- Fix SWIG interface to correctly manage things
- PythonPolicyManager isn't a singleton anymore. Since I did this in
a bit of a hurry, check the modifications are right
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@725 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Delete the "serialize()" method of concrete_environment, since useless
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@721 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Work on step_forward() a little more
- Add extra check inside dynamic_thread()::decrease_remaining_time()
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@717 3ecf2c5c-341e-0410-92b4-d18e462d057c
from configure. This is mainly there for buggy compilers / binutils.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@715 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Commented a lot of code to make it compile-able. But still it doesn`t link (at least for me...)
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@710 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Delegate plugin loading to PluginManager so that it can
be tested
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@709 3ecf2c5c-341e-0410-92b4-d18e462d057c
if they want to schedule Threads or Processes altogether
- Move setter methods for last_acquisition/last_release from
DynamicSchedulable to DynamicThread
- Rewrite aforesaid methods, along with the respective getter
methods, into Dynamic(Thread|Process)
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@708 3ecf2c5c-341e-0410-92b4-d18e462d057c
rather thant ${prefix}/include/sgpemv2: users will
be able to use directives like "#include <sgpemv2/scheduler.hh>"
in their headers, which reduces name clashes.
- Make _mutex a protected member of Singleton, so that
inheriters can recycle it :-)
- Add exclusive access control for step_forward()
- Add two states (future and exhausted) to Request
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@707 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Add get_request() method to (Dynamic)SubRequest.
- Implement most of ConcreteHistory. It is a fairly complex class,
with some real evilness in it. It should be thouroughly documented
ASAP.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@699 3ecf2c5c-341e-0410-92b4-d18e462d057c
- TODO Integrate this new plugin management system into the application
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@697 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Fixed all Dynamic and Static entities lacking proper destructors:
Dynamic entities need to delete Dynamic children, Static entities
need only to delete them from the list of their siblings
- Added methods to get Static "core" from Dynamic*.
- Now get_core() is a pure virtual function into DynamicSchedulable.
A quite nice solution(?), really, if I can say that myself... ;-)
- ConcreteHistory is half-implemented. It's the other half that
worries me.
- TODO: finish off ConcreteHistory, and check that things get destroyed
properly (no leaks allowed, use valgrind).
- TODO: rework Simulation
- TODO: Scheduler rewrite
- *A side note*: this code is becoming a mess. I prefer references over
pointers, but someone other prefer pointers over references, and what
happens is that you've to continously pass from one to another when
invoking other methods... this is bad.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@694 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Fix ReadyQueue constructor
- Change DynamicSubRequest to take an int as a parameter
- Implement ConcreteEnvironment::get_request_queue() (my word, it's ugly!)
- Please note that it still doesn't compile right: ConcreteHistory
and Scheduler need to be radically changed
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@692 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Add interfaces for Process and Thread into the "sgpem.i" SWIG
interface file, change DynamicSchedulable into Schedulable
- Add dynamic_cast for the return value of ReadyQueue::get_item_at()
into the pyloader "sgpem.i" SWIG interface, so that a Schedulable
can be either recognized as a Thread or a Process
- TODO: wrap STL exceptions in SWIG interface
- Please note that code won't compile until the new History and
Scheduler::step_forward() will be in place. This is a known issue.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@689 3ecf2c5c-341e-0410-92b4-d18e462d057c
(when appropriate)
- FIXME: see if const std::vector<Something*>& let you invoke
a non-const method of *Something. Else copy these values to
const std::vector<const Something*> before returning.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@686 3ecf2c5c-341e-0410-92b4-d18e462d057c
placeholder for the ready queue, needed by environment, and factorized
stubs for tests in a separate directory. Updated makefile including environment, concrete_environment, ready_queue, but NOT the stubs and/or the tests.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@685 3ecf2c5c-341e-0410-92b4-d18e462d057c
- Explicitly instantiate Singleton templates to be exported from libbackend.so
- Install only header files that are backend interfaces to be exposed to the user
- Don't use full path for including templates in header files
- Instantiate a couple of smart_ptr templates to have their symbols exported outside the DSO. This happens in history.cc. FIXME: the interface for History will definitely need to be reworked, and the two smart_ptr explicit instantiations removed.
- Change SWIG exported interface to make use of Schedulable instead of (Dynamic|Static)Schedulable
- Fix provided policies to make use of the new interface
- TODO: limit the use of smart_ptrs.
git-svn-id: svn://svn.gna.org/svn/sgpemv2/trunk@653 3ecf2c5c-341e-0410-92b4-d18e462d057c
tchernobog
paolo
elvez
matrevis