github/ROCm
1
1
Fork 0
mirror of https://github.com/ROCm/ROCm.git synced 2026-01-09 14:48:06 -05:00
Code Issues Packages Projects Releases Wiki Activity

Fix hip7 rn (#523)

Browse Source 
* Update RELEASE.md

Update per LRT meeting notes

* Update RELEASE.md

move warpSize change as requested

* Update RELEASE.md

update warpSize change wording.

* Update RELEASE.md

* Update RELEASE.md

Why either?

* Update RELEASE.md

Add content from HIP 7 Changelog

* Update RELEASE.md

looks good

* Update RELEASE.md

Co-authored-by: Julia Jiang <56359287+jujiang-del@users.noreply.github.com>

---------

Co-authored-by: Julia Jiang <56359287+jujiang-del@users.noreply.github.com>
This commit is contained in:
randyh62
2025-08-26 16:02:49 -07:00
committed by GitHub
parent 59afdef1fb
commit a7edb17538
1 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
RELEASE.md
View File

@@ -884,11 +884,12 @@ functions added for logical reduction. For details, see [Warp cross-lane functio
- HIP APIs for `FP4`/`FP6`/`FP8`, which are compatible with corresponding CUDA APIs.
- HIP Extensions APIs for microscaling formats, which are supported on AMD GPUs.
* New `wptr` and `rptr` values in `ClPrint`, for better logging in dispatch barrier methods.
* New debug mask, to print precise code object information for logging.
* The `_sync()` version of crosslane builtins such as `shfl_sync()` are enabled by default. These can be disabled by setting the preprocessor macro `HIP_DISABLE_WARP_SYNC_BUILTINS`.
* Added `constexpr` operators for `fp16`/`bf16`.
* Added warp level primitives: `__syncwarp` and reduce intrinsics (e.g. `__reduce_add_sync()`)
* Extended fine grained system memory pool.
* Support for the flags in APIs as following, now allows uncached memory allocation.
- `hipExtHostRegisterUncached`, used in `hipHostRegister`.
- `hipHostMallocUncached` and `hipHostAllocUncached`, used in `hipHostMalloc` and `hipHostAlloc`.
* `num_threads` total number of threads in the group. The legacy API size is alias.
* Added PCI CHIP ID information as the device attribute.
* Added new tests applications for OCP data types `FP4`/`FP6`/`FP8`.
@@ -898,6 +899,8 @@ functions added for logical reduction. For details, see [Warp cross-lane functio
* Some unsupported GPUs such as gfx9, gfx8 and gfx7 are deprecated on Microsoft Windows.
* Removal of beta warnings in HIP Graph APIs
All Beta warnings in usage of HIP Graph APIs are removed, they are now officially and fully supported.
* `warpSize` has changed.
In order to match the CUDA specification, the `warpSize` variable is no longer `constexpr`. In general, this should be a transparent change; however, if an application was using `warpSize` as a compile-time constant, it will have to be updated to handle the new definition. For more information, see the discussion of `warpSize` within the [HIP C++ language extensions](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_cpp_language_extensions.html#warpsize).
* Behavior changes
- `hipGetLastError` now returns the error code which is the last actual error caught in the current thread during the application execution.
- Cooperative groups in `hipLaunchCooperativeKernelMultiDevice` and `hipLaunchCooperativeKernel` functions, additional input parameter validation checks are added.
@@ -999,9 +1002,6 @@ In order to match the CUDA runtime behavior more closely, HIP APIs with streams
- Event Management Related APIs
* `hipEventRecord`
* `hipEventRecordWithFlags`
* `warpSize` Change
In order to match the CUDA specification, the `warpSize` variable is no longer `constexpr`. In general, this should be a transparent change; however, if an application was using `warpSize` as a compile-time constant, it will have to be updated to handle the new definition. For more information, see either the discussion of `warpSize` within the [HIP C++ language extensions](https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_cpp_language_extensions.html#warpsize).
#### Optimized
@@ -1021,7 +1021,6 @@ HIP runtime has the following functional improvements which improves runtime per
Developers can now use the environment variable `HSA_SCRATCH_SINGLE_LIMIT_ASYNC` to change the default allocation size with expected scratch limit in ROCR runtime. On top of it, this value can also be overwritten programmatically in the application using the HIP API `hipDeviceSetLimit(hipExtLimitScratchCurrent, value)` to reset the scratch limit value.
* HIP runtime now enables peer-to-peer (P2P) memory copies to utilize all available SDMA engines, rather than being limited to a single engine. It also selects the best engine first to give optimal bandwidth.
* Improved launch latency for `D2D` copies and `memset` on MI300 series.
* Memory manager was implemented to improve the efficiency of memory usage and speed-up memory allocation/free in memory pools.
* Introduced a threshold to handle the command submission patch to the GPU device(s), considering the synchronization with CPU, for performance improvement.
#### Resolved issues
@@ -1037,6 +1036,11 @@ HIP runtime has the following functional improvements which improves runtime per
* A numerical error/corruption found in Pytorch during graph replay. HIP runtime fixed the input sizes of kernel launch dimensions in hipExtModuleLaunchKernel for the execution of hipGraph capture.
* A crash during kernel execution in a customer application. The structure of kernel arguments was updated via adding the size of kernel arguments, and HIP runtime does validation before launch kernel with the structured arguments.
#### Known issues
* `hipLaunchHostFunc` returns an error during stream capture. Any application using `hipLaunchHostFunc` might fail to capture graphs during stream capture, instead, it returns `hipErrorStreamCaptureUnsupported`.
* Compilation failure in kernels via hiprtc when use option `std=c++11`.
### **hipBLAS** (3.0.0)
#### Added