vatai/mpk issues and pull requests

#42 - Implement having the results in the same memory location as the input

Issue - State: open - Opened by vatai almost 5 years ago

#41 - Investigate (and fix) why big matrices get stuck

Issue - State: open - Opened by vatai almost 5 years ago

#40 - Make proper results to compare communications for various matrices.

Issue - State: closed - Opened by utsavsinghal5 almost 5 years ago - 3 comments

#39 - Prove or verify that mirroring will give us same partitions as we get from METIS

Issue - State: open - Opened by utsavsinghal5 almost 5 years ago - 1 comment

#38 - Make a separate method for phase processing loop in `comm_comp_patterns`

Issue - State: closed - Opened by utsavsinghal5 almost 5 years ago - 1 comment

#37 - Reading pattern and symmetric matrices

Issue - State: closed - Opened by vatai almost 5 years ago - 1 comment

#36 - Output files have always the same name

Issue - State: closed - Opened by vatai almost 5 years ago - 1 comment

#35 - pdmpk testing (LAPJV testing)

Issue - State: closed - Opened by vatai almost 5 years ago - 5 comments

#34 - Get same partition at the top and bottom (Symmetry, Up-down mirror idea)

Issue - State: open - Opened by utsavsinghal5 almost 5 years ago - 1 comment

#33 - Implement LAPJV (input generaton) in PDMPK

Issue - State: closed - Opened by vatai almost 5 years ago - 7 comments

#32 - [pdmpk] Optimise `init_idcs`

Issue - State: closed - Opened by vatai almost 5 years ago - 1 comment

#31 - [pdmpk,dmpk] Implement a faster search for the optimal permutation of new partition labels

Issue - State: closed - Opened by vatai almost 5 years ago

#30 - [pdmpk] Unify `init_dict` and `comm_dict`

Issue - State: closed - Opened by vatai almost 5 years ago

#29 - Implement partial vertices

Issue - State: closed - Opened by vatai about 5 years ago - 2 comments

#28 - Communication hiding

Issue - State: open - Opened by vatai over 5 years ago - 1 comment

#27 - Apply optimisations pernode/hybrid optimisation

Issue - State: open - Opened by vatai over 5 years ago - 1 comment

#26 - For graph bcsstk28 to program does not produce the correct results

Issue - State: closed - Opened by vatai over 5 years ago - 9 comments

#25 - Metis does not support loops

Issue - State: closed - Opened by vatai over 5 years ago - 2 comments

#24 - Remove code duplication in makefiles using include

Issue - State: closed - Opened by vatai over 5 years ago

#23 - Crash with relatively small size

Issue - State: closed - Opened by vatai over 5 years ago - 2 comments

#22 - Reduce the size of comm_table

Issue - State: closed - Opened by vatai over 5 years ago - 2 comments
Labels: wontfix

#21 - Remove the redundant communication of the vertices which can be computed in the partition.

Issue - State: open - Opened by utsavsinghal5 over 5 years ago - 1 comment

#20 - Measure performance penalty for unbalanced communication.

Issue - State: open - Opened by vatai over 5 years ago

#19 - Optimize communication (more verbose `store_part[]`)

Issue - State: open - Opened by vatai over 5 years ago

#18 - Figure out number of phases backward which can do communication with current phase.

Issue - State: closed - Opened by utsavsinghal5 over 5 years ago - 3 comments

#17 - Redesign `comm_data_t`

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment
Labels: enhancement

#16 - Avoid copying to/from the send/receive buffers

Issue - State: closed - Opened by vatai over 5 years ago - 6 comments

#15 - Run it on the university cluster

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment

#14 - Compare the speed MPI with the OpenMP version

Issue - State: open - Opened by vatai over 5 years ago - 2 comments

#13 - Implement measurements

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment

#12 - Implement `nphase` equals `0` case for mpi

Issue - State: closed - Opened by vatai over 5 years ago

#11 - Record commit checksum in log files

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment
Labels: enhancement, wontfix

#10 - Implement skirt for MPI

Issue - State: closed - Opened by vatai over 5 years ago

#9 - Implement alltoall MPI calls and execution

Issue - State: closed - Opened by vatai over 5 years ago

#8 - Debug `MPI_alltoallv()` parameter arrays

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment

#7 - Debug the `comm_table[]` generation

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment
Labels: enhancement

#6 - Diagonal elements

Issue - State: closed - Opened by vatai over 5 years ago - 3 comments

#5 - Custom matrix values

Issue - State: closed - Opened by vatai over 5 years ago - 1 comment

#4 - Automate testing for the new MPI code

Issue - State: closed - Opened by vatai over 5 years ago

#3 - Create optimal partition label selecting algorithm

Issue - State: closed - Opened by vatai over 5 years ago - 3 comments

#2 - Figure out PETSc

Issue - State: open - Opened by vatai over 5 years ago

#1 - Doccument skirt code

Issue - State: closed - Opened by vatai over 5 years ago

GitHub / vatai/mpk issues and pull requests