Code

Auto TensorCore CodeGen#4234

Code

Merged

tqchen merged 21 commits intoapache:masterapache/tvm:masterfrom minminsun:AutoTensorCoreCodeGenCopy head branch name to clipboardNov 9, 2019

+1,920

Commits

Commits on Oct 31, 2019

• Add Auto TensorCore TensorCore Unit Test

  minminsun

  committed
  bfffc34

  Copy full SHA for bfffc34

• Rebase to tvm master branch & Add auto tensor core

  

  jcf94

  authored and

  minminsun

  committed
  8772a03

  Copy full SHA for 8772a03

• Code Refine

  

  jcf94

  authored and

  minminsun

  committed
  ebb01b7

  Copy full SHA for ebb01b7

• Add tensor core switch by pragma

  

  jcf94

  authored and

  minminsun

  committed
  626166b

  Copy full SHA for 626166b

• Add pragma in tensor core example code

  

  jcf94

  authored and

  minminsun

  committed
  2f7ab5b

  Copy full SHA for 2f7ab5b

• Get real tile size to replace hard coded 16

  minminsun

  committed
  e0cbc0e

  Copy full SHA for e0cbc0e

• support more than 2 dimensions (e.g. batchmatmul) for buffer bind scope

  minminsun

  committed
  af0824b

  Copy full SHA for af0824b

• support batch matmul

  

  Orion34-lanbo

  authored and

  minminsun

  committed
  a76e4e7

  Copy full SHA for a76e4e7

• Move cuda env check to tensor_core.cc

  

  jcf94

  authored and

  minminsun

  committed
  4b0d18d

  Copy full SHA for 4b0d18d

• Coderefine for tensor_core.cc

  

  jcf94

  authored and

  minminsun

  committed
  f80fad8

  Copy full SHA for f80fad8

• Refine comments

  minminsun

  committed
  afa9061

  Copy full SHA for afa9061

Commits on Nov 6, 2019

• Some refinements of code and comment

  

  jcf94

  authored and

  minminsun

  committed
  7c6b880

  Copy full SHA for 7c6b880

• Update TensorCore UT to pass the CPU test

  jcf94

  committed
  d5134f2

  Copy full SHA for d5134f2

Commits on Nov 7, 2019

• remove redundant code

  minminsun

  committed
  4868685

  Copy full SHA for 4868685

• matmul's storage align for different layout

  Orion34-lanbo

  committed
  022adb9

  Copy full SHA for 022adb9

• Add support for differenct position of type cast

  jcf94

  committed
  17a08bb

  Copy full SHA for 17a08bb

• Add formal tutorial for auto tensorcore codegen

  minminsun

  committed
  a975d6f

  Copy full SHA for a975d6f

• move tensorcore check up to tutorial code

  minminsun

  committed
  9fe5363

  Copy full SHA for 9fe5363

Commits on Nov 8, 2019

• code and doc refine

  minminsun

  committed
  9baf969

  Copy full SHA for 9baf969

• comment out tune_and_evaluate in tutorial

  minminsun

  committed
  f61f362

  Copy full SHA for f61f362

• fix cpplint error

  minminsun

  committed
  4824579

  Copy full SHA for 4824579