Skip to content

GPU: Unify usage of half precision float into half_t typedef in GPUCommonMath.h - #16149

Open
davidrohr wants to merge 2 commits into
AliceO2Group:devfrom
davidrohr:PR
Open

davidrohr wants to merge 2 commits into
AliceO2Group:devfrom
davidrohr:PR

Conversation

@davidrohr

Copy link
Copy Markdown
Collaborator

@ChSonnabend : Could you check if that works correctly for your ONNXRuntime code?

@davidrohr
davidrohr requested review from a team and sgorbuno as code owners November 18, 2026 10:41
@davidrohr

Copy link
Copy Markdown
Collaborator Author

@ktf : Could you check the metal changes?

if (IsNegative()) {
return -o.f;
}
#elif __cplusplus >= 202302L && defined(__STDCPP_FLOAT16_T__)

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
#elif __cplusplus >= 202302L && defined(__STDCPP_FLOAT16_T__)
#if __cplusplus >= 202302L && defined(__STDCPP_FLOAT16_T__)

no?

@ktf

ktf commented Nov 18, 2026

Copy link
Copy Markdown
Member

Apart from the inline comment, everything else is bit-to-bit compatible on metal.

@alibuild

Copy link
Copy Markdown
Collaborator

Error while checking build/O2/fullCI_slc9 for f147c03 at 2026-11-18 16:13:

## sw/BUILD/O2Physics-latest/log
/sw/SOURCES/O2Physics/slc9_x86-64-slc9_x86-64/0/Tutorials/ML/benchmarkOnnxModel.cxx:19:10: fatal error: ML/3rdparty/GPUORTFloat16.h: No such file or directory
ninja: build stopped: subcommand failed.


## sw/BUILD/o2checkcode-latest/log
--
========== List of errors found ==========
++ echo 'Found 0 errors and 0 warnings.'
Found 0 errors and 0 warnings.
++ [[ 0 -gt 0 ]]
++ [[ 0 -gt 0 ]]
++ [[ 0 -gt 0 ]]
++ mkdir -p /sw/INSTALLROOT/80eff5f0f77ea9d6e3cb212127adf57f2e5889b8/slc9_x86-64/o2checkcode/1.0-local62/etc/modulefiles
++ alibuild-generate-module
+ HAS_BIN=
+ HAS_LIB=
+ HAS_CMAKE=
+ HAS_ROOT=
+ '[' 0 '!=' 0 ']'
+ cat
+ printf 'if ![ is-loaded '\''BASE/1.0'\'' ] {\n module load BASE/1.0\n}'
+ echo 'BUILD_REQUIRES=CMake alibuild-recipe-tools defaults-release'
BUILD_REQUIRES=CMake alibuild-recipe-tools defaults-release
+ FULL_BUILD_REQUIRES='defaults-release FairCMakeModules UUID safe_int ninja-fortran Alice-GRID-Utils ninja date double-conversion Python-modules-list CMake pytorch_cpuinfo json-c googlebenchmark alibuild-recipe-tools bz2 O2-customization MPFR'
+ echo 'FULL_BUILD_REQUIRES=defaults-release FairCMakeModules UUID safe_int ninja-fortran Alice-GRID-Utils ninja date double-conversion Python-modules-list CMake pytorch_cpuinfo json-c googlebenchmark alibuild-recipe-tools bz2 O2-customization MPFR'
FULL_BUILD_REQUIRES=defaults-release FairCMakeModules UUID safe_int ninja-fortran Alice-GRID-Utils ninja date double-conversion Python-modules-list CMake pytorch_cpuinfo json-c googlebenchmark alibuild-recipe-tools bz2 O2-customization MPFR
++ env
++ cut -f1 -d=
++ grep -v '^DEFAULT_'
++ grep -v PKGREVISION
++ grep -v ALIBUILD_RECIPE_TOOLS
++ grep REVISION
++ sed -e s/_REVISION//
+ for x in $(env | cut -f1 -d= | grep -v "^DEFAULT_" | grep -v PKGREVISION | grep -v ALIBUILD_RECIPE_TOOLS | grep REVISION | sed -e 's/_REVISION//')
++ eval 'echo $LIBFFI_REVISION'
+++ echo 14
+ REVISION_VALUE=14
++ eval 'echo $LIBFFI_VERSION'
+++ echo v3.2.1-alice1
+ VERSION_VALUE=v3.2.1-alice1
++ eval 'echo $LIBFFI_ROOT'
+++ echo /sw/slc9_x86-64/libffi/v3.2.1-alice1-14
+ ROOT_PATH_VALUE=/sw/slc9_x86-64/libffi/v3.2.1-alice1-14
+ '[' -z 14 ']'
+ echo 'defaults-release FairCMakeModules UUID safe_int ninja-fortran Alice-GRID-Utils ninja date double-conversion Python-modules-list CMake pytorch_cpuinfo json-c googlebenchmark alibuild-recipe-tools bz2 O2-customization MPFR'
+ tr '[:lower:]' '[:upper:]'
+ tr - _
+ tr ' ' '\n'
[0 more errors; see full log]

Full log here.

@alibuild

Copy link
Copy Markdown
Collaborator

Error while checking build/O2/fullCI_slc9 for f5bad33 at 2026-11-18 16:14:

## sw/BUILD/O2Physics-latest/log
/sw/SOURCES/O2Physics/slc9_x86-64-slc9_x86-64/0/Tutorials/ML/benchmarkOnnxModel.cxx:207:38: error: request for member 'ToFloat' in '(& halves)->std::vector<_Float16>::operator[](i)', which is of non-class type '__gnu_cxx::__alloc_traits<std::allocator<_Float16>, _Float16>::value_type' {aka '_Float16'}
/sw/SOURCES/O2Physics/slc9_x86-64-slc9_x86-64/0/Tutorials/ML/benchmarkOnnxModel.cxx:302:74: error: request for member 'ToFloat' in '*(fp16Data + ((sizetype)(i * 2)))', which is of non-class type 'const _Float16'
ninja: build stopped: subcommand failed.

Full log here.

@alibuild

Copy link
Copy Markdown
Collaborator

Error while checking build/O2/fullCI_slc9 for 20e707d at 2026-11-18 17:45:

## sw/BUILD/O2Physics-latest/log
/sw/SOURCES/O2Physics/slc9_x86-64-slc9_x86-64/0/Tutorials/ML/benchmarkOnnxModel.cxx:207:38: error: request for member 'ToFloat' in '(& halves)->std::vector<_Float16>::operator[](i)', which is of non-class type '__gnu_cxx::__alloc_traits<std::allocator<_Float16>, _Float16>::value_type' {aka '_Float16'}
/sw/SOURCES/O2Physics/slc9_x86-64-slc9_x86-64/0/Tutorials/ML/benchmarkOnnxModel.cxx:302:74: error: request for member 'ToFloat' in '*(fp16Data + ((sizetype)(i * 2)))', which is of non-class type 'const _Float16'
ninja: build stopped: subcommand failed.

Full log here.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

3 participants