fix: Add Linux NPU discovery through sysfs accel devices

[the0cp]

Description

This PR adds Linux NPU discovery through sysfs accel devices

Currently, DeviceDiscovery::DiscoverDevicesForPlatform() on Linux discovers CPU and GPU devices, but NPU discovery is still missing. As a result, plugin execution providers that filter devices by OrtHardwareDeviceType_NPU do not receive any NPU hardware devices on Linux, even when the NPU is present and exposed by the kernel.

This change scans /sys/class/accel for accelN devices and creates OrtHardwareDevice entries with:

• type = OrtHardwareDeviceType_NPU
• PCI vendor_id
• PCI device_id
• accel_idx metadata
• pci_bus_id metadata when available

This enables Linux systems with NPUs exposed through the accel subsystem, such as AMD Ryzen AI / XDNA devices, to be reported through ORT device discovery and made available to plugin EP factories.

Changes

• Add Linux sysfs discovery for NPU devices under /sys/class/accel.
• Read NPU PCI vendor and device IDs from the underlying sysfs device path.
• Add NPU metadata including accel_idx and pci_bus_id.
• Include discovered NPU devices in DeviceDiscovery::DiscoverDevicesForPlatform().
• Add a kSysfsAccelPath constant for the Linux accel sysfs path.

Motivation

Linux plugin EPs that target NPUs rely on ORT passing OrtHardwareDeviceType_NPU devices into GetSupportedDevices(). Without Linux NPU discovery, those EPs cannot claim NPU devices and provider selection policies such as PREFER_NPU silently fall back to CPU.

Fixes #28660.

[the0cp]

Gentle ping. This PR has been open for about two weeks and the CI looks green from my side.

cc: @chilo-ms — would you be the right person to review or help route this change?

[Copilot]

Pull request overview

This PR extends ONNX Runtime’s Linux DeviceDiscovery::DiscoverDevicesForPlatform() to enumerate NPU devices by scanning the sysfs accel subsystem (/sys/class/accel) and emitting OrtHardwareDeviceType_NPU entries so plugin EP factories can match/claim NPUs on Linux.

Changes:

• Add sysfs scanning for accelN entries under /sys/class/accel and translate them into OrtHardwareDeviceType_NPU devices.
• Read PCI vendor_id/device_id and optionally attach pci_bus_id, plus add accel_idx metadata.
• Wire NPU enumeration into DeviceDiscovery::DiscoverDevicesForPlatform() and introduce kSysfsAccelPath.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chilo-ms]

Could you help add test for the changes?
We might need to create a fake device for the test as there is no real NPU device in the pipeline environment.
Also just wondering have you tested the change in a system that has a NPU device?

[the0cp]

Could you help add test for the changes? We might need to create a fake device for the test as there is no real NPU device in the pipeline environment. Also just wondering have you tested the change in a system that has a NPU device?

Thanks for the review!

I updated the PR to add unit tests for the NPU sysfs discovery path using a fake sysfs accel tree, following the same style as the existing PCI device discovery tests. I also parameterized the accel sysfs base path so the tests can pass a temporary directory while production discovery still uses /sys/class/accel.
I do not have access to a physical NPU system locally, so I have not tested this on real NPU hardware. My validation is based on the simulated sysfs layout matching the issue’s reported Linux exposure of the NPU.

I verified the targeted tests locally:

./build/Linux/Debug/onnxruntime_test_all --gtest_filter="*NpuDeviceDiscoveryTest*:*PciDeviceDiscoveryTest*"

12 tests from 2 test suites passed.