Repository navigation

Implements `dpctl.tensor.logsumexp` and `dpctl.tensor.reduce_hypot` - #1446

Merged

ndgrigorian merged 14 commits into

logsumexp-reduce-hypot-impl

Oct 26, 2023

Merged

Implements `dpctl.tensor.logsumexp` and `dpctl.tensor.reduce_hypot`#1446
ndgrigorian merged 14 commits into
masterfrom
logsumexp-reduce-hypot-impl

Commits on Oct 25, 2023
- Implements logsumexp and reduce_hypot
  ndgrigorian
  committed
  c2d04a1
- Implements dedicated kernels for temp reductions over axes 1 and 0 in contiguous arrays
  ndgrigorian
  committed
  7b4a835
- logsumexp and reduce_hypot no longer use atomics
  ndgrigorian
  committed
  aa749b4
- Adds tests for reduce_hypot and logsumexp
  ndgrigorian
  committed
  6b07501
- Arithmetic reductions no longer use atomics for inexact types
  ndgrigorian
  committed
  1247e9b
- Removed support of atomic reduction for max and min
  ndgrigorian
  committed
  9d99b71
- Adds new tests for reductions
  ndgrigorian
  committed
  10ede9a
- Split reductions into multiple source files
  ndgrigorian
  committed
  7d243d8
- Remove unneccessary imports of reduction init functions
  ndgrigorian
  committed
  4638025
- Added functions for querying reduction atomic support per type per function
  ndgrigorian
  committed
  f2babe8
- Corrected ``min`` contig variant typo
  ndgrigorian
  committed
  4e2789d
- Removes _tree_reduction_over_axis
  ndgrigorian
  committed
  4dd054f
- Always use atomic implementation for min/max if available
  
  oleksandr-pavlyk
  authored and
  ndgrigorian
  committed
  448a7f1
- ``logaddexp`` implementation moved to math_utils
  ndgrigorian
  committed
  d88e78f