Description:
- Added a new API `amdsmi_gpu_driver_reload()` to reload the AMD GPU driver independently.
- Updated CLI (`sudo amd-smi reset -r`) and Python bindings to support driver reload functionality.
- Removed automatic driver reload from `amdsmi_set_gpu_memory_partition()` and `amdsmi_set_gpu_memory_partition_mode()`.
- Enhanced CLI and test cases to allow users to control when the driver reload occurs.
- Updated documentation and changelog to reflect the new driver reload process.
- Improved error handling and logging for driver reload operations.
- Added progress bar and user confirmation prompts for driver reload commands.
* Update build/test strategy to only allow one test execution at a time
* Modify API verbage + modify systemctl error output
- Systemctl is typically not enabled on docker.
- And is an edge case for gpu being active process/etc for display devices.
* Remove AMDSMI_STATUS_AMDGPU_RESTART_ERR from the return values
* Move driver reload to after we save original compute partitions
---------
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
Description:
- Added a new API `amdsmi_gpu_driver_reload()` to reload the AMD GPU driver independently.
- Updated CLI (`sudo amd-smi reset -r`) and Python bindings to support driver reload functionality.
- Removed automatic driver reload from `amdsmi_set_gpu_memory_partition()` and `amdsmi_set_gpu_memory_partition_mode()`.
- Enhanced CLI and test cases to allow users to control when the driver reload occurs.
- Updated documentation and changelog to reflect the new driver reload process.
- Improved error handling and logging for driver reload operations.
- Added progress bar and user confirmation prompts for driver reload commands.
* Update build/test strategy to only allow one test execution at a time
* Modify API verbage + modify systemctl error output
- Systemctl is typically not enabled on docker.
- And is an edge case for gpu being active process/etc for display devices.
* Remove AMDSMI_STATUS_AMDGPU_RESTART_ERR from the return values
* Move driver reload to after we save original compute partitions
---------
Signed-off-by: Charis Poag <Charis.Poag@amd.com>
[ROCm/amdsmi commit: d24dc7ef89]
Add support to Query UBB/OAM temperature.
* Updated Python API with new temperature metrics enum
---------
Co-authored-by: Bill Liu <shuzhliu@amd.com>
Co-authored-by: gabrpham_amdeng <Gabriel.Pham@amd.com>
Add support to Query UBB/OAM temperature.
* Updated Python API with new temperature metrics enum
---------
Co-authored-by: Bill Liu <shuzhliu@amd.com>
Co-authored-by: gabrpham_amdeng <Gabriel.Pham@amd.com>
[ROCm/amdsmi commit: abd3c02a3c]
Problem with original test:
- Created circular dependencies between queues:
* Queue1: Kernel A → Barrier(waits for signal_2) → Kernel C
* Queue2: Barrier(waits for signal_1) → Kernel B → sets signal_2
- With strict "one kernel at a time" serialization, this created deadlock:
* Queue1 executed Kernel A, then blocked on barrier waiting for signal_2
* Serializer switched to Queue2, but Queue2 was blocked waiting for signal_1
* Neither queue could proceed: Queue1 needed Queue2's Kernel B to complete,
but Queue2 couldn't start until Queue1 finished completely
- Test would hang indefinitely at hsa_signal_wait_relaxed() for signal_2
Solution implemented:
- Reordered packet submission to eliminate circular dependencies
- Ensured signal producers execute before consumers need them:
* Kernel A produces signal_1 before Queue2's barrier needs it
* Kernel B produces signal_2 before Queue1's continuation needs it
- Dependencies now flow forward without cycles, allowing serializer progress
Refactoring changes:
- Extract common functionality into helper functions:
* create_completion_signal() for signal creation
* create_queue() for queue creation
* submit_kernel_packet() for kernel dispatch packets
* submit_barrier_packet() for barrier packets
- Add comprehensive documentation explaining expected execution pattern
- Simplify main() function making the dependency flow more readable
Co-authored-by: Benjamin Welton <bewelton@amd.com>
[ROCm/rocprofiler-sdk commit: b5e1645a14]
Problem with original test:
- Created circular dependencies between queues:
* Queue1: Kernel A → Barrier(waits for signal_2) → Kernel C
* Queue2: Barrier(waits for signal_1) → Kernel B → sets signal_2
- With strict "one kernel at a time" serialization, this created deadlock:
* Queue1 executed Kernel A, then blocked on barrier waiting for signal_2
* Serializer switched to Queue2, but Queue2 was blocked waiting for signal_1
* Neither queue could proceed: Queue1 needed Queue2's Kernel B to complete,
but Queue2 couldn't start until Queue1 finished completely
- Test would hang indefinitely at hsa_signal_wait_relaxed() for signal_2
Solution implemented:
- Reordered packet submission to eliminate circular dependencies
- Ensured signal producers execute before consumers need them:
* Kernel A produces signal_1 before Queue2's barrier needs it
* Kernel B produces signal_2 before Queue1's continuation needs it
- Dependencies now flow forward without cycles, allowing serializer progress
Refactoring changes:
- Extract common functionality into helper functions:
* create_completion_signal() for signal creation
* create_queue() for queue creation
* submit_kernel_packet() for kernel dispatch packets
* submit_barrier_packet() for barrier packets
- Add comprehensive documentation explaining expected execution pattern
- Simplify main() function making the dependency flow more readable
Co-authored-by: Benjamin Welton <bewelton@amd.com>
Init and fini kernel needs to be launched when we load and unload code object. Avoid looping through all kernels within a code object just to run the init and fini kernels. Compiler currently only generates 1 init and fini kernel.
[ROCm/clr commit: cd46294b31]
Init and fini kernel needs to be launched when we load and unload code object. Avoid looping through all kernels within a code object just to run the init and fini kernels. Compiler currently only generates 1 init and fini kernel.
GabrPham
Pham, Gabriel
Maisam Arif
systems-assistant[bot]
vedithal-amd
xuchen-amd
Poag, Charis
Liu, Shuzhou (Bill)
Saeed, Oosman
srawat
Galantsev, Dmitrii
Pryor, Adam
Welton, Benjamin
Joseph Macaranas
Joseph Macaranas
Xie, Pengda
Kandula, Venkateshwar reddy
Bhardwaj, Gopesh
Baraldi, Giovanni