Skip to content

Bump hipTensor version and update CHANGELOG for ROCm 7.14 #8782

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
evedovelli merged 3 commits into from
Jun 25, 2026
+9 −21
Merged

Bump hipTensor version and update CHANGELOG for ROCm 7.14 #8782

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
af34c58
Bump hipTensor version and update CHANGELOG for ROCm 7.14
evedovelli Jun 24, 2026
0ac980e
Merge new changes with "Since last release" section
evedovelli Jun 24, 2026
899aefa
Bump version in test
evedovelli Jun 24, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 7 additions & 19 deletions projects/hiptensor/CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,28 +2,21 @@

Full documentation for hipTensor is available at [rocm.docs.amd.com/projects/hiptensor](https://rocm.docs.amd.com/projects/hipTensor/en/latest/index.html).

## Since last release ROCm 7.13

### Added

* Added support for new GPU target gfx1250.
* Added YAML-driven CTest test filter standardization, applying `quick`/`standard`/`comprehensive`/`full` tier labels to the installed test tree so tests can be run by tier with `ctest -L <tier>`.

### Optimized

* Re-selected `HIPTENSOR_ALGO_ACTOR_CRITIC` winning kernels for all contraction type/rank/layout combinations to reflect current Composable Kernel tile parameters.

## Since last release ROCm 7.12
## hipTensor 2.3.0 for ROCm 7.14

### Added
* Added Windows support.
* Added contraction support with FP16 and BF16 data and compute types for gfx11 and gfx12 targets.
* Added support for the following new GPU targets:
* gfx11: gfx1100, gfx1101, gfx1102, gfx1103, gfx1150, gfx1151, gfx1152, gfx1153.
* gfx12: gfx1200, gfx1201.
* gfx12: gfx1200, gfx1201, gfx1250.
* Added unary element-wise operators to contraction, including the new `BilinearUnary` class, dedicated instances, samples, and tests.
* Added Dockerfiles (prebuilt and full build) and documentation to streamline hipTensor build environment setup.
* Added the `CREATE_TEST_APP_LOCAL_DEPLOY` CMake option to stage required ROCm DLLs on Windows, and updated the Windows build documentation accordingly.
* Added YAML-driven CTest test filter standardization, applying `quick`/`standard`/`comprehensive`/`full` tier labels to the installed test tree so tests can be run by tier with `ctest -L <tier>`.
* Added support for trinary contractions.
* Added hipTensor to the TheRock build system, enabling source builds, artifact distribution, and CI testing on both Linux and Windows.
* Added native Linux packages with HPC SDK metapackages: `amdrocm-hiptensor` (runtime), `amdrocm-hiptensor-devel` (headers and CMake config), and `amdrocm-hiptensor-test` (CTest binaries).

### Changed
* Replaced numeric UID-based actor-critic kernel lookup with platform-stable string-based kernel name comparison to enable cross-platform compatibility.
Expand All @@ -38,17 +31,12 @@ Full documentation for hipTensor is available at [rocm.docs.amd.com/projects/hip
### Optimized
* Improved column-major contraction performance by applying CK-style stride reordering for column-major inputs.
* Achieved 2x–3x speedup in contraction TFLOPS/s by using switch-case dispatch in `HiptensorUnaryOp` instead of static table lookup.
* Re-selected `HIPTENSOR_ALGO_ACTOR_CRITIC` winning kernels for all contraction type/rank/layout combinations to reflect current Composable Kernel tile parameters.

### Resolved issues
* Fixed use-after-free bug where `hiptensorCreatePlan` held dangling pointers to user-provided descriptors; the plan now deep-copies all descriptors.
* Fixed incorrect BF16 results in contraction with unary ops caused by silent `bhalf_t`-to-float integer promotion in cross-type overloads.

### Known issues
* Unary operations in contraction are not currently supported with `HIPTENSOR_ALGO_ACTOR_CRITIC` for problems with both F16 datatypes and compute types, or both BF16 datatypes and compute types.

### Upcoming changes
* Add support for trinary contraction.

## hipTensor 2.2.0 for ROCm 7.2.0

### Added
Expand Down
2 changes: 1 addition & 1 deletion projects/hiptensor/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -229,7 +229,7 @@ if(HIPTENSOR_BUILD_SAMPLES)
endif()

# Versioning via rocm-cmake
set (VERSION_STRING "2.2.0")
set (VERSION_STRING "2.3.0")
rocm_setup_version(VERSION ${VERSION_STRING})

# Configure a header file to pass the CMake version settings to the source
Expand Down
2 changes: 1 addition & 1 deletion projects/hiptensor/test/00_unit/util_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@ TEST(CheckApiParamsTest, UtilTest)

TEST(hiptensorGetVersionTest, UtilTest)
{
EXPECT_EQ(hiptensorGetVersion(), 2002000);
EXPECT_EQ(hiptensorGetVersion(), 2003000);
}

TEST(logLevelToStringTest, UtilTest)
Expand Down
Loading