[Users] Carpet error: strides disagree

[Erik Schnetter]

Luca

To allow more error checking, and to allow more generic parallel
operations in Carpet (e.g. supporting symmetry conditions), the grid
operators now take an additional argument "srcregbox" that describes
where an operator should take its data from. Previously, each operator
determined that on its own, and the high-level infrastructure that
needs to provide the operator with these data (e.g. copy from another
process) re-determined this independently.

Apparently this new code has an error. I see that the failing copy
operation receives a dstregbbox with a smaller stride. Most likely,
this copy operation is part of communicating data to a prolongation
operation, and this copy operation receives the wrong dstregbox
argument.

As a short-term remedy, you can check out a version of Carpet from a
few days ago. I'll have a look at this problem.

-erik

On Thu, Jan 12, 2012 at 1:03 AM, Luca Baiotti <baiotti at ile.osaka-u.ac.jp> wrote:
> Hello,
>
> with a fresh ET checkout I ran into the error:
>
> copy_3d.cc:
> srcbbox=([0,0,0]:[160,160,160]:[2,2,2]/[0,0,0]:[80,80,80]/[81,81,81]/531441)
> dstbbox=([98,58,58]:[112,102,102]:[2,2,2]/[49,29,29]:[56,51,51]/[8,23,23]/4232)
> srcregbbox=([98,58,58]:[112,102,102]:[2,2,2]/[49,29,29]:[56,51,51]/[8,23,23]/4232)
> dstregbbox=([101,81,60]:[109,100,100]:[1,1,1]/[101,81,60]:[109,100,100]/[9,20,41]/7380)
> WARNING level 0 in thorn CarpetLib processor 0 host
>   (line 58 of
> /Users/baiotti/Cactus-checkpoit-schedule-update/Cactus/arrangements/Carpet/CarpetLib/src/copy_3d.cc):
>
>   -> Internal error: strides disagree
> Assertion failed: (0), function Abort, file
> /Users/baiotti/Cactus-checkpoit-schedule-update/Cactus/arrangements/Carpet/Carpet/src/helpers.cc,
> line 235.
>
>
> What does it mean and what could the cause be?
>
> I am using very simple parameters for carpet:
>
>
> grid::type          = "coordbase"
> grid::avoid_origin  = "no"
> grid::domain        = "full"
>
> coordbase::xmin =    -1.
> coordbase::xmax =     1.
> coordbase::ymin =    -1.
> coordbase::ymax =     1.
> coordbase::zmin =    -1.
> coordbase::zmax =     1.
>
> coordbase::dx   = 0.025
> coordbase::dy   = 0.025
> coordbase::dz   = 0.025
>
> Carpet::verbose                  = yes
> Carpet::veryverbose              = yes
>
> carpet::use_buffer_zones         = yes
> carpet::num_integrator_substeps  = 3
>
> Carpet::domain_from_coordbase    = "yes"
> Carpet::enable_all_storage       = "no"
>
> methodoflines::ode_method               =       "rk2"
> methodoflines::mol_intermediate_steps   =       2
> methodoflines::initial_data_is_crap     =       "yes"
> InitBase::initial_data_setup_method     =       "init_all_levels"
> Carpet::init_3_timelevels        = "no"
> Carpet::init_each_timelevel      = "no"
>
> Carpet::prolongation_order_space = 3
> Carpet::prolongation_order_time  = 2
>
> Carpet::regrid_in_level_mode         = yes
> Carpet::regrid_during_initialisation = no
>
> Carpet::max_refinement_levels           = 2
> CarpetRegrid2::num_centres              = 1
>
> CarpetRegrid2::num_levels_1             = 2
> CarpetRegrid2::position_x_1             = 0.0
> CarpetRegrid2::radius_1             [1] = 0.25
>
> Carpet::poison_new_timelevels           = yes
> Carpet::check_for_poison                = yes
> CarpetLib::poison_new_memory            = yes
> CarpetLib::poison_value                 = 222
>
>
>
> Cheers,
>
> Luca
> _______________________________________________
> Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users



-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/