User:LeonKwek - Revision history

LeonKwek: /* Progress Log */

2011-08-10T07:02:05Z

Progress Log

@@ Line 58: / Line 58: @@
 == Progress Log ==
 '''2 Aug'''
 * Strange errors seen when running bballPlot.py, no bounce is being observed! Why is that? Is the bouncing being turned off somehow?
 * Possibility of a virtual machine as another tool for tracking down these problems

Jpye: /* Progress Log */

2011-08-02T07:24:24Z

Progress Log

@@ Line 58: / Line 58: @@
 == Progress Log ==
 '''22 Jun'''
 Had the bouncing ball model written up a while ago, but finally got it to simulate correctly today:

LeonKwek at 05:44, 22 June 2011

2011-06-22T05:44:11Z

← Older revision		Revision as of 05:44, 22 June 2011
Line 61:		Line 61:
	Had the bouncing ball model written up a while ago, but finally got it to simulate correctly today:		Had the bouncing ball model written up a while ago, but finally got it to simulate correctly today:
	[[File:Bball.png]]		[[File:Bball.png]]

	the hitch was a sneaky combination of two errors, one of which has a certain random element that was confusing the beejeesus out of me. First though, I was stringing up a second derivative as		the hitch was a sneaky combination of two errors, one of which has a certain random element that was confusing the beejeesus out of me. First though, I was stringing up a second derivative as
	<source lang="a4c">		<source lang="a4c">

LeonKwek at 05:42, 22 June 2011

2011-06-22T05:42:44Z

@@ Line 23: / Line 23: @@
 '''High / immediate priority'''
-* 20/05/2011: Get a conditional "bouncing ball" model running. Want to rigorously test stopping, reporting and restarting after boundary crossings Expect issues with IDACalcIC when using very fine time steps / boundaries that occur on or very close to time steps : 2 hours
 '''Medium priority'''
 * 20/05/2011: Fix the issue of non-incident var with incident derivative : unknown
 '''Low priority'''
@@ Line 39: / Line 39: @@
 (Date Added -- Date Completed): (Task) : (Estimated time -- Actual time)
 Comments:
 * 25/05 -- 19/06: Sort the issue of IDA stopping short of boundaries, switching the system, then actually crossing the boundary and switching the system back to the wrong equation set.
@@ Line 56: / Line 58: @@
 == Progress Log ==
 ''' 21 Jun '''
 Spent hours last week trying to get to the bottom of the SUNDIALS_VERSION_MINOR business but again to no effect. Decided in the end to do a complete fresh install of ubuntu 11 to go along with the installation of a new CPU. On the new system with the sundials 2.4.0 package the scons TryRun of IDA now fails altogether. The test works fine when I run it separately from the ascend built environment but I haven't been able to pin down the issue. I tried changing the sundials_version_text test code to just main(){return 0} but no dice.

LeonKwek at 05:41, 21 June 2011

2011-06-21T05:41:47Z

← Older revision		Revision as of 05:41, 21 June 2011
Line 34:		Line 34:
	== Stuff to discuss with John ==		== Stuff to discuss with John ==
	* scons help with SUNDIALS_VERSION_MINOR		* scons help with SUNDIALS_VERSION_MINOR
			* By reporting a point whenever we cross a boundary, the size of the data set is unknown until after integration. Might cause issues with plotting scripts?

	== Completed Tasks ==		== Completed Tasks ==

LeonKwek at 05:32, 21 June 2011

2011-06-21T05:32:37Z

← Older revision		Revision as of 05:32, 21 June 2011
Line 38:		Line 38:
	(Date Added -- Date Completed): (Task) : (Estimated time -- Actual time)		(Date Added -- Date Completed): (Task) : (Estimated time -- Actual time)
	Comments:		Comments:
	* 25/05 -- 19/06: Sort the issue of IDA stopping short of boundaries, switching the system, then actually crossing the boundary and switching the system back to the wrong equation set.		* 25/05 -- 19/06: Sort the issue of IDA stopping short of boundaries, switching the system, then actually crossing the boundary and switching the system back to the wrong equation set.
	~~Finally~~

	* 20/05/2011 -- 25/05/2011: Move the flagfn stuff over to the IDA integrator data struct to clean up the prepare_integrator functions : 10 min, 10 min		* 20/05/2011 -- 25/05/2011: Move the flagfn stuff over to the IDA integrator data struct to clean up the prepare_integrator functions : 10 min, 10 min

LeonKwek at 05:31, 21 June 2011

2011-06-21T05:31:19Z

← Older revision		Revision as of 05:31, 21 June 2011
Line 23:		Line 23:

	'''High / immediate priority'''		'''High / immediate priority'''
	* 25/05/2011: Sort the issue of IDA stopping short of boundaries, switching the system, then actually crossing the boundary and switching the system back to the wrong equation set.
	* 20/05/2011: Get a conditional "bouncing ball" model running. Want to rigorously test stopping, reporting and restarting after boundary crossings Expect issues with IDACalcIC when using very fine time steps / boundaries that occur on or very close to time steps : 2 hours		* 20/05/2011: Get a conditional "bouncing ball" model running. Want to rigorously test stopping, reporting and restarting after boundary crossings Expect issues with IDACalcIC when using very fine time steps / boundaries that occur on or very close to time steps : 2 hours
			* 21/06/2011: Remove the need to analyse twice at start up to set the equations correctly

	'''Medium priority'''		'''Medium priority'''
Line 33:		Line 33:

	== Stuff to discuss with John ==		== Stuff to discuss with John ==
	* ~~Non incident var~~ with ~~non incident derivative thing~~		* scons help with SUNDIALS_VERSION_MINOR
	* Check if OTHERWISE to sort default boolean_var values is ok. Should the complier pick up if CASE list is non-exhaustive?

	== Completed Tasks ==		== Completed Tasks ==
	(Date Added -- Date Completed): (Task) : (Estimated time -- Actual time)		(Date Added -- Date Completed): (Task) : (Estimated time -- Actual time)
	Comments:		Comments:
			* 25/05 -- 19/06: Sort the issue of IDA stopping short of boundaries, switching the system, then actually crossing the boundary and switching the system back to the wrong equation set.
			Finally

	* 20/05/2011 -- 25/05/2011: Move the flagfn stuff over to the IDA integrator data struct to clean up the prepare_integrator functions : 10 min, 10 min		* 20/05/2011 -- 25/05/2011: Move the flagfn stuff over to the IDA integrator data struct to clean up the prepare_integrator functions : 10 min, 10 min
Line 55:		Line 56:

	== Progress Log ==		== Progress Log ==
			''' 21 Jun '''
			Spent hours last week trying to get to the bottom of the SUNDIALS_VERSION_MINOR business but again to no effect. Decided in the end to do a complete fresh install of ubuntu 11 to go along with the installation of a new CPU. On the new system with the sundials 2.4.0 package the scons TryRun of IDA now fails altogether. The test works fine when I run it separately from the ascend built environment but I haven't been able to pin down the issue. I tried changing the sundials_version_text test code to just main(){return 0} but no dice.

			Everything compiles and runs fine with the check disabled, so I'll leave it at that for now. IDASetRootDirection does the job fine and issues with double crossings seem to be fixed. Noticed some strange results while testing however: found the issue stems again from problem analyzing(.c) without conditional systems in mind. Boolean vars get assigned default to true so that the initial analysis of the WHENs will generally choose the wrong set of initial equations. As a temporary workaround can just do a LRSolve if boundaries are present in the model and re-analyse if some_disvars_change. A better solution will come about along side dealing with the non-incident var issue, which will have to deal with the deeper design issue of somehow flagging the model as conditional during analysis.

			Also begs some design questions about "assuming IDA is always correct" when it comes to boundary crossings. If the wrong set of equations mange to slip in at some point, then IDA happily continues through the whole integration while sending back nonsense data. The user has to have responsibility for running sanity checks on any results they get, but as my experience with the wrong initial equations has shown, these errors can be very hard to pin down because the output looks otherwise reasonable.

	'''8 Jun'''		'''8 Jun'''

Jpye: /* Progress Log */

2011-06-08T03:13:14Z

Progress Log

← Older revision		Revision as of 03:13, 8 June 2011
Line 57:		Line 57:

	'''8 Jun'''		'''8 Jun'''

	Discussion with John, Leon. Agreed that the IDA boundary direction function from SUNDIALS 2.4.0 is needed to resolve issue of possible repeated boundary detection. Need to resolve issue of stray values of SUNDIALS_VERSION_MINOR from SCons as part of that. Decided that the problem of solving systems with non-incident x but incident x_dot will need revisiting of the system/analyze.c code; there may not be an easier way. So we will leave that issue for later. Aiming to get the bouncing ball model solved for next week!		Discussion with John, Leon. Agreed that the IDA boundary direction function from SUNDIALS 2.4.0 is needed to resolve issue of possible repeated boundary detection. Need to resolve issue of stray values of SUNDIALS_VERSION_MINOR from SCons as part of that. Decided that the problem of solving systems with non-incident x but incident x_dot will need revisiting of the system/analyze.c code; there may not be an easier way. So we will leave that issue for later. Aiming to get the bouncing ball model solved for next week!

Jpye: /* Progress Log */

2011-06-08T03:13:07Z

Progress Log

← Older revision		Revision as of 03:13, 8 June 2011
Line 55:		Line 55:

	== Progress Log ==		== Progress Log ==

			'''8 Jun'''

			Discussion with John, Leon. Agreed that the IDA boundary direction function from SUNDIALS 2.4.0 is needed to resolve issue of possible repeated boundary detection. Need to resolve issue of stray values of SUNDIALS_VERSION_MINOR from SCons as part of that. Decided that the problem of solving systems with non-incident x but incident x_dot will need revisiting of the system/analyze.c code; there may not be an easier way. So we will leave that issue for later. Aiming to get the bouncing ball model solved for next week!

	'''26th May'''		'''26th May'''
	Spent the whole day unsuccessfully trying to track down why my SUNDIALS_VERSION_MINOR remains equal to 3 after updating. Scons appears to detect the version change, namely if I print minor right before the line		Spent the whole day unsuccessfully trying to track down why my SUNDIALS_VERSION_MINOR remains equal to 3 after updating. Scons appears to detect the version change, namely if I print minor right before the line

Jpye: /* Current task list */

2011-06-08T02:44:40Z

Current task list

← Older revision		Revision as of 02:44, 8 June 2011
Line 27:		Line 27:

	'''Medium priority'''		'''Medium priority'''
	* 20/05/2011: Fix the issue of ~~Non~~ incident var with ~~non~~ incident derivative : unknown		* 20/05/2011: Fix the issue of non-incident var with incident derivative : unknown

	'''Low priority'''		'''Low priority'''
	* 14/04/2011: Figure out a better / the correct solution to the LRSlv messing with the compiler issue. Possibility: work with Ben to properly file a "design document" + follow through with the solution: unknown		* 14/04/2011: Figure out a better / the correct solution to the LRSlv messing with the compiler issue. Possibility: work with Ben to properly file a "design document" + follow through with the solution: unknown

	== Stuff to discuss with John ==		== Stuff to discuss with John ==