Re: [FLASH-USERS] IDL problems + dump times

Mateusz, the coordinate information in the plotfile that you sent me is
really messed up. THe problem appears to be in how the plotfile was
generated, and not in the IDL routines. For example, dumping out the
coordinate information in that file produces:

[zingale@nan ~/fidlr_2.1]$ hdp dumpsds -i 16
windtunnel_4lev_hdf_plt_cnt_0001 | head -50
File name: windtunnel_4lev_hdf_plt_cnt_0001

Variable Name = coordinates
         Index = 16
         Type= 32-bit floating point
         Ref. = 34
         Rank = 2
         Number of attributes = 0
         Dim0: Name=fakeDim18
                 Size = 2115
                 Scale Type = number-type not set
                 Number of attributes = 0
         Dim1: Name=fakeDim19
                 Size = 2
                 Scale Type = number-type not set
                 Number of attributes = 0
         Data :
                -0.000000 1.450000
                -0.000000 1.325000
                2.000000 1.525000
                -0.000000 1.325000
                2.000000 1.525000
                0.000000 1.625000
                -2.000000 1.600000
                2.000000 1.650000
                -0.000000 1.637500
                -2.000000 1.600000
                -0.000000 1.637500
                36893488147419103232.000000 1.675000
                -2.000000 1.662500
                0.000000 1.656250
                -0.000000 1.668750
                0.000000 1.656250
                -0.000000 1.668750
                2.000000 1.681250
                0.000000 1.687500
                -2.000000 1.693750
                2.000000 1.681250
                -2.000000 1.693750
                0.000000 1.656250
                -2.000000 1.662500
                -0.000000 1.668750
                0.000000 1.656250
                -0.000000 1.668750
                2.000000 1.681250
                0.000000 1.687500
                -2.000000 1.693750
                2.000000 1.681250
                -2.000000 1.693750
                0.000000 1.625000
[zingale@nan ~/fidlr_2.1]$

Note, the very large coordinate value present there (such values are
repeast throughout that record).

I would guess that there is a mismatch between the datatype of the
coordinate data (and all of the other data) in FLASH and the type
specifier that we give the HDF library when writing it out. Please note
that FLASH is supposed to be compiled in double precision, using the
compiler flags to do the promotion. For the plotfiles, we cast things
back down to single precision before doing the writing, to save on
space. It is possible that either one of these parts is not working right
on your machine.

Can you remind me what platform, compilers, and build options you are
using?

Mike

------------------------------------------------------------------------------
Michael Zingale
UCO/Lick Observatory
UCSC
Santa Cruz, CA 95064

phone: (831) 459-5246
fax: (831) 459-5265
e-mail: zingale@ucolick.org
web: http://www.ucolick.org/~zingale

On Mon, 7 Oct 2002, Mateusz Ruszkowski wrote:

>
> Jonathan,
>
> Thanks for your help.
> In the problem I am currently running, checkpoint files are dumped *less*
> often than trstrt ... The first checkpoint file corresponds to time = 0.0
> (of course) but the second one corresponds to time = 0.174 .... I used:
> trstrt = 1.0e-2, nrstrt = 1000000, wall_clock_checkpoint = 1.0e30. The
> second checkpoint file should have been generated at 0.01, so there is
> still some problem...
>
>
> Perhaps using plot files would help. However, the only reason I did not
> use plot files instead of checkpoint files is that IDL generates error
> message (even for test cases) when I attempt to display plot files
> (checkpoint files look fine). I am getting the following output from IDL
> for the windtunnel case (i.e. different problem than the one mentioned
> above):
>
>
> [mr@hevelius windtunnel]$ idl
> IDL Version 5.4 (linux x86). (c) 2000, Research Systems, Inc.
> Installation number: 100.
> Licensed for use by: hevelius.Colorado.EDU
>
> IDL> xflash
> % Compiled module: XFLASH.
> initializing xflash
> ... IDL version = 5.4
> ... PNG graphics will be used
> ... xflash directory is/scratch/temp/FLASH2.1/tools/fidlr2/
> % Compiled module: XFLASH_DEFAULTS.
> % Compiled module: VAR_INDEX.
> % Compiled module: COLOR_INDEX.
> % Compiled module: CW_FIELD.
> % Compiled module: CW_BGROUP.
> % Compiled module: XMANAGER.
> % Compiled module: DETERMINE_FILE_TYPE.
> % Compiled module: GET_XFLASH_PATH.
> % Loaded DLM: HDF.
> % Compiled module: GET_DIMENSIONALITY.
> % Compiled module: FHDF_READ.
> % Compiled module: GET_VAR_LIST.
> % Compiled module: GET_PARTICLE_NUMBER.
> ^^^^ old filename =
> ^^^^ new filename =
> /usr/users/mr/FLASHY/FLASH2.1/setups/windtunnel/windtunnel_4lev_hdf_plt_cnt_0001
> ^^^^ need to read again
> % Compiled module: XPLOT_AMR_NEW.
> % Compiled module: LOADCT.
> % Compiled module: COLOR.
> multi_plots_per_page = 0
> ## plot info: 0 1 1
> working on
> /usr/users/mr/FLASHY/FLASH2.1/setups/windtunnel/windtunnel_4lev_hdf_plt_cnt_0001
> % Compiled module: READ_AMR.
> var_name = none
> preparing to read in 2115 blocks from 1 file(s)
> working on file named
> /usr/users/mr/FLASHY/FLASH2.1/setups/windtunnel/windtunnel_4lev_hdf_plt_cnt_0001
> % Compiled module: UNDEFINE.
> number of blocks: 2115
> number of top level zones: 75008
> fraction of uniform grid: 0.457812
> % Compiled module: CREATE_VARIABLE.
> looking to create variabledens
> all else failed, var_index = 0
> dens extrema: -0.00000 3.68935e+19
> main plot window = 0
> % XMANAGER: Caught unexpected error from client application. Message
> follows...
> % PLOT: Data coordinate system not established.
> % Execution halted at: XPLOT_AMR_NEW 524
> /scratch/temp/FLASH2.1/tools/fidlr2/xplot_amr_new.pro
> % XFLASH_EVENT 704
> /scratch/temp/FLASH2.1/tools/fidlr2/xflash.pro
> % XMANAGER_EVLOOP_STANDARD 478
> /usr/local/rsi/idl_5.4/lib/xmanager.pro
> % XMANAGER 708
> /usr/local/rsi/idl_5.4/lib/xmanager.pro
> % XFLASH 1481
> /scratch/temp/FLASH2.1/tools/fidlr2/xflash.pro
> % $MAIN$
> IDL>
>
>
>
>
> > > I ran windtunnel test with trstrt = 0.5 in flash.par (default), but some
> > > of the checkpoint files were not generated in dt = 0.5 second intervals
> > > (e.g. time = (...), 1.0, 1.5, 2.0, 2.243, 2.5, 3.0, (...) ) This is a
> > > problem when one wants to generate movies. Is there a quick fix to it ?
> >
> > There are several triggers for generating restart files -- trstrt is
> > not the only one. nrstrt, for instance, makes checkpoints every so
> > many timesteps. There is also a check for a certain amount of wall
> > clock time having elapsed, at which point a checkpoint file is made.
> > These are all desired behaviours. If you really want to, you can
> > disable these other triggers by setting the number of timesteps, or
> > the wallclock time, to be extremely large.
> >
> > Regardless, checkpoint files are complete dumps of all the data in the
> > domain, and are poor choices for generating movies. For movies, you're
> > better off using plot files, which not only have reduced amounts of data,
> > but have fewer triggers for creating a checkpoint file.
> >
> > - Jonathan
> > --
> > ljdursi@flash.uchicago.edu
>
Received on Mon Oct 7 15:23:51 2002