ChangeLog

2014-03-05 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Serial Simulator: Merging the serial simulator with the main ROOT-Sim kernel

2014-02-28 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Rendez Vous Subsystem: finished the preliminary implementation of the Rendez-Vous Subsystem

2014-02-21 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Shared Simulation State: bridging the Kernel Module with the simulation kernel.
		The system is now being modified to support LP suspension/resume.

2014-02-20 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* LP memory allocator: implemented a LP memory preallocator working with mmap.
		This is being coupled with the kernel module facilities via ioctl() calls.

2014-02-16 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* New Multithreaded Version: Today, after more than one month of debugging
		the new multithreaded version is almost completely stable.
		Debugging has made me insert several perfomance drops here and there, so
		now it starts the process of enhancing the performance of operations on the
		critical path of the simulator.


2014-01-04 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: The kernel has been patched to call a module hook
		upon process/thread rescheduling to correct the memory
		view the ROOT-Sim processes are getting


2014-01-28 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Rollback: Outgoing messages are correctly registered. During the
		rollback operation anti-messages are correctly sent and received.
		Still missing the logic associated with anti-messages handling in
		bottom halves management.
		
		
2014-01-28 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: The kernel module is correctly working. There is still a small
		bug which causes the OS kernel to hung after several executions. It's
		likely related to buddy system's corruption.
	 	

2013-12-17 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Rollback: Reimplemented the rollback operation. Still missing the antimessages
		sending.
	

2013-12-17 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Numerical Library: numerical initialization is fully contained within
		the numerical library. Some of the stale code in init.c has been moved.
	* Scheduler: reimplemented (static) LP <-> KLTs binding. Added a 'lid' field in
		the LP control block, used to translate KLT binding to local id.
		Added new LP states: SILENT_EXEC and ROLLBACK. The first is used
		to notify ScheduleNewEvent that messages must be discarded. The second
		is used to notify the scheduler that, before executing the next event
		a rollback operation must be executed.
	* Communication: output queue is now using a compact envelope type to keep
		information used to send antimessages. Created a type for the bottom halves.
	* Lists: added the logic to push an object on the list head. This breaks ordering,
		but can be used to implement LIFO queues.
	* Multithread: If a KLT calls system_exit due to an error, LPS is unmapped and other
		threads might incur in segfault. Adding an error flag, to silently
		shutdown the simulation in case of an error.
	* Rollback: identified the point where the rollback operation should be triggered.
	

2013-12-16 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Applications: ported all application examples to the new code base
	* Code base: fixed all memory leaks. Tested the simulator with all
		the available examples. There is a problem with Traffic:
		a call to malloc() returns NULL. This happens after that
		DyMeLoR calls __real_realloc. Maybe there is a bug there...
	* Scheduler: starting to reimplement LP binding towards KLTs


2013-12-15 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: Correctly accessing PDP entries
	

2013-12-13 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: Correctly accessing PML4 entries
	
	
2013-12-13 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Code base: fixing memory leaks
	* Kernel Module: Correctly accessing PML4 entries and PDP entries
	

2013-12-12 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* CCGS: Silent execution is now called to realign LPs' states to the freshly computed
		gvt value. All the applications have been modified to have 'LP id' as the first
		parameter. Finished rewriting the subsystem.
	* State Management: reimplemented silent execution, relying on the new LP scheduling paradigm.
	* Global: checking for memory leaks.
	* Lists: list_truncate was not updating the size of the list. list_pop() now just
		deletes the first element.
	

2013-12-11 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Today: it's a very cool date. 11/12/13. We must wait for year 2103 for such
		a cool date to come back. I'm looking forward to that entry log!!!
	* Code base: finished polishing warnings throughout the code.
	* Shell: there is still a warning in yy_scan_bytes: warning: comparison between
		signed and unsigned integer expressions. There is bug file 140 addressing
		this issue: http://sourceforge.net/p/flex/bugs/140/
		It appears that the bug has been fixed, and the patch will appear in versions
		> 2.5.37. Current release is 2.5.37, so this should be fixed in future
		releases.
	* Numerical library: signedness error in manual initialization of master seed
	* GVT: the reduction is computed correctly. Fossil collection is correctly removing
		data structures. 
	* main: fixed a small bug that might entail a wrong kid number printing
	* CCGS: Starting to reintegrate CCGS for termination detection.



2013-12-10 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: now the first part of the module is working, even when running
		multiple threads.


2013-12-10 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Kernel Module: worked with Francesco, see his changelog entry
	* Overall Code Base: corrected about 100 warnings in the whole code. Added a lot
		of warning flags in the development CFLAGS macro, to avoid bad debugging
		surprises
		

2013-12-09 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* GVT: Rewriting the overall GVT. I think there is a stack corruption somewhere,
		cause the debugging trace is very stange. I have activated all the warning
		levels at compile time. There is a segfault in the fossil collection,
		but in a place where the function should not be called...


2013-12-05 Francesco Quaglia <quaglia@dis.uniroma1.it>
	* Kernel Module: added the code to inspect the content of the PML4 page directory. mmapping is
		working fine, and shut down of the process as well.

2013-12-06 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* ULT: the subsystem has been ported to Cygwin and Windows, using Windows Fibers. Testing
		has not been done very thoroughly. A small bug was introduced in the Linux version
		during the porting, and this has been fixed.
	* GVT: rewriting the subsystem. Up to now, only one KLT per kernel is allowed to run
		the distributed protocol. I should check whether it's possible to rely on more cores
		to perform the reduction in parallel. System is actually not stable.
	* Kernel Module: added support to print CR registers, to inspect the current execution mode.


2013-12-05 Francesco Quaglia <quaglia@dis.uniroma1.it>

	* Kernel Module: added code to inspect the content of a pgd (either original or replaced)


2013-12-05 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Kernel Module: fixed header inclusions. It now compiles on kernel 3.* as well, and
		the stub is correctly working.header inclusions.
		Added a compile time check to determine whether we have at least kernel 2.6.25
		as I'm using pagemap functions.
	* State Saving: added a function to force state saving, independently of checkpointing
		interval. This is used upon the invocation of SetState() during INIT event,
		to create a first snapshot in the simulation. It will be useful during the
		reimplementation of the incremental checkpointing, and the autonomic checkpointing.
		Periodic State Saving and Copy State Saving are correctly working, when using
		1 KLT and 1024 LPs. There was a small (historic!) bug that caused an off-by-one
		error in the computation of the checkpointing interval.
	* GVT Subsystem: starting to completely rewrite the GVT subsystem...
		Module snapshots.c as been removed (there was only one function from the CCGS
		subsystem there). Moved to main gvt.c module, and marked static. This could
		be moved to a separate ccgs.c module, if there is enough code to justify this.
	* Timers: the macro DECLARE_TIMER() has been superseded by the new timer_t type.
		Other macros have been case-lowered to resemble functions. Additionally, they
		have been restructured to be is-safe.


2013-12-04 Francesco Quaglia <quaglia@dis.uniroma1.it>

	* Kernel Module: fixed a bug that caused a general protection error when updating CR3


2013-12-04 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* Lists: fixed a small bug that caused priority inversion in the input event queue
	* Scheduler: rewritten functions to retrieve next event's timestamp and make bound
		advance during the execution.
	* Communication: fixed small bugs that prevented correct insertion of newly scheduled
		events.
	* Topology: micro optimizations in some checks on boundaries, in the case of Hexagonal
		or Square topology. Addedd a sanity check that prevents the finder, when
		only 1 LP is in the system, to loop indefinitely.
	* Global Execution: after this commit, if we run 1 LP on 1 KLT, the flow of (forward)
		events is correctly restored! There is still no computation of the GVT,
		so no output is printed during the execution, the memory exhausts soon,
		and there is no termination detection yet.
		The forward execution has been tested using more than 1 LP and it works.
	* State Management: I have reinserted calls to LogState. Using 1 LP 1 KLT, it
		segfaults when executing subsequent events.


2013-12-03 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>
	* User Level Threads: Porting ULT to cygwin (more for development reasons).
		sigaltstack is missing, so a different way must be used to create
		the thread's stack.


2013-12-02 Francesco Quaglia <quaglia@dis.uniroma1.it>

	* Kernel Module: Carrying on the development of the module. The page tables are
		now swapped and CR3 register is updated. There is still some bug, as the
		kernel panics when the application exits.


2013-11-29 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Communication: most of the (local) communication subsystem has
		been rewritten. Still has to debug local message enqueueing,
		as there are problems when scheduling a new event. Only
		INIT is correctly scheduled and processed. Still running
		with only one kernel-level thread and one user-level
		thread (LP).
	* User Level Threads: ULT has been moved to the arch dependend
		part of the simulator. Current code for creating/scheduling
		ULT has been flagged as Linux-dependent code.


2013-11-28 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Scheduler: it is now the scheduler duty to decide which LP must
		be scheduled, and to actually schedule it. A whole lot
		of code has been changed to support this refactoring


2013-11-27 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Simulator Codebase: rewritten most of the portion of the code on the critical path.
		Everything is still unstable, as INIT event is not place in the input queue,
		and actual user-level thread does nothing. This is just a starting point
		for a runtime testing of the new codebase.
	* Backtrace: merged code from an older branch to show a call backtrace upon segfault.
		This can be enabled using the new --backtrace runtime option.
	* Queues: now queues are correctly initialized
	* INIT event: INIT event is now generated and placed into the input queue of every LP


2013-11-26 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Kernel Module: cleaned up kernel module code. Now we have /sys endpoints correctly created.
		If errors happen upon module initialization, everything should be cleaned up
		before failing


2013-11-26 Francesco Quaglia <quaglia@dis.uniroma1.it>

	* Kernel Module: implmented a stub of the module's pgd allocation


2013-11-21 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Linked Lists: fixed minor bugs
	* State Management: reimplementation concluded. Still unstable, have to test and debug
	* Message Queue Management: reimplementing message queues according to the new list implementation
	* LP execution state: an LP can now be ready, blocked or active


2013-11-20 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Kernel Module: Sketching the kernel module to manage multiple page tables


2013-11-19 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* State Management: Rewritten code to rely on new linked list implementation. Still unstable.


2013-11-13 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Page-Table Kernel Module: Created the skeleton for the new kernel module, which will
		support different per-thread page tables.
	* Queues: Continuing the migration to the new list implementation
	* Memory Management: state handling subsystem is being merged into memory management subsystem
	* Execution Flow: reimplementing the logic behind rollback operation


2013-11-11 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Scheduler: minor bugfixes
	* Event Processing: The code to invoke ProcessEvent() callback has been refactored. A single
		entry point is provided, so that when a LP must be activated, independently if it's
		in forward execution, in coasting forward or in initialization, the internal scheduler's
		API activate_LP() can be invoked. This, in turn, activates the LP's user-level thread.
	* Queues: Starting migrating the queues management code base to the new list implementation
	* Data Types: Modifying datatypes names: suffixes _type are being changed to _t. time_type
		has been changed to simtime_t, to avoid collision with standard time_t. queue_type
		has been discarded, as its role was only to support linked lists, which is in turn
		done by relying on list(msg_t) new supported type.



2013-11-08 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Userspace LP threads: created support for creating userspace threads for LP scheduling
	* Scheduler reorganization: reorganizing scheduler organization to cope with LP user threads



2013-11-06 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* LP execution state revisited: restructuring LPs' execution state, to simplify accessing
		data and support events' preemption



2013-11-05 Alessandro Pellegrini <pellegrini@dis.uniroma1.it>

	* Linked Lists reimplemented: this new version will simplify handling events, and will
		allow the insertion of "fake" events to support intra-/inter-kernel management
		operations.
	* ChangeLog: compliant changelog created. It's impossible to track all the changes made so far...



2011-05-15 ROOT-Sim Team  <rootsim@egooglegroups.com>

	* Version 0.9.0-RC1: this is the first official release of ROOT-Sim
	* Dynamic Memory Logger Subsystem: memory management subsystem incorporated
	* Incremental Dynamic Memory Manager Subsystem
	* Smallest Timestamp First O(n) Scheduler
	* Smallest Timestamp First O(1) Scheduler
	* Consistent Committed Global State Subsystem