Code

Lockfree Ripser#31

Code

Open

mrzv wants to merge 42 commits intoRipser:masterRipser/ripser:masterfrom mrzv:lockfreemrzv/ripser:lockfreeCopy head branch name to clipboard

+6,396-129

Commits

Commits on Feb 26, 2020

• Switch compressed_sparse_matrix to vector<vector<ValueType>>

  mrzv

  committed
  eaf5900

  Copy full SHA for eaf5900

• Add USE_DENSE_PIVOTS

  mrzv

  committed
  7d2babc

  Copy full SHA for 7d2babc

• Fix compilation with coefficients

  mrzv

  committed
  35ee0dd

  Copy full SHA for 35ee0dd

Commits on Mar 9, 2020

• Make cofacet_entries static thread_local

  mrzv

  committed
  10a06fa

  Copy full SHA for 10a06fa

Commits on Mar 12, 2020

• Add trivial_concurrent_hash_map; for now doesn't work with coefficients

  mrzv

  committed
  adeb456

  Copy full SHA for adeb456

• Make entry_t = index_t, when using coefficients, to enable atomic_ref

  mrzv

  committed
  1eca2b0

  Copy full SHA for 1eca2b0

Commits on Mar 15, 2020

• Undo over-reserving pivot_column_index

  mrzv

  committed
  44f6581

  Copy full SHA for 44f6581

• Switch compressed_sparse_matrix to a vector of atomic pointers to vector

  mrzv

  committed
  d923928

  Copy full SHA for d923928

• Read column and check correctness of the read

  mrzv

  committed
  caec39e

  Copy full SHA for caec39e

• Store coefficient in the value of the hash map, not the key

  mrzv

  committed
  cbfcf29

  Copy full SHA for cbfcf29

• Restructure the code for parallelism

  mrzv

  committed
  0882d65

  Copy full SHA for 0882d65

• Add i/atomic_ref_serial.hpp, included via a flag

  mrzv

  committed
  1047cb2

  Copy full SHA for 1047cb2

• Add USE_PARALLEL_STL for-loop parallelization

  mrzv

  committed
  7e21cf8

  Copy full SHA for 7e21cf8

Commits on Mar 16, 2020

• Add logic to move reduction to the right

  mrzv

  committed
  7ddbf8c

  Copy full SHA for 7ddbf8c

• Add shuffled serial and hand-rolled chunking in foreach

  mrzv

  committed
  8c5764d

  Copy full SHA for 8c5764d

• Don't add diagonal element to working_reduction_column, when continuing reduction

  mrzv

  committed
  f04a19b

  Copy full SHA for f04a19b

• Make simplex_coboundary_enumerator::neighbor_{it,end} cache thread_local + add thread sanitizer to debug build

  mrzv

  committed
  7711c40

  Copy full SHA for 7711c40

• Disable emergent pair check in parallel

  mrzv

  committed
  eb20913

  Copy full SHA for eb20913

• trivial_concurrent_hash_map::find() guarantees that a value has been written

  mrzv

  committed
  665ff4f

  Copy full SHA for 665ff4f

• Don't output progress and pairs in parallel

  mrzv

  committed
  f62afd4

  Copy full SHA for f62afd4

• Add --threads argument + indicate progress in parallel

  mrzv

  committed
  7ff922f

  Copy full SHA for 7ff922f

• Enable emergent pairs optimization in parallel

  mrzv

  committed
  de75624

  Copy full SHA for de75624

• Add memory reclamation (for hand-rolled parallel foreach)

  mrzv

  committed
  780ff3b

  Copy full SHA for 780ff3b

• Add pair output in the parallel case

  mrzv

  committed
  5eea68e

  Copy full SHA for 5eea68e

• Add MemoryManager to serial versions of foreach

  mrzv

  committed
  d9e0120

  Copy full SHA for d9e0120

• Add MemoryManager to parallel STL version of foreach

  mrzv

  committed
  d391d5c

  Copy full SHA for d391d5c

Commits on Mar 17, 2020

• Pass memory manager by reference

  mrzv

  committed
  4a1ec6a

  Copy full SHA for 4a1ec6a

• Add parallel assembly of columns

  mrzv

  committed
  2de8ca1

  Copy full SHA for 2de8ca1

• Add parallel_sort from Boost

  mrzv

  committed
  a1ece3d

  Copy full SHA for a1ece3d

• Fix compilation with parallel STL

  mrzv

  committed
  fdb9a41

  Copy full SHA for fdb9a41

• Add missing INDICATE_PROGRESS guards

  mrzv

  committed
  fbdc42e

  Copy full SHA for fbdc42e

Commits on Apr 7, 2020

• Indicate empty columns by nullptr, rather than empty vector

  mrzv

  committed
  a793274

  Copy full SHA for a793274

• Fix resize bug in trivial_concurrent_hash_map + bump multiplier to 10

  mrzv

  committed
  0c7e27a

  Copy full SHA for 0c7e27a

• Add necessary function to hash_map based on unordered_map

  mrzv

  committed
  0ae052b

  Copy full SHA for 0ae052b

Commits on Apr 8, 2020

• Add USE_TBB_HASHMAP option

  mrzv

  committed
  fa93b2b

  Copy full SHA for fa93b2b

• Don't allocate second hash map, if not printing pairs

  mrzv

  committed
  cb34f33

  Copy full SHA for cb34f33

• Use the next power of 2 for size in trivial_concurrent_hash_map + bump multiplier down to 8

  mrzv

  committed
  f2bcbbd

  Copy full SHA for f2bcbbd

Commits on Apr 11, 2020

• Add USE_TBB option

  mrzv

  committed
  b283b1f

  Copy full SHA for b283b1f

• Tweak the Makefile: add more targets + automatic dependencies

  mrzv

  committed
  4bf9361

  Copy full SHA for 4bf9361

• USE_CONCURRENT_PIVOTS -> USE_TRIVIAL_CONCURRENT_HASHMAP + reorder #elif to actually use TBB hashmap, when requested

  mrzv

  committed
  ba9be8f

  Copy full SHA for ba9be8f

Commits on Apr 19, 2020

• Fix memory reclamation by extending the wait by an extra epoch

  mrzv

  committed
  1264577

  Copy full SHA for 1264577

Commits on Jun 24, 2020

• Add a link to the paper that describes the algorithm

  mrzv

  committed
  625be2c

  Copy full SHA for 625be2c