Skip to content

Cherry-pick AMDSMI commits into ROCm 10.0 - #9734

Closed
marbre wants to merge 3 commits into
release/therock-10.0from
users/marbre/therock-10.0-AIRDEL-17
Closed

Cherry-pick AMDSMI commits into ROCm 10.0#9734
marbre wants to merge 3 commits into
release/therock-10.0from
users/marbre/therock-10.0-AIRDEL-17

Conversation

@marbre

@marbre marbre commented Aug 5, 2026

Copy link
Copy Markdown
Member

Motivation

Cherry-picks amdsmi commits into the ROCm 10.0 release branch.

Technical Details

Issue Tracking

JIRA ID: AIRDEL-17

@github-actions github-actions Bot added documentation Improvements or additions to documentation project: amdsmi github actions Pull requests that update GitHub Actions code labels Aug 5, 2026
@marbre marbre changed the title Users/marbre/therock 10.0 airdel 17 Cherry-pick AMDSMI commits into ROCm 10.0 Aug 5, 2026
@marbre
marbre marked this pull request as ready for review August 5, 2026 20:32
@marbre
marbre requested review from a team August 5, 2026 20:32
@marbre
marbre requested review from a team as code owners August 5, 2026 20:32
This was referenced Aug 7, 2026
Closed
Merged
marifamd and others added 3 commits August 8, 2026 12:51
@marifamd
feat(amdsmi): reland packaging rework with CLI shebang fixes (#9340)
ae7eeba 
Relands the manylinux wheel CI job (#7913) and the Python packaging rework
(#7914) after they were reverted in #9339, together with the follow-up
amd-smi CLI shebang fixes so the rework and its corrections land as one
change.

Packaging rework (reintroduced):
- Install the amdsmi module into the system python's site-packages via
  install(DIRECTORY) instead of pip + a .pth file; drop the pip plumbing and
  python3-pip/setuptools/wheel deps from the DEBIAN/RPM scriptlets.
- Replace the ROCm-path-walking resolver in amdsmi_wrapper.py (and
  tools/generator.py, regenerated in lockstep) with a flat loader:
  AMDSMI_LIB_OVERRIDE, a bundled libamd_smi_python.so, then the system
  libamd_smi.so; a _MissingLibrary sentinel keeps import tolerant.
- Add BUILD_PYTHON_WHEEL (default OFF): build a SONAME-isolated
  libamd_smi_python.so and a wheel that disables the loader's system fallback.
- Stage the module under share/amd_smi for ROCM_PATH consumers and TheRock.
- Add test_abi_compat.py and the manylinux wheel CI job.

CLI shebang fixes:
- Use a portable "#!/usr/bin/env python3" shebang for the amd-smi CLI so it
  runs on relocatable installs with no /usr/bin/python3 (TheRock containers)
  and on wheel runners where /usr/bin/python3 lacks the amdsmi module.
- Stop pinning the CLI shebang to a fixed interpreter: remove pin_shebang.cmake
  and the AMDSMI_CLI_SHEBANG_INTERP derivation. The pinned absolute path became
  a hard RPM "Requires:" (e.g. /usr/libexec/platform-python3.6) that RHEL10 and
  SLES16 could not satisfy; the RPM now depends only on python(abi).
- Resolve the env-form shebang in the install-verify harness instead of running
  "/usr/bin/env -c".

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
@amd-juwillia @marifamd
feat(amdsmi): [ROCM-3941] Add SONAME conflict test and build harness (#…
ac51a15 
…7912)

Add tests/run_amdsmi_pkg_conflict_test.py, which asserts that two AMD SMI
packages built with different SOVERSIONs do not co-own the same
libamd_smi.so.<major>, so installing one over the other cannot silently
break the other's loader.

Wire the ABI-compat output into the build harness: summarize_results in
run_amdsmi_build.py now scans abi_compat_output.txt alongside the existing
integration/unit/perf test outputs so an ABI regression surfaces in the CI
summary.

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
@marifamd
fix(amdsmi): correctness and coverage fixes for the Python packaging …
91aea1d 
…rework (#9033)

## Description

Correctness and coverage follow-ups to the AMD SMI Python packaging
rework (relanded in #9340). This branch is the packaging-hardening layer
on top of the SONAME conflict test + build harness (#7912).

Rebased onto `develop` after #7912 and #9340 merged; the now-upstream
shebang and self-locate-loader commits were dropped, and the surviving
relocatable-loader change was reconciled with develop's version (keeps
develop's `len(parents) > 3` guard and adds a `try/except OSError`
fall-through so a present-but-unloadable relocatable library does not
shadow the system SONAME).

### Loader / wrapper
- Fall through to the system library when the relocatable TheRock
library exists but fails to load (missing deps).
- Dedupe the wrapper `_libraries` init (the ctypesgen preamble already
initializes it).
- Read and write the staged wrapper as UTF-8 in
`disable_system_fallback.py` so the wheel flag flip is deterministic
regardless of build locale; make the staging idempotent so a rebuild
that reuses the flipped copy succeeds.

### Packaging correctness
- Stop the RPM from co-owning system Python directories.
- Restrict package `preun`/`prerm` cleanup to a full erase (not
upgrades).
- Quote the install prefix when writing `ld.so.conf.d` on RPM.
- Pin permissions on the installed Python module; make the operational
log group-writable (not world-writable) and stop the logger from
restoring world-writable permissions.
- Only strip release libraries when a strip tool is available; fail the
build when the system `python3` cannot be detected.

### Wheel build
- Build one wheel per platform (not per interpreter), tag it
`manylinux_2_28`, and verify the bundled `.so`.
- Use a clean PyPI version and expose the commit as metadata.
- Reject regenerating the wrapper against the wheel library; skip the
system site-packages install for wheel builds.

### Tests
- Loader-contract tests in `test_abi_compat.py`: preferred bundled
`.so`, wheel refuses system fallback when the bundle is missing, the
relocatable OSError fall-through, and an ambiguous-wrapper hard failure.
Assert the idempotent no-op in the fallback double-run guard.
- Guard against drift between the two installed module copies; install
the dual-copy runner so its guard tests run in CI; run the wheel
install-location check without a GPU.

### Verification
- `cmake -DBUILD_TESTS=ON` + `make`: 0 errors, `libamd_smi.so.27` +
`amdsmitst` built.
- `test_abi_compat.py` (10) and `test_dual_copy_guard.py` (4): all pass.

## JIRA ID

ROCM-3941

---------

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
@marifamd
marifamd force-pushed the users/marbre/therock-10.0-AIRDEL-17 branch from 36175f9 to 91aea1d Compare August 8, 2026 17:51
@marbre

marbre commented Aug 11, 2026 •
edited
Loading

Copy link
Copy Markdown
Member Author

  • rocprofiler-compute test is failing with

    2:    INFO    |-> [rocprofiler-sdk] Finished executing kernel
2:    INFO    |-> [rocprofiler-sdk] Finished executing kernel
2:    INFO    |-> [rocprofiler-sdk] Finished executing kernel
2:    INFO    |-> [rocprofiler-sdk] Finished copying the output vector from the GPU to the CPU
2:    INFO    |-> [rocprofiler-sdk] Error at position i 1, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 2, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 3, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 4, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 5, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 6, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 7, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 8, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 9, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 10, Expected: 0.000000, Found: 0.000000
2:    INFO    |-> [rocprofiler-sdk] Error at position i 11, Expected: 0.000000, Found: 0.000000
[..]

    while this passes on the release branch in TheRock: https://github.com/ROCm/TheRock/actions/runs/31371289759/job/93605017563

  • rocprofiler-sdk test is timing out.

@marifamd

marifamd commented Aug 11, 2026

Copy link
Copy Markdown
Contributor

@marbre Is this issue related: ROCm/rocm-libraries#10056 I believe it was merged after 10.0 branch

@marbre

marbre commented Aug 11, 2026

Copy link
Copy Markdown
Member Author

@marbre Is this issue related: ROCm/rocm-libraries#10056 I believe it was merged after 10.0 branch

The error message is a complete different one and the test is passing on the release branch, so there is a chance it is related to the cherry-pick. Can't merge with that red signal.

@marbre

marbre commented Aug 12, 2026

Copy link
Copy Markdown
Member Author

@marifamd any update on this?

@marifamd

marifamd commented Aug 12, 2026

Copy link
Copy Markdown
Contributor

After reviewing with @dmitrii-galantsev there are some patches that haven't been in a nightly yet, they were merged last night. This change is too risky for 10.0 and I'm closing this out

@marifamd marifamd closed this Aug 12, 2026
marbre pushed a commit that referenced this pull request Aug 18, 2026
@marifamd @gabrpham
@amd-josnarlo @dmitrii-galantsev @lamikr @nileshnegi @cursoragent
refactor(amdsmi): backport unified ABI changes to release/therock-10.0 (
4c86383 
#9864)

## Motivation

Bring the unified ABI changes onto the `release/therock-10.0` release
branch. These already-merged `develop` commits change the amd_smi ABI
(64-bit metrics counters, driver-reload removal, unified struct sync)
and must land together so the 10.0 header, wrapper and interface stay
consistent with each other — and so 10.0 does not ship an ABI that
differs from `develop`.

The two earlier attempts are closed: #9852 stacked these changes on the
Python packaging series, and #9734 was that packaging series. With #9734
closed, the packaging rework is not coming to 10.0, so this PR applies
the ABI changes to the raw release branch.

## Technical Details

**Five commits** on the `release/therock-10.0` tip (`6ccde5dbc27`). Four
are cherry-picks of merged `develop` commits (`-x` provenance
preserved); one is a changelog retarget.

| # | Commit | Cherry-picked from | Notes |
|---|--------|--------------------|-------|
| 1 | `15432e923f` fix(amdsmi): Bump GPU metrics accumulator counters
(#8210) | `aeb9da985b` | byte-identical |
| 2 | `7ff1686439` refactor(amdsmi): remove driver reload API and
related functions (#8575) | `706209f4f5` | byte-identical |
| 3 | `23ec5edc37` refactor(amdsmi): Sync with unified with known abi
changes (#7844) | `923767a953` | 3 conflicts, resolved below |
| 4 | `9030ac095c` docs(amdsmi): set changelog version to 10.0 | — |
retargets the changelog heading from `7.15.0` to `10.0` |
| 5 | `94cec8fda6` fix(rccl): correct UALoE fabric detection and
diagnostics on the amdsmi path (#9464) | `0f674c16e8` | byte-identical —
**required**, see below |

### Why the RCCL commit is part of this PR

Commit 5 is not incidental: **#7844 does not build without it.** `#7844`
flattens `amdsmi_fabric_info_t`, deleting `amdsmi_fabric_info_ver_t` and
moving the version to a top-level `fabric_version`. RCCL on the release
branch hardcodes the old nested spelling:

```cpp
// projects/rccl/src/misc/amdsmi_wrap.cc @ 6ccde5d
fabricInfo.fabric_info.version                   // deleted by #7844
fabricInfo.fabric_info.fabric_version.v1         // deleted by #7844
```

Both identifiers cease to exist after commit 3, so RCCL fails to compile
against the new header. `#9464` replaces those accesses with a
`amdSmiFabricInfoIsFlat` SFINAE trait that compiles against **both** the
flat and nested layouts, and fixes the UALoE detection bugs described in
that PR. It is a large diff (8 files, +1140/−29, including new RCCL
tests), so it is called out explicitly here.

**These five commits must land as a unit.** Splitting the PR, or
reverting the RCCL commit on its own, breaks the RCCL build.

### Conflicts resolved in #7844

Three conflicts, all because the release branch lacks changes `#7844`
was written on top of:

- **`py-interface/CMakeLists.txt`** — kept the release version.
`#7844`'s only hunk here reworks the packaging site-packages install
block, which does not exist on this branch (it came from the packaging
series in the now-closed #9734).
- **`py-interface/amdsmi_interface.py`** — keep-both: added the
`AMDSMI_GPU_CUID_SIZE` alias (context on `develop`, which has #9273)
alongside the release branch's `_AMDSMI_STRING_LENGTH` constants.
- **`CHANGELOG.md`** — dropped the packaging-rework entry, which
describes work not on this branch. `#7844`'s ABI entries merged cleanly.

## Issue Tracking

JIRA ID: AILITOOLS-270, ROCM-1553, ROCM-26658
JIRA ID: AIDREL-13
JIRA ID: AIDREL-19

## Test Plan

- Cherry-picked #8210 → #8575 → #7844 → #9464 onto the
`release/therock-10.0` tip (`6ccde5dbc27`)
- Full amd_smi build (library + CLI + `amdsmitst`)
- Regenerated `amdsmi_wrapper.py` and diffed against the committed file
- Diffed every cherry-pick against its `develop` original
- Verified the `amdsmi_fabric_info_t` flattening is ABI-preserving, by
compiling both headers and comparing `sizeof`/`offsetof`
- Checked every in-repo consumer of the changed/removed types for
breakage

## Test Result

- **Build:** clean (library + CLI + tests, 0 errors)
- **Wrapper:** byte-identical after regeneration (0-line diff)
- **Cherry-pick fidelity:** commits 1, 2 and 5 diff byte-identically
against their `develop` originals (modulo hunk offsets); commit 3
differs only in the three conflicts documented above
- **Fabric ABI is preserved:** `sizeof(amdsmi_fabric_info_t) = 320`
before and after, with every offset unchanged — `bdf=0`, version `=8`,
`v1=12`, `v1.accel_state=252`, `reserved=256`
- **Removed symbols:** `amdsmi_gpu_driver_reload` has no remaining
references outside the CHANGELOG; none of the 16 removed/renamed symbols
has a stale consumer anywhere under `projects/`
- **In-repo consumers:**
- `projects/rdc` reads `amdsmi_gpu_metrics_t` and the widened
accumulators and is *not* modified here — a `-fsyntax-only` A/B of
`RdcMetricFetcherImpl.cc` against the old and new headers produces the
identical error set (pre-existing, ESMI-gated), so nothing regresses.
The fields only widen, and they feed a `uint64_t`-valued map.
- `projects/rocshmem` keeps a private dlopen mirror of
`amdsmi_fabric_info_t`; it does not include `amd_smi/amdsmi.h`, and the
layout is unchanged, so it is unaffected.
- `projects/rocprofiler-systems` uses `amdsmi_gpu_metrics_t` and builds
green in this PR's TheRock CI.
- `projects/rocm-smi-lib` has its own independent `rocm_smi.h`;
unaffected.

## Known issues inherited from `develop`

Faithful cherry-picks, so these come along unchanged; they are present
on `develop` today and are not introduced by this backport. Listed so
the release branch owners know what is shipping. Fixes belong on
`develop` first.

- **`gpu_metrics` accumulators lose their "unset" sentinel** (from
#8210). The counters widened to 64-bit but the amdgpu table still
reports 32, so the `0xFFFFFFFF` sentinel zero-extends and stops matching
the 64-bit sentinel consumers check. `amdsmi_get_gpu_metrics_info()`
reports `gfx_activity_acc`/`mem_activity_acc` as a literal `4294967295`
instead of `N/A` on hardware that does not report them (reproduced on
gfx1030). Unchanged on parts that do report real values.
- **`tests/python/cli/test_set.py` fails instead of skipping** (from
#7844). `cmds_need_permission` is assigned in `setUpClass` after the
`_initialized` early-return, so `TestSet` raises `AttributeError`; and
the filter that replaced the old blanket skip covers only 3 of the
privileged `set` subcommands, so an unprivileged run also fails on
`--perf-level`, `--profile`, `--perf-determinism`, `--clk-level` and
others.
- **`amd-smi set --json` and `amd-smi static --partition --json`
disagree on one key** (from #7844). `set` now emits
`accelerator_partition_mem_alloc_mode` while `static` and the
`--compute-partition-mem-alloc-mode` flag keep
`compute_partition_mem_alloc_mode`.
- **Dead constants**: the keep-both resolution leaves
`_AMDSMI_STRING_LENGTH` and `_AMDSMI_MAX_STRING_LENGTH` with zero
references (every call site moved to the public
`AMDSMI_MAX_STRING_LENGTH`).

## Submission Checklist

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

---------

Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>
Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
Signed-off-by: gabrpham <Gabriel.Pham@amd.com>
Signed-off-by: josnarlo <joseph.narlo@amd.com>
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
Co-authored-by: gabrpham <Gabriel.Pham@amd.com>
Co-authored-by: Joseph Narlo <joseph.narlo@amd.com>
Co-authored-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
Co-authored-by: Mika Laitio <mika.laitio@amd.com>
Co-authored-by: Nilesh M Negi <Nilesh.Negi@amd.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@astyrrian1 astyrrian1 astyrrian1 approved these changes
@HereThereBeDragons HereThereBeDragons HereThereBeDragons approved these changes
@marifamd marifamd marifamd approved these changes

Assignees

@marifamd marifamd

Labels

documentation Improvements or additions to documentation github actions Pull requests that update GitHub Actions code organization: ROCm project: amdsmi

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@marbre @marifamd @astyrrian1 @HereThereBeDragons @amd-juwillia