 Samuel Thibault
|
804e6f8233
properly test for task termination
|
11 gadi atpakaļ |
 Marc Sergent
|
0026768555
Adding the handling of multiple fathers in hierarchical schedulers,
|
11 gadi atpakaļ |
|
Marc Sergent
|
e998e008c6
Merge from trunk @11714:11788
|
11 gadi atpakaļ |
 Andra Hugo
|
2769aaa8cb
hypervisor: correct computation of idle
|
11 gadi atpakaļ |
|
Samuel Thibault
|
4d81325ef7
fix missing initialization
|
11 gadi atpakaļ |
|
Samuel Thibault
|
b8062a9fba
keep pushed back tasks in proper order
|
12 gadi atpakaļ |
|
Andra Hugo
|
20906d8d77
hyp only: optimise resize (use idle and speed for the triggering policy)
|
12 gadi atpakaļ |
|
Andra Hugo
|
6266721e89
do not take the mutex if the current worker is calling the code
|
12 gadi atpakaļ |
|
Andra Hugo
|
39b06ba608
do not take changing_ctx_mutex when finishing executing the tasks (in decrement_nsubmitted_tasks), just check if the counter is 1, do what you have to do, and then decrement it (decrement the counter while verifying it wakes up the the thread calling wait_for_all)
|
12 gadi atpakaļ |
|
Marc Sergent
|
fe685563e7
Adding several features which needs each other to compute correctly :
|
12 gadi atpakaļ |
|
Samuel Thibault
|
b70f84e342
Properly find the task which optimizes computation time benefit
|
12 gadi atpakaļ |
|
Samuel Thibault
|
a62e0dba32
better comment
|
12 gadi atpakaļ |
|
Samuel Thibault
|
a06ea29938
Use separate expected ends, to avoid systematically preferring small tasks, possibly leading to starvation
|
12 gadi atpakaļ |
|
Samuel Thibault
|
4f82d35ce9
Avoid keeping the lock held while scheduling tasks, to let scheduling happen in parallel. This means several processors might take 5 tasks each, thus emptying the fifo, but at least we don't have a huge contention point
|
12 gadi atpakaļ |
|
Samuel Thibault
|
23b0ea64b9
Add a heft variant which draws tasks from its own fifo
|
12 gadi atpakaļ |
|
Samuel Thibault
|
30c9f2732c
Plug perfmodel_select children normally from the scheduling strategy, instead of internally within perfmodel_select. This makes the room call go through the MCT node
|
12 gadi atpakaļ |
|
Samuel Thibault
|
f7684ba8e2
keep the original data transfer time for the beta penalty, otherwise it seems we do not take communications into account enough to achieve overlapping. Proper communication scheduling would allow to replace this heuristic
|
12 gadi atpakaļ |
|
Samuel Thibault
|
b693f565e0
Handle case when there is no worker any more for a given scheduling node
|
12 gadi atpakaļ |
|
Samuel Thibault
|
ae4b92a838
Add missing files
|
12 gadi atpakaļ |
|
Samuel Thibault
|
11266adce9
FIXME
|
12 gadi atpakaļ |
|
Samuel Thibault
|
2441190259
factorize MCT code
|
12 gadi atpakaļ |
|
Samuel Thibault
|
231fa5c8d4
fix spacing
|
12 gadi atpakaļ |
|
Samuel Thibault
|
2240d43c8f
more cleanup
|
12 gadi atpakaļ |
|
Samuel Thibault
|
716c571d26
clean up MCT computations
|
12 gadi atpakaļ |
|
Marc Sergent
|
7a983bdaa4
Correcting copyrights on src/sched_policies/
|
12 gadi atpakaļ |
 Nathalie Furmento
|
0d7c8ed948
merge trunk@9680, trunk@9683: tests/sched_policies/prio.c: use portable random function, drand48 returns values between 0 and 1
|
12 gadi atpakaļ |
|
Marc Sergent
|
496771f13c
src/sched_policies/node_worker.c : Bug fix concerning the way to get a combined worker's id.
|
12 gadi atpakaļ |
|
Marc Sergent
|
db2bcd07f6
Merge from trunk @11646:11713
|
12 gadi atpakaļ |
|
Nathalie Furmento
|
ed41317844
examples/pi/SobolQRNG/sobol_gold.c: include strings.h
|
12 gadi atpakaļ |
|
Marc Sergent
|
76691542fa
Adding a prefetch field in the task' structure to optimize unneeded prefetching calls, and adapting default threshold values for some predefined hierarchical schedulers.
|
12 gadi atpakaļ |
