mirror of
https://github.com/ROCm/ROCm.git
synced 2026-04-27 03:01:52 -04:00
7.0.0 release notes feedback updated [Batch 6] (#550)
* RN changes updated * Changelog synced and release notes updated * Compatibility changes added
This commit is contained in:
Pratik Basyal
41
CHANGELOG.md
41
CHANGELOG.md
@@ -33,20 +33,19 @@ for a complete overview of this release.
|
||||
- Increased available JPEG engines to 40. Current ASICs might not support all 40. These are indicated as `UINT16_MAX` or `N/A` in CLI.
|
||||
|
||||
* Bad page threshold count.
|
||||
- Added `amdsmi_get_gpu_bad_page_threshold` to Python API and CLI; root/sudo permissions required to display the count.
|
||||
- Added `amdsmi_get_gpu_bad_page_threshold` to Python API and CLI; root/sudo permissions are required to display the count.
|
||||
|
||||
* CPU model name for RDC.
|
||||
- Added new C and Python API `amdsmi_get_cpu_model_name`.
|
||||
- Not sourced from esmi library.
|
||||
|
||||
* Added `amdsmi_get_cpu_affinity_with_scope()`.
|
||||
* New API `amdsmi_get_cpu_affinity_with_scope()`.
|
||||
|
||||
* `socket power` to `amdsmi_get_power_info`
|
||||
- Previously the C API had the value in the `amdsmi_power_info` structure, but was unused
|
||||
- Now we populate the value in both C and Python APIs
|
||||
- Previously, the C API had the value in the `amdsmi_power_info` structure, but was unused.
|
||||
- The value is representative of the socket's power agnostic of the the GPU version.
|
||||
|
||||
* New event notification types to `amdsmi_evt_notification_type_t`.
|
||||
* New event notification types to `amdsmi_evt_notification_type_t`.
|
||||
The following values were added to the `amdsmi_evt_notification_type_t` enum:
|
||||
- `AMDSMI_EVT_NOTIF_EVENT_MIGRATE_START`
|
||||
- `AMDSMI_EVT_NOTIF_EVENT_MIGRATE_END`
|
||||
@@ -58,7 +57,7 @@ for a complete overview of this release.
|
||||
- `AMDSMI_EVT_NOTIF_PROCESS_START`
|
||||
- `AMDSMI_EVT_NOTIF_PROCESS_END`
|
||||
|
||||
- Power cap to `amd-smi monitor`.
|
||||
- Power cap to `amd-smi monitor`.
|
||||
- `amd-smi monitor -p` will display the power cap along with power.
|
||||
|
||||
#### Changed
|
||||
@@ -66,7 +65,7 @@ for a complete overview of this release.
|
||||
* Separated driver reload functionality from `amdsmi_set_gpu_memory_partition()` and
|
||||
`amdsmi_set_gpu_memory_partition_mode()` APIs -- and from the CLI `amd-smi set -M <NPS mode>`.
|
||||
|
||||
* Disabled `amd-smi monitor --violation` on guest. Modified `amd-smi metric --throttle` to alias to `amd-smi metric --violation`.
|
||||
* Disabled `amd-smi monitor --violation` on guests. Modified `amd-smi metric -T/--throttle` to alias to `amd-smi metric -v/--violation`.
|
||||
|
||||
* Updated `amdsmi_get_clock_info` in `amdsmi_interface.py`.
|
||||
- The `clk_deep_sleep` field now returns the sleep integer value.
|
||||
@@ -87,13 +86,17 @@ for a complete overview of this release.
|
||||
- `acc_low_utilization`, `per_low_utilization`, `active_low_utilization`
|
||||
- Python API and CLI now report these expanded fields.
|
||||
|
||||
* The char arrays in the following structures have been changed.
|
||||
* The char arrays in the following structures have been changed.
|
||||
- `amdsmi_vbios_info_t` member `build_date` changed from `AMDSMI_MAX_DATE_LENGTH` to `AMDSMI_MAX_STRING_LENGTH`.
|
||||
- `amdsmi_dpm_policy_entry_t` member `policy_description` changed from `AMDSMI_MAX_NAME` to `AMDSMI_MAX_STRING_LENGTH`.
|
||||
- `amdsmi_name_value_t` member `name` changed from `AMDSMI_MAX_NAME` to `AMDSMI_MAX_STRING_LENGTH`.
|
||||
|
||||
* For backwards compatibility, updated `amdsmi_bdf_t` union to have an identical unnamed struct.
|
||||
|
||||
* Updated `amdsmi_get_temp_metric` and `amdsmi_temperature_type_t` with new values.
|
||||
- Added new values to `amdsmi_temperature_type_t` representing various baseboard and GPU board temperature measures.
|
||||
- Updated `amdsmi_get_temp_metric` API to be able to take in and return the respective values for the new temperature types.
|
||||
|
||||
#### Removed
|
||||
|
||||
- Unnecessary API, `amdsmi_free_name_value_pairs()`
|
||||
@@ -106,9 +109,9 @@ for a complete overview of this release.
|
||||
|
||||
- Unused member `year` in struct `amdsmi_version_t`.
|
||||
|
||||
- `amdsmi_io_link_type_t` and replaced with `amdsmi_link_type_t`.
|
||||
- `amdsmi_io_link_type_t` has been replaced with `amdsmi_link_type_t`.
|
||||
- `amdsmi_io_link_type_t` is no longer needed as `amdsmi_link_type_t` is sufficient.
|
||||
- `amdsmi_link_type_t` enum has changed.
|
||||
- `amdsmi_link_type_t` enum has changed; primarily, the ordering of the PCI and XGMI types.
|
||||
- This change will also affect `amdsmi_link_metrics_t`, where the link_type field changes from `amdsmi_io_link_type_t` to `amdsmi_link_type_t`.
|
||||
|
||||
- `amdsmi_get_power_info_v2()`.
|
||||
@@ -133,7 +136,7 @@ for a complete overview of this release.
|
||||
|
||||
- Removed partition information from the default `amd-smi static` CLI command.
|
||||
- Users can still retrieve the same data by calling `amd-smi`, `amd-smi static -p`, or `amd-smi partition -c -m`/`sudo amd-smi partition -a`.
|
||||
- Reading ``current_compute_partition`` may momentarily wake the GPU up. This is due to reading XCD registers, which is expected behavior. Changing partitions is not a trivial operation, `current_compute_partition` SYSFS controls this action.
|
||||
- Reading `current_compute_partition` may momentarily wake the GPU up. This is due to reading XCD registers, which is expected behavior. Changing partitions is not a trivial operation, `current_compute_partition` SYSFS controls this action.
|
||||
|
||||
- Optimized CLI command `amd-smi topology` in partition mode.
|
||||
- Reduced the number of `amdsmi_topo_get_p2p_status` API calls to one fourth.
|
||||
@@ -144,6 +147,10 @@ for a complete overview of this release.
|
||||
|
||||
- Fixed `amd-smi monitor` decoder utilization (`DEC%`) not showing up on MI300 series ASICs.
|
||||
|
||||
#### Known issues
|
||||
|
||||
- `amd-smi monitor` on Linux Guest systems triggers an attribute error.
|
||||
|
||||
```{note}
|
||||
See the full [AMD SMI changelog](https://github.com/ROCm/amdsmi/blob/release/rocm-rel-7.0/CHANGELOG.md) for details, examples, and in-depth descriptions.
|
||||
```
|
||||
@@ -653,7 +660,7 @@ HIP runtime has the following functional improvements which improves runtime per
|
||||
|
||||
#### Added
|
||||
|
||||
* Support for OCP `FP8` on AMD Instinct MI350X accelerators.
|
||||
* Support for OCP `FP8` on AMD Instinct MI350X GPUs.
|
||||
* Support for PyTorch 2.7 via Torch-MIGraphX.
|
||||
* Support for the Microsoft ONNX Contrib Operators (Self) Attention, RotaryEmbedding, QuickGelu, BiasAdd, BiasSplitGelu, SkipLayerNorm.
|
||||
* Support for Sigmoid and AddN TensorFlow operators.
|
||||
@@ -1027,7 +1034,7 @@ Review the [README](https://github.com/ROCm/rocm_bandwidth_test/blob/amd-mainlin
|
||||
* L2 to EA stalls
|
||||
* L2 to EA stalls per channel
|
||||
|
||||
* Roofline support for AMD Instinct MI350 series accelerators.
|
||||
* Roofline support for AMD Instinct MI350 series GPUs.
|
||||
|
||||
##### Textual User Interface (TUI) (beta version)
|
||||
|
||||
@@ -1037,9 +1044,9 @@ Review the [README](https://github.com/ROCm/rocm_bandwidth_test/blob/amd-mainlin
|
||||
|
||||
##### PC Sampling (beta version)
|
||||
|
||||
* Stochastic (hardware-based) PC sampling has been enabled for AMD Instinct MI300X series and later accelerators.
|
||||
* Stochastic (hardware-based) PC sampling has been enabled for AMD Instinct MI300X series and later GPUs.
|
||||
|
||||
* Host-trap PC Sampling has been enabled for AMD Instinct MI200 series and later accelerators.
|
||||
* Host-trap PC Sampling has been enabled for AMD Instinct MI200 series and later GPUs.
|
||||
|
||||
* Support for sorting of PC sampling by type: offset or count.
|
||||
|
||||
@@ -1200,7 +1207,7 @@ See the full [ROCm SMI changelog](https://github.com/ROCm/rocm_smi_lib/blob/rele
|
||||
|
||||
#### Added
|
||||
|
||||
- Support for AMD Instinct MI350X and MI355X accelerators.
|
||||
- Support for AMD Instinct MI350X and MI355X GPUs.
|
||||
- Introduced rotating buffer mechanism for GEMM operations.
|
||||
- Support for read and write tests in Babel.
|
||||
- Support for AMD Radeon RX9070 and RX9070GRE graphics cards.
|
||||
@@ -1316,7 +1323,7 @@ The previous default accumulator types could lead to situations in which unexpec
|
||||
#### Added
|
||||
|
||||
- Support for [rocJPEG](https://rocm.docs.amd.com/projects/rocJPEG/en/latest/index.html) API Tracing.
|
||||
- Support for AMD Instinct MI350X and MI355X accelerators.
|
||||
- Support for AMD Instinct MI350X and MI355X GPUs.
|
||||
- `rocprofiler_create_counter` to facilitate adding custom derived counters at runtime.
|
||||
- Support in `rocprofv3` for iteration based counter multiplexing.
|
||||
- Perfetto support for counter collection.
|
||||
|
||||
60
RELEASE.md
60
RELEASE.md
@@ -71,11 +71,11 @@ matrix](../../docs/compatibility/compatibility-matrix.rst) for the complete list
|
||||
|
||||
#### Updated framework support
|
||||
|
||||
ROCm 7.0 introduces several newly supported versions of Deep learning and AI frameworks:
|
||||
ROCm 7.0.0 introduces several newly supported versions of Deep learning and AI frameworks:
|
||||
|
||||
##### PyTorch
|
||||
|
||||
ROCm 7.0 enables the following PyTorch features:
|
||||
ROCm 7.0.0 enables the following PyTorch features:
|
||||
|
||||
* Support for PyTorch 2.7.
|
||||
* Integrated Fused Rope kernels in APEX.
|
||||
@@ -84,7 +84,7 @@ ROCm 7.0 enables the following PyTorch features:
|
||||
|
||||
##### JAX
|
||||
|
||||
ROCm 7.0 enables support for JAX 0.6.0.
|
||||
ROCm 7.0.0 enables support for JAX 0.6.0.
|
||||
|
||||
##### Megatron-LM
|
||||
|
||||
@@ -98,11 +98,15 @@ Megatron-LM for ROCm now supports:
|
||||
|
||||
##### TensorFlow
|
||||
|
||||
ROCm 7.0 enables support for TensorFlow 2.19.1.
|
||||
ROCm 7.0.0 enables the following TensorFlow support:
|
||||
|
||||
* Support for TensorFlow 2.19.1.
|
||||
* MX data type support for AMD Instinct MI350 series GPUs.
|
||||
* Triton autotuner.
|
||||
|
||||
##### ONNX Runtime
|
||||
|
||||
ROCm 7.0 enables support for ONNX Runtime 1.22.0.
|
||||
ROCm 7.0.0 enables support for ONNX Runtime 1.22.0.
|
||||
|
||||
##### vLLM
|
||||
|
||||
@@ -111,7 +115,7 @@ ROCm 7.0 enables support for ONNX Runtime 1.22.0.
|
||||
|
||||
##### Triton
|
||||
|
||||
ROCm 7.0 enables support for Triton 3.3.0.
|
||||
ROCm 7.0.0 enables support for Triton 3.3.0.
|
||||
|
||||
#### New frameworks
|
||||
|
||||
@@ -122,9 +126,9 @@ AMD ROCm has officially added support for the following Deep learning and AI fra
|
||||
* llama.cpp is an open-source framework for Large Language Model (LLM) inference that runs on both central processing units (CPUs) and graphics processing units (GPUs). It is written in plain C/C++, providing a simple, dependency-free setup. It is currently supported on ROCm 6.4.0. For more information, see [llama.cpp compatibility](https://advanced-micro-devices-rocm-internal--500.com.readthedocs.build/en/500/compatibility/ml-compatibility/llama-cpp-compatibility.html).
|
||||
|
||||
|
||||
### Instinct Driver/ROCm packaging separation
|
||||
### AMD GPU Driver/ROCm packaging separation
|
||||
|
||||
The Instinct Driver is now distributed separately from the ROCm software stack and is stored under in its own location ``/amdgpu/`` in the package repository at [repo.radeon.com](https://repo.radeon.com/amdgpu/). The first release is designated as Instinct Driver version 30.10. See the [ROCm Gets Modular: Meet the Instinct Datacenter GPU Driver](https://rocm.blogs.amd.com/ecosystems-and-partners/instinct-gpu-driver/README.html) blog and [User and kernel-space support matrix](https://rocm.docs.amd.com/projects/install-on-linux-internal/en/latest/reference/user-kernel-space-compat-matrix.html) for more information.
|
||||
The AMD GPU Driver is now distributed separately from the ROCm software stack and is stored under in its own location ``/amdgpu/`` in the package repository at [repo.radeon.com](https://repo.radeon.com/amdgpu/). The first release is designated as AMD GPU Driver version 30.10. See the [User and kernel-space support matrix](https://rocm.docs.amd.com/projects/install-on-linux-internal/en/latest/reference/user-kernel-space-compat-matrix.html) for more information.
|
||||
|
||||
[AMD SMI](https://github.com/ROCm/amdsmi) continues to stay with the ROCm software stack under the ROCm organization repository.
|
||||
|
||||
@@ -139,7 +143,7 @@ Other ROCm libraries are also in the process of migration along with ROCm tools
|
||||
|
||||
### HIP API compatibility improvements
|
||||
|
||||
To improve code portability between AMD ROCm and other programming models, HIP API has been updated in ROCm 7.0 to simplify cross-platform programming. These changes are incompatible with prior ROCm releases and might require recompiling existing HIP applications for use with ROCm 7.0. For more information, see the [HIP API 7.0 changes](https://rocm.docs.amd.com/projects/HIP/en/docs-develop/hip-7-changes.html) and the [HIP changelog](#hip-7-0-0) below.
|
||||
To improve code portability between AMD ROCm and other programming models, HIP API has been updated in ROCm 7.0.0 to simplify cross-platform programming. These changes are incompatible with prior ROCm releases and might require recompiling existing HIP applications for use with ROCm 7.0.0. For more information, see the [HIP API 7.0.0 changes](https://rocm.docs.amd.com/projects/HIP/en/docs-develop/hip-7-changes.html) and the [HIP changelog](#hip-7-0-0) below.
|
||||
|
||||
### HIP runtime updates
|
||||
|
||||
@@ -160,7 +164,7 @@ Additionally, the HIP runtime includes functional improvements, which improve fu
|
||||
|
||||
### Compiler changes and improvements
|
||||
|
||||
ROCm 7.0 introduces the AMD Next-Gen Fortran compiler. ``llvm-flang`` (sometimes called ``new-flang`` or ``flang-18``) is a re-implementation of the Fortran frontend. It is a strategic replacement for ``classic-flang`` and is developed in LLVM’s upstream repo at [llvm/llvm-project](https://github.com/llvm/llvm-project/tree/main/flang).
|
||||
ROCm 7.0.0 introduces the AMD Next-Gen Fortran compiler. ``llvm-flang`` (sometimes called ``new-flang`` or ``flang-18``) is a re-implementation of the Fortran frontend. It is a strategic replacement for ``classic-flang`` and is developed in LLVM’s upstream repo at [llvm/llvm-project](https://github.com/llvm/llvm-project/tree/main/flang).
|
||||
|
||||
Key compiler enhancements include:
|
||||
|
||||
@@ -194,7 +198,7 @@ Key compiler enhancements include:
|
||||
|
||||
#### New data type support
|
||||
|
||||
MX-compliant data types bring microscaling support to ROCm. For more information, see the [OCP Microscaling (MX) Formats Specification](https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf). ROCm 7.0 enables functional support for MX data types `FP4`, `FP6`, and `FP8` on AMD Instinct MI350 series GPUs in these ROCm libraries:
|
||||
MX-compliant data types bring microscaling support to ROCm. For more information, see the [OCP Microscaling (MX) Formats Specification](https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf). ROCm 7.0.0 enables functional support for MX data types `FP4`, `FP6`, and `FP8` on AMD Instinct MI350 series GPUs in these ROCm libraries:
|
||||
|
||||
* Composable Kernel (`FP4`, `FP6`, and `FP8` only)
|
||||
* hipBLASLt
|
||||
@@ -241,7 +245,7 @@ have been refined for improved usability. See the [AMD SMI changelog](#amd-smi-2
|
||||
|
||||
#### ROCgdb
|
||||
|
||||
The micro-scaling (MX) data types now support `FP4`, `FP6`, and `FP8`.
|
||||
ROCgdb now supports `FP4`, `FP6`, and `FP8` micro-scaling (MX) data types with AMD Instinct MI350 series GPUs.
|
||||
|
||||
See the [ROCgdb changelog](#rocgdb-16-3) for more details.
|
||||
|
||||
@@ -262,7 +266,7 @@ See the [ROCm Compute Profiler changelog](#rocm-compute-profiler-3-2-3) for more
|
||||
|
||||
#### ROCm Data Center (RDC) improvements
|
||||
|
||||
The ROCm Data Center tool (RDC) streamlines the administration of AMD GPUs in cluster data center environments. ROCm 7.0 introduces new data center management and monitoring tools for system administrators. For more information, see [ROCm Data Center (RDC) tool documentation](https://rocm.docs.amd.com/projects/rdc/en/latest/index.html).
|
||||
The ROCm Data Center tool (RDC) streamlines the administration of AMD GPUs in cluster data center environments. ROCm 7.0.0 introduces new data center management and monitoring tools for system administrators. For more information, see [ROCm Data Center (RDC) tool documentation](https://rocm.docs.amd.com/projects/rdc/en/latest/index.html).
|
||||
|
||||
#### ROCm Systems Profiler
|
||||
|
||||
@@ -275,7 +279,7 @@ ROCm Systems Profiler includes the following key changes:
|
||||
See the [ROCm Systems Profiler changelog](#rocm-systems-profiler-1-1-0) for more details.
|
||||
|
||||
#### ROCm Validation Suite
|
||||
In ROCm 7.0, ROCm Validation Suite includes support for the AMD Instinct MI355X and MI350X GPUs in the IET (Integrated Execution Test), GST (GPU Stress Test), and Babel (memory bandwidth test) modules.
|
||||
In ROCm 7.0.0, ROCm Validation Suite includes support for the AMD Instinct MI355X and MI350X GPUs in the IET (Integrated Execution Test), GST (GPU Stress Test), and Babel (memory bandwidth test) modules.
|
||||
|
||||
See the [ROCm Validation Suite changelog](#rocm-validation-suite-1-2-0) for more details.
|
||||
|
||||
@@ -283,7 +287,7 @@ See the [ROCm Validation Suite changelog](#rocm-validation-suite-1-2-0) for more
|
||||
|
||||
##### Core SDK enhancements
|
||||
|
||||
* ROCprofiler-SDK is now compatible with the HIP 7.0 API.
|
||||
* ROCprofiler-SDK is now compatible with the HIP 7.0.0 API.
|
||||
* ROCprofiler-SDK adds support for AMD Instinct MI350X and MI355X GPUs.
|
||||
* The stochastic and host-trap PC sampling support has been added for all AMD Instinct MI300 and MI350 series GPUs, which
|
||||
provides information particularly useful for understanding stalls during kernel execution.
|
||||
@@ -410,9 +414,9 @@ ROCm documentation continues to be updated to provide clearer and more comprehen
|
||||
|
||||
## User space, driver, and firmware dependent changes
|
||||
|
||||
GPU Software for AMD datacenter GPU products requires you to maintain a hardware and software stack with interdependencies between the GPU and baseboard firmware, AMD GPU drivers, and the ROCm user space software. Starting ROCm 7.0 release, we are publicly documenting these interdependencies. Note that while AMD publishes drivers and ROCm user space, your server or infrastructure provider publishes the GPU and baseboard firmware by bundling AMD’s firmware releases via AMD's Platform Level Data Model (PLDM) bundle (Firmware), which includes Integrated Firmware Image (IFWI).
|
||||
GPU Software for AMD datacenter GPU products requires you to maintain a hardware and software stack with interdependencies between the GPU and baseboard firmware, AMD GPU drivers, and the ROCm user space software. Starting ROCm 7.0.0 release, we are publicly documenting these interdependencies. Note that while AMD publishes drivers and ROCm user space, your server or infrastructure provider publishes the GPU and baseboard firmware by bundling AMD’s firmware releases via AMD's Platform Level Data Model (PLDM) bundle (Firmware), which includes Integrated Firmware Image (IFWI).
|
||||
|
||||
The GPU and baseboard firmware releases numbering may vary by GPU family. Note that, ROCm 7.0 release is the first release where the AMD GPU driver is versioned independently of ROCm.
|
||||
The GPU and baseboard firmware releases numbering may vary by GPU family. Note that, ROCm 7.0.0 release is the first release where the AMD GPU driver is versioned independently of ROCm.
|
||||
|
||||
<div class="pst-scrollable-table-container">
|
||||
<table class="table" align="left" valign="middle">
|
||||
@@ -510,23 +514,23 @@ The GPU and baseboard firmware releases numbering may vary by GPU family. Note t
|
||||
|
||||
### New feature details
|
||||
|
||||
#### AMD SMI changes dependent on PLDM bundles
|
||||
#### AMD SMI changes dependent on PLDM bundles (firmware)
|
||||
|
||||
New APIs introduced in AMD SMI for ROCm 7.0 provide additional data for the AMD Instinct products. To support these features, the following firmware for each GPUs are required:
|
||||
New APIs introduced in AMD SMI for ROCm 7.0.0 provide additional data for the AMD Instinct products. To support these features, the following firmware for each GPUs are required:
|
||||
|
||||
* AMD Instinct MI355x - PLDM bundle 01.25.13.04
|
||||
* AMD Instinct MI355X - PLDM bundle 01.25.13.04
|
||||
|
||||
* AMD Instinct MI350x - PLDM bundle 01.25.13.04
|
||||
* AMD Instinct MI350X - PLDM bundle 01.25.13.04
|
||||
|
||||
* AMD Instinct MI325x - PLDM bundle 01.25.04.00
|
||||
* AMD Instinct MI325X - PLDM bundle 01.25.04.00
|
||||
|
||||
* AMD Instinct MI300x - PLDM bundle 01.25.03.12
|
||||
* AMD Instinct MI300X - PLDM bundle 01.25.03.12
|
||||
|
||||
If ROCm 7.0 is applied on system with prior version of PLDM bundles (firmware), the new APIs will return `N/A` to indicate lack of support for these items.
|
||||
If ROCm 7.0.0 is applied on system with prior version of PLDM bundles (firmware), the new APIs will return `N/A` to indicate lack of support for these items.
|
||||
|
||||
#### Enhanced temperature telemetry introduced in AMD SMI for MI355X and MI350X GPUs
|
||||
|
||||
AMD SMI in ROCm 7.0 provides support for enhanced temperature metrics and temperature anomaly detection for AMD Instinct MI350X and MI355X GPUs when paired with: AMD Instinct MI355x/MI350X - PLDM bundle 01.25.13.04.
|
||||
AMD SMI in ROCm 7.0.0 provides support for enhanced temperature metrics and temperature anomaly detection for AMD Instinct MI350X and MI355X GPUs when paired with: PLDM bundle 01.25.13.04.
|
||||
|
||||
For more information on these features, see [AMD SMI changelog](https://github.com/ROCm/amdsmi/blob/release/rocm-rel-7.0/CHANGELOG.md).
|
||||
|
||||
@@ -536,7 +540,7 @@ KVM SR-IOV support for all Instinct GPUs require the open source AMD GPU Virtual
|
||||
|
||||
#### GPU partitioning support for AMD Instinct MI355X and MI350X GPUs
|
||||
|
||||
NPS2 and DPX partitioning on bare metal is enabled on AMD Instinct MI355X and MI350X GPUs on ROCm 7.0 when paired with: AMD Instinct MI355x/MI350X - PLDM bundle 01.25.13.04.
|
||||
NPS2 and DPX partitioning on bare metal is enabled on AMD Instinct MI355X and MI350X GPUs on ROCm 7.0.0 when paired with: PLDM bundle 01.25.13.04.
|
||||
|
||||
## ROCm components
|
||||
|
||||
@@ -2566,6 +2570,10 @@ Starting with GCC 5.1, GNU `libstdc++` introduced a dual Application Binary Inte
|
||||
|
||||
Default batched General Matrix Multiplications (GEMM) operations for rocBLAS and hipBLAS on gfx1200 and gfx1201 may have a decline in performance in comparison with non-batched and strided_batched GEMM operations. By default, the batched GEMM uses hipBLASLT kernels, and switching to the Tensile kernel resolves the performance decline issue. The issue will be fixed in a future ROCm release. As a workaround, you can set the environment variable `ROCBLAS_USE_HIPBLASLT=0` before the batched GEMM operation is performed on gfx1200 and gfx1201. After completing the batched operation, reset the variable to `ROCBLAS_USE_HIPBLASLT=1` before calling non-batched or strided_batched operations.
|
||||
|
||||
### Failure to declare out-of-bound CPERs for bad memory page
|
||||
|
||||
Exceeding of bad memory page threshold fails to declare Out-Of-Band Common Platform Error Records (CPERs). This issue affects all MI300 series GPUs, including MI300X, MI325, MI350X, and MI355X. This issue will be fixed in a future AMD GPU Driver releases.
|
||||
|
||||
## ROCm resolved issues
|
||||
|
||||
The following are previously known issues resolved in this release. For resolved issues related to
|
||||
|
||||
@@ -31,7 +31,7 @@ ROCm Version,7.0.0,6.4.3,6.4.2,6.4.1,6.4.0,6.3.3,6.3.2,6.3.1,6.3.0,6.2.4,6.2.2,6
|
||||
,,,,,,,,,,,,,,,,,,,
|
||||
FRAMEWORK SUPPORT,.. _framework-support-compatibility-matrix-past-60:,,,,,,,,,,,,,,,,,,
|
||||
:doc:`PyTorch <../compatibility/ml-compatibility/pytorch-compatibility>`,"2.7, 2.6, 2.5, 2.4, 2.3","2.6, 2.5, 2.4, 2.3","2.6, 2.5, 2.4, 2.3","2.6, 2.5, 2.4, 2.3","2.6, 2.5, 2.4, 2.3","2.4, 2.3, 2.2, 1.13","2.4, 2.3, 2.2, 1.13","2.4, 2.3, 2.2, 1.13","2.4, 2.3, 2.2, 2.1, 2.0, 1.13","2.3, 2.2, 2.1, 2.0, 1.13","2.3, 2.2, 2.1, 2.0, 1.13","2.3, 2.2, 2.1, 2.0, 1.13","2.3, 2.2, 2.1, 2.0, 1.13","2.1, 2.0, 1.13","2.1, 2.0, 1.13","2.1, 2.0, 1.13","2.1, 2.0, 1.13","2.1, 2.0, 1.13","2.1, 2.0, 1.13"
|
||||
:doc:`TensorFlow <../compatibility/ml-compatibility/tensorflow-compatibility>`,"2.19.1, 2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.14.0, 2.13.1, 2.12.1","2.14.0, 2.13.1, 2.12.1"
|
||||
:doc:`TensorFlow <../compatibility/ml-compatibility/tensorflow-compatibility>`,"2.19.1, 2.18.1","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.17.0, 2.16.2, 2.15.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.16.1, 2.15.1, 2.14.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.15.0, 2.14.0, 2.13.1","2.14.0, 2.13.1, 2.12.1","2.14.0, 2.13.1, 2.12.1"
|
||||
:doc:`JAX <../compatibility/ml-compatibility/jax-compatibility>`,0.6.0,0.4.35,0.4.35,0.4.35,0.4.35,0.4.31,0.4.31,0.4.31,0.4.31,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26,0.4.26
|
||||
:doc:`verl <../compatibility/ml-compatibility/verl-compatibility>` [#verl_compat]_,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,0.3.0.post0,N/A,N/A,N/A,N/A,N/A,
|
||||
:doc:`Stanford Megatron-LM <../compatibility/ml-compatibility/stanford-megatron-lm-compatibility>`,N/A,N/A,N/A,N/A,N/A,85f95ae,85f95ae,85f95ae,85f95ae,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,N/A,
|
||||
|
||||
|
@@ -56,7 +56,7 @@ compatibility and system requirements.
|
||||
,,,
|
||||
FRAMEWORK SUPPORT,.. _framework-support-compatibility-matrix:,,
|
||||
:doc:`PyTorch <../compatibility/ml-compatibility/pytorch-compatibility>`,"2.7, 2.6, 2.5, 2.4, 2.3","2.6, 2.5, 2.4, 2.3","2.4, 2.3, 2.2, 2.1, 2.0, 1.13"
|
||||
:doc:`TensorFlow <../compatibility/ml-compatibility/tensorflow-compatibility>`,"2.19.1, 2.18.1, 2.17.1, 2.16.2","2.18.1, 2.17.1, 2.16.2","2.17.0, 2.16.2, 2.15.1"
|
||||
:doc:`TensorFlow <../compatibility/ml-compatibility/tensorflow-compatibility>`,"2.19.1, 2.18.1","2.18.1, 2.17.1, 2.16.2","2.17.0, 2.16.2, 2.15.1"
|
||||
:doc:`JAX <../compatibility/ml-compatibility/jax-compatibility>`,0.6.0,0.4.35,0.4.31
|
||||
:doc:`verl <../compatibility/ml-compatibility/verl-compatibility>` [#verl_compat]_,N/A,N/A,N/A
|
||||
:doc:`Stanford Megatron-LM <../compatibility/ml-compatibility/stanford-megatron-lm-compatibility>`,N/A,N/A,85f95ae
|
||||
@@ -164,7 +164,7 @@ compatibility and system requirements.
|
||||
.. [#az-mi300x] Starting ROCm 6.4.0, Azure Linux 3.0 is supported only on AMD Instinct MI300X and AMD Radeon PRO V710.
|
||||
.. [#RDNA-OS] Radeon AI PRO R9700, Radeon RX 9070 XT (gfx1201), Radeon RX 9060 XT (gfx1200), Radeon PRO W7700 (gfx1101), and Radeon RX 7800 XT (gfx1101) are supported only on Ubuntu 24.04.2, Ubuntu 22.04.5, RHEL 9.6, and RHEL 9.4.
|
||||
.. [#7700XT-OS] Radeon RX 7700 XT (gfx1101) is supported only on Ubuntu 24.04.2 and RHEL 9.6.
|
||||
.. [#kfd_support] As of ROCm 6.4.0, forward and backward compatibility between the AMD Kernel-mode GPU Driver (KMD) and its user space software is provided up to a year apart. For earlier ROCm releases, the compatibility is provided for +/- 2 releases. The tested user space versions on this page were accurate as of the time of initial ROCm release. For the most up-to-date information, see the latest version of this information at `User and kernel-space support matrix <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/user-kernel-space-compat-matrix.html>`_.
|
||||
.. [#kfd_support] As of ROCm 6.4.0, forward and backward compatibility between the AMD Kernel-mode GPU Driver (KMD) and its user space software is provided up to a year apart. For earlier ROCm releases, the compatibility is provided for +/- 2 releases. The supported user space versions on this page were accurate as of the time of initial ROCm release. For the most up-to-date information, see the latest version of this information at `User and kernel-space support matrix <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/user-kernel-space-compat-matrix.html>`_.
|
||||
.. [#ROCT-rocr] Starting from ROCm 6.3.0, the ROCT Thunk Interface is included as part of the ROCr runtime package.
|
||||
|
||||
|
||||
@@ -255,6 +255,6 @@ Expand for full historical view of:
|
||||
.. [#verl_compat] verl is only supported on ROCm 6.2.0.
|
||||
.. [#dgl_compat] DGL is only supported on ROCm 6.4.0.
|
||||
.. [#taichi_compat] Taichi is only supported on ROCm 6.3.2.
|
||||
.. [#kfd_support-past-60] As of ROCm 6.4.0, forward and backward compatibility between the AMD Kernel-mode GPU Driver (KMD) and its user space software is provided up to a year apart. For earlier ROCm releases, the compatibility is provided for +/- 2 releases. The tested user space versions on this page were accurate as of the time of initial ROCm release. For the most up-to-date information, see the latest version of this information at `User and kernel-space support matrix <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/user-kernel-space-compat-matrix.html>`_.
|
||||
.. [#kfd_support-past-60] As of ROCm 6.4.0, forward and backward compatibility between the AMD Kernel-mode GPU Driver (KMD) and its user space software is provided up to a year apart. For earlier ROCm releases, the compatibility is provided for +/- 2 releases. The supported user space versions on this page were accurate as of the time of initial ROCm release. For the most up-to-date information, see the latest version of this information at `User and kernel-space support matrix <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/user-kernel-space-compat-matrix.html>`_.
|
||||
.. [#ROCT-rocr-past-60] Starting from ROCm 6.3.0, the ROCT Thunk Interface is included as part of the ROCr runtime package.
|
||||
|
||||
|
||||
Reference in New Issue
Block a user