derek.holloway
7502bf64a4
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
391 lines
14 KiB
Markdown
391 lines
14 KiB
Markdown
# Changelog for hipSPARSE
|
|
|
|
Documentation for hipSPARSE is available at
|
|
[https://rocm.docs.amd.com/projects/hipSPARSE/en/latest/](https://rocm.docs.amd.com/projects/hipSPARSE/en/latest/).
|
|
|
|
|
|
## (Unreleased) hipSPARSE 4.7.0
|
|
|
|
### Added
|
|
* Added Blocked ELL format support to the `hipsparseDenseToSparse` routine, along with the new `hipsparseBlockedEllSetPointers` function.
|
|
* Added the `HIPSPARSE_SPMV_CSR_ALG3` algorithm to `hipsparseSpMV`, which exposes the rocSPARSE CSR nnz split algorithm (`rocsparse_spmv_alg_csr_nnzsplit`).
|
|
* Added CSC format support to `hipsparseSpSV`.
|
|
* Added CSC format support to `hipsparseSpSM`.
|
|
|
|
### Resolved issues
|
|
* Fixed an issue with `hipsparseSpMM`, which produced incorrect results for the Blocked ELL sparse format.
|
|
|
|
## hipSPARSE 4.6.0 for ROCm 7.14
|
|
|
|
### Added
|
|
* Added `hipsparseCreateBsr` and `hipsparseCreateConstBsr` in order to enable BSR format support in generic routines.
|
|
* Added BSR format support to `hipsparseSpMV`.
|
|
* Added BSR format support to `hipsparseSpMM`.
|
|
|
|
### Resolved issues
|
|
* Fixed an issue where calling `hipsparseSpMV` multiple times with different `hipsparseOperation_t`, `hipsparseSpMVAlg_t`, or compute-datatypes using the same sparse matrix descriptor resulted in errors.
|
|
|
|
### Upcoming changes
|
|
* The routines `hipsparseXcsrgeamNnz`, `hipsparseScsrgeam`, `hipsparseDcsrgeam`, `hipsparseCcsrgeam`, and `hipsparseZcsrgeam` have been deprecated and will be removed in a future release. Use `hipsparseScsrgeam2_bufferSizeExt`, `hipsparseDcsrgeam2_bufferSizeExt`, `hipsparseCcsrgeam2_bufferSizeExt`, `hipsparseZcsrgeam2_bufferSizeExt`, `hipsparseXcsrgeam2Nnz`, `hipsparseScsrgeam2`, `hipsparseDcsrgeam2`, `hipsparseCcsrgeam2`, and `hipsparseZcsrgeam2` instead.
|
|
|
|
## hipSPARSE 4.5.0 for ROCm 7.12.0
|
|
|
|
### Added
|
|
* For sparse matrix vector product (`hipsparseSpMV`), when using `HIPSPARSE_SPMV_ALG_DEFAULT`, the routine now automatically falls back to a supported algorithm depending on the sparse matrix format and requested operation. For example, CSR format with transposed operations or CSC format with non-transposed operations will fall back to an appropriate algorithm.
|
|
|
|
### Resolved issues
|
|
* Fixed an issue where out-of-bounds memory reads can occur in the single precision bsrxmv kernels when `block_dim` equals `5` or `8`.
|
|
|
|
## hipSPARSE 4.4.0 for ROCm 7.11.0
|
|
|
|
### Added
|
|
* Added brain half float mixed precision to `hipsparseSpMV` where A, X, and Y use bfloat16 and the compute type uses float.
|
|
* Added half float mixed precision to `hipsparseSpMV` where A, X, and Y use float16 and the compute type uses float.
|
|
* Added brain half float mixed precision to `hipsparseSpMM` where A, B, and C use bfloat16 and the compute type use float.
|
|
* Added half float mixed precision to `hipsparseSpMM` where A, B, and C use float16 and the compute type use float.
|
|
* Added the `debian`, `almalinux`, `rockylinux`, and `oraclelinux` OS names to install script.
|
|
* Added sliced ELL format support to the `hipsparseSpMV` routine.
|
|
|
|
### Resolved issues
|
|
* In `hipsparseSpSM_solve()`, the external buffer is passed as a parameter, which does not match the NVIDIA CUDA cuSPARSE API. The `hipsparseSpSM_solve_ex()` routine has been added to properly match the
|
|
NVIDIA CUDA cuSPARSE `cusparseSpSM_solve()` API. The original `hipsparseSpSM_solve()` routine has been
|
|
deprecated and will be removed in a future release.
|
|
|
|
### Upcoming changes
|
|
* The routine `hipsparseSpSM_solve()` has been deprecated and will be removed in a future release.
|
|
Users should use `hipsparseSpSM_solve_ex()` instead.
|
|
|
|
## hipSPARSE 4.2.0 for ROCm 7.2.0
|
|
|
|
### Added
|
|
|
|
* Added `--clients-only` option to the `install.sh` and `rmake.py` scripts to allow building only the clients while using an already installed version of hipSPARSE.
|
|
|
|
### Optimized
|
|
|
|
* Improved the user documentation.
|
|
|
|
### Resolved issues
|
|
|
|
* Fixed a memory leak in `hipsparseCreate` functions.
|
|
|
|
## hipSPARSE 4.1.0 for ROCm 7.1.0
|
|
|
|
### Added
|
|
|
|
* Added brain half float mixed precision to `hipsparseAxpby` where X and Y use bfloat16 and result and the compute type use float.
|
|
* Added brain half float mixed precision to `hipsparseSpVV` where X and Y use bfloat16 and result and the compute type use float.
|
|
* Added brain half float mixed precision to `hipsparseSpMV` where A and X use bfloat16 and Y and the compute type use float.
|
|
* Added brain half float mixed precision to `hipsparseSpMM` where A and B use bfloat16 and C and the compute type use float.
|
|
* Added brain half float mixed precision to `hipsparseSDDMM` where A and B use bfloat16 and C and the compute type use float.
|
|
* Added brain half float mixed precision to `hipsparseSDDMM` where A and B and C use bfloat16 and the compute type use float.
|
|
* Added half float mixed precision to `hipsparseSDDMM` where A and B and C use float16 and the compute type use float.
|
|
* Added brain half float uniform precision to `hipsparseScatter` and `hipsparseGather` routines.
|
|
* Added documentation for installing and building hipSPARSE on Microsoft Windows.
|
|
|
|
## hipSPARSE 4.0.1 for ROCm 7.0.0
|
|
|
|
### Added
|
|
|
|
* Add the `int8`, `int32`, and `float16` data types to `hipDataTypeToHCCDataType` so that sparse matrix descriptors can be used with them.
|
|
* Added half float mixed precision to `hipsparseAxpby` where X and Y use float16 and result and the compute type use float.
|
|
* Added half float mixed precision to `hipsparseSpVV` where X and Y use float16 and result and the compute type use float.
|
|
* Added half float mixed precision to `hipsparseSpMM` where A and B use float16 and C and the compute type use float.
|
|
* Added half float mixed precision to `hipsparseSDDMM` where A and B use float16 and C and the compute type use float.
|
|
* Added half float uniform precision to `hipsparseScatter` and `hipsparseGather` routines.
|
|
* Added half float uniform precision to `hipsparseSDDMM` routine.
|
|
* Add `int8` precision to `hipsparseCsr2cscEx2` routine.
|
|
* Add the `almalinux` OS name to correct the gfortran dependency.
|
|
|
|
### Changed
|
|
|
|
* Switch to defaulting to C++17 when building hipSPARSE from source. Previously hipSPARSE was using C++14 by default.
|
|
|
|
### Resolved issues
|
|
|
|
* Fixed a compilation [issue](https://github.com/ROCm/hipSPARSE/issues/555) related to using `std::filesystem` and C++14.
|
|
* Fixed the empty clients-common package by moving the `hipsparse_clientmatrices.cmake` and `hipsparse_mtx2csr` files to it.
|
|
|
|
### Known issues
|
|
|
|
* In `hipsparseSpSM_solve()`, the external buffer is passed as a parameter. This does not match the NVIDIA CUDA cuSPARSE API. This extra external buffer parameter will be removed in a future release. For now, this extra parameter can be ignored and nullptr passed because it is unused internally by `hipsparseSpSM_solve()`.
|
|
|
|
## hipSPARSE 3.2.0 for ROCm 6.4.0
|
|
|
|
### Added
|
|
|
|
* Added build dependencies for CentOS/RHEL 9 in install script
|
|
|
|
### Changed
|
|
|
|
* Moved the `hipsparse_clientmatrices.cmake` and `hipsparse_mtx2csr` files from the `hipsparse-tests` package to the `hipsparse-clients-common` package
|
|
|
|
### Optimized
|
|
|
|
* Removed unused `GTest` dependency from `hipsparse-bench`
|
|
|
|
### Known issues
|
|
|
|
* In `hipsparseSpSM_solve()`, the external buffer is passed as a parameter. This does not match the NVIDIA CUDA cuSPARSE API. This extra external buffer parameter will be removed in a future release. For now this extra parameter can be ignored and nullptr passed as it is unused internally by `hipsparseSpSM_solve()`.
|
|
|
|
## hipSPARSE 3.1.2 for ROCm 6.3.0
|
|
|
|
### Added
|
|
|
|
* Added an alpha version of the hipsparse-bench executable to facilitate comparing NVIDIA CUDA cuSPARSE and rocSPARSE backends
|
|
|
|
### Changed
|
|
|
|
* Changed the default compiler from hipcc to amdclang in the install script and cmake files.
|
|
|
|
### Optimized
|
|
|
|
* Improved the user documentation
|
|
|
|
### Resolved issues
|
|
|
|
* Fixed the gfortran dependency for the `azurelinux` operating system.
|
|
|
|
### Known issues
|
|
|
|
* In `hipsparseSpSM_solve()`, the external buffer is passed as a parameter. This does not match the NVIDIA CUDA cuSPARSE API. This extra external buffer parameter will be removed in a future release. For now this extra parameter can be ignored and nullptr passed as it is unused internally by `hipsparseSpSM_solve()`.
|
|
|
|
## hipSPARSE 3.1.1 for ROCm 6.2.0
|
|
|
|
### Additions
|
|
|
|
* Added missing `hipsparseCscGet()` routine
|
|
|
|
### Changes
|
|
|
|
* All internal hipSPARSE functions now exist inside a namespace
|
|
* Match deprecations found in NVIDIA CUDA cuSPARSE 12.x.x when using NVIDIA CUDA cuSPARSE backend.
|
|
|
|
### Fixes
|
|
|
|
* Fixed SpGEMM and SpGEMM_reuse routines which were not matching NVIDIA CUDA cuSPARSE behaviour
|
|
|
|
### Optimizations
|
|
|
|
* Improved user manual
|
|
* Improved contribution guidelines
|
|
|
|
### Known issues
|
|
|
|
* In `hipsparseSpSM_solve()`, we currently pass the external buffer as a parameter. This does not match the NVIDIA CUDA cuSPARSE API and this extra external buffer parameter will be removed in a future release. For now this extra parameter can be ignored and nullptr passed as it is unused internally by `hipsparseSpSM_solve()`.
|
|
|
|
## hipSPARSE 3.0.1 for ROCm 6.1.0
|
|
|
|
### Fixes
|
|
* Fixes to the build chain
|
|
|
|
## hipSPARSE 3.0.0 for ROCm 6.0.0
|
|
|
|
### Additions
|
|
|
|
* Added `hipsparseGetErrorName` and `hipsparseGetErrorString`
|
|
|
|
### Changes
|
|
|
|
* Changed the `hipsparseSpSV_solve()` API function to match the NVIDIA CUDA cuSPARSE API
|
|
* Changed generic API functions to use const descriptors
|
|
* Improved documentation
|
|
|
|
## hipSPARSE 2.3.8 for ROCm 5.7.0
|
|
|
|
### Fixes
|
|
|
|
* Compilation failures when using the NVIDIA CUDA cuSPARSE 12.1.0 and 12.0.0 backends
|
|
* Compilation failures when using the NVIDIA CUDA cuSPARSE 10.1 (non-update version) backend
|
|
|
|
## hipSPARSE 2.3.7 for ROCm 5.6.1
|
|
|
|
### Fixes
|
|
|
|
* Reverted an undocumented API change in hipSPARSE 2.3.6 that affected the `hipsparseSpSV_solve`
|
|
function
|
|
|
|
## hipSPARSE 2.3.6 for ROCm 5.6.0
|
|
|
|
### Additions
|
|
|
|
* Added SpGEMM algorithms
|
|
|
|
### Changes
|
|
|
|
* blockDim == 0 now returns `HIPSPARSE_STATUS_INVALID_SIZE` for `hipsparseXbsr2csr` and
|
|
`hipsparseXcsr2bsr`
|
|
|
|
## hipSPARSE 2.3.5 for ROCm 5.5.0
|
|
|
|
### Fixes
|
|
|
|
* Fixed an issue where the `rocm` folder was not removed after upgrading meta packages
|
|
* Fixed a compilation issue with the NVIDIA CUDA cuSPARSE backend
|
|
* Added more detailed messages for unit test failures due related to missing input data
|
|
* Improved documentation
|
|
* Fixed a bug with deprecation messages when using gcc9
|
|
|
|
## hipSPARSE 2.3.3 for ROCm 5.4.0
|
|
|
|
### Additions
|
|
|
|
* Added `hipsparseCsr2cscEx2_bufferSize` and `hipsparseCsr2cscEx2` routines
|
|
|
|
### Changes
|
|
|
|
* `HIPSPARSE_ORDER_COLUMN` has been renamed to `HIPSPARSE_ORDER_COL` in order to match
|
|
NVIDIA CUDA cuSPARSE
|
|
|
|
## hipSPARSE 2.3.1 for ROCm 5.3.0
|
|
|
|
### Additions
|
|
|
|
* Added SpMM and SpMM batched for CSC format
|
|
|
|
## hipSPARSE 2.2.0 for ROCm 5.2.0
|
|
|
|
### Additions
|
|
|
|
* New packages for test and benchmark executables on all supported operating systems using CPack
|
|
|
|
## hipSPARSE 2.1.0 for ROCm 5.1.0
|
|
|
|
### Additions
|
|
|
|
* Added `gtsv_interleaved_batch` and `gpsv_interleaved_batch` routines
|
|
* Added `SpGEMM_reuse`
|
|
|
|
### Changes
|
|
|
|
* Changed `BUILD_CUDA` with `USE_CUDA` in the install script and CMake files
|
|
* Updated GoogleTest to 11.1
|
|
|
|
### Fixes
|
|
|
|
* Fixed a bug in SpMM Alg versioning
|
|
|
|
## hipSPARSE 2.0.0 for ROCm 5.0.0
|
|
|
|
### Additions
|
|
|
|
* Added (conjugate) transpose support for `csrmv`, `hybmv`, and `spmv` routines
|
|
|
|
## hipSPARSE 1.11.2 for ROCm 4.5.0
|
|
|
|
### Additions
|
|
|
|
* Triangular solve for multiple right-hand sides using BSR format
|
|
* SpMV for BSRX format
|
|
* Enhanced SpMM in CSR format to work with transposed A
|
|
* Matrix coloring for CSR matrices
|
|
* Batched tridiagonal solve (`gtsv_strided_batch`)
|
|
* SpMM for BLOCKED ELL format
|
|
* Generic routines for SpSV and SpSM
|
|
* Beta support for Windows 10
|
|
* Additional atomic-based algorithms for SpMM in COO format
|
|
* Additional algorithm for SpMM in CSR format
|
|
|
|
### Changes
|
|
|
|
* Packaging has been split into a runtime package (`hipsparse`) and a development package
|
|
(`hipsparse-devel`):
|
|
The development package depends on the runtime package. When installing the runtime package,
|
|
the package manager will suggest the installation of the development package to aid users
|
|
transitioning from the previous version's combined package. This suggestion by package manager is
|
|
for all supported operating systems (except CentOS 7) to aid in the transition. The `suggestion`
|
|
feature in the runtime package is introduced as a deprecated feature and will be removed in a future
|
|
ROCm release.
|
|
|
|
* GTest dependency has been updated to v1.10.0
|
|
|
|
### Fixes
|
|
|
|
* Fixed a bug with `gemvi` on Navi21
|
|
|
|
### Optimizations
|
|
|
|
* Optimization for pivot-based GTSV
|
|
|
|
## hipSPARSE 1.10.7 for ROCm 4.3.0
|
|
|
|
### Additions
|
|
|
|
* Tridiagonal solve with and without pivoting (batched)
|
|
* Dense matrix sparse vector multiplication (`gemvi`)
|
|
* Sampled dense-dense matrix multiplication (`sddmm`)
|
|
|
|
## hipSPARSE 1.10.6 for ROCm 4.2.0
|
|
|
|
### Additions
|
|
|
|
* Generic API support, including SpMM
|
|
|
|
## hipSPARSE 1.10.4 for ROCm 4.1.0
|
|
|
|
### Additions
|
|
|
|
* Generic API support, including Axpby, Gather, Scatter, Rot, SpVV, SpMV, SparseToDense,
|
|
DenseToSparse, and SpGEMM
|
|
|
|
## hipSPARSE 1.9.6 for ROCm 4.0.0
|
|
|
|
### Additions
|
|
|
|
* Changelog file
|
|
* `csr2gebsr`
|
|
* `gebsr2csr`
|
|
* `gebsr2gebsc`
|
|
* `gebsr2gebsr`
|
|
|
|
### Changes
|
|
|
|
* Updates to Debian package name.
|
|
|
|
## hipSPARSE 1.9.4 for ROCm 3.9
|
|
|
|
### Additions
|
|
|
|
* `prune_csr2csr, prune_dense2csr_percentage` and `prune_csr2csr_percentage`
|
|
* `bsrilu0`
|
|
|
|
## hipSPARSE 1.8.1 for ROCm 3.8
|
|
|
|
### Additions
|
|
|
|
* `bsric0`
|
|
|
|
## hipSPARSE 1.7.1 for ROCm 3.7
|
|
|
|
### Additions
|
|
|
|
* Fortran bindings
|
|
* Triangular solve for BSR format (bsrsv)
|
|
* CentOS 6 support
|
|
|
|
## hipSPARSE 1.7.1 for ROCm 3.6
|
|
|
|
### Additions
|
|
|
|
* Fortran bindings
|
|
* Triangular solve for BSR format (bsrsv)
|
|
* CentOS 6 support
|
|
|
|
## hipSPARSE 1.6.5 for ROCm 3.5
|
|
|
|
### Additions
|
|
|
|
* Switched to HIP-Clang as default compiler
|
|
* `csr2dense`, `csc2dense`, `csr2csr_compress`, `nnz_compress`, `bsr2csr`, `csr2bsr`, `bsrmv`, and
|
|
`csrgeam`
|
|
* static build
|
|
* New examples
|
|
|
|
### Optimizations
|
|
|
|
* `dense2csr`, `dense2csc`
|
|
|
|
### Fixes
|
|
|
|
* Installation process
|