The GRACOS Release Notes

gracos-9211-AA

• the bug that caused some one billion particle runs to crash during PM force computation is fixed
• the issue in which statement [format "%.4f" [varputs a]] kept expanding into [format "0.0000" [varputs a]] is fixed.

gracos-1.0.1a10

• Introduced variation of minimal allowed epsilon (previously 0.1) with ntab.
• Sections added into documentation about commands accel (acceleration) and lbal (runtime load balancing).
• The use of empiric model of load balancing is implemented via the -empiric flag to lbal.
• MPI error code is added to error messages, simplified MPI and FFTW call wrappers.
• With new text extraction tool 'texispell.py' fixed 36 spelling typos in the documentation; our apologies if this has caused you any trouble.
• Incorporated WSHAPE package; so far just for s2s2 force computation.
• rmax is now computed by linear fit rmax = 2.78 + (3.21-2.78)(eta-3.3) (this replaces the crossing point method used earlier).
• Added the conformal, proper cosmic and code integration time (tconf, tcosm, and tcode) arguments for gracos-rw (thanks Gary Felder for suggestion).
• Renamed fast mode of bitwise variable smode into the opposite of thorough.
• Speedup for PM-only (e.g. caustic) runs: epsilon is no longer used. Effective epsilon is found from pm_eta hence timestep is increased.
• Introduced the --enable-devel flag to configure, for development mode.
• Fixed non-blocking communication issue in portability to OpenMPI (and MPICH?); thanks to Francisco Serrano for the diagnostics and the fix.

gracos-1.0.1a9

• Improved documentation, posting to asro-ph/0711.4655.
• There were reports of a problem running gracos on MPICH implementation of MPI. While planning to address this issue in future version of GRACOS we kindly ask MPICH users to temporarily switch to MPI LAM.

gracos-1.0.1a8

• Default value initialization dropped for variables out_restart,out_analyze, and out_backup.
• Scheduled administration method via sched.txt is introduced.
• New command: checkpoint; see new section in the manual.
• Fixed a bug: the value of preprocessing directive MODE_REPEATABLE is corrected to 0. This bug caused workload measurement to be based on the number of particles in the cell instead of timers. Additionally, it turned off adaptive P3M effectively (thanks to Neal Dalal who pointed out to the large load imbalance for a 1024^3 run). The bug is only effective for version gracos-1.0.1a7.
• New options -vars, -bits and -toks introduced for bitvar. The bitvar syntax is changed (changed the order of first two arguments). Tcl procedure bitvars is eliminated: use bitvar -vars instead.
• Fixed error message when trying to load udf format on number of processes less than nshr while writing.
• Options save_dt, reset_econ , and inverse_gravity to command integ are renamed into disable-timestep-truncation reset-econ , and inverse-gravity
• Fixed cross-reference labels in DVI+PDF documentation.
• Splines implemented for rw_aexp instead of Newton-Raphson iteration used earlier. (Fixes the bug found by Neal Dalal and Olivier Dore: GRACOS can now run even for Om=0.01 and Ov=0.99).
• Fixed scale-free.sh, bbks.sh, restart.sh, and lingers.sh scripts in gracos-examples by adding the -sort flag.
• Mathematical and statistical tools (e.g. histograms, splines, etc) are grouped under a new mathmeth subpackage; packages spline and mathstat disappear. Mathematics from rw package (Robertson-Walker Universe) goes into mathmeth.

gracos-1.0.1a7

• Suboption nfh added for mwhite data format.
• Variables out_restart , out_backup , and out_analyze loose the default write protection. In addition, they now indicate the complete command name to be executed, with @MODE@, @LABEL@, and @OUTID@ patterns to be replaces at the runtime.
• Updated velocity normalization in Santa-Barbara Project Data Format. (Fixes the Santa Barbara format implementation bug found by Neal Dalal).
• Allowing only even nx,ny,nz in the PM-mesh based grids. Using odd values will now result in error message and program termination.
• Plenty of other changes since version llpm-hc-1729.

llpm-hc-1729

• Parallel domain decomposition with Hilbert curve
• Adaptive method of force computation with P3M