fix(amdsmi): [ROCM-1553] Bump GPU metrics accumulator counters by bkanango · Pull Request #8210 · ROCm/rocm-systems

bkanango · 2026-07-07T16:32:21Z

Motivation

Align GPU metrics accumulator counter types with the driver's 64-bit MetricsTable_t definition in the latest pmfw header to prevent overflow and maintain compatibility.

Technical Details

Changed below five GPU metrics accumulator counter fields from uint32_t to uint64_t to match amdgpu pmfw header:

gfx_activity_acc
mem_activity_acc
pcie_nak_sent_count_acc
pcie_nak_rcvd_count_acc
pcie_lc_perf_other_end_recovery

JIRA ID

JIRA ID : ROCM-1553

Test Plan

amd-smi metric

Test Result

Submission Checklist

Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

therock-pr-bot · 2026-07-07T16:33:48Z

❌ PR Check — Action Required

Check	Status	Details
📝 PR Title/Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	❌ Fail	Error: Source/code files changed without an accompanying unit test. Expected: add at least one test file named like `test_<name>.py` / `test_<name>.cpp` (or `<name>_test.`). Current:* code file(s) changed: `projects/amdsmi/include/amd_smi/amdsmi.h`, `projects/amdsmi/py-interface/amdsmi_interface.py`, `projects/amdsmi/py-interface/amdsmi_wrapper.py`, `projects/amdsmi/rocm_smi/include/rocm_smi/rocm_smi.h`, `projects/amdsmi/rocm_smi/include/rocm_smi/rocm_smi_dyn_gpu_metrics.h` (+2 more); no test file found
🔎 pre-commit	⏳ Pending	⏳ Still running…
🚫 Draft PR	🔜 To Be Enabled	—
🚩 Feature Flag	🔜 To Be Enabled	—
📊 Code Coverage	🔜 To Be Enabled	—

⚠️ 1 policy check(s) failed. Please address the issues above before this PR can be Reviewed.

🚫 Please fix the failed policies

❌ Unit Test

The Not ready to Review label was added to this PR. Once all policies pass, the label is removed automatically.

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

therock-pr-bot · 2026-07-07T16:33:50Z

🚫 Please fix the failed policies before requesting reviews.

The following policy checks failed:

❌ Unit Test

The Not ready to Review label has been added to this PR.
Once all policies pass, the label will be removed automatically.

bkanango · 2026-07-15T13:45:48Z

Ticket for CI failure - ROCm/TheRock#6501

amd-josnarlo

Do any of the m_gfx_activity_acc members located in rocm_smi_gpu_metrics.cc have to be bumped from uint32_t to uint64_t?

bkanango · 2026-07-27T16:06:36Z

Do any of the m_gfx_activity_acc members located in rocm_smi_gpu_metrics.cc have to be bumped from uint32_t to uint64_t?

These changes are only for dynamic metrics.

Changed below five GPU metrics accumulator counter fields from uint32_t to uint64_t to match amdgpu pmfw header: - gfx_activity_acc - mem_activity_acc - pcie_nak_sent_count_acc - pcie_nak_rcvd_count_acc - pcie_lc_perf_other_end_recovery Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

#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>

…mulators #8210 widened five `amdsmi_gpu_metrics_t` accumulators from 32-bit to 64-bit. The amdgpu metrics table they mirror still reports 32 bits, and the internal `m_*_acc` members correctly stayed `uint32_t`, so copying a table value into the public struct now zero-extends it. The table's "unset" sentinel `0xFFFFFFFF` becomes `0x00000000FFFFFFFF`, which no longer equals the 64-bit maximum that every consumer treats as "not reported": `_validate_if_max_uint(..., MaxUIntegerTypes.UINT64_T)` in `amdsmi_interface.py`, and the `numeric_limits<decltype(...)>::max()` comparisons in `rsmi_utilization_count_get()`. On a GPU that does not report these counters (verified on a W6800): before #8210 after #8210 gfx_activity_acc = N/A gfx_activity_acc = 4294967295 mem_activity_acc = N/A mem_activity_acc = 4294967295 Consumers cannot distinguish that from a real reading, and 4294967295 is not a plausible activity percentage accumulator. Add `widen_keep_sentinel<D>(value)`, which maps an all-ones source to an all-ones destination when the destination is wider, and is a plain cast otherwise. Apply it at the 27 static-table copy sites for the five widened fields and in the dynamic path's `assign_by_type`, whose variant carries the source at its real width and so hits the same defect. Same-width and narrowing conversions are unaffected, so no other field changes behavior. Verified on hardware: `gfx_activity_acc` and `mem_activity_acc` report `N/A` again, and building the same tree without the change reproduces `4294967295`. Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>

…mulators #8210 widened five `amdsmi_gpu_metrics_t` accumulators from 32-bit to 64-bit. The amdgpu metrics table they mirror still reports 32 bits, and the internal `m_*_acc` members correctly stayed `uint32_t`, so copying a table value into the public struct now zero-extends it. The table's "unset" sentinel `0xFFFFFFFF` becomes `0x00000000FFFFFFFF`, which no longer equals the 64-bit maximum that every consumer treats as "not reported": `_validate_if_max_uint(..., MaxUIntegerTypes.UINT64_T)` in `amdsmi_interface.py`, and the `numeric_limits<decltype(...)>::max()` comparisons in `rsmi_utilization_count_get()`. On a GPU that does not report these counters (verified on a W6800): before #8210 after #8210 gfx_activity_acc = N/A gfx_activity_acc = 4294967295 mem_activity_acc = N/A mem_activity_acc = 4294967295 Consumers cannot distinguish that from a real reading, and 4294967295 is not a plausible activity percentage accumulator. Add `widen_keep_sentinel<D>(value)`, which maps an all-ones source to an all-ones destination when the destination is wider, and is a plain cast otherwise. Apply it at the 27 static-table copy sites for the five widened fields and in the dynamic path's `assign_by_type`, whose variant carries the source at its real width and so hits the same defect. Same-width and narrowing conversions are unaffected, so no other field changes behavior. Verified on hardware: `gfx_activity_acc` and `mem_activity_acc` report `N/A` again, and building the same tree without the change reproduces `4294967295`. Tests: `GpuUnit.MetricsSentinel*` cover the helper. Two more feed an unset 32-bit accumulator through the static v1.4 reader and the dynamic v1.9 reader and check the public field; both fail without the call-site change. Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>

bkanango requested a review from a team July 7, 2026 16:32

bkanango force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 0ab2a18 to 1c95396 Compare July 7, 2026 16:32

github-actions Bot added the project: amdsmi label Jul 7, 2026

therock-pr-bot Bot added the Not ready to Review label Jul 7, 2026

github-actions Bot added the MINOR ABI BREAKAGE label Jul 7, 2026

systems-assistant Bot added the organization: ROCm label Jul 7, 2026

bkanango changed the title ~~fix(amdsmi): [ROCM-AMDSMI] Bump GPU metrics accumulator counters~~ fix(amdsmi): [ROCM-1553] Bump GPU metrics accumulator counters Jul 7, 2026

bkanango force-pushed the users/bkanango/PR-ROCM-metric32to64 branch 4 times, most recently from d4ab083 to 95f0b9c Compare July 13, 2026 17:54

marifamd added MAJOR ABI BREAKAGE and removed MINOR ABI BREAKAGE labels Jul 13, 2026

marifamd force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 95f0b9c to f177eff Compare July 15, 2026 21:23

github-actions Bot added MINOR ABI BREAKAGE and removed MAJOR ABI BREAKAGE labels Jul 15, 2026

bkanango force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from f177eff to 347f79c Compare July 16, 2026 20:45

amd-josnarlo requested changes Jul 16, 2026

View reviewed changes

bkanango force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 347f79c to 7699ac4 Compare July 24, 2026 21:29

marifamd requested a review from amd-josnarlo July 24, 2026 22:49

marifamd approved these changes Jul 24, 2026

View reviewed changes

marifamd force-pushed the users/bkanango/PR-ROCM-metric32to64 branch 2 times, most recently from 31354a4 to 49f7f06 Compare July 27, 2026 06:37

marifamd requested a review from a team as a code owner July 27, 2026 06:37

github-actions Bot added documentation Improvements or additions to documentation github actions Pull requests that update GitHub Actions code labels Jul 27, 2026

marifamd changed the base branch from develop to users/marifamd/bump-version-27.0.0 July 27, 2026 07:08

marifamd force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 49f7f06 to e235232 Compare July 27, 2026 07:56

amd-josnarlo force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from e235232 to 8bf64e2 Compare July 27, 2026 14:13

amd-josnarlo approved these changes Jul 27, 2026

View reviewed changes

Base automatically changed from users/marifamd/bump-version-27.0.0 to develop July 27, 2026 20:21

erman-gurses force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 8bf64e2 to 55b9056 Compare July 27, 2026 20:22

bkanango and others added 3 commits July 27, 2026 22:53

docs(amdsmi): add 7.15 changelog entry for 64-bit metrics counters

362288b

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

chore(amdsmi): regenerate Rust wrapper for 64-bit metrics counters

0c02484

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

marifamd force-pushed the users/bkanango/PR-ROCM-metric32to64 branch from 55b9056 to 0c02484 Compare July 28, 2026 03:53

gabrpham approved these changes Jul 28, 2026

View reviewed changes

marifamd merged commit aeb9da9 into develop Jul 28, 2026
65 of 70 checks passed

marifamd deleted the users/bkanango/PR-ROCM-metric32to64 branch July 28, 2026 15:20

This was referenced Aug 7, 2026

refactor(amdsmi): backport unified ABI changes to therock-10.0 #9852

Closed

refactor(amdsmi): backport unified ABI changes to release/therock-10.0 #9864

Merged

dmitrii-galantsev mentioned this pull request Aug 12, 2026

fix(amdsmi): three follow-up fixes for the unified ABI series #10091

Draft

1 task done

Navigation Menu

Repository navigation

fix(amdsmi): [ROCM-1553] Bump GPU metrics accumulator counters - #8210

fix(amdsmi): [ROCM-1553] Bump GPU metrics accumulator counters#8210
marifamd merged 3 commits into
developfrom
users/bkanango/PR-ROCM-metric32to64

bkanango commented Jul 7, 2026 •

edited by amd-josnarlo

Loading

Uh oh!

therock-pr-bot Bot commented Jul 7, 2026 •

edited

Loading

Uh oh!

therock-pr-bot Bot commented Jul 7, 2026 •

edited

Loading

Uh oh!

bkanango commented Jul 15, 2026

Uh oh!

amd-josnarlo left a comment

Uh oh!

bkanango commented Jul 27, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

bkanango commented Jul 7, 2026 • edited by amd-josnarlo Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Technical Details

JIRA ID

Test Plan

Test Result

Submission Checklist

Uh oh!

therock-pr-bot Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ PR Check — Action Required

Uh oh!

therock-pr-bot Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bkanango commented Jul 15, 2026

Uh oh!

amd-josnarlo left a comment

Choose a reason for hiding this comment

Uh oh!

bkanango commented Jul 27, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

bkanango commented Jul 7, 2026 •

edited by amd-josnarlo

Loading

therock-pr-bot Bot commented Jul 7, 2026 •

edited

Loading

therock-pr-bot Bot commented Jul 7, 2026 •

edited

Loading