Commits

Show all
Author Commit Message Labels Comments Date
Matthew Turk
Removed the bad timestepping information. Said -- as a result of equilibrium species wreaking havoc -- that non-HI,H2I and ge fields can have up to but not including 30% correction factor without manually inserting a timestep (the timestepping algorithm will still catch any relevant problems, but it will take more rejected steps before doing so) and changed tiny to 1e-35 in fortran.def. Added optimizatiom but no symbol-stripping to the orange makefile. Made sure that even below 0.8 GeV, we calculate a k1 rate. It is small, but this avoids a sharp cutoff.
Branches
popiii_colorfield
Matthew Turk
Moved high-density solver to the non-photon object list.
Branches
popiii_colorfield
Matthew Turk
Moved some global data definitions outside an ifdef TRANSFER block.
Branches
popiii_colorfield
Matthew Turk
Added new intel compiler makefile for triton, updated old triton makefile to have optimization settings.
Branches
popiii_colorfield
Matthew Turk
Merging in changes from WoC repo
Branches
popiii_colorfield
Matthew Turk
Removed spurious check-for-exit that occurred before the check-for-write.
Branches
popiii_colorfield
Matthew Turk
Removed the NO_ITERATE_POTENTIAL ifdef (which I missed -- and which should get picked up by PotentialIterations = 0) and changed the style of ifdef/ifndef to reflect what is done elsewhere in the routine. This code now has two modes: bitwise-yes: This transforms us back to fully blocking calls in the gravity, and runs are identical. bitwise-no : This allows the non-blocking calls in gravity, and runs are not guaranteed to be bitwise identical, as it can depend on the order of arrival for the gravity solver. Thanks very VERY much to both John Wise a…
Branches
popiii_colorfield
Matthew Turk
Fixing one of my ifdefs in PrepareDensityField, adding bitwise-yes and bitwise-no to Make.config.*. I'd like to note, however, that as of this revision, I can no longer reproduce the problems. So I am going to investigate further.
Branches
popiii_colorfield
Matthew Turk
Added slightly more useful information for group opening failure in the old IO routines
Branches
popiii_colorfield
Matthew Turk
Adding BITWISE_IDENTICALITY option that basically brings back blocking gravity. This patch is largely John's, but I did the reinsertion of blocking for the potential iterations. In my (somewhat limited, but increasing) tests this brings back bitwise identicality even with potential iterations on.
Branches
popiii_colorfield
John Wise
Fixed typo in SBC that caused crashes in the communication (wrong number of loops!)
Matthew Turk
Removed debugging statement
Branches
popiii_colorfield
John Wise
Forgot to remove scale_fields.o from Make objects.
Matthew Turk
Made the config.assemble not use the single-open HDF5 read style
Branches
popiii_colorfield
John Wise
Matthew Turk
Merging from week-of-code, and added the new grid io as a config-time parameter called NEW_GRID_IO or newgridio in the config system: make newgridio-yes to enable it. Checkpoints won't work without it, and I think it is overall a much nicer-to-read system. It disables the difference between IO type and internal representation.
Branches
popiii_colorfield
John Wise
Putting scale_fields back to F77. Will make the F90 version another branch.
John Wise
John Wise
Clean-up of WalkPhotonPackage, taking some ideas from Peter. Converted scale_fields to FORTRAN90 to take advantage of a simple vectorization.
Matthew Turk
New triton makefile to reflect the new projects directory...
John Wise
Put Fen's changes into SetBoundaryConditions() and re-deleted EVRO. These changes weren't included in the executable because EVRO wasn't in the object compile list.
John Wise
Merged with tip.
John Wise
cic_deposit wasn't called on the finest particle refinement level. smooth_deposit was called instead on the finest level, which wasn't necessary and slowed things down a lot when many grids and particles existed on the finest level.
John Wise
Changed some abs() to ABS() and moved SyncPhotonCount from main object list to photon_obj list.
fenzhao
merge
John Wise
Changes to the adaptive timestepping to have more consistent timesteps. Added some slight optimizations to WalkPhotonPackage. In PhotonTest, you can specify species fractions in spheres. Fixed some other bugs with the communication utilities, which didn't affect the runs.
John Wise
John Wise
SGIs don't like the drand48() declaration. Removed.
John Wise
Merged with tip.
bor...@client65-88.sdsc.edu
Adding src/lcaperf directory of low-level lcaperf post-processing utilities
  1. Prev
  2. Next