------------------------------------------------------------------------------------------------------------------------- AvailMem: Largest = 29230.83 Mb (on task= 24), Smallest = 29201.10 Mb (on task= 12), Average = 29215.28 Mb Total Mem: Largest = 32213.01 Mb (on task= 0), Smallest = 32213.01 Mb (on task= 0), Average = 32213.01 Mb Committed_AS: Largest = 3011.91 Mb (on task= 12), Smallest = 2982.18 Mb (on task= 24), Average = 2997.73 Mb SwapTotal: Largest = 23436.99 Mb (on task= 0), Smallest = 23436.99 Mb (on task= 0), Average = 23436.99 Mb SwapFree: Largest = 22588.00 Mb (on task= 0), Smallest = 21600.56 Mb (on task= 12), Average = 22020.50 Mb AllocMem: Largest = 3011.91 Mb (on task= 12), Smallest = 2982.18 Mb (on task= 24), Average = 2997.73 Mb ------------------------------------------------------------------------------------------------------------------------- Task=0 has the maximum committed memory and is host: sandy-022 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~------- **ENABLE_HEALTHTEST** When this is enabled, the code tries to assess upon start-up whether all CPU cores are freely available and show the same execution speed. Also, the MPI bandwidth both inside nodes and between nodes is tested. ------- **FORCETEST** = 0.001 Calculates for the specified fraction of particles direct summation forces, which can then be compared to the forces computed by the Tree/PM/FMM algorithms of GADGET-4 in order to check or monitor the force accuracy of the code. This is only included as a testing and debugging option. The value of the option should be set to a number between 0 and 1 (e.g. 0.001), and this number gives the fraction of randomly chosen particles at each timestep for which forces by direct summation are computed. The normal tree-forces and the exact direct summation forces are then collected in a file `forcetest.txt` for later inspection. Note that the simulation itself is unaffected by this option, but it will of course run much(!) slower, particularly if `FORCETEST * NumPart * NumPart >> NumPart` Note: The particle IDs must be set to numbers >= 1 for this option to work. ------- **FORCETEST_TESTFORCELAW** = 1 Special option for measuring the effective force law. Can be set to 1 or 2 for checking with TreePM/FMM-PM, or TreePM/FMM-PM + PLACEHIGHRESREGION. The option `FORCETEST` must be activated as well. The simulation needs to be fed with a special initial conditions file for which only one particle has mass, the others are massless test particles. The code will then go through cycles in which the particle with the mass is randomly placed, and the other particles are randomly placed around it, with distance spacing uniform in log(r). After 40 cycles are carried out, the code terminates, and the force-law accuracy can be examined by analysing the file `forcetest.txt` . ------- **FORCETEST_FIXEDPARTICLESET** This always checks the same particle IDs if force accuracy is checked during a run. ------- **VTUNE_INSTRUMENT** This option creates additional instrumentation instructions for the Intel VTune code performance tool, based on the internal timing routines. This can be used for a performance analysis based on this tool. ------- **DEBUG_MD5** This option can be used to compute MD5 checksums of the P[] and SphP[] arrays regularly in the code, with the results being written to the log-file memory.txt. Using this, one can check for binary invariance of the code when the code is interrupted and resumed from restart files. ------- **TILING** = 2 Replicates the read-in ICs the specified number of times in each dimension. This can be used for scaling tests. ------- **SQUASH_TEST** Squeezes the ICs on read-in on order to create a distortion from spherical symmetry in certain force calculation tests. ------- **DOMAIN_SPECIAL_CHECK** Outputs test data to check the balancing algorithms. ------- **EWALD_TEST** A development test for testing the accuracy of the Ewald table lookup. ------- **RECREATE_UNIQUE_IDS** This option can be used to reinitialize the particle IDs upon start-up. Useful if one has to deal with a broken IC file. ------- **NO_STOP_BELOW_MINTIMESTEP** Do not stop when the code wants to adopt a timestep below the specified minimum timestep, but rather enforce this step size. ------- **DO_NOT_PRODUCE_BIG_OUTPUT** This special option allows one to refrain from writing large output files (restart files, snapshots, and group catalogues), which can be useful for scaling tests. ------- **STOP_AFTER_STEP** = 8 After the corresponding step has been completed, the simulation ends. This is meant to simplify certain performance and scalability tests. ------- **MEASURE_TOTAL_MOMENTUM** This option computes the total conjugate momentum after every step. Can be used to check for manifest momentum conservation of different force computation schemes. ------- **TREE_NO_SAFETY_BOX** When enabled, this disables the geometric 'near node' protection, i.e. for the one-sided tree, one may then be closer to a node's center than 1.5 times the node size, and for FMM, adjacent nodes may interact.