feat(amdsmi): add NPM node power limit setting and node power reading by bill-shuzhou-liu · Pull Request #11471 · ROCm/rocm-systems

bill-shuzhou-liu · 2026-09-10T20:53:37Z

Motivation

Adds the ability to set a per-node power cap for NPM (Node Power Management) devices and read back the current instantaneous node power, per the "Set npm power limit" design doc.

Technical Details

Public API (include/amd_smi/amdsmi.h, src/amd_smi/amd_smi.cc):

Add amdsmi_set_npm_limit() to set the per-node power cap
Rename node_power → current_node_power on amdsmi_npm_info_t for reading instantaneous node power

RocM SMI layer (rocm_smi/src/rocm_smi.cc, rocm_smi/src/rocm_smi_npm.cc, rocm_smi/include/rocm_smi/rocm_smi_npm.h):

Implement the sysfs-backed set/get plumbing for the node power limit and current node power

Python bindings (py-interface/amdsmi_interface.py, py-interface/amdsmi_wrapper.py):

Expose amdsmi_set_npm_limit() and the renamed current_node_power field

CLI (amdsmi_cli/amdsmi_parser.py, amdsmi_cli/amdsmi_helpers.py, amdsmi_cli/subcommands/node.py, amdsmi_cli/subcommands/set_value.py):

Add amd-smi set --node-power-limit/-n to set the cap
Update amd-smi node -p output to report current_node_power

Rust bindings (rust-interface/src/amdsmi_wrapper.rs):

Bind the new/renamed API

Docs (docs/how-to/amdsmi-cli-tool.md, docs/reference/amdsmi-py-api.md, CHANGELOG.md):

Document the new CLI flag and renamed field

Issue Tracking

JIRA ID: ROCM-26304

Test Plan

tests/amd_smi_test/unit/gpu/amdsmi_npm_limit_test.cc, rocm_smi_npm_test.cc (C++ GTest)
tests/python/unit/gpu/test_cli_set_node_power_limit.py, test_cli_current_node_power_management.py, test_npm_info_current_node_power.py, test_npm_set_limit.py (Python unit)
Built locally with BUILD_TESTS=ON, installed, and exercised amd-smi set --node-power-limit / amd-smi node -p against real hardware

Test Result

All new and existing unit tests pass; pre-commit passes on all changed files.

Submission Checklist

Look over the contributing guidelines at https://github.com/ROCm/rocm-systems/blob/develop/CONTRIBUTING.md.

therock-pr-bot · 2026-09-10T20:54:12Z

✅ All Policy Checks Passed

Check	Status	Details
📝 PR Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	✅ Pass	—
🚫 Draft PR	🔜 To Be Enabled	—
🚩 Feature Flag	🔜 To Be Enabled	—
📊 Code Coverage	🔜 To Be Enabled	—

🎉 All policy checks passed!

📖 Need help? See the Policy FAQ for details on every check and how to fix failures.

🙋 Wish to Override Policy?

bill-shuzhou-liu · 2026-09-24T13:42:36Z

Tested on MI450X and works. please refer jira ticket for screen capture.

Adds amdsmi_set_npm_limit()/amd-smi node --set-power-limit for setting the per-node power cap, and a current_node_power field on amdsmi_npm_info_t/amd-smi node -p for reading instantaneous node power (renamed from node_power per the "Set npm power limit" design doc update). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

- validate_and_set_node_power_limit() now checks amdsmi_npm_info_t::status up front and rejects the request with AmdSmiInvalidParameterValueException when NPM is disabled on the node, before evaluating the max-bound check or calling amdsmi_set_npm_limit() - Add test_npm_disabled_rejects and thread a status key through the existing fake amdsmi_get_npm_info() fixtures in test_cli_set_node_power_limit.py - Document the new rejection case in CHANGELOG.md and docs/reference/amdsmi-py-api.md Signed-off-by: Bill Liu <shuzhliu@amd.com>

oliveiradan

@bill-shuzhou-liu , I just added a few comments for you to check

- Convert the parenthesized 3.10+-only `with (...)` blocks in test_cli_set_node_power_limit.py to a form valid back to Python 3.7, and pin `target-version = "py37"` in pyproject.toml so ruff format stops re-converting it on every pre-commit run - Reject amdsmi_set_npm_limit() requests at the C-library layer (rsmi_dev_npm_limit_set()) when NPM is disabled on the node, so this is enforced for every caller rather than only amd-smi's CLI; document the new rejection case in amdsmi.h, the Python interface, docs, and CHANGELOG.md - Fix the reserved[] doxygen comment to say `reserved[5]` and mention both max_node_power_limit and current_node_power - Add SetNpmLimitRootRejectsWhenNpmDisabled GTest coverage for the new NPM-disabled rejection path Signed-off-by: Bill Liu <shuzhliu@amd.com>

Question answered and resolved.

Migrate leftover manual sys.modules save/restore (_SAVED_MODULE_NAMES + tearDownClass) to the shared stub_modules() helper, and complete the fake amdsmi.amdsmi_wrapper stub in test_cli_set_node_power_limit.py with the two status constants amdsmi_helpers.AMDSMIHelpers reads at import time (AMDSMI_STATUS_UNEXPECTED_SIZE/_UNEXPECTED_DATA). Without these, setUpClass raised before stub_modules' cleanup was registered, leaking the fake modules into sys.modules for the rest of the process and causing 66+ unrelated failures in test_cli_exit_codes.py. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

systems-assistant · 2026-10-05T04:19:58Z

TheRock Submodule Bump Activity

Newest first

2026-10-05 — Included in TheRock via ROCm/TheRock#8732.

…#12942) ## Motivation The ``AMDSMI ASAN`` workflow fails on ``develop``. Its ``ASAN unit tests`` job runs ``--gtest_filter='*Unit*'`` in a container with no amdgpu driver, and 16 ``GpuUnit`` NPM tests assert that ``amdsmi_init()`` succeeds. It returns ``AMDSMI_STATUS_DRIVER_NOT_LOADED`` there. The tests (#11471) and the workflow (#12363) each passed CI alone and broke the job once both were on ``develop``. ## Technical Details **Tests** (``tests/amd_smi_test/unit/gpu/amdsmi_npm_limit_test.cc``): - Add ``REQUIRE_AMDSMI_INIT``: skips the test when ``amdsmi_init()`` returns ``AMDSMI_STATUS_DRIVER_NOT_LOADED``, and still asserts ``AMDSMI_STATUS_SUCCESS`` for any other status - Replace the 17 ``ASSERT_EQ(init.status(), AMDSMI_STATUS_SUCCESS)`` calls with it No library, header, wrapper, CLI, or docs change, so no cascade. ## JIRA ID JIRA ID: ROCM-999999 ## Test Plan - Build ``amdsmitst`` (Debug, ``-DADDRESS_SANITIZER=ON``, ``-DBUILD_TESTS=ON``) and run ``--gtest_filter='*Unit*'`` in an ``ubuntu:24.04`` container with ``/sys/module/amdgpu`` and ``/sys/class/kfd`` masked, once with the ``develop`` version of the test file and once with this change - Run the NPM and power-info ``GpuUnit`` cases on a host with the amdgpu driver loaded ## Test Result - Driverless, ``develop`` version: 180 passed, 16 failed (the 16 ``GpuUnit`` NPM tests in the CI log) - Driverless, this change: 180 passed, 0 failed, 16 skipped - Driver loaded: 24 passed, 10 skipped, unchanged by this change - Built with ``ENABLE_ESMI_LIB=OFF``, unlike CI; the NPM tests do not use ESMI Signed-off-by: Oliveira, Daniel <daniel.oliveira@amd.com> Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

## Motivation ROCm 10.1 has branched to `release/therock-10.1`. This opens the 10.2 development cycle: it bumps the library minor version to 27.2.0 and fills the `10.2.0` changelog section. While opening it, the `10.1.0` section on `develop` was audited against the release branch, and entries that will not ship in 10.1 were moved to `10.2.0`. ## Technical Details **Version** (`include/amd_smi/amdsmi.h`, `rust-interface/src/amdsmi_wrapper.rs`): - Bump `AMDSMI_LIB_VERSION_MINOR` `1 -> 2` (27.2.0). SONAME stays `libamd_smi.so.27` (`SOVERSION` is `MAJOR` only). - Regenerate the Rust binding constant to match. The Python wrapper and Go shim carry no minor-version constant. **Changelog** (`CHANGELOG.md`): - Extend the existing `## amd_smi_lib for ROCm 10.2.0` section. - Move the `10.1.0` entries whose source PR is not on `release/therock-10.1`, and add entries for three develop-only PRs whose fixes merged without one: | Action | Source PR | Entries | |--------|-----------|---------| | Move `10.1.0` → `10.2.0` | #8624 | CLI exit-code rework: 5 Changed, 1 Removed, 2 Resolved Issues | | Move `10.1.0` → `10.2.0` | #11494 | `AMDSMI_VRAM_TYPE_HBM4` (Added) | | Move `10.1.0` → `10.2.0` | #11066 | fwupd UMA carveout (Changed) | | Move `10.1.0` → `10.2.0` | #11471 | NPM node power limit: 1 Added, 2 Changed | | Move `10.1.0` → `10.2.0` | #10678 | Version-absent `amd-smi metric` fields omitted from human output (Changed; was under Removed, but it changes output) | | Move `10.1.0` → `10.2.0` | #12144 | xGMI data counters reading as unavailable on MI450 (Resolved Issues) | | Move `10.1.0` → `10.2.0` | #12804 | NPM balancing mode get/set (Added) | | New in `10.2.0` | #11474 | Topology APIs returned uninitialized NUMA data (Resolved Issues) | | New in `10.2.0` | #11045 | `amdsmi_get_gpu_fabric_info()` rejected partial UALink sysfs data (Resolved Issues) | | New in `10.2.0` | #12999 | NULL output pointers, CPER paging after a large record, `amd-smi static --clock` traceback (3 Resolved Issues) | - Drop the `### Removed` heading left empty in `10.1.0`. - Keep the #11628 entry in `10.1.0` (cherry-picked to the release branch as #12032) and the `10.0.0` section as is (it differs from `release/therock-10.0` only by the #11046 placement fixes). **Docs** (`docs/how-to/amdsmi-cli-tool.md`): - Update the `amd-smi --help` banner and default-view header examples to 27.2.0 / ROCm 10.2.0 (stale since 27.0.0). ## Issue Tracking N/A ## Test Plan - Configure and build with `-DBUILD_TESTS=ON`; confirm CMake reports 27.2.0. - Run `pre-commit` (amdsmi config) on the changed files. - Check `10.2.0`, `10.1.0` and `10.0.0` against changelog conventions: allowed subsection headings, two-space line breaks on headline bullets, no JIRA IDs in entry text. - Diff the `10.1.0` entries and headings against `release/therock-10.1`. - No GPU run or new unit test: no runtime code changed. ## Test Result - Build: pass (`amdsmitst` links), CMake reports 27.2.0 - `pre-commit`: all hooks passed - Changelog conventions: clean for `10.2.0`, `10.1.0`, `10.0.0` - `10.1.0` vs `release/therock-10.1`: identical entries and headings --------- Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

## Motivation The ABI compliance check never flags a break. `check_report_content()` matches text that abi-compliance-checker 2.3 does not print, so every report reads as clean, and major mode passes with a warning even when the checker exits 1. It passed #8575 (removed `amdsmi_gpu_driver_reload`) and #11821 (grew `amdsmi_driver_info_t` from 768 to 1792 bytes, reverted in #13926). ## Technical Details **Report parser** (`tests/abi_check/abi_check.py`): - Read the counts from the report's section headings (`<h2>Removed Symbols <span class='failed'> 1 </span>`) - Fail on removed symbols, High or Medium severity problems, and any `Size of this type has been changed` row, naming the type - The checker rates struct growth Low because it assumes the library allocates the struct; amd-smi callers allocate their output structs, so growth breaks binaries built against the old header - Fail major mode when the checker exits non-zero; it used to finish "with warnings" and pass **Baseline** (`tests/abi_check/abi_check.py`): - Compare against `git merge-base <base> HEAD` instead of the base branch tip. Without this, once the check can fail, a PR that forked before an API landed on `develop` is reported as removing it **Tests and docs**: - `tests/abi_check/abi_check_test.py`: fixtures are fragments of real abi-compliance-checker 2.3 reports; adds merge-base and pass/fail tests - `.pre-commit-config.yaml`: run `abi_check_test.py` when `tests/abi_check/` changes, so the LSTT Formatting job runs it; no workflow ran it before - `tests/README.md`: describe both **Not changed**: `abi-compliance-check.yml`. The check stays non-blocking and still reports through the `MAJOR ABI BREAKAGE` / `MINOR ABI BREAKAGE` labels. The workflow runs the base branch's `abi_check.py`, so this PR's own ABI jobs still use the old parser. ## Issue Tracking Related: #11821, #13926 ## Test Plan - `abi_check_test.py` on Python 3.6, 3.10 and 3.14, against `develop`'s `abi_check.py` and this branch - `develop`'s and this branch's `abi_check.py` with abi-compliance-checker 2.3 on Ubuntu 22.04 (the CI job's tool and OS), major and minor mode, on real `amdsmi.h` pairs, each in a throwaway git repo so `git merge-base` and `git show` behave as in CI - `pre-commit run --from-ref origin/develop --to-ref HEAD` from `projects/amdsmi`, as the LSTT Formatting job runs it ## Test Result - `abi_check_test.py`: the new tests fail on `develop`'s `abi_check.py` (3 failures, 5 errors) and all 16 pass with this change on every Python version Real abi-compliance-checker 2.3 runs, exit code (major / minor): | `amdsmi.h` change | `develop` | This PR | This PR reports | |---|---|---|---| | #8575: removed `amdsmi_gpu_driver_reload` | 0 / 1 | **1** / 1 | Removed Symbols: 1 | | #11821: grew `amdsmi_driver_info_t` | 0 / 1 | **1** / 1 | `amdsmi_driver_info_t` 768 to 1792 bytes | | `release/therock-10.0` to `release/therock-10.1` | 0 / 1 | **1** / 1 | `amdsmi_enumeration_info_t` 276 to 280 bytes | | `release/therock-10.1` to `release/therock-10.2` | 0 / 1 | **1** / 1 | Medium data type problems; `amdsmi_driver_info_t`, `amdsmi_fabric_info_t` and its `fabric_info_` union grew | | #12804: added APIs | 0 / 0 | 0 / 0 | | | #11471: added APIs | 0 / 1 | 0 / 1 | | | `AMDSMI_LIB_VERSION_MINOR` bump | 0 / 1 | 0 / 1 | | | No change | 0 / 0 | 0 / 0 | | | PR forked before #12804, base branch then gained it | 0 / 1 | 0 / 0 | | Minor mode keeps its meaning (any change the checker reports with `-strict`); only its baseline changed. Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

bill-shuzhou-liu requested a review from a team September 10, 2026 20:53

github-actions Bot added documentation Improvements or additions to documentation project: amdsmi labels Sep 10, 2026

github-actions Bot added the MINOR ABI BREAKAGE label Sep 10, 2026

systems-assistant Bot added the organization: ROCm label Sep 10, 2026

bill-shuzhou-liu and others added 2 commits September 28, 2026 07:41

bill-shuzhou-liu force-pushed the users/shuzhliu/feature-set-npm-power-limit branch from d352a61 to e2e5d5e Compare September 28, 2026 13:01