Code

Make way for aarch64 #2

Code

Merged

jerinphilip merged 38 commits intomainjerinphilip/slimt:mainfrom aarch64+refactorjerinphilip/slimt:aarch64+refactorCopy head branch name to clipboardAug 14, 2023

+1,362-613

Commits

Commits on Aug 12, 2023

• Add ruy as a dependency

  jerinphilip

  committed
  a7a3e8d

  Copy full SHA for a7a3e8d

• Moving things around, refactor begins

  jerinphilip

  committed
  3f6c1e7

  Copy full SHA for 3f6c1e7

• Add android-ndk build script

  jerinphilip

  committed
  08afe7f

  Copy full SHA for 08afe7f

• Setting i8 specific header for time being

  jerinphilip

  committed
  015bf8a

  Copy full SHA for 015bf8a

• Have we regained compilation?

  jerinphilip

  committed
  290a51f

  Copy full SHA for 290a51f

• Using templates for ODR co-existence

  jerinphilip

  committed
  3e2ed1b

  Copy full SHA for 3e2ed1b

• Ruy code in place, now I find the segfault

  jerinphilip

  committed
  1a93363

  Copy full SHA for 1a93363

• B_cols changes to indices.size(), but not sure if this sticks?

  jerinphilip

  committed
  d0391d9

  Copy full SHA for d0391d9

• Remove hacks around Wemb_QuantMultA

  jerinphilip

  committed
  7132b44

  Copy full SHA for 7132b44

• Only selected_B_cols, not B_cols

  jerinphilip

  committed
  7a8372b

  Copy full SHA for 7a8372b

• Templating vectors...

  jerinphilip

  committed
  a171aef

  Copy full SHA for a171aef

• Move vectorized_ to simd.hh

  jerinphilip

  committed
  234005b

  Copy full SHA for 234005b

• w8 -> w4

  jerinphilip

  committed
  f1c1149

  Copy full SHA for f1c1149

• Finishing up android ndk compile

  jerinphilip

  committed
  ecf4968

  Copy full SHA for ecf4968

Commits on Aug 13, 2023

• MacOS M1 patches

  jerinphilip

  committed
  8cf0bc9

  Copy full SHA for 8cf0bc9

• Fix formatting after M1

  jerinphilip

  committed
  06206c7

  Copy full SHA for 06206c7

• Add neon.h

  jerinphilip

  committed
  9611c50

  Copy full SHA for 9611c50

• Updates to README.md

  jerinphilip

  committed
  c387d33

  Copy full SHA for c387d33

• Add submodule clone instructions in the README.md

  jerinphilip

  committed
  2714f58

  Copy full SHA for 2714f58

• Add Apple Accelerate BLAS

  jerinphilip

  committed
  7a234c3

  Copy full SHA for 7a234c3

• Refine as a cmake-target?

  jerinphilip

  committed
  01d2c72

  Copy full SHA for 01d2c72

• Fix linking on actual M1

  jerinphilip

  committed
  5e5c74f

  Copy full SHA for 5e5c74f

• Fix formatting

  jerinphilip

  committed
  5c57a51

  Copy full SHA for 5c57a51

• Indirect CI to sh via scripts

  jerinphilip

  committed
  cc339e6

  Copy full SHA for cc339e6

• Update android paths and try

  jerinphilip

  committed
  1c8e507

  Copy full SHA for 1c8e507

• fail-fast false

  jerinphilip

  committed
  cc2e477

  Copy full SHA for cc2e477

• Add an 03-test.sh placeholder

  jerinphilip

  committed
  d9143fa

  Copy full SHA for d9143fa

• Empty, trigger CI

  jerinphilip

  committed
  7c27280

  Copy full SHA for 7c27280

• Fix path from CI

  jerinphilip

  committed
  1dcdcbd

  Copy full SHA for 1dcdcbd

• Quote prefix

  jerinphilip

  committed
  1c5b502

  Copy full SHA for 1c5b502

• Simplify

  jerinphilip

  committed
  f20023b

  Copy full SHA for f20023b

• Use matrix.name

  jerinphilip

  committed
  70a4a10

  Copy full SHA for 70a4a10

• Remove the i8xi8 ugliness

  jerinphilip

  committed
  e39e995

  Copy full SHA for e39e995

• Tensori8Ops -> QMM

  jerinphilip

  committed
  ec14a6f

  Copy full SHA for ec14a6f

• Few leftover rename updates

  jerinphilip

  committed
  58e39db

  Copy full SHA for 58e39db

Commits on Aug 14, 2023

• Default ruy to off

  jerinphilip

  committed
  1d955a6

  Copy full SHA for 1d955a6

• Update android cmake configure args

  jerinphilip

  committed
  6aab8c8

  Copy full SHA for 6aab8c8

• Remove __SSE__

  jerinphilip

  committed
  d4f406c

  Copy full SHA for d4f406c