Samuel Thibault
|
01c27bfb70
Fix MCT time estimation: avoid including transfer time twice
|
8 år sedan |
Samuel Thibault
|
8ff5a532f4
Fix modular-heft-prio: tasks should be accumulated before the eager scheduler
|
8 år sedan |
Samuel Thibault
|
a94eec3bcb
Start factorizing queue completion times between worker component and dmda
|
8 år sedan |
Samuel Thibault
|
0e0c061054
Fix estimated ends of components, and use average ends by default
|
8 år sedan |
Samuel Thibault
|
bfc1b8ff4e
Fix paje trace when pipelining tasks and transfers
|
8 år sedan |
Samuel Thibault
|
3f440da88f
mention new scheduler
|
8 år sedan |
Samuel Thibault
|
e9dce8bf50
Add modular-heft-prio scheduler
|
8 år sedan |
Nathalie Furmento
|
a1f9243dd5
examples/sched_ctx/gpu_partition.c: temporary fix: skip test as cuda computation fails
|
8 år sedan |
Corentin Salingue
|
5d3f3a018a
Update doc with MPI Master-Slave support + fix visibility of 2 sections in MPI doc
|
8 år sedan |
Samuel Thibault
|
882ad99086
Do not assume that -fPIC means static linking, nowadays distributions compile all code with fPIC
|
8 år sedan |
Andra Hugo
|
149d06b650
do not even compile the gpu_partiton example if we don't have cuda
|
8 år sedan |
Samuel Thibault
|
3ff56dcc26
fix indent
|
8 år sedan |
Samuel Thibault
|
3b6ce806ec
Fix warnings
|
8 år sedan |
Nathalie Furmento
|
0e7aeb8008
src/Makefile.am: do not check _starpu_spin_init as it has 2 implementations, one being not inline
|
8 år sedan |
Samuel Thibault
|
cbe6f98e73
Make sure that all static inlines are indeed inlined
|
8 år sedan |
Andra Hugo
|
64107a11ad
gpu_partitioning - remove comments from cpu implementation
|
8 år sedan |
Nathalie Furmento
|
219bcae91d
doc:
|
8 år sedan |
Samuel Thibault
|
de8bed4c48
Add openmpi suppression
|
8 år sedan |
Samuel Thibault
|
2b6e2c26e5
fix memory leak: only really parallel tasks get freed
|
8 år sedan |
Andra Hugo
|
dc923e3168
allocate CPUs to take care of the gpu streams and do not use them as CPU workers + use a single gpu for the gpu_partitionong example
|
8 år sedan |
Nathalie Furmento
|
7585b099a8
tools/valgrind/valgrind.sh: do not need -v option when running valgrind
|
8 år sedan |
Nathalie Furmento
|
723c5e350f
tools/valgrind/hwloc.suppr: add suppression rules
|
8 år sedan |
Nathalie Furmento
|
0e31ebcd3f
mpi/tests: minor changes
|
8 år sedan |
Samuel Thibault
|
5543c45dc5
fix memleak
|
8 år sedan |
Corentin Salingue
|
2c1dcade11
add MPI Master Slave case in schedulers
|
8 år sedan |
Nathalie Furmento
|
cf5f930104
fix warnings detected by cppcheck
|
8 år sedan |
Corentin Salingue
|
87b1d4738b
MPI Master-Slave does not support Redux
|
8 år sedan |
Corentin Salingue
|
f96050ded6
give the configuration to StarPU
|
8 år sedan |
Samuel Thibault
|
64115beec5
Use asynchronous task input fetching for cuda and opencl too
|
8 år sedan |
Samuel Thibault
|
9362a8dd6c
Use MSG_process_join starting from simgrid 3.14
|
8 år sedan |