Minor changes on Maximum thread values.
Add granularity on variables types and Marsaglia RNG versions. Add licence
Add Cecill v2 licence on source code
Add granularity on variable types and Marsaglia RNG generators.
Minor change about check.
Extend granularity on size and Marsaglia RNG generators. Add both asynchrone and synchrone MPI calls.
Add granularity choice on type of counters and type of Marsaglia generator.
Support for Intel Xeon Phi
Replace synchrone to asynchrone MPI calls as in Hybrid version.
Replace synchrone to asynchrone MPI Send/Receive. At the beginning only to avoid distribution of tasksbut it was a problem on OpenIB (mlx4_core.log_mtts_per_seg=5 to add in GRUB)
Modify output to provide rates.
Convert CUDA implementation as OpenCL one.
Split MainLoop* by calls on one MainLoop
Add different Marsaglia RNG.
Minor modifications.
Add vendor print and strip output on device name.
Add print of Platform vendor on startup
Suppress extra explorations to keep only atomic ones...
Suppress function with no use.
Add Distributed splutter version with atomic version of increment.
Minor changes
Add Hybrid MPI/OpenMP version
Add CUDA version for Sparse a Dense exploration.
Add Sparse mode.
Correct size of allocation.
« Précédent 1 ... 8 9 10 11 12 13 Suivant » (226-250/310) | Par page : 25, 50, 100
Formats disponibles : Atom