Partitioned/clustered scheduling

2013-11-19 Thread Sebastian Huber
Hello, we would like to implement partitioned/clustered scheduling for SMP RTEMS. The reasons for this are highlighted in Björn B. Brandenburg, Scheduling and Locking in Multiprocessor Real-Time Operating Systems, 2011 Partitioned/clustered scheduling means that the set of processors of a

[PATCH 1/3] scheduler/EDF: Use unprotected insert and extract

2013-11-19 Thread Sebastian Huber
Interrupts are disabled by the caller _Thread_Change_priority() or _Thread_Set_transient() or directly in the scheduler operation. Thus there is no need to use protected variants. --- cpukit/score/src/scheduleredfenqueue.c |2 +- cpukit/score/src/scheduleredfextract.c |2 +- cpukit/score/

[PATCH 3/3] score/rbtree: Remove "unprotected" from API

2013-11-19 Thread Sebastian Huber
--- cpukit/posix/src/keyfreememory.c | 10 +- cpukit/posix/src/keygetspecific.c |2 +- cpukit/posix/src/keyrundestructors.c |2 +- cpukit/posix/src/keysetspecific.c |2 +- cpukit/sapi/include/rtems/rbtree.h| 60 +++

[PATCH 2/3] score/rbtree: Delete protected operations

2013-11-19 Thread Sebastian Huber
The user of the red-black tree container must now ensure that at most one thread at once can access an instance. --- cpukit/sapi/include/rtems/rbtree.h| 112 --- cpukit/score/Makefile.am |4 +- cpukit/score/include/rtems/score/rbtree.h | 141 ++---

Re: Partitioned/clustered scheduling

2013-11-19 Thread Gedare Bloom
On Tue, Nov 19, 2013 at 8:39 AM, Sebastian Huber wrote: > Hello, > > we would like to implement partitioned/clustered scheduling for SMP RTEMS. > The reasons for this are highlighted in > > Björn B. Brandenburg, Scheduling and Locking in Multiprocessor Real-Time > Operating Systems, 2011 > > Parti

Re: [PATCH 1/3] scheduler/EDF: Use unprotected insert and extract

2013-11-19 Thread Gedare Bloom
Looks good to me. On Tue, Nov 19, 2013 at 10:54 AM, Sebastian Huber wrote: > Interrupts are disabled by the caller _Thread_Change_priority() or > _Thread_Set_transient() or directly in the scheduler operation. Thus > there is no need to use protected variants. > --- > cpukit/score/src/scheduler

Re: [PATCH 2/3] score/rbtree: Delete protected operations

2013-11-19 Thread Gedare Bloom
OK On Tue, Nov 19, 2013 at 10:54 AM, Sebastian Huber wrote: > The user of the red-black tree container must now ensure that at most > one thread at once can access an instance. > --- > cpukit/sapi/include/rtems/rbtree.h| 112 --- > cpukit/score/Makefile.am

Re: [PATCH 3/3] score/rbtree: Remove "unprotected" from API

2013-11-19 Thread Gedare Bloom
This is OK. I have one concern which is that any code that was developed against the "protected" api will compile. We'll need to make a note of this change so that users can beware. On Tue, Nov 19, 2013 at 10:54 AM, Sebastian Huber wrote: > --- > cpukit/posix/src/keyfreememory.c |

Re: Partitioned/clustered scheduling

2013-11-19 Thread Gedare Bloom
On Tue, Nov 19, 2013 at 12:36 PM, Gedare Bloom wrote: > On Tue, Nov 19, 2013 at 8:39 AM, Sebastian Huber > wrote: >> Hello, >> >> we would like to implement partitioned/clustered scheduling for SMP RTEMS. >> The reasons for this are highlighted in >> >> Björn B. Brandenburg, Scheduling and Lockin

Re: Partitioned/clustered scheduling

2013-11-19 Thread Chris Johns
On 20/11/2013 12:39 am, Sebastian Huber wrote: Hello, we would like to implement partitioned/clustered scheduling for SMP RTEMS. The reasons for this are highlighted in What time frame are you looking at for this ? Björn B. Brandenburg, Scheduling and Locking in Multiprocessor Real-Time Op