Code

Improve Dense stride handling#774

Code

Merged

upsj merged 17 commits intodevelopginkgo-project/ginkgo:developfrom dense_stridesginkgo-project/ginkgo:dense_stridesCopy head branch name to clipboardJun 3, 2021

+2,962-623

Commits

Commits on Jun 1, 2021

• avoid modifying stride on Dense convert_to

  upsj

  committed
  8ca33bd

  Copy full SHA for 8ca33bd

• add stride handling to Dense precision conversion

  upsj

  committed
  494a413

  Copy full SHA for 494a413

• add out-parameter versions to Dense functions

  upsj

  committed
  7bedb7f

  Copy full SHA for 7bedb7f

• improve Dense reference tests

  upsj

  committed
  550e400

  Copy full SHA for 550e400

• add Reference tests for Dense output parameters

  upsj

  committed
  7319687

  Copy full SHA for 7319687

• add device tests for Dense output functions

  upsj

  committed
  0173df2

  Copy full SHA for 0173df2

• remove Identity kernel calls from core tests

  upsj

  committed
  7d04012

  Copy full SHA for 7d04012

• fix icc compilation issues

  upsj

  committed
  4d7ff23

  Copy full SHA for 4d7ff23

• remove support for Intel 2017

  upsj

  committed
  84b2f77

  Copy full SHA for 84b2f77

• review update

  Show description for cc30d8f
  

  3 peoplecommitted
  cc30d8f

  Copy full SHA for cc30d8f

• improve dense convert_to

  Show description for 65228ef
  upsj

  committed
  65228ef

  Copy full SHA for 65228ef

• add make_temporary_output_clone

  Show description for f455156
  upsj

  committed
  f455156

  Copy full SHA for f455156

Commits on Jun 2, 2021

• fix ref dense temporary clone test without omp

  upsj

  committed
  040749f

  Copy full SHA for 040749f

• actually avoid copies in temporary_output_clone

  upsj

  committed
  8b62161

  Copy full SHA for 8b62161

• unify permutation core implementations

  upsj

  committed
  caa8595

  Copy full SHA for caa8595

• avoid copies on empty matrices

  Show description for 4c847f4
  upsj

  committed
  4c847f4

  Copy full SHA for 4c847f4

Commits on Jun 3, 2021

• fix single-precision Dense copy on dpcpp

  Show description for 4677856
  

  upsj

  and

  Thomas Grützmacher

  committed
  4677856

  Copy full SHA for 4677856