Commits

Branch selector

 main

User selector

bythew3i

Datepicker

All time

Commit History

Commits on Sep 30, 2024

• [Mosaic TPU] Force offset to 0 when inferring input has offset out of the first tile.

  Show description for 4a596ae
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  4a596ae

  Copy full SHA for 4a596ae

• [Mosaic TPU] Canonicalize arith.select's condition to vector if other types are vector.

  Show description for 7e2f487
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  7e2f487

  Copy full SHA for 7e2f487

Commits on Sep 25, 2024

• Roll back cl/678765762 [Mosaic TPU] Support bitcast without forcing retiling.

  Show description for e4ca4f5
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  e4ca4f5

  Copy full SHA for e4ca4f5

• [Mosaic TPU] Support bitcast without forcing retiling.

  Show description for 37641dd
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  37641dd

  Copy full SHA for 37641dd

Commits on Sep 24, 2024

• [Mosaic TPU] Support all cases for extui.

  Show description for 407dc77
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  407dc77

  Copy full SHA for 407dc77

Commits on Sep 20, 2024

• [Mosaic:TPU] Efficient relayout with internal scratch

  Show description for 6b93b35
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  6b93b35

  Copy full SHA for 6b93b35

Commits on Sep 19, 2024

• [Mosaic TPU][NFC] Remove FailureOr in getNativeVregOrVmaskTypeImpl

  Show description for 47b177b
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  47b177b

  Copy full SHA for 47b177b

Commits on Sep 17, 2024

• [Pallas TPU] Fix dtype_bitwidth for int in util.

  Show description for d27fce6
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  d27fce6

  Copy full SHA for d27fce6

Commits on Sep 16, 2024

• [Pallas TPU] Refactor ref indexers to transforms and support ref bitcast.

  Show description for 839ce9a
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  839ce9a

  Copy full SHA for 839ce9a

Commits on Sep 13, 2024

• [Mosaic TPU] Propagate the memory space change for memref bitcast and reshape.

  Show description for 8d93e10
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  8d93e10

  Copy full SHA for 8d93e10

Commits on Sep 12, 2024

• [Mosaic TPU] Better error message when shape of memref bitcast is invalid.

  Show description for 178fb03
  

  bythew3i

  authored and

  Google-ML-Automation

  committed
  178fb03

  Copy full SHA for 178fb03

Commits on Sep 5, 2024

• [Mosaic TPU] Fix operands order in try canonicalize add of matmul.

  Show description for dba6741
  

  bythew3i

  authored and

  jax authors

  committed
  dba6741

  Copy full SHA for dba6741

Commits on Sep 4, 2024

• [Mosaic TPU] Fix mosaic alignment check in concatenate rule.

  Show description for c1d3c2d
  

  bythew3i

  authored and

  jax authors

  committed
  c1d3c2d

  Copy full SHA for c1d3c2d

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

Pagination

Previous

Next