Commits

Branch selector

 master

User selector

blakehechtman

Datepicker

All time

Commit History

Commits on Nov 12, 2021

• [XLA] add an accessor methods for the mutable operands of an HLO

  Show description for 4061b99
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  4061b99

Commits on Nov 1, 2021

• [XLA] Add an option to set the operand precision on any HLO and not just dot/conv

  Show description for 23984dc
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  23984dc

Commits on Oct 14, 2021

• [XLA:ALGEBRAIC_SIMPLIFIER] cleanup the IsConvertPairNoOp to use the primitive util formulation.

  Show description for 08fcfae
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  08fcfae

Commits on Sep 26, 2021

• [XLA] Correctly handle the types of zeros in dot and convolution since they natively support mixed precision.

  Show description for 37cacc4
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  37cacc4

Commits on Sep 24, 2021

• Internal change

  Show description for 65da5e0
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  65da5e0

Commits on Sep 20, 2021

• [XLA:LAYOUT_ASSIGNMENT] Ignore propagation direction for reshapes in layout

  Show description for bcf7164
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  bcf7164

Commits on Sep 13, 2021

• [XLA:ALGEBRAIC_SIMPLIFIER] Do not swap convolution dimensions equivalent to a rhs non contracting dot dimensions.

  Show description for 7c73ab8
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  7c73ab8

• [XLA:SPACE_TO_BATCH] Don't crash trying to get the rank of a variadic reduce

  Show description for 0ec8c25
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  0ec8c25

Commits on Aug 30, 2021

• [XLA:LAYOUT_ASSIGNMENT] Correctly return 0 for log(pow(0,0))

  Show description for f8cd67e
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  f8cd67e

Commits on Aug 12, 2021

• [XLA] NFC, change the CallContextToString

  Show description for 8b347d8
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  8b347d8

• [XLA] NFC rename call contexts to be more clear

  Show description for 5cbc6f7
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  5cbc6f7

Commits on Aug 6, 2021

• [XLA:LAYOUT_ASSIGNMENT] propagate reshape layouts depth first if there are unmodifed dimensions.

  Show description for a99f1df
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  a99f1df

Commits on Jul 31, 2021

• [XLA] Do not scan instructions in a channel dependency group if there is a single computation or the module is SPMD

  Show description for dc9bb6d
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  dc9bb6d

Commits on Jul 30, 2021

• [XLA:LAYOUT_ASSIGNMENT] propagate reshape layouts depth first if there are unmodifed dimensions.

  Show description for 8ed02f6
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  8ed02f6

Commits on Jul 24, 2021

• [XLA] Don't run a pass if it has not been changed in more than 3 fixed pass iterations.

  Show description for fbaadb4
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  fbaadb4

Commits on Jul 22, 2021

• [XLA] Don't run a pass if it has not been changed in more than 3 fixed pass iterations.

  Show description for ca2c450
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  ca2c450

Commits on Jun 30, 2021

• [XLA] Generalize ConvertOperandFolding to handle all pure upcasts.

  Show description for 8859409
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  8859409

Commits on Jun 29, 2021

• [XLA:RESHAPE_MOVER] Don't move scalar reshapes as it easily causes infinite

  Show description for 2dbbb61
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  2dbbb61

Commits on Jun 25, 2021

• [XLA:LAYOUT_ASSIGNMENT] Deprioritize the layout propagation of slicing instruction from output to input.

  Show description for 6c34306
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  6c34306

Commits on Jun 23, 2021

• [XLA] Prefer the layout of compuations with arguments.

  Show description for 7edc828
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  7edc828

Commits on Jun 21, 2021

• [XLA] Do not reserve large hash maps when there are many small computations.

  Show description for 491ea16
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  491ea16

Commits on Jun 15, 2021

• [XLA] Parse invalid{} layout without crashing.

  Show description for 501b7f4
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  501b7f4

Commits on Jun 2, 2021

• [XLA] Only fold transpose into reduce if all uses are reduce

  Show description for 1a8563e
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  1a8563e

Commits on May 24, 2021

• [XLA:ALGEBRAIC_SIMPLIFIER] Rewrite Reduce(Transpose(X)) Transpose(Reduce(X)) for all ranks.

  Show description for afa644b
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  afa644b

Commits on Apr 8, 2021

• [XLA] If dumping all passes is enabled, only dump the HLO if a pass changed the HLO.

  Show description for d063eae
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  d063eae

Commits on Jan 30, 2021

• [XLA] Use array all to all in the client frontend

  Show description for 88b90a0
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  88b90a0

Commits on Jan 20, 2021

• [XLA] Preserve metadata when sinking broadcasts.

  Show description for 43f5aec
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  43f5aec

Commits on Jan 7, 2021

• [XLA] Transform more cases of reshape(iota) to a mixed radix calculation with

  Show description for b18eefe
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  b18eefe

Commits on Oct 30, 2020

• [XLA] Grouped dims do not need to be modified when being swapped as they will

  Show description for ec23155
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  ec23155

Commits on Oct 20, 2020

• [XLA] Add select and scatter to pattern matchers.

  Show description for efebd64
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  efebd64

Commits on Oct 16, 2020

• [XLA] Add select and scatter to pattern mattcher

  Show description for a92f7f9
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  a92f7f9

Commits on Oct 15, 2020

• [XLA] No-op reshapes should not return true for inserting or deleting one-sized

  Show description for 3825d32
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  3825d32

Commits on Oct 14, 2020

• [XLA] Add pattern matcher for select and scatter.

  Show description for 7e54bf3
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  7e54bf3

Commits on Oct 8, 2020

• [XLA] Preserve the layout of the slice shape during algebraic_simplification.

  Show description for c51c793
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  c51c793

• [XLA] log a WARNING for out of range F64 constants instead of a check failure.

  Show description for 886229e
  

  blakehechtman

  authored and

  tensorflower-gardener

  committed
  886229e

Pagination

PreviousNext