Commits

Breadcrumbs

History for

ondev

User selector

All users

All time

Commit History

Commits on Dec 27, 2025

• GPU: Change default CUDA architectures to 80-real 86-real 89-real 120-real 75-virtual

  davidrohr

  committed
  0bc34ef

  Copy full SHA for 0bc34ef

Commits on Nov 29, 2025

• GPU Standalone: Add GPUCA_BUILD_DEBUG_HOSTONLY debug option

  davidrohr

  committed
  3f8a384

  Copy full SHA for 3f8a384

Commits on Nov 17, 2025

• GPU CMake: Add option to override deterministic mode

  davidrohr

  committed
  ea2972e

  Copy full SHA for ea2972e

Commits on Nov 12, 2025

• Find Thrust + fix compilation on CUDA13 (#14801)

  Show description for 72e6d01
  mpuccio

  authored
  72e6d01

  Copy full SHA for 72e6d01

Commits on Oct 24, 2025

• GPU: Fix architecture name, we do not use MI2xx

  davidrohr

  committed
  e0c17bc

  Copy full SHA for e0c17bc

Commits on Sep 19, 2025

• GPU CMake: Bump default CUDA architecture to 75 to support CUDA 13

  davidrohr

  committed
  17da522

  Copy full SHA for 17da522

Commits on Aug 25, 2025

• GPU CMake: Force using alidist GCC for host compilation of CUDA/HIP code

  davidrohr

  committed
  012946b

  Copy full SHA for 012946b

Commits on Jun 16, 2025

• GPU CUDA CMake: Do not print warnings that architectures will be depreacted

  davidrohr

  committed
  fe8111b

  Copy full SHA for fe8111b

• GPU CMake: Switch from deprecated AMDGPU_TARGETS to GPU_TARGETS

  davidrohr

  committed
  d4fb131

  Copy full SHA for d4fb131

Commits on Jun 4, 2025

• Add a version number to FindO2GPU.cmake and remark to keep it in sync with alidist

  davidrohr

  committed
  0c08a1f

  Copy full SHA for 0c08a1f

Commits on May 29, 2025

• GPU CMake: Some adaptions for gpu-systems package

  davidrohr

  committed
  49882f7

  Copy full SHA for 49882f7

Commits on May 27, 2025

• GPU RTC: Check that current compiler versions are the same as those used to create RTC code in cache

  davidrohr

  committed
  a7534db

  Copy full SHA for a7534db

Commits on Apr 23, 2025

• GPU CMake: Another attempt to silence the CMake warnings about architectures

  davidrohr

  committed
  1f6767c

  Copy full SHA for 1f6767c

Commits on Apr 22, 2025

• GPU CMake: Set GPU architecture before checking the language, to suppress warnings about architecture detection failure

  davidrohr

  committed
  4204d44

  Copy full SHA for 4204d44

Commits on Apr 18, 2025

• GPU CMake: Fix compile flags, particularly HIP deterministic mode was missing -ffp-contract=off

  davidrohr

  committed
  b856a63

  Copy full SHA for b856a63

Commits on Mar 27, 2025

• GPU: Make launch bounds of GPUDefParamters available as C++ struct at runtime

  davidrohr

  committed
  6d54cfc

  Copy full SHA for 6d54cfc

• GPU HIP CMake: Remove deprecated option

  davidrohr

  committed
  d1fcd96

  Copy full SHA for d1fcd96

• GPU CMake: Cleanup and do not use Clang x86 denormals flag for device compilation

  davidrohr

  committed
  d9d78bc

  Copy full SHA for d9d78bc

Commits on Mar 26, 2025

• GPU Compile Flags: Set denormals to zero unconditionally

  Show description for 78d0f9c
  davidrohr

  committed
  78d0f9c

  Copy full SHA for 78d0f9c

Commits on Mar 24, 2025

• GPU: Replace GPUCA_NO_FAST_MATH by more fine-grain GPUCA_DETERMINISTIC_MODE

  davidrohr

  committed
  e836e3f

  Copy full SHA for e836e3f

• GPU CMake: Shuffle stuff a bit and move all NO_FAST_MATH stuff to GPU/... and FindO2GPU.cmake

  davidrohr

  committed
  ee497d7

  Copy full SHA for ee497d7

Commits on Mar 14, 2025

• GPU CMake: Auto-detect ROCm even if CMake prefix path not set

  davidrohr

  committed
  b183c55

  Copy full SHA for b183c55

Commits on Mar 7, 2025

• GPU: Bump required LLVM version for OpenCL

  davidrohr

  committed
  f34afc2

  Copy full SHA for f34afc2

Commits on Jan 17, 2025

• GPU: Rename OpenCL2 to OpenCL, now that we have only 1 OpenCL implementation

  davidrohr

  committed
  843479b

  Copy full SHA for 843479b

• GPU:Remove OpenCL 1.2, now that OpenCL 2 supports all its functionality

  davidrohr

  committed
  90ca34c

  Copy full SHA for 90ca34c

Commits on Sep 11, 2024

• GPU: GPUCA_BUILD_DEBUG makes sure all possible source files are compiled for all kernels

  davidrohr

  committed
  cc64785

  Copy full SHA for cc64785

• GPU CMake: Add check for minimum ROCm version

  davidrohr

  committed
  158e6f8

  Copy full SHA for 158e6f8

Commits on Sep 7, 2024

• GPU: More WORKAROUNDs for CUDA bug exposing device code via host symbols

  davidrohr

  committed
  b1438a9

  Copy full SHA for b1438a9

• GPU CMake: Raise required CUDA version to 12.6

  davidrohr

  committed
  1a27e5d

  Copy full SHA for 1a27e5d

Commits on Aug 23, 2024

• GPU Compilation: print CUDA resource usage only if requested by GPUCA_KERNEL_RESOURCE_USAGE_VERBOSE flag, also print HIP resource usage

  davidrohr

  committed
  3e82457

  Copy full SHA for 3e82457

• GPU CMake: With new ROCm/Cmake, GCC usage seems to work, no need any more to enforce gcc-toolchain

  davidrohr

  committed
  9b6e47a

  Copy full SHA for 9b6e47a

Commits on Aug 20, 2024

• GPU CMake: Remove unsed O2_HIP_CMAKE_LINK_FLAGS and create HIP --offload-arch arguments only when needed

  davidrohr

  committed
  c481ad4

  Copy full SHA for c481ad4

Commits on Aug 9, 2024

• GPU CMake: Set amdgpu-function-calls=true, in case not set by the ROCm scripts / cmake files

  davidrohr

  committed
  d40ccfa

  Copy full SHA for d40ccfa

• GPU Cmake: allow unsupported GCC only if we don't autodetect CUDA

  davidrohr

  committed
  ded3e55

  Copy full SHA for ded3e55

Commits on Aug 8, 2024

• GPU CMake: Make compatible to ROCm >= 6.2

  davidrohr

  committed
  6c44a3f

  Copy full SHA for 6c44a3f