Commits

Branch selector

 main

User selector

bythew3i

Datepicker

All time

Commit History

Commits on Aug 28, 2024

• [Pallas TPU] Fix itemsize check for int4 in bitcast lowering.

  Show description for 2c11a91
  

  bythew3i

  authored and

  jax authors

  committed
  2c11a91

  Copy full SHA for 2c11a91

• [Mosaic TPU] Print expected Mosaic version after finding unsupported version.

  Show description for a3cccd3
  

  bythew3i

  authored and

  jax authors

  committed
  a3cccd3

  Copy full SHA for a3cccd3

• [Mosaic TPU] Support memref bitcast.

  Show description for b010750
  

  bythew3i

  authored and

  jax authors

  committed
  b010750

  Copy full SHA for b010750

Commits on Aug 13, 2024

• [Mosaic:TPU] Refactor relayout helper functions to take ctx instead of only target shape.

  Show description for 8f23392
  

  bythew3i

  authored and

  jax authors

  committed
  8f23392

  Copy full SHA for 8f23392

• [Mosaic:TPU] Add shuffled load and store.

  Show description for 2dea3d6
  

  bythew3i

  authored and

  jax authors

  committed
  2dea3d6

  Copy full SHA for 2dea3d6

Commits on Jul 25, 2024

• [XLA:Mosaic] Support any int type upcast.

  Show description for b1b7d04
  

  bythew3i

  authored and

  jax authors

  committed
  b1b7d04

  Copy full SHA for b1b7d04

Commits on Jul 23, 2024

• [XLA:Mosaic] Pass rewrite ctx of apply-vector-layout pass to relayout function.

  Show description for 59e944d
  

  bythew3i

  authored and

  jax authors

  committed
  59e944d

  Copy full SHA for 59e944d

• [XLA:Mosaic] Create apply layout pass with ctx instead of config list.

  Show description for 7e2107b
  

  bythew3i

  authored and

  jax authors

  committed
  7e2107b

  Copy full SHA for 7e2107b

Commits on Jul 19, 2024

• [XLA:Mosaic] Simplify the logic in converting dynamic roll to Log(N) static ops.

  Show description for faf89ab
  

  bythew3i

  authored and

  jax authors

  committed
  faf89ab

  Copy full SHA for faf89ab

Commits on Jul 17, 2024

• [XLA:Mosaic] Extend support of tpu bitcast with offsets and implicit dim.

  Show description for 63a3e67
  

  bythew3i

  authored and

  jax authors

  committed
  63a3e67

  Copy full SHA for 63a3e67

Commits on Jul 13, 2024

• [XLA:Mosaic] Support memref shapecast.

  Show description for aa16485
  

  bythew3i

  authored and

  jax authors

  committed
  aa16485

  Copy full SHA for aa16485

Commits on Jul 2, 2024

• [Pallas][Mosaic] Relax dynamic index on 2nd minor dim in load/store.

  Show description for 484d09f
  

  bythew3i

  authored and

  jax authors

  committed
  484d09f

  Copy full SHA for 484d09f

Commits on Jun 27, 2024

• [XLA:Mosaic] Clean up pass dump and make debug easier.

  Show description for af8bdd1
  

  bythew3i

  authored and

  jax authors

  committed
  af8bdd1

  Copy full SHA for af8bdd1

Commits on Jun 19, 2024

• [Pallas] Expose internal scratch size config.

  Show description for 103d620
  

  bythew3i

  authored and

  jax authors

  committed
  103d620

  Copy full SHA for 103d620

Commits on Jun 18, 2024

• [Pallas] Add support for dynamic shift in roll op.

  Show description for 3a3014b
  

  bythew3i

  authored and

  jax authors

  committed
  3a3014b

  Copy full SHA for 3a3014b

• [XLA:Mosaic] Support dynamic roll

  Show description for cac1791
  

  bythew3i

  authored and

  jax authors

  committed
  cac1791

  Copy full SHA for cac1791

• [XLA:Mosaic] Fix ext rule with large native tile.

  Show description for c180b86
  

  bythew3i

  authored and

  jax authors

  committed
  c180b86

  Copy full SHA for c180b86

• [XLA:Mosaic] Add internal scratch VMEM

  Show description for ed4958c
  

  bythew3i

  authored and

  jax authors

  committed
  ed4958c

  Copy full SHA for ed4958c

Commits on Jun 11, 2024

• [XLA:Mosaic] No need to assume a multiple of tile if tile dim size is 1.

  Show description for 5b38549
  

  bythew3i

  authored and

  jax authors

  committed
  5b38549

  Copy full SHA for 5b38549

Commits on Jun 10, 2024

• [XLA:Mosaic] Fix infer layout for nested loop.

  Show description for 53daa0c
  

  bythew3i

  authored and

  jax authors

  committed
  53daa0c

  Copy full SHA for 53daa0c

Commits on Jun 6, 2024

• [Pallas] Fix typo in test.

  Show description for 7a5975e
  

  bythew3i

  authored and

  jax authors

  committed
  7a5975e

  Copy full SHA for 7a5975e

Commits on May 31, 2024

• [XLA:Mosaic] Fix infer/apply vector layout rule for terminators (scf::yieldOp, scf::conditionOp).

  Show description for 389bf93
  

  bythew3i

  authored and

  jax authors

  committed
  389bf93

  Copy full SHA for 389bf93

Commits on May 21, 2024

• [XLA:Mosaic] Support retiling from (8, 128, -2) to (8, 128) for 32-bit data.

  Show description for ccaf466
  

  bythew3i

  authored and

  jax authors

  committed
  ccaf466

  Copy full SHA for ccaf466

Commits on May 7, 2024

• [Pallas] Fix some typos.

  Show description for 79f11d5
  

  bythew3i

  authored and

  jax authors

  committed
  79f11d5

  Copy full SHA for 79f11d5

Commits on Apr 19, 2024

• [XLA:Mosaic] Support trunc/ext op for 1D vector with any implicit dim.

  Show description for 1671617
  

  bythew3i

  authored and

  jax authors

  committed
  1671617

  Copy full SHA for 1671617

Commits on Apr 17, 2024

• [XLA:Mosaic] Generalize (8,128) -> (8 * packing,128) retiling for packed type.

  Show description for d44b16c
  

  bythew3i

  authored and

  jax authors

  committed
  d44b16c

  Copy full SHA for d44b16c

Commits on Apr 10, 2024

• [Pallas][Mosaic] Expose semaphore read.

  Show description for e3018db
  

  bythew3i

  authored and

  jax authors

  committed
  e3018db

  Copy full SHA for e3018db

• [Pallas] Fix typo in semaphore_wait error messages.

  Show description for 763c6ff
  

  bythew3i

  authored and

  jax authors

  committed
  763c6ff

  Copy full SHA for 763c6ff

Commits on Mar 29, 2024

• [Pallas] Fix a typo in error message of swap rule.

  Show description for 7137b25
  

  bythew3i

  authored and

  jax authors

  committed
  7137b25

  Copy full SHA for 7137b25

Commits on Mar 28, 2024

• [XLA:Mosaic] Remove duplicate headers in debug assert insertion.

  Show description for 67f4f60
  

  bythew3i

  authored and

  jax authors

  committed
  67f4f60

  Copy full SHA for 67f4f60

Commits on Mar 15, 2024

• [Pallas] Raise NotImplementedError for strided load/store in interpret mode.

  Show description for 69795eb
  

  bythew3i

  authored and

  jax authors

  committed
  69795eb

  Copy full SHA for 69795eb

Commits on Mar 14, 2024

• [Pallas] Add stride in Pallas dynamic slice and support strided load/store.

  Show description for 2048e3c
  

  bythew3i

  authored and

  jax authors

  committed
  2048e3c

  Copy full SHA for 2048e3c

• [XLA:Mosaic] Support dynamic indices in strided load/store.

  Show description for 7578e10
  

  bythew3i

  authored and

  jax authors

  committed
  7578e10

  Copy full SHA for 7578e10

Commits on Mar 12, 2024

• [XLA:Mosaic] Support strided load/store memref with arbitrary shape as long as last dim size is 128 and dtype is 32bit.

  Show description for 30208fa
  

  bythew3i

  authored and

  jax authors

  committed
  30208fa

  Copy full SHA for 30208fa

Commits on Mar 6, 2024

• [XLA:Mosaic] Support input offset (replicated, 0) in shapecast.

  Show description for 75f2f75
  

  bythew3i

  authored and

  jax authors

  committed
  75f2f75

  Copy full SHA for 75f2f75

Pagination

Previous

Next