CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
11 KiB
11 KiB
Changelog for HIPIFY
Documentation for HIPIFY is available at https://rocmdocs.amd.com/projects/HIPIFY/en/latest/.
HIPIFY for ROCm 7.2.0
Added
- Partial support for CUDA 13.0.0 support
- cuDNN 9.14.0 support
- cuTENSOR 2.3.1.0 support
- LLVM 21.1.6 support
- Full
hipFFTwsupport - #2062 Partial hipification support for a particular CUDA API
- #2073 Detect CUDA version before hipification
- New options:
--local-headersto enable hipification of quoted local headers (non-recursive)--local-headers-recursiveto enable hipification of quoted local headers recursively
Resolved issues
- #2088 Missing support of
cuda_bf16.himport in hipification
HIPIFY for ROCm 7.0.0
Added
- CUDA 12.9.1 support
- cuDNN 9.11.0 support
- cuTENSOR 2.2.0.0 support
- LLVM 20.1.8 support
Resolved issues
hipDNNsupport is removed by default- #1859[hipify-perl] Fix warnings on unsupported Driver or Runtime APIs which were erroneously not reported
- #1930 Revise
JIT API - #1962 Support for cuda-samples helper headers
- #2035 Remove
const_cast<const char**>inhiprtcCreateProgramandhiprtcCompileProgram
HIPIFY for ROCm 6.4.0
Added
- CUDA 12.6.3 support
- cuDNN 9.7.0 support
- cuTENSOR 2.0.2.1 support
- LLVM 19.1.7 support
- Full support for direct hipification of
cuRANDintorocRANDunder the--rocoption - #1617 Support for
fp8math device/host API
Resolved issues
MIOpensupport in hipify-perl under the-miopenoption- Use
const_cast<const char**>for the last arguments in thehiprtcCreateProgramandhiprtcCompileProgramfunction calls, as in CUDA, they are of theconst char* const*type - #1769 Support for
fp16device/host API - #1800 Fix instructions on building LLVM for HIPIFY on Linux
Known issues
- #833
hipify-clangbuild failure against LLVM 15-18 onUbuntu,CentOS, andFedora
HIPIFY for ROCm 6.3.1
Added
- CUDA 12.6.2 support
- cuDNN 9.5.1 support
- LLVM 19.1.3 support
- Full
hipBLAS64-bit APIs support - Full
rocBLAS64-bit APIs support
Resolved issues
- Added missing support for device intrinsics and built-ins:
__all_sync,__any_sync,__ballot_sync,__activemask,__match_any_sync,__match_all_sync,__shfl_sync,__shfl_up_sync,__shfl_down_sync, and__shfl_xor_sync
HIPIFY for ROCm 6.3.0
Added
- CUDA 12.6.1 support
- cuDNN 9.5.0 support
- LLVM 19.1.1 support
rocBLAS64-bit APIs support- Initial support for direct hipification of
cuDNNintoMIOpenunder the--rocoption - Initial support for direct hipification of
cuRANDintorocRANDunder the--rocoption - #1650 Added a filtering ability for the supplementary hipification scripts
Resolved issues
- Correct
rocheader files support
Known issues
- #1617 Support for
fp8data types
HIPIFY for ROCm 6.2.4
Added
- cuDNN 9.3.0 support
Resolved issues
- Removed some post HIP 6.2 APIs from support
- Added hipification support for HIP functions
hipSetValidDevices,hipMemcpy2DArrayToArray,hipMemcpyAtoA,hipMemcpyAtoD,hipMemcpyAtoA,hipMemcpyAtoHAsync, andhipMemcpyHtoAAsync - Fixed an issue with
Skipped some replacementswhen hipification didn't occur at all
HIPIFY for ROCm 6.2.1
Added
- CUDA 12.5.1 support
- cuDNN 9.2.1 support
- LLVM 18.1.8 support
hipBLAS64-bit APIs support- Support for Math Constants
math_constants.h
HIPIFY for ROCm 6.2.0
Added
- CUDA 12.4.1 support
- cuDNN 9.1.1 support
- LLVM 18.1.6 support
- Full
hipBLASLtsupport
Resolved issues
- Apply
reinterpret_castfor an explicit conversion betweenpointer-to-functionandpointer-to-object; affected functions:hipFuncGetAttributes,hipFuncSetAttribute,hipFuncSetCacheConfig,hipFuncSetSharedMemConfig,hipLaunchKernel, andhipLaunchCooperativeKernel
HIPIFY for ROCm 6.1.2
Added
- cuDNN 9.0.0 support
- LLVM 18.1.2 support
- New options:
--clang-resource-directoryto specify the clang resource path - the path to the parent folder for theincludefolder that contains__clang_cuda_runtime_wrapper.hand other header files used during the hipification process
Resolved issues
- Clang resource files used during hipification are now searchable and also can be specified by the
--clang-resource-directoryoption
HIPIFY for ROCm 6.1.0
Added
- CUDA 12.3.2 support
- cuDNN 8.9.7 support
- LLVM 17.0.6 support
- Full
hipSOLVERsupport - Full
rocSPARSEsupport - New options:
--amapto hipify as much as possible, ignoring--default-preprocessorbehavior
Resolved issues
- Code blocks skipped by the Preprocessor are not hipified anymore under the
--default-preprocessoroption
HIPIFY for ROCm 6.0.2
Resolved issues
- Use the new locations of header files of some HIP and ROCm libraries (
hipRAND,hipFFT,rocSOLVER)
HIPIFY for ROCm 6.0.0
Added
- CUDA 12.2.2 support
- cuDNN 8.9.5 support
- LLVM 17.0.3 support
- Improved support for Windows and Visual Studio 2019 and 2022
- More rocSPARSE support
- ABI changes are shown in the 'C' ('Changed') column for CUDA, HIP, and ROC API
Known issues
- #837 Added a new function to call transformation type "additional non-const arg"
- #1014 Added a new function to call transformation type "replace argument with a const"
HIPIFY for ROCm 5.7.0
Added
- CUDA 12.2.0 support
- cuDNN 8.9.2 support
- LLVM 16.0.6 support
- Initial rocSPARSE support
- Initial
CUDA2ROCdocumentation generation for rocBLAS, rocSPARSE, and MIOpen:- In separate files:
hipify-clang --md --doc-format=full --doc-roc=separate - In one file:
hipify-clang --md --doc-format=full --doc-roc=joint
- In separate files:
- New options:
--use-hip-data-types(UsehipDataTypeinstead ofhipblasDatatype_torrocblas_datatype)--doc-roc=\<value\>(ROC documentation generation:skip(default),separate, andjoint; the--mdor--csvoption must be included)
Known issues
- #822 Added a new function to call transformation type "additional const by value arg"
- #830 Added a new function to call transformation type "move arg from place X to place Y"
HIPIFY for ROCm 5.6.0
Added
- CUDA 12.1.0 support
- cuDNN 8.8.1 support
- LLVM 16.0.0 support
- New options:
--default-preprocessor(synonymous with--skip-excluded-preprocessor-conditional-blocks)--no-undocumented-features--no-warnings-on-undocumented-features--versions
Resolved issues
- Accessing
half2 structmembers (undocumented feature) - Retargeted
INSTALLto thebinsubfolder
HIPIFY for ROCm 5.5.0
Added
- Partial CUDA 12.0.0 support
- cuDNN 8.7.0 support
- Initial MIOpem support
- cuBLAS 64-bit API (CUDA 12.0) initial support
- rocBLAS and MIOpen synthetic tests
- LLVM 15.0.7 support
Changed
- Synthetic unit tests for
cuBLAS2rocBLASandcuDNN2MIOpen
HIPIFY for ROCm 5.4.1
Added
- CUDA 11.8.0 support
- cuDNN 8.6.0 support
- Device types support
- LLVM 15.0.4 support
Resolved issues
- Removed
RPATHdefinitions (Linux)
HIPIFY for ROCm 5.4.0
Added
- hipRTC support
- Error handling API support
- hipBLAS synthetic tests
- LLVM 15.0.0 support
HIPIFY for ROCm 5.3.0
Added
- CUDA 11.7.0 support
- cuDNN 8.4.1 support
- CUB initial support
- More synthetic tests
- New options:
--hip-kernel-execution-syntax
Changed
- LLVM 3.8.0 is out of support
- HIPIFY-specific options support in unit testing
Resolved issues
- Patches for LLVM 14.0.x (Windows only)
- Add
GNUInstallDirsfor CMake on Linux
HIPIFY for ROCm 5.2.0
Added
- CUDA 11.6.0 support
- cuDNN 8.3.3 support
- LLVM 14.0.0 support
HIPIFY for ROCm 5.1.0
Added
- CUDA 11.5.1 support
- cuDNN 8.3.2 support
Resolved issues
- Hipification of
cuOccupancyMaxPotentialBlockSizeandcuOccupancyMaxPotentialBlockSizeWithFlags
HIPIFY for ROCm 5.0.0
Added
- CUDA 11.4.2 support
- cuDNN 8.3.2 support
- Initial hipRTC support
- GNU C/C++ 11.2 support
- Visual Studio 2022 support
- LLVM 13.0.0 support
- New options:
--experimental--cuda-kernel-execution-syntax
Changed
- Support for different formats of locally generated documentation
- Experimentally supported APIs
Resolved issues
- Packaging for Debian and RPM Linux distributions
- Undo argument typecasting for four driver API functions (
cuStreamWaitValue32,cuStreamWaitValue64,cuStreamWriteValue32, andcuStreamWriteValue64) because the arguments in the corresponding HIP functions are nowuint
HIPIFY for ROCm 4.5.0
Added
- cuDNN 8.2.4 support
- Initial graph API support
- GNU C/C++ 11.1 support
- LLVM 12.0.1 support
Changed
- Synthetic unit tests
-std=c++14by default
Resolved issues
- Abandoned
HIP_DYNAMIC_SHARED
HIPIFY for ROCm 4.3.0
Added
- CUDA 11.3.0 support
- cuDNN 8.2.0 support
- LLVM 12.0.0 support
Resolved issues
- Added missing type casting arguments for
cuStreamWaitValue32(64)andcuStreamWriteValue32(64)
HIPIFY for ROCm 4.2.0
Added
- CUDA 11.2.2 support
- cuDNN 8.1.1 support
- Initial device API support
- LLVM 11.1.0 support
- New options:
--doc-format=<value>, withfull(default),strict, andcompactoptions
Changed
- Tests on kernel launch syntax
HIPIFY for ROCm 4.1.0
Added
- CUDA 11.2.0 support
- Stream-ordered memory API support
- cuDNN 8.1.1 support
- LLVM 11.0.1 support
Changed
- Initial support for API versioning
Resolved issues
- Patches for LLVM 10.0.x and 11.0.0 (Windows and Linux)
HIPIFY for ROCm 4.0.0
Added
- LLVM 11.0.0 support
HIPIFY for ROCm 3.10.0
Changed
- Revised CUDA and HIP API and data type versioning
- Revised and removed deprecated CUDA and HIP APIs and data types
HIPIFY for ROCm 3.9.0
Added
-
CUDA 11.0.1 support
-
CUDA2HIPdocumentation generation in Markdown and CSV formats -
Versioning support for CUDA and HIP APIs
-
New options:
--md(generate Markdown documentation)--csv(generate CSV documentation)
Changed
-
Improved
hipify-perlgeneration
HIPIFY for ROCm 3.8.0
Added
- cuDNN 8.0.2 support
compile_commands.jsonsupport (-p <build-path>)
Changed
- Improved
hipify-perlgeneration
HIPIFY for ROCm 3.7.0
Added
- CUDA 11.0.0 support
- Linux packaging
- LLVM 10.0.1 support
HIPIFY for ROCm 3.6.0
Added
deprecatedflag for all corresponding CUDA and HIP APIs
Changed
- Added warning for all deprecated APIs
HIPIFY for ROCm 3.5.0
Added
- CUDA 10.2.0 support
- cuDNN 7.6.5 support
- LLVM 10.0.0 support
Changed
hipify-clangandclangoptions separator (--) support