rocBLAS 5.1.0 for ROCm 7.1.0

Added

• Sample for clients using OpenMP threads calling rocBLAS functions.
• gfx1103, gfx1150, and gfx1151 enabled.

Changed

• By default, the Tensile build is no longer based on tensile_tag.txt but uses the same commit from shared/tensile in the rocm-libraries repository. The rmake or install -t option can build from another local path with a different commit.

Optimized

• Improved the performance of Level 2 gemv transposed (TransA != N) for the problem sizes where m is small and n is large on gfx90a and gfx942.

rocBLAS 5.0.2 for ROCm 7.0.2

Added

• Enabled gfx1150 and gfx1151.
• The ROCBLAS_USE_HIPBLASLT_BATCHED variable to independently control the batched hipblaslt backend. Set ROCBLAS_USE_HIPBLASLT_BATCHED=0 to disable batched GEMM use of the hipblaslt backend.

rocBLAS 4.4.1 for ROCm 6.4.4

rocBLAS code for ROCm 6.4.4 did not change. The library was rebuilt for the updated ROCm 6.4.4 stack.

rocblas 5.0.0 for ROCm 7.0.1

rocBLAS code for ROCm 7.0.1 did not change. The library was rebuilt for the updated ROCm 7.0.1 stack.

rocBLAS 5.0.0 for ROCm 7.0.0

Added

• gfx950 support
• ROCBLAS_LAYER = 8 internal API logging for gemm debugging
• Support for AOCL 5.0 gcc build as a client reference library
• Allow PkgConfig for client reference library fallback detection

Changed

• CMAKE_CXX_COMPILER is now passed on during compilation for a Tensile build
• Change default atomics mode from allowed to not allowed

Removed

• Support code for non-production gfx targets
• rocblas_hgemm_kernel_name, rocblas_sgemm_kernel_name, and rocblas_dgemm_kernel_name API functions
• Use of warpSize as a constexpr
• Use of deprecated behavior of hipPeekLastError
• rocblas_float8.h and rocblas_hip_f8_impl.h files
• rocblas_gemm_ex3, rocblas_gemm_batched_ex3, rocblas_gemm_strided_batched_ex3 API functions

Optimized

• Optimized gemm by using gemv kernels when applicable
• Optimized gemv for small m and n with a large batch count on gfx942
• Improved the performance of Level 1 dot for all precisions and variants when N > 100000000 on gfx942
• Improved the performance of Level 1 asum and nrm2 for all precisions and variants on gfx942
• Improved the performance of Level 2 sger (single precision) on gfx942
• Improved the performance of Level 3 dgmm for all precisions and variants on gfx942

Resolved issues

• Fixed environment variable path-based logging to append multiple handle output to the same file
• Support numerics when trsm is running with rocblas_status_perf_degraded
• Fixed the build dependency installation of joblib on some operating systems
• Return rocblas_status_internal_error when rocblas_[set,get]_ [matrix,vector] is called with a host pointer in place of a device pointer
• Reduced the default verbosity level for internal GEMM backend information
• Updated from the deprecated rocm-cmake to ROCmCMakeBuildTools
• Corrected AlmaLinux gfortran package dependencies

Upcoming changes

• Deprecated the use of negative indices to indicate the default solution is being used for gemm_ex with rocblas_gemm_algo_solution_index

rocBLAS 4.4.1 for ROCm 6.4.3

rocBLAS code for ROCm 6.4.3 did not change. The library was rebuilt for the updated ROCm 6.4.3 stack.

rocBLAS 4.4.1 for ROCm 6.4.2

Resolved issues

• Zero imaginary portion of diagonal of C matrix for cherk/zherk for gfx90a/gfx942 with problem sizes k > 500

rocBLAS 4.4.0 for ROCm 6.4.1

rocBLAS code for ROCm 6.4.1 did not change. The library was rebuilt for the updated ROCm 6.4.1 stack.

rocBLAS 4.4.0 for ROCm 6.4.0

Added

• rocTX support in rocBLAS (not available on Windows or in the static library version on Linux)
• On gfx12, all functions now support full rocblas_int dynamic range for batch_count
• --ninja build option
• Support for GPU_TARGETS cmake variable

Changed

• rocblas-test client removes the stress tests unless YAML-based testing or gtest_filter adds them
• rocblas clients OpenMP default threading is reduced to be less than the logical core count
• gemm_ex testing and timing reuses device memory
• gemm_ex timing initializes matrices on device

Optimized

• Significantly reduced workspace memory requirements for Level 1 ILP64: iamax and iamin
• Reduced workspace memory requirements for Level 1 ILP64: dot, asum, nrm2
• Improved the performance of Level 2 gemv for the problem sizes (TransA == N && m > 2*n) and (TransA == T)
• Improved the performance of Level 3 syrk and herk for the problem size (k > 500 && n < 4000)

Resolved issues

• gfx12: ger, geam, geam_ex, dgmm, trmm, symm, hemm, ILP64 gemm, and larger data support
• Added a gfortran package dependency for Azure Linux OS
• Outdated SLES OS package dependencies (cxxtools and joblib) in install.sh -d
• Code object stripping for RPM packages

Upcoming changes

• Deprecated the cmake variable AMDGPU_TARGETS. Use GPU_TARGETS instead.

rocBLAS 4.3.0 for ROCm 6.3.3

rocBLAS code for ROCm 6.3.3 did not change. The library was rebuilt for the updated ROCm 6.3.3 stack.