balint Joo | 17 November 2009, 01:17:54 UTC | Committing to pull up to laptop | 17 November 2009, 01:17:54 UTC |
773832e | Balint Joo | 16 November 2009, 16:56:42 UTC | Restructured MapObject as a simple finite state machine. There are 3 states: INIT, READ, WRITE + an error state. openWrite() takes from INIT to WRITE, otherwise to error openRead() takes either from INIT, READ and WRITE to READ destructor takes from READ or WRITE to finalized, or from INIT to error lookup works only in READ insert works only in WRITE exists, size and dump() work in all modes. I am ready to push this out to the great big world after just a little more testing... (Robert for example should check correctness), while I should check the memory footprint on a qcdeval() machine. | 16 November 2009, 16:56:42 UTC |
faa8e4a | edwards | 14 November 2009, 21:56:05 UTC | Periodic update. SKIPPED: ChangeLog | 14 November 2009, 21:56:05 UTC |
9676b56 | eneil | 14 November 2009, 20:18:56 UTC | Updated changelog. SKIPPED: ChangeLog | 14 November 2009, 20:18:56 UTC |
76e03cd | eneil | 14 November 2009, 20:01:43 UTC | Made an assortment of small fixes that allow Chroma to compile properly with Nc=2. Also fixed and tested the SU(2) exponentiation in eesu2.cc. | 14 November 2009, 20:01:43 UTC |
6a02afa | Balint Joo | 14 November 2009, 03:33:04 UTC | Another induced commit. This time from autoheader | 14 November 2009, 03:33:04 UTC |
671624b | Balint Joo | 14 November 2009, 03:20:18 UTC | Induced changes: configure | 14 November 2009, 03:20:18 UTC |