starpu merge requestshttps://gitlab.inria.fr/starpu/starpu/-/merge_requests2020-06-08T16:37:54+02:00https://gitlab.inria.fr/starpu/starpu/-/merge_requests/1Add STARPU_DISPLAY_BINDINGS env var to display all bindings2020-06-08T16:37:54+02:00Philippe SWARTVAGHERAdd STARPU_DISPLAY_BINDINGS env var to display all bindings3 cases to test:
- an MPI program
- a non-MPI program, with StarPU built with MPI
- a non-MPI program, with StarPU built without MPI3 cases to test:
- an MPI program
- a non-MPI program, with StarPU built with MPI
- a non-MPI program, with StarPU built without MPIhttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/2Draft: WIP: Nmad coop coll dynamic interface2022-06-29T13:53:30+02:00Philippe SWARTVAGHERDraft: WIP: Nmad coop coll dynamic interfaceA big one !
- add dynamic broadcasts in nmad backend
- add better clock synchronization for traces (when used with nmad backend)
- add some benchmarks related to distributed StarPU
- add priority and type (collective or point-to-point) ...A big one !
- add dynamic broadcasts in nmad backend
- add better clock synchronization for traces (when used with nmad backend)
- add some benchmarks related to distributed StarPU
- add priority and type (collective or point-to-point) of communications in comms.rec trace file
- and some other little stuff I don't remember :smiley:
Remaining TODOs:
- [ ] documentation
- [ ] ChangeLog
- [ ] (comments of this merge request)
- [ ] test branch with backend MPI on buildbot
- [ ] test branch with backend MPI on ci.inria.frstarpu 1.4Nathalie FurmentoNathalie Furmentohttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/3A codelet can now define a callback function pointer which will be called whe...2020-10-19T15:02:47+02:00Nathalie FurmentoA codelet can now define a callback function pointer which will be called when...A codelet can now define a callback function pointer which will be called when the task does not define itself a callback function.
Fix issue #5A codelet can now define a callback function pointer which will be called when the task does not define itself a callback function.
Fix issue #5https://gitlab.inria.fr/starpu/starpu/-/merge_requests/5FxT2021-05-13T13:38:18+02:00Nathalie FurmentoFxTFix issue #4
Remaining TODOs:
- [x] check performances tasks_size_overhead with scheduler lwsFix issue #4
Remaining TODOs:
- [x] check performances tasks_size_overhead with scheduler lwsstarpu 1.4https://gitlab.inria.fr/starpu/starpu/-/merge_requests/6Add env var STARPU_WORKERS_COREID2020-11-03T16:31:55+01:00Philippe SWARTVAGHERAdd env var STARPU_WORKERS_COREIDThe environmenent variable `STARPU_WORKERS_COREID` acts like `STARPU_WORKERS_CPUID` but binds workers to cores instead of hyperthread.
Please make a review of my changes.The environmenent variable `STARPU_WORKERS_COREID` acts like `STARPU_WORKERS_CPUID` but binds workers to cores instead of hyperthread.
Please make a review of my changes.https://gitlab.inria.fr/starpu/starpu/-/merge_requests/7Python autotools2020-11-05T10:26:53+01:00Nathalie FurmentoPython autotoolsCompile and test Python with the autotoolsCompile and test Python with the autotoolshttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/8Fxt Trace trace name2020-11-05T18:35:53+01:00Nathalie FurmentoFxt Trace trace namecould we merge the following change ?
src: write fxt event for trace name earlier so that a graph generated before the task execution will contain the task names
it does not seem to break anything, @thibault can you please confirm ?could we merge the following change ?
src: write fxt event for trace name earlier so that a graph generated before the task execution will contain the task names
it does not seem to break anything, @thibault can you please confirm ?https://gitlab.inria.fr/starpu/starpu/-/merge_requests/10mpi: add clock synchronization routines and benchmark2021-01-27T11:16:54+01:00Philippe SWARTVAGHERmpi: add clock synchronization routines and benchmarkAdd distributed clock synchronization routines.
@thibault @furmento @adenis can you please have a look on this MR, especially regarding the integration of the file [mpi_sync_clocks.h](https://gitlab.inria.fr/pswartva/starpu/-/blob/2fc51...Add distributed clock synchronization routines.
@thibault @furmento @adenis can you please have a look on this MR, especially regarding the integration of the file [mpi_sync_clocks.h](https://gitlab.inria.fr/pswartva/starpu/-/blob/2fc51875df55db41c15d7661ccb774285a659924/mpi/src/mpi_sync_clocks/mpi_sync_clocks.h):
- the `typedef` in [starpu_mpi.h](https://gitlab.inria.fr/pswartva/starpu/-/blob/2fc51875df55db41c15d7661ccb774285a659924/mpi/include/starpu_mpi.h#L809) (maybe there is better solution than a `typedef` to a `void*` ?)
- the wrappers in [starpu_mpi_sync_clocks.c](https://gitlab.inria.fr/pswartva/starpu/-/blob/2fc51875df55db41c15d7661ccb774285a659924/mpi/src/mpi_sync_clocks/starpu_mpi_sync_clocks.c): they are really basic wrappers, maybe we can make something cleaner/simpler with some `#define` ?https://gitlab.inria.fr/starpu/starpu/-/merge_requests/11mpi: perform reduction only over contributing nodes2021-02-19T11:34:07+01:00Antoine Jegompi: perform reduction only over contributing nodes- added a long to describe a redux_map of contributing nodes
- added a test program to assess the behaviour
fixes #3- added a long to describe a redux_map of contributing nodes
- added a test program to assess the behaviour
fixes #3https://gitlab.inria.fr/starpu/starpu/-/merge_requests/12starpupy adapts to MPI master/slave model2021-02-01T20:24:21+01:00HE Kunstarpupy adapts to MPI master/slave modelhttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/13STARPU_MPI_REDUX starpu_data_access_mode to only redux between MPI and not wi...2021-02-16T20:09:47+01:00Antoine JegoSTARPU_MPI_REDUX starpu_data_access_mode to only redux between MPI and not within nodesThis allows to perform a reduction between processes (typically inter-node) but not within workers of multiple nodes (no intra-node reduction). This is required when the data is huge, and we do not want to replicate as many times as ther...This allows to perform a reduction between processes (typically inter-node) but not within workers of multiple nodes (no intra-node reduction). This is required when the data is huge, and we do not want to replicate as many times as there are workers, but only as many times as there are MPI processes.https://gitlab.inria.fr/starpu/starpu/-/merge_requests/14Fix bugs in starpu_fxt_tool2021-02-14T20:07:25+01:00Lucas NesiFix bugs in starpu_fxt_toolThis fixes:
- Duplicated entries on paje caused by resetting the symbol_list at each MPI trace file.
- Workers` unknown names because of a missing starpu_drivers_preinit.This fixes:
- Duplicated entries on paje caused by resetting the symbol_list at each MPI trace file.
- Workers` unknown names because of a missing starpu_drivers_preinit.https://gitlab.inria.fr/starpu/starpu/-/merge_requests/15Distributed CG2021-02-22T17:10:20+01:00Philippe SWARTVAGHERDistributed CGHere is a distributed implementation of the Conjugate Gradient. It's based on the basic example already in StarPU examples/cg/cg.c, but distributes data among MPI processes and submits tasks taking into account MPI.
Any review is welcom...Here is a distributed implementation of the Conjugate Gradient. It's based on the basic example already in StarPU examples/cg/cg.c, but distributes data among MPI processes and submits tasks taking into account MPI.
Any review is welcomed. :)
*Background*: I'm looking for an application memory-bound with communications overlapped by computations. The Conjugate Gradient seems to be a good candidate. Since I want communications, I'm not especially looking to optimize/reduce them.https://gitlab.inria.fr/starpu/starpu/-/merge_requests/16Add more priorities/options to data requests2021-03-04T02:45:24+01:00Lucas NesiAdd more priorities/options to data requestsThis:
- Do changes on data requests traces
- Add priorities on data requests from MPI
- By adding priority to starpu_data_acquire_on_node_cb_sequential_consistency_sync_jobids
- By adding new function starpu_mpi_irecv_detached_prio
-...This:
- Do changes on data requests traces
- Add priorities on data requests from MPI
- By adding priority to starpu_data_acquire_on_node_cb_sequential_consistency_sync_jobids
- By adding new function starpu_mpi_irecv_detached_prio
- Add option (STARPU_MPI_EARLYDATA_ALLOCATE) to MPI driver to do early data request allocations and do not block too much.
- Add option (STARPU_CUDA_ONLY_FAST_ALLOC_OTHER_MEMNODES) to CUDA workers do not do slow allocations on other memnodes (RAM pinned memory allocations)
- During the beginning of the execution the CUDA workers will not be slowed down.
- Removes datawizard_progress from fetch_data_on_node as it can fail.
- Add priorities on data requests from _starpu_fetch_task_inputhttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/18Update/auto heteroprio2021-06-25T11:44:51+02:00Nathalie FurmentoUpdate/auto heteropriohttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/20Record in traces on which NUMA nodes are allocated buffers2022-02-25T10:44:38+01:00Philippe SWARTVAGHERRecord in traces on which NUMA nodes are allocated buffershttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/21Mpi redux tree2021-06-21T18:56:29+02:00Antoine JegoMpi redux treeExtend `starpu_mpi_redux_data` to allow the reduction to proceed through a n-ary tree.
The default behaviour corresponds to n being the number of contributors to the reduction.Extend `starpu_mpi_redux_data` to allow the reduction to proceed through a n-ary tree.
The default behaviour corresponds to n being the number of contributors to the reduction.Antoine JegoAntoine Jegohttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/22Remove useless AM_CONDITIONAL STARPU_USE_PAPI2021-05-12T10:42:18+02:00Philippe SWARTVAGHERRemove useless AM_CONDITIONAL STARPU_USE_PAPIWhile using StarPU's configure.ac as an example to use autotools in another project, I noticed this `STARPU_USE_PAPI` is used nowhere in the project, so I think we can delete this line from the configure.ac.
Since I'm not completely sur...While using StarPU's configure.ac as an example to use autotools in another project, I noticed this `STARPU_USE_PAPI` is used nowhere in the project, so I think we can delete this line from the configure.ac.
Since I'm not completely sure about this change (maybe this line is really necessary !) I make a MR rather than directly committing to master.https://gitlab.inria.fr/starpu/starpu/-/merge_requests/23Plugin2021-05-29T13:30:34+02:00Nathalie FurmentoPluginAdd eclipse plugin for StarPUAdd eclipse plugin for StarPUhttps://gitlab.inria.fr/starpu/starpu/-/merge_requests/24WIP: transactions2022-06-27T22:08:32+02:00Nathalie FurmentoWIP: transactionsMerge branch transactions
cc @aumage
Remaining TODOs:
- [x] documentation
- [x] ChangeLogMerge branch transactions
cc @aumage
Remaining TODOs:
- [x] documentation
- [x] ChangeLogstarpu 1.4