Commits

Breadcrumbs

History for

ondev

User selector

All users

All time

Commit History

Commits on Mar 8, 2023

• GPU: Support multiple HIP target architectures in one build, and switch from deprecated --amdgpu-target to --offload-arch

  davidrohr

  committed
  49fe708

  Copy full SHA for 49fe708

Commits on Feb 7, 2023

• GPU: Add printout of HIP version to CMake

  davidrohr

  committed
  a2bc07f

  Copy full SHA for a2bc07f

Commits on Feb 4, 2023

• GPU: Pass correct GCC toolchain to hipcc, since autodetection seems not to work

  davidrohr

  committed
  e69dc4b

  Copy full SHA for e69dc4b

Commits on Jan 16, 2023

• GPU: Workaround disabling fast-math on AMD GPUs, since it crashes

  davidrohr

  committed
  50ded79

  Copy full SHA for 50ded79

Commits on Jan 10, 2023

• GPU: Fix ROCm path auto-detection

  davidrohr

  committed
  ad83bbe

  Copy full SHA for ad83bbe

Commits on Jan 6, 2023

• CMake: fix FindO2GPU syntax

  

  aphecetche

  authored and

  davidrohr

  committed
  9563509

  Copy full SHA for 9563509

Commits on Jan 2, 2023

• GPU: Improve ROCm path auto-detection, if not set but /opt/rocm exists, just use it

  davidrohr

  committed
  74be470

  Copy full SHA for 74be470

• GPU: Do not auto-detect AMD clang-ocl, since it has compiler problems, and can break O2 compiltion if auto-detected

  davidrohr

  committed
  64d58ec

  Copy full SHA for 64d58ec

Commits on Dec 26, 2022

• GPU: Add cmake option to disable all fast-math and non ieee denormal treatments

  davidrohr

  committed
  e5cb596

  Copy full SHA for e5cb596

Commits on Oct 24, 2022

• GPU: Fix CUDA compiler detection with unsupported GCC verison under CMake >=3.24

  davidrohr

  committed
  628c068

  Copy full SHA for 628c068

Commits on Jul 29, 2022

• GPU: Fix CMake part when allowing non-supported GCC version for CUDA

  davidrohr

  committed
  c3d8ab8

  Copy full SHA for c3d8ab8

Commits on May 17, 2022

• GPU: Adapt hipcc compile options to ROCm 5.x

  

  davidrohr

  authored and

  TimoWilken

  committed
  7062c80

  Copy full SHA for 7062c80

Commits on May 2, 2022

• Revert rocm 5.x fix

  

  mconcas

  authored and

  davidrohr

  committed
  33066f9

  Copy full SHA for 33066f9

• Fix FindO2GPU for rocm 5

  

  mconcas

  authored and

  davidrohr

  committed
  1ea1605

  Copy full SHA for 1ea1605

Commits on Apr 25, 2022

• Make CMake fatalise if hipify-perl is missing

  

  mconcas

  authored and

  davidrohr

  committed
  f5d53e0

  Copy full SHA for f5d53e0

Commits on Jan 27, 2022

• GPU: Was not using clang and llvm-spirv from the alidist clang installation

  davidrohr

  committed
  f3bedb4

  Copy full SHA for f3bedb4

Commits on Jan 25, 2022

• GPU: Add flag requested by AMD to avoid a memory issue

  

  David Rohr

  authored and

  davidrohr

  committed
  67c193c

  Copy full SHA for 67c193c

• GPU: Add -fgpu-defer-diag option to hipcc

  davidrohr

  committed
  597d5e2

  Copy full SHA for 597d5e2

Commits on Sep 11, 2021

• Some CMake improvements for GPU detection

  davidrohr

  committed
  f77e8a1

  Copy full SHA for f77e8a1

Commits on Jun 23, 2021

• Updated copyright headers in all files with new text … (#6430)

  Show description for 03608ff
  ihrivnac

  authored
  03608ff

  Copy full SHA for 03608ff

Commits on Jun 7, 2021

• GPU: Enforce the CUDA Architecture we set, otherwise some other CMake module might override it

  davidrohr

  committed
  fbe5289

  Copy full SHA for fbe5289

Commits on May 14, 2021

• GPU: AMD HIP creates a cache variable, overriding our settings if not FORCEd

  davidrohr

  committed
  8acc0c8

  Copy full SHA for 8acc0c8

Commits on Apr 25, 2021

• Fix CUDA version check, apparently the CMake way has changed meanwhile

  davidrohr

  committed
  6395e6b

  Copy full SHA for 6395e6b

Commits on Apr 21, 2021

• GPU: Require clang >= 13 for OpenCL2 compilation

  davidrohr

  committed
  7263003

  Copy full SHA for 7263003

• GPU: Require at least CUDA 11.3

  davidrohr

  committed
  759c329

  Copy full SHA for 759c329

Commits on Mar 4, 2021

• Add missing find_package if OCL2 is requested but OCL1 is not

  davidrohr

  committed
  e651938

  Copy full SHA for e651938

Commits on Jan 27, 2021

• GPU: Apparently one must not do find_package(ROCM)

  davidrohr

  committed
  592ae04

  Copy full SHA for 592ae04

Commits on Jan 9, 2021

• GPU: Disable CMake warnings already earlier

  davidrohr

  committed
  b5dc383

  Copy full SHA for b5dc383

Commits on Jan 7, 2021

• GPU: Disable some ROCm CMake warnings

  davidrohr

  committed
  6bbd168

  Copy full SHA for 6bbd168

Commits on Nov 10, 2020

• Use 61 as minimum CUDA compute capability to support half precision

  davidrohr

  committed
  79250b7

  Copy full SHA for 79250b7

Commits on Aug 27, 2020

• GPU: Use CMake 3.18 style to set CUDA GPU Targets

  davidrohr

  committed
  bca6140

  Copy full SHA for bca6140

Commits on Aug 26, 2020

• GPU: Disable CUDA host compiler check

  davidrohr

  committed
  2f10fe8

  Copy full SHA for 2f10fe8

• GPU: Use C++17 in CUDA compilation with CUDA 11/CMake 3.18

  davidrohr

  committed
  63e6af7

  Copy full SHA for 63e6af7

Commits on Aug 25, 2020

• GPU: Switch to CUDA 11

  davidrohr

  committed
  1dad6ed

  Copy full SHA for 1dad6ed

• GPU: Adjust CMake for hip-clang

  davidrohr

  committed
  24bb8a1

  Copy full SHA for 24bb8a1