9a6a850 | Francois Trahay | 13 June 2008, 12:12:37 UTC | Fix missing flag | 13 June 2008, 12:12:37 UTC |
f5c630e | Francois Trahay | 09 June 2008, 13:33:54 UTC | Add some (well many) debug messages and logs | 09 June 2008, 13:33:54 UTC |
e9cf4a7 | Francois Trahay | 13 May 2008, 12:43:43 UTC | Fix sampling-prog output | 13 May 2008, 12:43:43 UTC |
fb283cf | Francois Trahay | 13 May 2008, 12:42:53 UTC | Fix sampling on infiniband | 13 May 2008, 12:42:53 UTC |
191e6b8 | Elisabeth Brunet | 06 May 2008, 15:26:48 UTC | balise manquante pour quand on a pas marcel dans la flavor | 06 May 2008, 15:26:48 UTC |
5b9dff4 | Elisabeth Brunet | 06 May 2008, 15:22:33 UTC | le format des perfs était incorrect | 06 May 2008, 15:22:33 UTC |
7eb4a5a | Elisabeth Brunet | 06 May 2008, 13:48:45 UTC | finalement ce sont bien des entiers qu'on voulait pour l'échantillonnage | 06 May 2008, 13:48:45 UTC |
67b09c3 | Elisabeth Brunet | 06 May 2008, 13:44:54 UTC | manquait un bout | 06 May 2008, 13:44:54 UTC |
359c455 | Elisabeth Brunet | 06 May 2008, 08:54:16 UTC | le changement de type des échantillons n'avait pas été répercuté partout | 06 May 2008, 08:54:16 UTC |
b329e1c | Elisabeth Brunet | 05 May 2008, 17:49:31 UTC | interpolation faite sur des entiers | 05 May 2008, 17:49:31 UTC |
54f3a99 | Francois Trahay | 05 May 2008, 16:13:56 UTC | Small fixes to run the split_all strategy | 05 May 2008, 16:13:56 UTC |
3177a1c | Elisabeth Brunet | 05 May 2008, 15:04:01 UTC | l'indirection était mal faite | 05 May 2008, 15:04:01 UTC |
1d372d0 | Elisabeth Brunet | 05 May 2008, 11:37:10 UTC | les échantillons de perf sont stockés en tant que flottant et non d'entier pour plus de précision | 05 May 2008, 11:37:10 UTC |
6545c72 | Francois Trahay | 05 May 2008, 09:47:06 UTC | Offloading message submission is now possible, even with the split_all strategy | 05 May 2008, 09:47:06 UTC |
c8b8a77 | Francois Trahay | 05 May 2008, 06:49:09 UTC | Fix warnings | 05 May 2008, 06:49:09 UTC |
cf764a2 | Francois Trahay | 29 April 2008, 16:12:00 UTC | Add a tasklet for each driver that aims at posting requests | 29 April 2008, 16:12:00 UTC |
4cda864 | Elisabeth Brunet | 29 April 2008, 15:56:10 UTC | assignation du wrapper à un driver dans la strategie | 29 April 2008, 15:56:10 UTC |
65fbfdb | Elisabeth Brunet | 29 April 2008, 15:32:11 UTC | les moyens sont maintenant coupés dans la tasklet et plus au try_and_commit - 1ere etape | 29 April 2008, 15:32:11 UTC |
eb11225 | Francois Trahay | 29 April 2008, 13:20:28 UTC | Fix warnings | 29 April 2008, 13:20:28 UTC |
1934e67 | Francois Trahay | 29 April 2008, 12:42:00 UTC | Crappy fix of locks. Now this should work with both multirails (not tested on homogenous multirails) and offloading. This may deadlock or yell "bad: scheduling while atomic" from time to time, but it's enough for the paper. | 29 April 2008, 12:42:00 UTC |
4fe5613 | Francois Trahay | 23 April 2008, 15:59:16 UTC | This seems to lead to a reentrance pb, so comment it for now | 23 April 2008, 15:59:16 UTC |
b1ae247 | Francois Trahay | 23 April 2008, 15:19:44 UTC | Very ugly fix | 23 April 2008, 15:19:44 UTC |
02b2442 | Francois Trahay | 23 April 2008, 10:08:32 UTC | Fix locks in case of simultaneaous multirail on non-thread-safe drivers | 23 April 2008, 10:08:32 UTC |
bd65ec6 | Francois Trahay | 23 April 2008, 09:36:55 UTC | Fix locks | 23 April 2008, 09:36:55 UTC |
be0577d | Elisabeth Brunet | 22 April 2008, 19:09:46 UTC | pour l'instant on desactive l'echantillonage (les fichiers de perf sont en dur dans la base) car pm2-which ne retrouve pas le programme de test | 22 April 2008, 19:09:46 UTC |
b9901be | Elisabeth Brunet | 22 April 2008, 19:07:37 UTC | la forme des fichiers étaient incorrectes | 22 April 2008, 19:07:37 UTC |
24a9d56 | Elisabeth Brunet | 22 April 2008, 19:05:08 UTC | modification de la stratégie pour que les ack multiples soient correctement traités | 22 April 2008, 19:05:08 UTC |
82118b0 | Elisabeth Brunet | 22 April 2008, 19:01:06 UTC | nettoyage | 22 April 2008, 19:01:06 UTC |
84ddcaf | Elisabeth Brunet | 22 April 2008, 11:05:09 UTC | sauvegarde des echantillonages tant que les scripts ne marchent plus (cf pm2-which) | 22 April 2008, 11:05:09 UTC |
dbb2c9f | Elisabeth Brunet | 22 April 2008, 11:04:21 UTC | les signatures des prises de locks avaient changées | 22 April 2008, 11:04:21 UTC |
4e100f0 | Elisabeth Brunet | 22 April 2008, 09:13:44 UTC | include du sampling manquant + binding des tasklets des cores séparés | 22 April 2008, 09:13:44 UTC |
bdabe26 | Francois Trahay | 22 April 2008, 09:11:25 UTC | minor typo | 22 April 2008, 09:11:25 UTC |
a491da7 | Francois Trahay | 21 April 2008, 15:09:14 UTC | Separated locks in nmad. Now we should be able to process things in parallel. (very experimental !) | 21 April 2008, 15:09:14 UTC |
8f010b7 | Francois Trahay | 21 April 2008, 14:53:46 UTC | Fix missing dependency | 21 April 2008, 14:53:46 UTC |
bcb9c8d | Elisabeth Brunet | 21 April 2008, 14:38:15 UTC | ajout de la strategie all-split pour ezflavor | 21 April 2008, 14:38:15 UTC |
b379df4 | Elisabeth Brunet | 08 April 2008, 17:16:55 UTC | 1er depot qui splite les moyens suivant un ratio determiné par une interpolation lineaire sur l'echantillonnage | 08 April 2008, 17:16:55 UTC |
bc008bd | Elisabeth Brunet | 08 April 2008, 14:30:46 UTC | creation de la branche nmad-split2 (la même que namd-split mais qui avait été mal créée) à partir du trunk | 08 April 2008, 14:30:46 UTC |
2ffbb15 | Nathalie Furmento | 08 April 2008, 13:57:39 UTC | test MPI_Pack and MPI_Unpack | 08 April 2008, 13:57:39 UTC |
896a8cf | Nathalie Furmento | 08 April 2008, 13:51:09 UTC | new functions MPI_Pack and MPI_Unpack | 08 April 2008, 13:51:09 UTC |
94eaf03 | Nathalie Furmento | 08 April 2008, 12:49:30 UTC | define the datatype MPI_PACKED | 08 April 2008, 12:49:30 UTC |
82bbb9d | Nathalie Furmento | 08 April 2008, 12:24:43 UTC | define datatype MPI_REAL4 | 08 April 2008, 12:24:43 UTC |
13c0a01 | Nathalie Furmento | 08 April 2008, 12:22:28 UTC | MPI_REAL8 is not 8 reals but 1 real on 8 bytes | 08 April 2008, 12:22:28 UTC |
0bf9229 | Nathalie Furmento | 08 April 2008, 11:43:57 UTC | define the type MPI_REAL8 | 08 April 2008, 11:43:57 UTC |
b11760f | Alexandre Denis | 08 April 2008, 10:33:05 UTC | on definit MPI_ANY_TAG aussi en Fortran | 08 April 2008, 10:33:05 UTC |
51c4530 | Samuel Thibault | 01 April 2008, 16:02:57 UTC | documentation | 01 April 2008, 16:02:57 UTC |
b98eabb | Samuel Thibault | 01 April 2008, 15:33:20 UTC | make sure spread starts at the top | 01 April 2008, 15:33:20 UTC |
57b03a0 | Francois Trahay | 01 April 2008, 15:33:13 UTC | Fix wrong idleness detection | 01 April 2008, 15:33:13 UTC |
b3f7eda | Francois Trahay | 01 April 2008, 15:28:10 UTC | When using remote_tasklet we try to shcedule tasklet localy to avoid an expensive signal | 01 April 2008, 15:28:10 UTC |
f16fc1c | Samuel Thibault | 01 April 2008, 15:20:28 UTC | permit to create internal-purpose threads before regular VPs are created | 01 April 2008, 15:20:28 UTC |
2d1eae2 | Samuel Thibault | 01 April 2008, 15:03:12 UTC | put back a correct vpset check on thread creation | 01 April 2008, 15:03:12 UTC |
f44a432 | Cecile Romo-Glinos | 28 March 2008, 10:11:12 UTC | Oubli de correction de la definition de la variable request | 28 March 2008, 10:11:12 UTC |
e0ff904 | Cecile Romo-Glinos | 28 March 2008, 09:03:08 UTC | MPI Standardisation | 28 March 2008, 09:03:08 UTC |
cc785fa | Mathieu Faverge | 20 March 2008, 12:45:46 UTC | MPI_Testany doit retourner un index undefined dés qu'il trouve une requete nulle ou inactive | 20 March 2008, 12:45:46 UTC |
95ceb35 | Francois Trahay | 19 March 2008, 16:49:26 UTC | mx_dir option require mx | 19 March 2008, 16:49:26 UTC |
9fc0b5b | Nathalie Furmento | 18 March 2008, 09:02:14 UTC | provides a default value for MX_DIR in the case the option mx_dir is not enabled | 18 March 2008, 09:02:14 UTC |
8bef885 | Nathalie Furmento | 17 March 2008, 13:31:36 UTC | remove temporary printouts | 17 March 2008, 13:31:36 UTC |
6375874 | Nathalie Furmento | 17 March 2008, 13:30:28 UTC | new option -x which allows to expand the flavor name given as a parameter, e.g pm2-create-sample-flavors -x nmad will create all the flavors whose names contain nmad | 17 March 2008, 13:30:28 UTC |
6af21bc | Alexandre Denis | 17 March 2008, 13:14:16 UTC | warning | 17 March 2008, 13:14:16 UTC |
56bdad1 | Nathalie Furmento | 17 March 2008, 13:09:07 UTC | allows to specify another root directory for MX | 17 March 2008, 13:09:07 UTC |
2aa4036 | Nathalie Furmento | 17 March 2008, 13:08:48 UTC | allows to specify another root directory for MX | 17 March 2008, 13:08:48 UTC |
91bd988 | Nathalie Furmento | 17 March 2008, 13:03:11 UTC | disables optional value for option mx | 17 March 2008, 13:03:11 UTC |
624d361 | Samuel Thibault | 14 March 2008, 16:12:54 UTC | update go_marcel_main usage | 14 March 2008, 16:12:54 UTC |
244e409 | Alexandre Denis | 14 March 2008, 16:01:57 UTC | toilettage | 14 March 2008, 16:01:57 UTC |
5e8703a | Alexandre Denis | 14 March 2008, 15:58:15 UTC | passage explicite de la fonction dans go_marcel_main | 14 March 2008, 15:58:15 UTC |
9b25689 | Alexandre Denis | 14 March 2008, 15:57:33 UTC | on linke newmad explicitement avec Puk quand il est externe | 14 March 2008, 15:57:33 UTC |
3d856b7 | Alexandre Denis | 14 March 2008, 14:42:19 UTC | suppression d'une ref Mad2 devenue inutile | 14 March 2008, 14:42:19 UTC |
1f4f7a5 | Francois Broquedis | 14 March 2008, 14:11:48 UTC | Debug messages are off by default in marcel_bubble_affinity.c | 14 March 2008, 14:11:48 UTC |
f554915 | Francois Broquedis | 14 March 2008, 14:05:02 UTC | Removing some annoying prints in marcel_bubble_affinity.c and marcel_bubble_sched.c | 14 March 2008, 14:05:02 UTC |
77c148c | Francois Broquedis | 14 March 2008, 13:08:08 UTC | Fixed the debug-dependent unused variable in marcel_bubble_affinity.c | 14 March 2008, 13:08:08 UTC |
100134e | Francois Broquedis | 14 March 2008, 13:04:38 UTC | Translate the comments into english + correct indent in marcel_bubble_affinity.c | 14 March 2008, 13:04:38 UTC |
e3601e3 | Francois Trahay | 13 March 2008, 16:59:04 UTC | remove useless options | 13 March 2008, 16:59:04 UTC |
8ba1116 | Francois Trahay | 13 March 2008, 16:54:40 UTC | Fix ld issues from previous commit. | 13 March 2008, 16:54:40 UTC |
d69aa5e | Francois Trahay | 13 March 2008, 16:08:50 UTC | Possibility to choose MX's driver in a flavor | 13 March 2008, 16:08:50 UTC |
d64a314 | Nathalie Furmento | 12 March 2008, 22:10:37 UTC | - Use the command id to get the user name - Re-insert num in the log filename | 12 March 2008, 22:10:37 UTC |
225b523 | Nathalie Furmento | 12 March 2008, 19:49:07 UTC | - pm2-load -l results into log files named /tmp/pm2log-$USER-rank rank being a value in the range [1..number of processes] - Note to whoever is 'responsible' : Check leo_loaders, as there is a nasty trick to pass the rank of the process to leo-load | 12 March 2008, 19:49:07 UTC |
f0efaf7 | Alexandre Denis | 11 March 2008, 10:50:59 UTC | correction de la table de gate_id dans le lanceur Mad3 | 11 March 2008, 10:50:59 UTC |
2205042 | Nathalie Furmento | 10 March 2008, 15:28:00 UTC | remove useless variable | 10 March 2008, 15:28:00 UTC |
2f139f1 | Nathalie Furmento | 10 March 2008, 15:25:38 UTC | correct previous commit | 10 March 2008, 15:25:38 UTC |
e6db2ad | Nathalie Furmento | 10 March 2008, 15:22:40 UTC | flavor nmad-mpi no longer enables pioman&marcel by default, new flavor nmad-mpi-pioman which enables pioman&marcel | 10 March 2008, 15:22:40 UTC |
8eff83a | Francois Trahay | 10 March 2008, 10:05:49 UTC | Fix tasklets behavior * preemption becomes optional * "remote" tasklets' signals are correctly sent/catched* | 10 March 2008, 10:05:49 UTC |
4ade5c8 | Olivier Aumage | 07 March 2008, 17:16:00 UTC | - remove obsolete entries | 07 March 2008, 17:16:00 UTC |
13d8ef5 | Olivier Aumage | 07 March 2008, 17:08:56 UTC | - remove console module from PM2 trunk | 07 March 2008, 17:08:56 UTC |
52895f4 | Olivier Aumage | 07 March 2008, 17:08:28 UTC | - no more console to start | 07 March 2008, 17:08:28 UTC |
70b7e9a | Olivier Aumage | 07 March 2008, 17:05:22 UTC | - no more PM2 console | 07 March 2008, 17:05:22 UTC |
401924a | Olivier Aumage | 07 March 2008, 17:02:26 UTC | - remove leony module from PM2 trunk | 07 March 2008, 17:02:26 UTC |
cff3110 | Olivier Aumage | 07 March 2008, 17:01:39 UTC | - no more leony in the files | 07 March 2008, 17:01:39 UTC |
59efedf | Olivier Aumage | 07 March 2008, 17:00:33 UTC | - suppress leony script links | 07 March 2008, 17:00:33 UTC |
1600e96 | Alexandre Denis | 07 March 2008, 17:00:03 UTC | utilisation du lanceur generique par Mad-MPI dans tous les cas (leonie/PadicoTM/autres) | 07 March 2008, 17:00:03 UTC |
32661c2 | Alexandre Denis | 07 March 2008, 16:59:02 UTC | encapsulation du lanceur de proto_mad3 dans un composants NewMad_Launcher | 07 March 2008, 16:59:02 UTC |
b8d5c22 | Olivier Aumage | 07 March 2008, 16:58:10 UTC | - remove sigmund module from trunk | 07 March 2008, 16:58:10 UTC |
6ef38bd | Alexandre Denis | 07 March 2008, 16:57:24 UTC | initialisation paresseuse un peu plus intelligente | 07 March 2008, 16:57:24 UTC |
0122563 | Olivier Aumage | 07 March 2008, 16:56:23 UTC | - remove sigmund from PM2 scripts | 07 March 2008, 16:56:23 UTC |
d6da07d | Alexandre Denis | 07 March 2008, 16:56:05 UTC | on rend publique la detection d'initialisation de tbx | 07 March 2008, 16:56:05 UTC |
9eb4ac3 | Olivier Aumage | 07 March 2008, 16:52:18 UTC | - suppress sigmund scripts links | 07 March 2008, 16:52:18 UTC |
31fec71 | Olivier Aumage | 07 March 2008, 16:47:21 UTC | - suppress mad1, mad2 and dsm modules from trunk | 07 March 2008, 16:47:21 UTC |
f262486 | Olivier Aumage | 07 March 2008, 16:42:58 UTC | - remove mad1 mad2 and dsm modules from PM2 scripts | 07 March 2008, 16:42:58 UTC |
f1f55ed | Alexandre Denis | 07 March 2008, 09:33:47 UTC | pas besoin de #ifdef PADICO, une initialisation paresseuse suffit | 07 March 2008, 09:33:47 UTC |
342471b | Alexandre Denis | 07 March 2008, 09:25:17 UTC | correction init MPI | 07 March 2008, 09:25:17 UTC |
d6e0fcc | Alexandre Denis | 07 March 2008, 09:24:30 UTC | correction du type de gate_id | 07 March 2008, 09:24:30 UTC |
2d53720 | Nathalie Furmento | 07 March 2008, 07:42:45 UTC | disable option nuioa (file numa.h is not available on william which runs the nightly tests) | 07 March 2008, 07:42:45 UTC |