xrt-smi: support platform device identification and aie2ps validate archive by wendyliang25 · Pull Request #10093 · Xilinx/XRT

wendyliang25 · 2026-10-02T22:39:56Z

Problem solved by the commit

Platform (non-PCI) devices such as xc2ve3858 (Versal Edge / aie2ps / npu12) lacked device identification in xrt-smi and failed during examine and validate. Specifically:

xrt-smi examine attempted to query rom_vbnv and pcie_id, which are not supported on platform devices, resulting in missing or fallback device names and unrecognized hardware types.
xrt-smi validate failed to resolve the validate runner archive path and hardware configuration for aie2ps platform devices.

Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered

Discovered when running xrt-smi examine and xrt-smi validate on platform SoC targets (e.g., xc2ve3858).

How problem was solved, alternative solutions (if any) and why they were rejected

Added device_id_str query (query_requests.h):
- Added xq::device_id_str to allow shims to report a human-readable part name (e.g. "xc2ve3858") sourced from device-tree compatible strings.
- Introduced xrt_core::platform::get_device_name() in info_platform.{h,cpp} to inspect device_id_str first, falling back to rom_vbnv and numeric PCIe device ID so existing devices remain unaffected.
- Updated XBUtilities.cpp to use get_device_name().
Platform device identification in smi_hardware_config (smi.{h,cpp}, smi_ryzen.cpp):
- Added overloads for get_hardware_type() and get_family() taking std::string devid_str and const xrt_core::device* dev.
- Checks device_id_str before falling back to pcie_id, mapping "xc2ve3858" to hardware_type::aie2ps / hardware_family::aie2ps.
- Configured get_validate_archive_path() to resolve hardware_family::aie2ps to Runner/xrt_smi_ve2.a.
- Updated smi_ryzen.cpp and XBUtilities.cpp to call get_hardware_type(device) instead of querying pcie_id directly.

Risks (if any) associated the changes in the commit

Low risk. Standard PCIe and Ryzen devices continue to fall back to rom_vbnv and pcie_id without behavioral changes.

What has been tested and how, request additional testing if necessary

Verified xrt-smi examine on xc2ve3858: device name displays as xc2ve3858 and architecture version displays as aie2ps.
Verified xrt-smi validate locates and unpacks the Runner/xrt_smi_ve2.a test suite archive.

Documentation impact (if any)

None

github-actions

clang-tidy made some suggestions

Copilot

Copilot review overview

🟡 Changes recommended

Advertised AIE2PS validation tests still directly query unsupported PCIe identifiers and can fail during default validation.

Review effort: Balanced
Findings: 1

Open (1)

Exclude unsupported pcie_id runners from the aie2ps validation suite · New

What changed in this PR

Adds platform-device identification and AIE2PS validation archive routing to xrt-smi.

Changes:

Adds string-based device identification with PCIe/VBNV fallbacks.
Maps AIE2PS hardware to the VE2 validation archive and Strix configuration.
Updates device listings and reports to use platform-aware identification.

File	Description
`XBUtilities.cpp`	Uses platform-aware names and hardware detection.
`smi.h`	Adds string/device hardware lookup APIs.
`smi.cpp`	Implements AIE2PS detection and archive routing.
`smi_ryzen.cpp`	Selects Strix configuration for AIE2PS.
`query_requests.h`	Defines the device part-name query.
`info_platform.h`	Declares platform-aware device naming.
`info_platform.cpp`	Implements device-name fallback behavior.

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

github-actions · 2026-10-03T00:57:39Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-10-03T02:22:50Z

clang-tidy review says "All clean, LGTM! 👍"

aktondak · 2026-10-06T16:21:06Z

  };
 }

+config_gen_aie2ps::


Is it captured somewhere in documentation, what all commands are applicable to aie2ps family of devices? Confluence?

Hi @aktondak , it is not in a dedicated page yet. The applicable command surface is defined in code in config_gen_aie2ps (smi_ryzen.cpp)

examine reports: aie-partitions/host/platform/telemetry/preemption/clocks/debug;

validate tests: latency/throughput/runlist-latency/runlist-throughput;

Copilot

Copilot review overview

🟡 Changes recommended

Validation still performs an unsupported direct rom_vbnv query before executing the AIE2PS test archive.

Review effort: Balanced
Findings: 1 · 1

Open (2)

Route validation platform lookup through the fallback · New
Update predicate documentation to include AIE2PS · New

Resolved since last review (1)

Exclude unsupported pcie_id runners from the aie2ps validation suite

github-actions · 2026-10-06T23:00:51Z

clang-tidy review says "All clean, LGTM! 👍"

github-actions · 2026-10-06T23:51:23Z

clang-tidy review says "All clean, LGTM! 👍"

Add a device_id_str query so shims can report a human-readable part number (e.g. xc2ve3858). xrt-smi examine now shows this string as the device Name, falling back to the ROM VBNV and then the numeric device id when the query is empty, so existing devices are unaffected. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Wendy Liang <wendy.liang@amd.com>

Support device identification for platform parts (e.g., xc2ve3858) by querying device_id_str before falling back to pcie_id in smi_hardware_config, mapping them to the aie2ps hardware family. Use the aie2ps family to resolve the validate runner archive to Runner/xrt_smi_ve2.a. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Wendy Liang <wendy.liang@amd.com>

Update validation test runners and diagnostic tools to determine hardware type via get_hardware_type(device) instead of querying pcie_id directly, preventing errors on platform devices that lack PCIe IDs. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Wendy Liang <wendy.liang@amd.com>

Provide a dedicated config_gen_aie2ps configuration generator that advertises only the validation tests supported on aie2ps devices (latency and throughput, with runlist tests hidden), avoiding failures from tests like GEMM which are not packaged in xrt_smi_ve2.a. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Wendy Liang <wendy.liang@amd.com>

Implement device hardware type resolution locally in XBUtilities to avoid requiring new symbol exports or shim-side implementations. Validation test runners and diagnostic tools now determine hardware type through XBUtilities::get_hardware_type(device), checking device_id_str first for platform parts before querying pcie_id, ensuring backward compatibility with existing shims. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> Signed-off-by: Wendy Liang <wendy.liang@amd.com>

github-actions · 2026-10-08T01:43:18Z

clang-tidy review says "All clean, LGTM! 👍"

* Remove validate tests from PFs (#10040) * [XRT-SMI] Cleanup logic into specific files and prune VF specific options (#10047) * update platform report (#10052) * CR-1272937: xrt::kernel::group_id() throw on missing memory connectivity (#10028) When a kernel argument has no memory bank connectivity, group_id() returned 65535 (no_memidx encoded as 0xFFFF). Passing that value to xrt::bo() caused a cryptic std::out_of_range crash instead of an actionable error. Validate the argument index and memory group assignment in group_id() and throw xrt_core::error(EINVAL) with a descriptive message. connectivity and no_memidx are public so group_id() can use them. * CR-1272937: throw from group_id() only for buffer arguments (#10059) group_id() must still throw when a global/constant argument has no memory connectivity so xrt::bo is not given bank 0xffff. Scalar, stream, and local arguments are queried by field applications and must keep the legacy encoded invalid group. * AIESW-41600 Remove skipped validate tests for npu3 (#10062) * update list of commands for vf (#10075) * xrt-smi: gate validate hidden options, align --pmode help, report each --loop iteration (#10076) Require --advanced to parse validate hidden options (matching examine), and use the same --pmode mode list as configure for Ryzen/NPU3. Run the validate --loop iterations from the validate subcommand instead of inside TestRunner::startTest, so each iteration's result is printed and recorded in the JSON output instead of only the last one. A warning in a later iteration no longer masks an earlier failure. * Support displaying soft realtime in aie-partitions report (#10085) * [XRT-SMI] Add warning for enabled categories after disable (#10081) * Use create_dtrace_handle_elf API for creating dtrace handle (#10084) * xrt-smi: support platform device identification and aie2ps validate archive (#10093) Signed-off-by: Wendy Liang <wendy.liang@amd.com> Signed-off-by: Akshay Tondak <aktondak@amd.com>

wendyliang25 requested review from AShivangi, chvamshi-xilinx, rbramand-xilinx and stsoe as code owners October 2, 2026 22:39

github-actions Bot reviewed Oct 2, 2026

View reviewed changes

Comment thread src/runtime_src/core/common/query_requests.h

wendyliang25 requested review from aktondak and a balanced review from Copilot and removed request for AShivangi, chvamshi-xilinx, rbramand-xilinx and stsoe October 2, 2026 22:48

Copilot started reviewing on behalf of wendyliang25 October 2, 2026 22:49 View session

Copilot AI reviewed Oct 2, 2026

View reviewed changes

Comment thread src/runtime_src/core/common/smi/smi_ryzen.cpp

wendyliang25 force-pushed the devid branch from 508fc02 to 1cad95c Compare October 3, 2026 00:48

wendyliang25 requested review from bisingha-xilinx and chvamshi-xilinx October 3, 2026 01:35

aktondak reviewed Oct 6, 2026

View reviewed changes

Comment thread src/runtime_src/core/common/smi/smi.cpp Outdated

aktondak reviewed Oct 6, 2026

View reviewed changes

Comment thread src/runtime_src/core/tools/common/XBUtilities.cpp Outdated

aktondak requested changes Oct 6, 2026

View reviewed changes

wendyliang25 force-pushed the devid branch from 3765ca9 to fc6b311 Compare October 6, 2026 22:51

wendyliang25 requested review from aktondak and a balanced review from Copilot October 6, 2026 22:52

Copilot started reviewing on behalf of wendyliang25 October 6, 2026 22:53 View session

Copilot AI reviewed Oct 6, 2026

View reviewed changes

Comment thread src/runtime_src/core/common/info_platform.cpp

Comment thread src/runtime_src/core/common/smi/smi.cpp

wendyliang25 force-pushed the devid branch from fc6b311 to 370c30f Compare October 6, 2026 23:42

chvamshi-xilinx approved these changes Oct 7, 2026

View reviewed changes

aktondak approved these changes Oct 7, 2026

View reviewed changes

wendyliang25 and others added 5 commits October 7, 2026 18:32

wendyliang25 force-pushed the devid branch from 370c30f to 4cd700f Compare October 8, 2026 01:33

chvamshi-xilinx merged commit 1932bdf into Xilinx:master Oct 8, 2026
26 checks passed

aktondak mentioned this pull request Oct 9, 2026

Cherry pick mcdm m1.0 xrt smi #10108

Merged

Conversation

wendyliang25 commented Oct 2, 2026

Problem solved by the commit

Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered

How problem was solved, alternative solutions (if any) and why they were rejected

Risks (if any) associated the changes in the commit

What has been tested and how, request additional testing if necessary

Documentation impact (if any)

Uh oh!

github-actions Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

github-actions Bot commented Oct 3, 2026

Uh oh!

github-actions Bot commented Oct 3, 2026

Uh oh!

Uh oh!

aktondak Oct 6, 2026

Choose a reason for hiding this comment

Uh oh!

wendyliang25 Oct 6, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

github-actions Bot commented Oct 6, 2026

Uh oh!

github-actions Bot commented Oct 6, 2026

Uh oh!

github-actions Bot commented Oct 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants