https://gitlab.inria.fr/pm2/pm2

sort by:
Revision Author Date Message Commit Date
211f8af Merge trunk -r15261:15509 in driver-query/ branch. These branches contain the same at this point. ********************************* * THIS BRANCH IS CLOSED FOR NOW * ********************************* 27 March 2007, 10:08:53 UTC
fb08201 useless TODO comment 26 March 2007, 09:57:59 UTC
11805f6 ping_multirails uses nm_core_drivers_load_init_some() in driver-query/ branch too 22 March 2007, 10:34:35 UTC
793fd01 print each driver url at the end of the loop in nm_core_drivers_load_init_some() so that we get MX url before quadrics does its barrier 22 March 2007, 10:33:47 UTC
0ae34e6 move nm_core_driver_load_init_some() to driver-query/ branch/ 22 March 2007, 10:19:34 UTC
d4d61de query is now required in the driver's ops 20 March 2007, 16:20:53 UTC
08c6a67 Port all drivers to query/init, uniformize and cleanup all this. query() initializes the driver's private structure in case there is a resource to pass to init(), and sets the common capabilities. 20 March 2007, 16:16:41 UTC
2f8756b we need at least to initialize mx in query() to be able to get some info 14 March 2007, 15:16:22 UTC
4f99901 Split driver.init() into: * driver.query() to choose the resource, register it, and set its capabilities * driver.init() to open the registered resource and initialize the nmad driver nm_core_driver_query() should be called by the application between load() and init(). But, nm_core_driver_init_load() takes care of doing all this for basic applications. It'll be easy to add an argument to query() so that we could choose which board should be open (first one, the one that has this capability, the one on this numa node, ...). The query() driver op is optional for now, the init() op can do everything if there is no point in doing something in query(). Only IB and MX have a query op for now. Not very nice for now, but should be enough to make numa binding decision across multiple networks before any driver actually allocates anything. 14 March 2007, 14:34:44 UTC
7f1307e Split nm_core_driver_init() into nm_core_driver_load() and nm_core_driver_init(). Add a basic helper nm_core_driver_load_init() doing both load+init to avoid having to change all existing programs too much (especially since they don't tweak anything between load and init). 14 March 2007, 10:10:27 UTC
653ae39 New branch to work on splitting driver.init into load+query+init 14 March 2007, 10:04:30 UTC
5d29f85 leonie sends local value of variable PM2_ROOT to nodes 14 March 2007, 09:57:57 UTC
39ac329 add option fortran_target_none for module common 14 March 2007, 09:27:35 UTC
c5ea3ae add option fortran_target_none for module common 14 March 2007, 09:27:09 UTC
3a2c70f also compares isend 13 March 2007, 16:25:14 UTC
cd4b52d quadrics is named qsnet in mad4 configuration files 13 March 2007, 16:24:39 UTC
0a8a255 manage several protocols when using leonie 13 March 2007, 16:23:58 UTC
e445285 optimisation IB et correction de la gestion des iovecs 13 March 2007, 16:12:56 UTC
03b0447 - Fortran compiler target query in pm2-config 13 March 2007, 15:18:26 UTC
072be14 - Fortran compiler-aware mpi_init 13 March 2007, 15:00:04 UTC
296dd4d - Fortran compiler targets, needed for selecting the right set of compiler dependent iargc/getarg routines 13 March 2007, 14:58:52 UTC
c61b661 - fixes and add in datatype mapping, new mpif.h Fortran header 13 March 2007, 11:10:18 UTC
ef7fc32 ajout d'un champ 'rdv_threshold' dans les capabilities des drivers 13 March 2007, 11:09:06 UTC
ed05506 print the bandwidth in ping_multirails since it matters a little bit more than the latency :) 13 March 2007, 11:03:42 UTC
ad0af20 r7789@implementation: samy | 2007-03-13 10:38:45 +0100 cette syntaxe n'est pas supportée par tous les shells 13 March 2007, 09:38:49 UTC
c09e349 force completed mode anyway when reaching maximum sequence number 12 March 2007, 17:23:47 UTC
edbffec force completed mode when reaching maximum sequence number 12 March 2007, 16:40:35 UTC
73f2a7e enregistrement dynamique avec pipeline dans le driver IB 12 March 2007, 16:37:45 UTC
6ca2b2b all2all 12 March 2007, 16:25:44 UTC
b1ff8d5 all2all 12 March 2007, 16:25:17 UTC
c3b1e5b avoid gcc warnings 12 March 2007, 15:42:22 UTC
d7a9ed1 compiler for C++ 12 March 2007, 15:37:48 UTC
6617ee1 - mpi_waitall Fortran binding 12 March 2007, 15:15:18 UTC
c73f123 - mpi_waitall 12 March 2007, 14:54:25 UTC
0397fe6 do not show output of leonie compilation 12 March 2007, 13:11:48 UTC
bf64bf6 new parameter -m to start application within a graphical console 12 March 2007, 13:10:54 UTC
2d237a6 r7787@implementation: samy | 2007-03-12 11:25:04 +0100 récupération des arguments à revoir 12 March 2007, 10:25:05 UTC
a5fe265 r7785@implementation: samy | 2007-03-12 11:02:49 +0100 oublie de déréférencement du pointeur 12 March 2007, 10:02:50 UTC
f423098 r7777@implementation: samy | 2007-03-12 10:28:51 +0100 fonctions Fortran pour étaler les bulles comme il faut 12 March 2007, 09:29:02 UTC
d6b845b r7767@implementation: samy | 2007-03-08 15:17:55 +0100 r7766@implementation (orig r1088): sthibault | 2007-03-08 15:16:14 +0100 augmenter les chances que ça passe 12 March 2007, 09:28:56 UTC
2697d2d reorganisation pour driver IB multi-methode 09 March 2007, 16:26:55 UTC
1805866 - double datatype for MIN reduce 09 March 2007, 15:19:26 UTC
b37e55d - Mad_MPI Fortran bindings: add reduce/allreduc, and some additional MPI data types 09 March 2007, 13:39:54 UTC
eb5fa56 - mpi bindings, ongoing work 09 March 2007, 10:28:58 UTC
e118a2e Correction du changement de priorité d'entité qd celle-ci a été déclarée avant le profile_activate 09 March 2007, 10:25:15 UTC
e2c7dc0 update definition of PM2_BUILD_DIR 08 March 2007, 20:31:28 UTC
17d3504 ignore build directory 08 March 2007, 19:06:48 UTC
c989c37 commentaires 08 March 2007, 17:07:11 UTC
ef08188 -- PM2_BUILD_DIR par défaut à $PM2_ROOT/build au lieu de $HOME/build -- compilation dans un répertoire dépendant de l'archi 08 March 2007, 17:03:16 UTC
8398304 premier jet d'un driver IB avec enregistrement dynamique 08 March 2007, 15:59:40 UTC
e0c48c1 add empty line after contents of mod.sh file (in case file does not end with a new line (cf. Brice ...) 08 March 2007, 15:45:58 UTC
320f0ec - alias fortran 08 March 2007, 14:18:15 UTC
bbdcc12 r7764@implementation: samy | 2007-03-08 15:09:47 +0100 variable inutilisée 08 March 2007, 14:09:53 UTC
54af4bb r7760@implementation: samy | 2007-03-08 15:02:07 +0100 correction vpset de la machine 08 March 2007, 14:02:12 UTC
424774a r7759@implementation: samy | 2007-03-08 15:01:54 +0100 correction dépendance 08 March 2007, 14:02:09 UTC
f6e7244 r7747@implementation: samy | 2007-03-08 14:26:15 +0100 correction bulles 08 March 2007, 13:41:41 UTC
8d0201e flavor no longer needed 08 March 2007, 12:58:11 UTC
2492b81 ibverbs is a prefered protocol 08 March 2007, 12:46:03 UTC
cb97225 improve error message 08 March 2007, 11:13:55 UTC
6f202b6 improve error message 08 March 2007, 11:07:06 UTC
d74f40d load sisci and ibverbs drivers 08 March 2007, 11:06:32 UTC
d2ae305 - mpif77/90 scripts 08 March 2007, 10:13:13 UTC
ae1376a support de whiptail (en plus de 'dialog') 08 March 2007, 09:41:54 UTC
7bc7339 merge update of marcel-bubblegum 08 March 2007, 08:22:01 UTC
8e2c8dc r7735@implementation: samy | 2007-03-07 19:38:28 +0100 correction détection multicore 07 March 2007, 18:39:47 UTC
7f4294a r7734@implementation: samy | 2007-03-07 19:28:08 +0100 correction détection multicore 07 March 2007, 18:39:45 UTC
c9b4e91 r7731@implementation: samy | 2007-03-07 18:06:41 +0100 r7729@implementation (orig r1085): sthibault | 2007-03-07 18:05:23 +0100 découpler la numérotation physique de la numérotation logique 07 March 2007, 18:39:43 UTC
1036668 r7730@implementation: samy | 2007-03-07 18:06:41 +0100 r7677@implementation (orig r1084): sthibault | 2007-03-01 19:26:22 +0100 binder les LWPs sur les CPUs ! 07 March 2007, 18:39:41 UTC
2e13216 r7728@implementation: samy | 2007-03-07 11:10:32 +0100 éviter les warnings 07 March 2007, 18:39:39 UTC
03b74d7 define path for dynamic libraries 07 March 2007, 15:00:58 UTC
0880f47 log messages 07 March 2007, 11:13:31 UTC
0ffd87f call make with generated Makefile 07 March 2007, 10:47:58 UTC
d76a5f4 call pm2-config 07 March 2007, 10:43:41 UTC
5af1f92 generates a Makefile to compile application 07 March 2007, 10:36:57 UTC
f15ae46 r7724@implementation: samy | 2007-03-07 10:53:23 +0100 la doc ça ne fait jamais de mal 07 March 2007, 09:53:27 UTC
6865a54 do not trigger automatic flushing when any source 07 March 2007, 09:20:14 UTC
e890c61 r7716@implementation: samy | 2007-03-06 08:08:37 +0100 trouver tout seul le chemin vers la trace par défaut 06 March 2007, 16:24:21 UTC
20317fb r7715@implementation: samy | 2007-03-05 18:40:31 +0100 correction dépendances 06 March 2007, 16:24:17 UTC
14f51bc fix bug (thanks to Jerome\!) 06 March 2007, 16:16:08 UTC
9f766d0 mx is the preferred protocol (for MadMPI) 06 March 2007, 13:39:49 UTC
89a9859 set flavor when calling pm2-config 06 March 2007, 13:38:44 UTC
d5c61c2 can specify flavor with parameter -flavor 06 March 2007, 13:14:51 UTC
28f051d new flavor marcel-bubblegum + profile autostart 06 March 2007, 11:46:52 UTC
178b8ac r7713@implementation: samy | 2007-03-05 18:18:50 +0100 guillemets pour protéger les mots 05 March 2007, 17:18:52 UTC
4f0cb12 r7702@implementation: samy | 2007-03-05 16:14:21 +0100 r7694@implementation (orig r499): sthibaul | 2007-03-05 16:09:28 +0100 maniere plus legere de verifier l'execution 05 March 2007, 15:17:00 UTC
7b211f1 r7701@implementation: samy | 2007-03-05 16:14:20 +0100 r7693@implementation (orig r498): sthibaul | 2007-03-05 14:26:06 +0100 support non-linux 05 March 2007, 15:16:58 UTC
3cf7399 r7700@implementation: samy | 2007-03-05 16:14:19 +0100 r7692@implementation (orig r497): sthibaul | 2007-03-05 14:19:13 +0100 support ppc 05 March 2007, 15:16:56 UTC
a1c49ca r7699@implementation: samy | 2007-03-05 16:14:18 +0100 r7691@implementation (orig r496): sthibaul | 2007-03-05 14:18:35 +0100 support ppc 05 March 2007, 15:16:54 UTC
ca37a0d r7698@implementation: samy | 2007-03-05 16:14:18 +0100 r7690@implementation (orig r495): sthibaul | 2007-03-05 14:18:16 +0100 support ppc 05 March 2007, 15:16:52 UTC
de870b8 r7697@implementation: samy | 2007-03-05 16:14:17 +0100 r7689@implementation (orig r494): sthibaul | 2007-03-05 14:04:22 +0100 support darwin 64 05 March 2007, 15:16:51 UTC
dfd7133 r7696@implementation: samy | 2007-03-05 16:14:16 +0100 r7688@implementation (orig r493): sthibaul | 2007-03-05 14:04:09 +0100 aligner la taille des variables locales 05 March 2007, 15:16:48 UTC
23efc77 r7695@implementation: samy | 2007-03-05 16:13:59 +0100 r7173@implementation (orig r490): sthibaul | 2007-01-25 21:49:29 +0100 include fixup for timercmp 05 March 2007, 15:16:45 UTC
76d1528 trigger automatic flushing when reaching maximum sequence number and last sequence not completed 05 March 2007, 13:27:04 UTC
db33136 r7683@implementation: samy | 2007-03-05 11:23:04 +0100 dire à l'utilisateur ce qu'il doit faire 05 March 2007, 10:23:08 UTC
666035b - missing export 05 March 2007, 10:18:07 UTC
8ebf136 r7681@implementation: samy | 2007-03-05 10:35:08 +0100 r7680@implementation (orig r109): thibault | 2007-03-05 10:37:05 +0100 tbx_timing pour ppc64 05 March 2007, 09:35:28 UTC
99fb953 update help 02 March 2007, 09:48:08 UTC
7ebc0b1 r7649@implementation: samy | 2007-03-01 13:55:34 +0100 oubli des options communes (build_dynamic notamment) 01 March 2007, 12:56:53 UTC
673bff4 Pour éviter les erreurs de compilation quand on n'utilise pas le profiling, on rajoute un ifdeef autour de profile_stop() 01 March 2007, 09:14:35 UTC
58b16bd Il manque un parametre dans le prototype de xpaul_callback_will_block 01 March 2007, 09:13:09 UTC
back to top