hipTensor 1.4.0 for ROCm 6.3.1

hipTensor code for ROCm 6.3.1 did not change. The library was rebuilt for the updated ROCm 6.3.1 stack.

hipTensor 1.4.0 for ROCm 6.3.0

Additions

• Added API support for tensor reduction of ranks 2, 3, 4, 5 and 6
• Added CPU reference for tensor reductions
• Added unit tests for tensor reductions
• Added documentation for tensor reductions

Changes

• Updated target archs for ASAN builds
• ASAN library builds now use -mcmodel=large to accommodate larger lib size
• Updated permute backend to accommodate changes to element-wise ops implementation
• Updated validation acceptance criteria to match CK backend tests

Fixes

• Fixed a bug in randomized tensor input data generation
• Various documentation formatting updates and fixes
• Split kernel instances to improve build times

hipTensor 1.3.0 for ROCm 6.2.4

hipTensor code for ROCm 6.2.4 did not change. The library was rebuilt for the updated ROCm 6.2.4 stack.

hipTensor 1.3.0 for ROCm 6.2.2

hipTensor code for ROCm 6.2.2 did not change. The library was rebuilt for the updated ROCm 6.2.2 stack.

hipTensor 1.3.0 for ROCm 6.2.1

hipTensor code for ROCm 6.2.1 did not change. The library was rebuilt for the updated ROCm 6.2.1 stack.

hipTensor 1.3.0 for ROCm 6.2.0

Additions

• Added support for tensor permutation of ranks of 2, 3, 4, 5 and 6
• Added tests for tensor permutation of ranks of 2, 3, 4, 5 and 6
• Added support for tensor contraction of M6N6K6: M, N, K up to rank 6
• Added tests for tensor contraction of M6N6K6: M, N, K up to rank 6
• Added new test YAML parsing to support sequential parameters ordering

Changes

• Documentation updates for installation, programmer's guide and API reference
• Prefer amd-llvm-devel package before system LLVM library
• Preferred compilers changed to CC=amdclang CXX=amdclang++
• Updated actor-critic selection for new contraction kernel additions

Fixes

• Fixed LLVM parsing crash
• Fixed memory consumption issue in complex kernels
• Work-around implemented for compiler crash during debug build
• Allow random modes ordering for tensor contractions

rocm-6.1.2

ROCm release v6.1.2

rocm-6.1.1

ROCm release v6.1.1

hipTensor 1.2.0 for ROCm 6.1.0

Additions

• API support for permutation of rank 4 tensors: f16 and f32
• New datatype support in contractions of rank 4: f16, bf16, complex f32, complex f64
• Added scale and bilinear contraction samples and tests for new supported data types
• Added permutation samples and tests for f16, f32 types

Fixes

• Fixed bug in contraction calculation with data type f32

rocm-6.0.2

ROCm release v6.0.2