test(hip-tests): detect XNACK via the pageable memory access attribute by iassiour · Pull Request #13430 · ROCm/rocm-systems

iassiour · 2026-10-14T01:32:34Z

Motivation

HipTest::isXnackOn() and five tests detect XNACK by searching gcnArchName for xnack+. GPUs whose ISA has no XNACK on/off target feature (e.g. gfx1250) never report xnack+, so these tests always skip there, even with HSA_XNACK=1.

Technical Details

isXnackOn() now returns hipDeviceAttributePageableMemoryAccess. Since fix(rocr): use runtime XNACK state for SVM_ACCESSIBLE_BY_DEFAULT query #12558 this reflects the process XNACK mode on every GPU, and it is the capability these tests actually need (GPU access to malloc/mmap memory).
The five tests that duplicated the gcnArchName check now call isXnackOn():
- hipMemCoherencyTstMProc.cc
- hipHmmOvrSubscriptionTst.cc
- hipHostRegisterStress.cc
- hipMemAdvise_old.cc
- hipGetProcAddressVmmApis.cc
On NVIDIA isXnackOn() returns false; the NVIDIA backend never fills gcnArchName, so the old check could not match there either.

Issue Tracking

JIRA ID: AIRUNTIME-2926

Test Plan

MI350X (gfx950), ROCr/CLR built from develop ae6a9a29dd. Ran all isXnackOn()-gated tests, unpatched develop vs this change, with HSA_XNACK unset and HSA_XNACK=1.

Test Result

Identical results with and without the change:

Test	`HSA_XNACK` unset	`HSA_XNACK=1`
`Unit_malloc_CoherentTst`, `Unit_malloc_CoherentTstWthAdvise`	skip	pass
`Unit_mmap_CoherentTst`, `Unit_mmap_CoherentTstWthAdvise`	skip	pass
`Unit_hipMemAdvise_TstAlignedAllocMem`	skip	pass
`Unit_hipMemAdvise_TstAlignedAllocMem_XNACK`	pass	pass
`Unit_hipGetProcAddress_VMM`	pass	pass
`Stress_HMM_OverSubscriptionTst`, `Stress_hipHostRegister_Oversubscription`	skip	not run

The two stress tests were not run with HSA_XNACK=1 because they oversubscribe host memory (~1.2× GPU memory). Their gate is the same isXnackOn() call as the tests above.

Not tested on gfx1250.

Submission Checklist

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

Copilot

🟢 Approval recommended

The centralized attribute maps directly to the runtime’s XNACK state, and all replacements preserve the existing test gating behavior.

0 open findings

What changed in this PR

Updates HIP tests to detect AMD XNACK through the pageable-memory-access attribute, supporting stall-only XNACK GPUs.

Changes:

Reimplements the shared XNACK detection helper.
Migrates affected memory, VMM, oversubscription, and coherency tests to the helper.

File	Description
`projects/hip-tests/catch/include/hip_test_common.hh`	Detects XNACK using pageable-memory support on AMD.
`projects/hip-tests/catch/unit/virtualMemoryManagement/hipGetProcAddressVmmApis.cc`	Uses shared XNACK detection for VMM validation.
`projects/hip-tests/catch/unit/memory/hipMemAdvise_old.cc`	Updates XNACK-dependent memory-advice tests.
`projects/hip-tests/catch/stress/memory/hipHostRegisterStress.cc`	Updates the oversubscription prerequisite check.
`projects/hip-tests/catch/stress/memory/hipHmmOvrSubscriptionTst.cc`	Updates the HMM stress-test prerequisite check.
`projects/hip-tests/catch/multiproc/hipMemCoherencyTstMProc.cc`	Updates coherency-test XNACK checks.

🧠 Review effort: Balanced

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

therock-pr-bot · 2026-10-14T01:38:10Z

✅ All Policy Checks Passed

Check	Status	Details
📝 PR Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	⚠️ Warning	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/hip-tests/catch/multiproc/hipMemCoherencyTstMProc.cc`, `projects/hip-tests/catch/stress/memory/hipHmmOvrSubscriptionTst.cc`, `projects/hip-tests/catch/stress/memory/hipHostRegisterStress.cc`, `projects/hip-tests/catch/unit/memory/hipMemAdvise_old.cc`, `projects/hip-tests/catch/unit/virtualMemoryManagement/hipGetProcAddressVmmApis.cc`; no test file found
🚫 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?

mangupta

🤖 AI-assisted pre-review (for @mangupta)

mangupta · 2026-10-14T11:07:58Z

🤖 AI-assisted pre-review (for @mangupta)

Re-reviewed the updated head (68dc38922ba4) — no new findings at or above the posting threshold.

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

litvaOo · 2026-10-21T08:23:50Z

ROCr test errors are a known HSA issue, overriding.

attribute (#13430) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ## Motivation `HipTest::isXnackOn()` and five tests detect XNACK by searching `gcnArchName` for `xnack+`. GPUs whose ISA has no XNACK on/off target feature (e.g. gfx1250) never report `xnack+`, so these tests always skip there, even with `HSA_XNACK=1`. ## Technical Details - `isXnackOn()` now returns `hipDeviceAttributePageableMemoryAccess`. Since #12558 this reflects the process XNACK mode on every GPU, and it is the capability these tests actually need (GPU access to `malloc`/`mmap` memory). - The five tests that duplicated the `gcnArchName` check now call `isXnackOn()`: - `hipMemCoherencyTstMProc.cc` - `hipHmmOvrSubscriptionTst.cc` - `hipHostRegisterStress.cc` - `hipMemAdvise_old.cc` - `hipGetProcAddressVmmApis.cc` - On NVIDIA `isXnackOn()` returns false; the NVIDIA backend never fills `gcnArchName`, so the old check could not match there either. ## Issue Tracking JIRA ID: AIRUNTIME-2926 ## Test Plan MI350X (gfx950), ROCr/CLR built from develop `ae6a9a29dd`. Ran all `isXnackOn()`-gated tests, unpatched develop vs this change, with `HSA_XNACK` unset and `HSA_XNACK=1`. ## Test Result Identical results with and without the change: | Test | `HSA_XNACK` unset | `HSA_XNACK=1` | |---|---|---| | `Unit_malloc_CoherentTst`, `Unit_malloc_CoherentTstWthAdvise` | skip | pass | | `Unit_mmap_CoherentTst`, `Unit_mmap_CoherentTstWthAdvise` | skip | pass | | `Unit_hipMemAdvise_TstAlignedAllocMem` | skip | pass | | `Unit_hipMemAdvise_TstAlignedAllocMem_XNACK` | pass | pass | | `Unit_hipGetProcAddress_VMM` | pass | pass | | `Stress_HMM_OverSubscriptionTst`, `Stress_hipHostRegister_Oversubscription` | skip | not run | The two stress tests were not run with `HSA_XNACK=1` because they oversubscribe host memory (~1.2× GPU memory). Their gate is the same `isXnackOn()` call as the tests above. Not tested on gfx1250. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/rocm-systems/blob/develop/CONTRIBUTING.md. [rocm-systems] ROCm/rocm-systems#13430 (commit 08c6e8f)

#13430) ## Motivation `HipTest::isXnackOn()` and five tests detect XNACK by searching `gcnArchName` for `xnack+`. GPUs whose ISA has no XNACK on/off target feature (e.g. gfx1250) never report `xnack+`, so these tests always skip there, even with `HSA_XNACK=1`. ## Technical Details - `isXnackOn()` now returns `hipDeviceAttributePageableMemoryAccess`. Since #12558 this reflects the process XNACK mode on every GPU, and it is the capability these tests actually need (GPU access to `malloc`/`mmap` memory). - The five tests that duplicated the `gcnArchName` check now call `isXnackOn()`: - `hipMemCoherencyTstMProc.cc` - `hipHmmOvrSubscriptionTst.cc` - `hipHostRegisterStress.cc` - `hipMemAdvise_old.cc` - `hipGetProcAddressVmmApis.cc` - On NVIDIA `isXnackOn()` returns false; the NVIDIA backend never fills `gcnArchName`, so the old check could not match there either. ## Issue Tracking JIRA ID: AIRUNTIME-2926 ## Test Plan MI350X (gfx950), ROCr/CLR built from develop `ae6a9a29dd`. Ran all `isXnackOn()`-gated tests, unpatched develop vs this change, with `HSA_XNACK` unset and `HSA_XNACK=1`. ## Test Result Identical results with and without the change: | Test | `HSA_XNACK` unset | `HSA_XNACK=1` | |---|---|---| | `Unit_malloc_CoherentTst`, `Unit_malloc_CoherentTstWthAdvise` | skip | pass | | `Unit_mmap_CoherentTst`, `Unit_mmap_CoherentTstWthAdvise` | skip | pass | | `Unit_hipMemAdvise_TstAlignedAllocMem` | skip | pass | | `Unit_hipMemAdvise_TstAlignedAllocMem_XNACK` | pass | pass | | `Unit_hipGetProcAddress_VMM` | pass | pass | | `Stress_HMM_OverSubscriptionTst`, `Stress_hipHostRegister_Oversubscription` | skip | not run | The two stress tests were not run with `HSA_XNACK=1` because they oversubscribe host memory (~1.2× GPU memory). Their gate is the same `isXnackOn()` call as the tests above. Not tested on gfx1250. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/rocm-systems/blob/develop/CONTRIBUTING.md. --------- Co-authored-by: Claude <noreply@anthropic.com>

systems-assistant · 2026-10-22T19:13:12Z

TheRock Submodule Bump Activity

Newest first

2026-10-22 — Included in TheRock via ROCm/TheRock#9425.

iassiour requested a balanced review from Copilot October 14, 2026 01:32

github-actions Bot added the project: hip-tests label Oct 14, 2026

Copilot started reviewing on behalf of iassiour October 14, 2026 01:33

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

systems-assistant Bot added the organization: ROCm label Oct 14, 2026

iassiour marked this pull request as ready for review October 14, 2026 07:51

iassiour requested a review from a team as a code owner October 14, 2026 07:51

iassiour force-pushed the users/iassiour/hip-tests-xnack-pageable branch from 7bf69cf to 81f253d Compare October 14, 2026 07:51

mangupta reviewed Oct 14, 2026

View reviewed changes

Comment thread projects/hip-tests/catch/unit/memory/hipMemAdvise_old.cc Resolved

Comment thread projects/hip-tests/catch/include/hip_test_common.hh Resolved

mangupta added the lrt-bot: ready label Oct 14, 2026

iassiour force-pushed the users/iassiour/hip-tests-xnack-pageable branch 2 times, most recently from dc41e31 to 91ac531 Compare October 19, 2026 21:36

iassiour and others added 2 commits October 20, 2026 17:34

test(hip-tests): detect XNACK via the pageable memory access attribute

885c9dc

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

test(hip-tests): drop redundant pageable query and document isXnackOn

00da8b4

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

iassiour force-pushed the users/iassiour/hip-tests-xnack-pageable branch from 91ac531 to 00da8b4 Compare October 20, 2026 16:34

iassiour requested a review from a team October 20, 2026 17:07

kjayapra-amd approved these changes Oct 20, 2026

View reviewed changes

litvaOo merged commit 08c6e8f into develop Oct 21, 2026
162 of 167 checks passed

litvaOo deleted the users/iassiour/hip-tests-xnack-pageable branch October 21, 2026 08:24

mangupta removed the lrt-bot: ready label Oct 21, 2026

Navigation Menu

Repository navigation

test(hip-tests): detect XNACK via the pageable memory access attribute - #13430

test(hip-tests): detect XNACK via the pageable memory access attribute#13430
litvaOo merged 2 commits into
developfrom
users/iassiour/hip-tests-xnack-pageable

iassiour commented Oct 14, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

therock-pr-bot Bot commented Oct 14, 2026

Uh oh!

mangupta left a comment

Uh oh!

Uh oh!

Uh oh!

mangupta commented Oct 14, 2026

Uh oh!

litvaOo commented Oct 21, 2026

Uh oh!

Uh oh!

systems-assistant Bot commented Oct 22, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

iassiour commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Technical Details

Issue Tracking

Test Plan

Test Result

Submission Checklist

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

therock-pr-bot Bot commented Oct 14, 2026

✅ All Policy Checks Passed

Uh oh!

mangupta left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

mangupta commented Oct 14, 2026

Uh oh!

litvaOo commented Oct 21, 2026

Uh oh!

Uh oh!

systems-assistant Bot commented Oct 22, 2026

TheRock Submodule Bump Activity

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

iassiour commented Oct 14, 2026 •

edited

Loading