Code

Improve gemm code and performance#1541

Code

Merged

oleksandr-pavlyk merged 10 commits intomasterIntelPython/dpctl:masterfrom improve-gemmIntelPython/dpctl:improve-gemmCopy head branch name to clipboardFeb 10, 2024

+2,900-4,138

Commits

Commits on Feb 9, 2024

• Annotate read-only vectors as const

  oleksandr-pavlyk

  committed
  69f17a3

  Copy full SHA for 69f17a3

• Generate implementations for accumulating into some wider types

  Show description for e6d3564
  oleksandr-pavlyk

  committed
  e6d3564

  Copy full SHA for e6d3564

• Refactoring of gemm, adding faster kernel

  Show description for a61e00b
  oleksandr-pavlyk

  committed
  a61e00b

  Copy full SHA for a61e00b

• Remove bool output gemm

  oleksandr-pavlyk

  committed
  3fb0608

  Copy full SHA for 3fb0608

• Added a test for matmul of medium size matrix

  oleksandr-pavlyk

  committed
  69ede62

  Copy full SHA for 69ede62

• Remove unused variable

  oleksandr-pavlyk

  committed
  d7fc376

  Copy full SHA for d7fc376

• Use tolerance of 32 epsilon for gemm result comparisons

  oleksandr-pavlyk

  committed
  74df338

  Copy full SHA for 74df338

Commits on Feb 10, 2024

• Changed selector of algo hyperparameters for complex

  Show description for 43bba09
  oleksandr-pavlyk

  committed
  43bba09

  Copy full SHA for 43bba09

• Fixed issue identified during PR review

  Show description for a885034
  oleksandr-pavlyk

  committed
  a885034

  Copy full SHA for a885034

• Added test for matmul based on discovered failure due to typo

  oleksandr-pavlyk

  committed
  61ec3d3

  Copy full SHA for 61ec3d3