degomme [Tue, 5 Jan 2016 13:57:19 +0000 (14:57 +0100)]
avoid some boost warning/errors on windows
degomme [Tue, 5 Jan 2016 13:55:00 +0000 (14:55 +0100)]
weirdly, DEPENDS clause seems to be ignored sometimes when using make -j1 (as in ci on some nodes)
Gabriel Corona [Mon, 4 Jan 2016 12:07:44 +0000 (13:07 +0100)]
[simix] Remove lots of simcalls, use run_kernel instead
We should be able to most of the "Func" and "Proc" simcalls without
handler. Most getters can be done directly in the application process
without kernel synchronization.
Martin Quinson [Tue, 5 Jan 2016 08:01:08 +0000 (09:01 +0100)]
reindent
Martin Quinson [Tue, 5 Jan 2016 00:07:12 +0000 (01:07 +0100)]
cleanup: don't initialize variables to dummy values right before giving them the real value
Martin Quinson [Tue, 5 Jan 2016 00:05:59 +0000 (01:05 +0100)]
convert some C bits into C++ in surf
Martin Quinson [Mon, 4 Jan 2016 23:58:14 +0000 (00:58 +0100)]
compile simdag with g++ so that Surf cleaned of its C interface
Martin Quinson [Mon, 4 Jan 2016 22:58:53 +0000 (23:58 +0100)]
inline two C calls in surf
Martin Quinson [Mon, 4 Jan 2016 20:57:57 +0000 (21:57 +0100)]
mv src/simdag/private.h src/private/simdag_private.h
Gabriel Corona [Mon, 4 Jan 2016 16:36:18 +0000 (17:36 +0100)]
[simix] Fix forwarding of function in run_kernel
Martin Quinson [Mon, 4 Jan 2016 16:18:19 +0000 (17:18 +0100)]
[cmake] fix make dist
Gabriel Corona [Mon, 4 Jan 2016 13:34:53 +0000 (14:34 +0100)]
[simix] Add assertion in run_kernel
Gabriel Corona [Mon, 4 Jan 2016 12:02:30 +0000 (13:02 +0100)]
[simix] Fix simgrid::simix::kernel for void return
Gabriel Corona [Mon, 4 Jan 2016 11:31:26 +0000 (12:31 +0100)]
[simix] simcall to run code in kernel mode
With this, we should be able to limit the number of necessary simcalls
dramatcially. In the long term, we might even get rid of simcalls
altogether and use only this mechanism for process/kernel switching.
Martin Quinson [Mon, 4 Jan 2016 11:05:14 +0000 (12:05 +0100)]
speed up travis: no need for compile_optimizations there
Martin Quinson [Mon, 4 Jan 2016 10:56:04 +0000 (11:56 +0100)]
tiny cosmetics
Martin Quinson [Mon, 4 Jan 2016 10:44:31 +0000 (11:44 +0100)]
Do the right thing in CpuL07::onSpeedChange
- That's very close to what we have in CpuCas01
- I guess that existing actions were not correctly updated when an
avail event occurred. That's fixed now.
- It make ptask07 model usable with DVFS (fix: #43) YUHU!!!
Martin Quinson [Mon, 4 Jan 2016 10:31:42 +0000 (11:31 +0100)]
Simplify the code taking Cpu::m_speedPeak changes into account
The DVFS change set failed to do the right thing and instead added the
current speed as a bound to any execution action (in addition, this
code was in simix while this performance modeling clearly belongs to
surf). I guess that this was not playing well with availability
changes due to a trace, since that extra bound was not recomputed
correctly.
Instead, factorize the code of trace availability events with the
m_speedPeak changes created by DVFS. This creates a new Cpu event
called onSpeedChange, which is still to be exported properly. For now,
only the tracing code listen to that event (it's hardcoded).
Martin Quinson [Mon, 4 Jan 2016 10:18:12 +0000 (11:18 +0100)]
add a comment explaining the current state of the code
Martin Quinson [Mon, 4 Jan 2016 09:29:14 +0000 (10:29 +0100)]
CpuTI cleanups: initialize fields, dont overide with same content, typos and cosmetics
Martin Quinson [Sun, 3 Jan 2016 21:45:46 +0000 (22:45 +0100)]
cleanups in CpuCas01: overrides, initialize fields, kill non-obviously dead code
Martin Quinson [Sun, 3 Jan 2016 21:24:52 +0000 (22:24 +0100)]
cleanups in CpuTi: overide markers, initialize fields, etc
Martin Quinson [Sun, 3 Jan 2016 21:12:41 +0000 (22:12 +0100)]
[CpuTI] don't override pstate handling with THROW_UNIMPLEMENTED
Martin Quinson [Sun, 3 Jan 2016 21:11:24 +0000 (22:11 +0100)]
kill dead code in Cpu constructors
Martin Quinson [Sat, 2 Jan 2016 21:41:37 +0000 (22:41 +0100)]
Mark another execution path for deletion
Martin Quinson [Fri, 1 Jan 2016 20:54:23 +0000 (21:54 +0100)]
some cleanups in the CPU constructor hierarchy
Martin Quinson [Thu, 31 Dec 2015 18:32:57 +0000 (19:32 +0100)]
initialize fields at declaration, to improve readability
Martin Quinson [Wed, 30 Dec 2015 07:44:26 +0000 (08:44 +0100)]
kill dead code
Martin Quinson [Tue, 29 Dec 2015 14:54:43 +0000 (15:54 +0100)]
Give a p_netcard to simgrid::Host instead of relying on extensions for that
Martin Quinson [Mon, 28 Dec 2015 15:52:36 +0000 (16:52 +0100)]
Compile src/instr with g++ so that we can use C++ constructs
Martin Quinson [Sun, 27 Dec 2015 21:27:18 +0000 (22:27 +0100)]
Rename RoutingEdge into NetCard
That's shorter and hopefully more explicit
Martin Quinson [Sun, 27 Dec 2015 21:15:17 +0000 (22:15 +0100)]
kill a redundent type
Martin Quinson [Sun, 27 Dec 2015 21:12:58 +0000 (22:12 +0100)]
this is C++
Martin Quinson [Sun, 27 Dec 2015 20:46:32 +0000 (21:46 +0100)]
give simgrid::Host a p_cpu field instead of relying on extensions for non-optional features
Martin Quinson [Sun, 27 Dec 2015 20:35:49 +0000 (21:35 +0100)]
kill the numerical SURF_HOST_LEVEL
Use the typesafe interface instead
Martin Quinson [Sun, 27 Dec 2015 20:05:21 +0000 (21:05 +0100)]
code simplification: directly give the host name to SD_workstation_create
Martin Quinson [Sun, 27 Dec 2015 17:52:48 +0000 (18:52 +0100)]
kill dead code
Martin Quinson [Sat, 26 Dec 2015 08:57:41 +0000 (09:57 +0100)]
empty surf::Host into simgrid::Host
Martin Quinson [Fri, 25 Dec 2015 21:50:22 +0000 (22:50 +0100)]
rename executeParallelTask into L07Action ctor
Martin Quinson [Fri, 25 Dec 2015 21:14:25 +0000 (22:14 +0100)]
give a default value at field declaration, not everywhere in subclasses
Martin Quinson [Fri, 25 Dec 2015 21:03:38 +0000 (22:03 +0100)]
kill methods overriding super to the exact same content
Martin Quinson [Fri, 25 Dec 2015 20:59:44 +0000 (21:59 +0100)]
[ptask] kill the last global: the maxmin system is now registered as the one of all 3 models
Martin Quinson [Fri, 25 Dec 2015 20:48:33 +0000 (21:48 +0100)]
[ptask] remove some static variables
Martin Quinson [Fri, 25 Dec 2015 20:39:32 +0000 (21:39 +0100)]
cosmetics and informative comments
Martin Quinson [Fri, 25 Dec 2015 18:31:14 +0000 (19:31 +0100)]
Make the energy plugin usable with ptask (but not DVFS) (fix #27)
- Move the DVFS mechanism higher in the CPU hierarchy
- Remove all static_casts to CpuCas01 in energy plugin
- Make sure that ptaskL07 can execute sequential tasks
DVFS is still not working: the speedPeak changes have no impact on
ptask actions. I don't understand maxmin enough and need the help of
@alegrand, sorry.
Martin Quinson [Thu, 24 Dec 2015 14:43:54 +0000 (15:43 +0100)]
make MC compile out of tree
Martin Quinson [Thu, 24 Dec 2015 14:19:09 +0000 (15:19 +0100)]
port NS3 to the new extension mechanism
Martin Quinson [Wed, 23 Dec 2015 10:18:08 +0000 (11:18 +0100)]
revert a cleanup in Storage that left the lmm_constraint uninitialized
Martin Quinson [Wed, 23 Dec 2015 10:11:16 +0000 (11:11 +0100)]
[cmake] fix outoftree build for Java
And don't include the outdated ChangeLog.java anymore
Martin Quinson [Wed, 23 Dec 2015 08:44:53 +0000 (09:44 +0100)]
cosmetics in Storage
Martin Quinson [Wed, 23 Dec 2015 08:53:29 +0000 (09:53 +0100)]
fix a stupid error: actually plug the CPU into the host...
Martin Quinson [Wed, 23 Dec 2015 01:34:44 +0000 (02:34 +0100)]
simplify how cpus are plugged into hosts at creation
Martin Quinson [Wed, 23 Dec 2015 01:07:30 +0000 (02:07 +0100)]
move createHost higher in the surf::HostModel hierarchy
There is no need to specialize that function, actually
Martin Quinson [Wed, 23 Dec 2015 01:05:29 +0000 (02:05 +0100)]
don't pack the library into the jar until the library is built
Martin Quinson [Wed, 23 Dec 2015 00:21:26 +0000 (01:21 +0100)]
cosmetics in VM13
- Mark as override some methods
- Kill a method redefinition to the same content
Martin Quinson [Wed, 23 Dec 2015 00:17:16 +0000 (01:17 +0100)]
give the VM a link to their PM as a simgrid::Host
Martin Quinson [Tue, 22 Dec 2015 23:46:09 +0000 (00:46 +0100)]
inline onliners with lambdas
Martin Quinson [Tue, 22 Dec 2015 22:17:42 +0000 (23:17 +0100)]
rename networkLinkCreatedCallbacks into Link::onCreation and other cleanups
Martin Quinson [Tue, 22 Dec 2015 21:47:40 +0000 (22:47 +0100)]
[energy plugin] inline the callbacks using lambda functions
Martin Quinson [Tue, 22 Dec 2015 21:28:49 +0000 (22:28 +0100)]
rename callback cpuActionStateChangedCallbacks into CpuAction::onStateChange
Martin Quinson [Tue, 22 Dec 2015 21:26:06 +0000 (22:26 +0100)]
Kill unused simgrid::surf::HostAction
Martin Quinson [Tue, 22 Dec 2015 21:13:39 +0000 (22:13 +0100)]
code simplification
Martin Quinson [Tue, 22 Dec 2015 21:13:22 +0000 (22:13 +0100)]
cosmetics in doxygen
Martin Quinson [Tue, 22 Dec 2015 11:00:19 +0000 (12:00 +0100)]
Kill the surf::Host subclasses (surf::HostModel remain subclassed)
Martin Quinson [Tue, 22 Dec 2015 10:38:42 +0000 (11:38 +0100)]
pull up another (useless) method in the surf::Host hierarchy
Martin Quinson [Tue, 22 Dec 2015 08:55:05 +0000 (09:55 +0100)]
pull up two more methods in the surf::Host hierarchy
Martin Quinson [Mon, 21 Dec 2015 21:55:56 +0000 (22:55 +0100)]
push up a method in the surf::Host hierarchy
Martin Quinson [Mon, 21 Dec 2015 21:53:43 +0000 (22:53 +0100)]
cosmetics
Martin Quinson [Mon, 21 Dec 2015 21:42:05 +0000 (22:42 +0100)]
kill unused code
Martin Quinson [Mon, 21 Dec 2015 21:40:07 +0000 (22:40 +0100)]
code simplification around Host::onDestruction
Martin Quinson [Mon, 21 Dec 2015 21:39:17 +0000 (22:39 +0100)]
give all machines the exact same energetic profil, and adapt the tests accordingly
Martin Quinson [Mon, 21 Dec 2015 16:08:31 +0000 (17:08 +0100)]
kill unused callbacks: cpuCreated, cpuDestructed and cpuActionStateChanged
Martin Quinson [Mon, 21 Dec 2015 16:07:03 +0000 (17:07 +0100)]
kill some MSC cruft
Martin Quinson [Mon, 21 Dec 2015 15:53:08 +0000 (16:53 +0100)]
Remove the ability to write internal plugins in Java
It may be reintroduced once the ongoing cleanup is done, but it was
really too difficult to maintain this unused feature during the change.
Martin Quinson [Sun, 20 Dec 2015 22:15:09 +0000 (23:15 +0100)]
rename host callbacks
Martin Quinson [Sun, 20 Dec 2015 22:09:03 +0000 (23:09 +0100)]
make the host callbacks static to surf::Host instead of globals
Martin Quinson [Sun, 20 Dec 2015 18:06:04 +0000 (19:06 +0100)]
add some missing 'override' markers in the surf::Host hierarchy
Martin Quinson [Fri, 18 Dec 2015 23:39:49 +0000 (00:39 +0100)]
rename facets to extensions for clarity sake
Gabriel Corona [Fri, 18 Dec 2015 15:17:09 +0000 (16:17 +0100)]
[surf] Remove signal-related #defines
Gabriel Corona [Fri, 18 Dec 2015 10:26:40 +0000 (11:26 +0100)]
[surf] Triggers the destructed callbacks on the full object
Doing it in the Cpu/Host destructor is not so great as the object is
not an instance of the derived classes anymore.
Gabriel Corona [Fri, 18 Dec 2015 11:18:20 +0000 (12:18 +0100)]
Call destructors of facets in reverse order of their registrations
The rationale for this, is that if a level B as been added after a
facet A, the subsystem of B might depend on the subsystem on A and a
facet of B might need to have the facet of A around when executing its
cleanup function/destructor.
Martin Quinson [Thu, 17 Dec 2015 23:52:35 +0000 (00:52 +0100)]
Load system headers before intercepting gettimeofday to not mess up
Martin Quinson [Thu, 17 Dec 2015 23:42:02 +0000 (00:42 +0100)]
fix broken html entity
Martin Quinson [Thu, 17 Dec 2015 23:18:48 +0000 (00:18 +0100)]
[surf] new (abstract) class: PropertyHolder, with only one purpose
Martin Quinson [Thu, 17 Dec 2015 22:04:03 +0000 (23:04 +0100)]
not all resources need a property set, only Host, Link and Storage
So dupplicate the property-related code in these classes, and remove
it from Resource. We may want to refactorize this code in a specific
class, but I'm not confident with multiple inheritance yet.
The reason of fixing that working feature is the bloat fight. A CPU or
RoutingEdge don't need no properties. The Resource class contains too
much things and should be split of.
Gabriel Corona [Thu, 17 Dec 2015 13:56:57 +0000 (14:56 +0100)]
[surf] Use the pointers to simgrid::Host instead of looking by name
Martin Quinson [Thu, 17 Dec 2015 16:05:12 +0000 (17:05 +0100)]
tiny cleanup
Martin Quinson [Thu, 17 Dec 2015 15:39:59 +0000 (16:39 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Gabriel Corona [Thu, 17 Dec 2015 15:10:26 +0000 (16:10 +0100)]
Fix CPU registration with Java models
Fix 'Move sg_host_surfcpu_register() out of createCpu()'
Gabriel Corona [Thu, 17 Dec 2015 13:38:31 +0000 (14:38 +0100)]
[surf] simgrid::surf::Host, add backlink and attach() method
Gabriel Corona [Thu, 17 Dec 2015 13:13:38 +0000 (14:13 +0100)]
[surf] Add backlinks from Cpu to surf::Host
By the way, we change sg_host_surfcpu_register(host, cpu) into
cpu->plug(host).
Gabriel Corona [Thu, 17 Dec 2015 12:53:05 +0000 (13:53 +0100)]
[surf] Move sg_host_surfcpu_register() out of createCpu()
Gabriel Corona [Thu, 17 Dec 2015 10:07:40 +0000 (11:07 +0100)]
[xbt] Type-safe Facetable levels
For compatibility with C, the type-safety is currently not strongly
enforced.
Gabriel Corona [Thu, 17 Dec 2015 09:27:06 +0000 (10:27 +0100)]
[xbt] Rewrite the doc about the motivation for simgrid::xbt::string
Martin Quinson [Wed, 16 Dec 2015 16:13:17 +0000 (17:13 +0100)]
cosmetics
Martin Quinson [Wed, 16 Dec 2015 11:37:45 +0000 (12:37 +0100)]
woops, fix Java after removing an unimplemented function
Martin Quinson [Wed, 16 Dec 2015 11:33:59 +0000 (12:33 +0100)]
kill an unimplemented function. Fix #14
Gabriel Corona [Tue, 15 Dec 2015 14:46:58 +0000 (15:46 +0100)]
[mc] Fix the MC wrt the simgrid::Host modification
Martin Quinson [Tue, 15 Dec 2015 22:03:18 +0000 (23:03 +0100)]
rename Lib to Facetable
Martin Quinson [Tue, 15 Dec 2015 21:39:39 +0000 (22:39 +0100)]
initialize fields at declaration, not everywhere in code