Code

Atomic min max GS operator#1704

Code

Merged

MartinKarp merged 14 commits intodevelopExtremeFLOW/neko:developfrom feat/atomic-min-maxExtremeFLOW/neko:feat/atomic-min-maxCopy head branch name to clipboardFeb 13, 2025

Commits

Commits on Feb 6, 2025

• atomicMin atomicMax and unit tests

  tuananhdao

  committed
  5998615

  Copy full SHA for 5998615

• add benchmarks for GS_OP_MIN and GS_OP_MAX

  tuananhdao

  committed
  8de2743

  Copy full SHA for 8de2743

Commits on Feb 7, 2025

• add atomicCAS for comparison

  tuananhdao

  committed
  9b275a2

  Copy full SHA for 9b275a2

• fix sp/dp compiling hip

  tuananhdao

  committed
  40e43cf

  Copy full SHA for 40e43cf

• fix sp,dp for atomicMax_CAS

  tuananhdao

  committed
  a272928

  Copy full SHA for a272928

• update cuda kernels

  tuananhdao

  committed
  fd96077

  Copy full SHA for fd96077

• fix check negative for long long presentation of double

  tuananhdao

  committed
  1170881

  Copy full SHA for 1170881

Commits on Feb 10, 2025

• add comment to remove atomicCAS for the PR

  tuananhdao

  committed
  2240af0

  Copy full SHA for 2240af0

• Merge branch 'develop' into feat/atomic-min-max

  tuananhdao

  committed
  2ddc1d8

  Copy full SHA for 2ddc1d8

Commits on Feb 12, 2025

• correct atomicMin atomicMax

  tuananhdao

  committed
  7ab320b

  Copy full SHA for 7ab320b

• improve performance for atomic Min Max

  tuananhdao

  committed
  4d0290c

  Copy full SHA for 4d0290c

Commits on Feb 13, 2025

• sync cuda with hip gs kernels

  tuananhdao

  committed
  46e4552

  Copy full SHA for 46e4552

• cast second argument of atomicMin to unsigned long long for GNU compilers

  tuananhdao

  committed
  0dad4bc

  Copy full SHA for 0dad4bc

• guard cuda compute capability >= 6.0 for 64-bit atomics

  tuananhdao

  committed
  69d39e7

  Copy full SHA for 69d39e7