Code

Faster sparse_dense on GPUs#6580

Code

Merged

jroesch merged 14 commits intoapache:masterapache/tvm:masterfrom tkonolige:faster_sparse_densetkonolige/incubator-tvm:faster_sparse_denseCopy head branch name to clipboardOct 9, 2020

+537-60

Commits

Commits on Sep 29, 2020

• Faster sparse_dense on GPUs.

  Show description for 02a5bed
  tkonolige

  committed
  02a5bed

  Copy full SHA for 02a5bed

• formatting

  tkonolige

  committed
  e9ccbb7

  Copy full SHA for e9ccbb7

• more formatting

  tkonolige

  committed
  b4df663

  Copy full SHA for b4df663

• Check that alteroplayout is definedbefore using it

  tkonolige

  committed
  b563b1c

  Copy full SHA for b563b1c

• check if FTVMAlterOpLayout exists before using it

  tkonolige

  committed
  13610bc

  Copy full SHA for 13610bc

• formatting

  tkonolige

  committed
  0f012d1

  Copy full SHA for 0f012d1

• restore message passing

  tkonolige

  committed
  7bb3343

  Copy full SHA for 7bb3343

Commits on Oct 1, 2020

• Fix sparse_dense and sparse_dense_padded docs

  tkonolige

  committed
  5c2989c

  Copy full SHA for 5c2989c

Commits on Oct 2, 2020

• Fix old sparse_dense, autotvm and sparse_dense dont play well together

  tkonolige

  committed
  cb8068a

  Copy full SHA for cb8068a

• Remove unused imports

  tkonolige

  committed
  144bb56

  Copy full SHA for 144bb56

• clarify warp count in cuda_transpose

  tkonolige

  committed
  e312cce

  Copy full SHA for e312cce

Commits on Oct 8, 2020

• Document multidimensional access

  tkonolige

  committed
  1a051c7

  Copy full SHA for 1a051c7

• Warn users not to use sparse_dense_padded

  tkonolige

  committed
  f02df33

  Copy full SHA for f02df33

• rename nn.sparse_dense_padded to nn.internal.sparse_dense_padded

  tkonolige

  committed
  fda4e49

  Copy full SHA for fda4e49