[rocprofiler-sdk] Platform specific topology information by bgopesh · Pull Request #6399 · ROCm/rocm-systems

bgopesh · 2026-05-25T18:10:23Z

Motivation

This PR decouples agent (CPU/GPU) discovery from `agent.cpp` by introducing per-platform free functions under `rocprofiler::platform::{gnulinux,wsl,windows}`, preserving the existing Linux/KFD enumeration verbatim while adding WSL (DXCore/D3DKMT) and native Windows (D3DKMT) backends.

Changes:

Directory structure

```
source/lib/rocprofiler-sdk/
├── platform/
│ ├── CMakeLists.txt
│ ├── agent.hpp # dispatch header: is_available / enumerate / name
│ ├── gnulinux/
│ │ ├── CMakeLists.txt
│ │ ├── agent.hpp
│ │ └── agent.cpp # byte-identical move of Linux/KFD sysfs logic from agent.cpp
│ ├── wsl/
│ │ ├── CMakeLists.txt
│ │ ├── agent.hpp
│ │ └── agent.cpp # DXCore enumeration via D3DKMTEnumAdapters3 + RAII DxcoreHandle
│ └── windows/
│ ├── CMakeLists.txt
│ ├── agent.hpp
│ └── agent.cpp # native Windows D3DKMT skeleton (stub, follow-up)
└── agent.cpp # runtime dispatch: gnulinux -> wsl -> windows
```

The earlier `topology/` subdirectory (with abstract `TopologyProvider`, `LinuxKfdProvider`, `DxgProvider`, `D3dkmtProvider` classes) has been removed in favour of this flat, namespace-per-platform layout per CONTRIBUTING.md §8.

Key design points

No abstract base class. Each platform exposes three free functions in its own namespace: `is_available()`, `enumerate()`, `name()`. `agent.cpp` dispatches at runtime in precedence order: `gnulinux → wsl → windows`.
`ROCPROFILER_FORCE_PLATFORM={gnulinux|wsl|windows}` env var overrides the runtime selection (replaces the earlier `ROCPROFILER_FORCE_DXG` / `ROCPROFILER_FORCE_D3DKMT` vars).
Linux behaviour unchanged. The gnulinux body is a verbatim move of `read_topology()` and helpers from `agent.cpp`.
WSL body moved from the old `dxg_provider`; uses `D3DKMTEnumAdapters3` + `D3DKMTQueryAdapterInfo` (PHYSICALADAPTERDEVICEIDS, ADAPTERADDRESS, ADAPTERREGISTRYINFO, GETSEGMENTSIZE); RAII `DxcoreHandle` wraps the dlopen'd library; filters AMD adapters, packs PCI BDF into `location_id`, converts wide names to UTF-8. Fields DXCore cannot supply (gfx_target_version, CU/SDMA counts, mem_banks/caches/io_links) are left zero/empty.
Windows body is a stub skeleton; full D3DKMT enumeration is a follow-up.
`agent_internal.hpp` removed. `update_agent_runtime_visibility` is now declared in `agent.hpp` directly.

Technical Details

`platform/gnulinux/agent.cpp`: byte-identical move of `read_topology()` and helpers (`parse_cpu_info`, `read_map`, `read_property`, etc.) out of `agent.cpp`. Linux behaviour is unchanged.
`platform/wsl/agent.cpp`: WSL backend using D3DKMT/DXCore. Probes `D3DKMTEnumAdapters3` symbol; also checks `/dev/dxg` presence for `is_available()` so a box with `libdxcore.so` but no `/dev/dxg` does not silently return an empty topology.
`platform/windows/agent.cpp`: native Windows backend skeleton. All D3DKMT calls are `#ifdef _WIN32` guarded so the class still links on Linux (stubs return empty + warning). Full enumeration body is a follow-up.
`agent.cpp`: `read_topology()` and its helpers deleted. `internal_refresh_topology()` dispatches through the platform functions. No other public function changed. Unused `fs` alias and `filesystem.hpp` include removed (was flagged by clang-tidy `misc-unused-alias-decls`).

JIRA ID

Test Plan

Test Result

Submission Checklist

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

github-actions · 2026-05-25T18:55:28Z

Code Coverage Report

Tests Only

Samples Only

Tests + Samples

jrmadsen · 2026-05-26T16:12:58Z

Please avoid polymorphic abstractions. If we need to have a design discussion, please let me know. I'd prefer an approach/organization like:

agent.hpp (platform neutral interface)
agent.cpp (handles calling platform-specific implementation)
linux/
- agent.hpp (Linux interface in linux namespace)
- agent.cpp (existing agent.cpp in linux namespace)
windows/
- agent.hpp (Windows interface in windows namespace)
- agent.cpp
wsl/
- agent.hpp (WSL interface in wsl namespace)
- agent.cpp

This pattern will keep Linux/Windows/WSL code separate, work well with our current source code organization (for example, we may need a linux/registration.cpp and a windows/registration.cpp), and the namespacing will make calls explicit, e.g. rocprofiler::linux::<some-agent-function>(...).

bgopesh · 2026-05-26T17:03:00Z

Please avoid polymorphic abstractions. If we need to have a design discussion, please let me know. I'd prefer an approach/organization like:

agent.hpp (platform neutral interface)

agent.cpp (handles calling platform-specific implementation)

linux/

agent.hpp (Linux interface in linux namespace)

agent.cpp (existing agent.cpp in linux namespace)

windows/

agent.hpp (Windows interface in windows namespace)

agent.cpp

wsl/

agent.hpp (WSL interface in wsl namespace)

agent.cpp

This pattern will keep Linux/Windows/WSL code separate, work well with our current source code organization (for example, we may need a linux/registration.cpp and a windows/registration.cpp), and the namespacing will make calls explicit, e.g. rocprofiler::linux::<some-agent-function>(...).

Thanks for looking into it.
ok. makes sense. Let me redesign this with the suggested approach. I think the suggested approach would be good in long term for all platform specific differences.

bgopesh · 2026-05-26T17:22:40Z

@jrmadsen : Before I refactor this, can you please approve this structure with namespace:

Tentative layout I'd implement once these are settled (using kfd as a placeholder instead of linux since linux (and unix) are predefined as 1 by GCC unless we compile in strict mode):

source/lib/rocprofiler-sdk/
agent.hpp
agent.cpp # dispatcher: kfd:: vs wsl:: at runtime, plus update_agent_runtime_visibility
kfd/
agent.{hpp,cpp} # namespace rocprofiler::kfd (old read_topology() body, verbatim move)
CMakeLists.txt # always built on UNIX
wsl/
agent.{hpp,cpp} # namespace rocprofiler::wsl (old DxgProvider body, verbatim move)
CMakeLists.txt # always built on UNIX (WSL is Linux)
windows/
agent.{hpp,cpp} # namespace rocprofiler::windows (D3DKMT skeleton)

meserve-amd

I have not reviewed dxg_provider yet, but have some concerns about the architecture so far that can be discussed.

bgopesh · 2026-05-27T04:33:43Z

@meserve-amd : sure. Most of your concerns will go away once I remove the polymorphic topology and us ethe platform specific directory structure as Jonathan mentioned above.

bgopesh · 2026-05-27T06:57:50Z

@jrmadsen : Before I refactor this, can you please approve this structure with namespace:

Tentative layout I'd implement once these are settled (using kfd as a placeholder instead of linux since linux (and unix) are predefined as 1 by GCC unless we compile in strict mode):

source/lib/rocprofiler-sdk/ agent.hpp agent.cpp # dispatcher: kfd:: vs wsl:: at runtime, plus update_agent_runtime_visibility kfd/ agent.{hpp,cpp} # namespace rocprofiler::kfd (old read_topology() body, verbatim move) CMakeLists.txt # always built on UNIX wsl/ agent.{hpp,cpp} # namespace rocprofiler::wsl (old DxgProvider body, verbatim move) CMakeLists.txt # always built on UNIX (WSL is Linux) windows/ agent.{hpp,cpp} # namespace rocprofiler::windows (D3DKMT skeleton)

@jrmadsen This part is done.

bgopesh · 2026-05-28T13:10:37Z

@meserve-amd and @ApoKalipse-V : reminder for re-review

meserve-amd

Conditional approval on addressing the one Major comment.

Thanks for reworking the architecture here; this looks much better now.

I understand gnulinux was copied code, but there are a few nits Claude found in there when I let it loose. I think it's a good time to fix up things, but I'll leave it up to you.

bgopesh · 2026-05-29T04:44:07Z

Conditional approval on addressing the one Major comment.

Thanks for reworking the architecture here; this looks much better now.

I understand gnulinux was copied code, but there are a few nits Claude found in there when I let it loose. I think it's a good time to fix up things, but I'll leave it up to you.

Thanks for the review. Addressed the Major comment. I want to keep Linux implementation verbatim as I want to limit the scope of this PR to just focus on getting the structure right. I know changes you requested are small but still would like to address them after this is merged.

ApoKalipse-V

Code looks okay, so pre-approving, but we are missing tests.
Just make sure to add unit tests ensuring the different options return or fail accordingly to get some coverage.
e.g. WSL is compiled in linux but not covered by our CI because it doesn't run standard tests in WSL.

bgopesh · 2026-06-02T11:53:23Z

Code looks okay, so pre-approving, but we are missing tests. Just make sure to add unit tests ensuring the different options return or fail accordingly to get some coverage. e.g. WSL is compiled in linux but not covered by our CI because it doesn't run standard tests in WSL.

Added tests

Factor GPU/CPU agent discovery behind a polymorphic TopologyProvider interface so the existing Linux/KFD path is preserved verbatim while WSL and native Windows gain dedicated backends. - topology/linux_kfd_provider: byte-identical move of read_topology() and its helpers (parse_cpu_info, read_map, read_property, etc.) out of agent.cpp. Linux behaviour is unchanged. - topology/dxg_provider: WSL backend skeleton. Dlopens libdxcore.so and probes CreateDXCoreAdapterFactory; full DXCore-to-agent enumeration is left as a follow-up. - topology/d3dkmt_provider: native Windows backend skeleton. All D3DKMT calls are #ifdef _WIN32 guarded so the class still links on Linux (stubs return empty + warning). The full enumeration body is left as a follow-up. - topology/topology_provider: factory make_default_provider() with runtime selection precedence: ROCPROFILER_FORCE_DXG -> ROCPROFILER_FORCE_D3DKMT -> sysfs -> libdxcore.so loadable -> _WIN32 -> LinuxKfd fallback. The ROCPROFILER_FORCE_* env vars are new and intended for triage. - agent.cpp: read_topology() and its helpers deleted. The singleton accessor and internal_refresh_topology() now go through the provider. No other public function changed. - agent_internal.hpp: minimal forward declaration of update_agent_runtime_visibility so providers can call it without touching the public header.

Fixes probe_libdxcore() to look up D3DKMTEnumAdapters3 (the symbol actually called) instead of CreateDXCoreAdapterFactory. Fills in DxgProvider::enumerate() using D3DKMTEnumAdapters3 + D3DKMTQueryAdapterInfo (PHYSICALADAPTERDEVICEIDS, ADAPTERADDRESS, ADAPTERREGISTRYINFO, GETSEGMENTSIZE). Filters AMD adapters only, packs PCI BDF into location_id, converts wide adapter names to UTF-8, and closes every adapter handle. Fields DXCore cannot supply (gfx_target_version, CU/SDMA counts, mem_banks/caches/io_links) are left zero/empty rather than fabricated. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

After the topology refactor moved all sysfs walking into LinuxKfdProvider, agent.cpp no longer references the fs namespace alias or lib/common/filesystem.hpp. clang-tidy on ubuntu 22.04 flagged this as misc-unused-alias-decls (warnings-as-errors). Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

- dxg_provider.cpp: drop redundant 'static' from constexpr defs inside anonymous namespace (readability-static-definition-in-anonymous-namespace). - d3dkmt_provider.hpp: mark stub helpers static; they don't touch instance state (readability-convert-member-functions-to-static). - topology/CMakeLists.txt: cmake-format normalization. - agent.cpp: drop stray blank line per clang-format. Verified locally with clang-format 11.1.0, clang-tidy 19.1.0, and cmake-format 0.6.13 from requirements.txt. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

- DxgProvider::is_available() now also checks /dev/dxg presence so selection matches the documented "libdxcore.so + /dev/dxg" intent. Without this, a box with libdxcore.so but no /dev/dxg would pick DxgProvider and then return an empty topology silently. - ROCPROFILER_FORCE_D3DKMT is now gated on _WIN32. On non-Windows builds the override is ignored with a warning instead of selecting the stub provider that always returns empty. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

Drops the abstract TopologyProvider base in favor of per-platform free functions under rocprofiler::platform::{gnulinux,wsl,windows}, per CONTRIBUTING.md item 8. Each platform owns is_available()/enumerate()/name; agent.cpp dispatches at runtime in precedence gnulinux -> wsl -> windows, with ROCPROFILER_FORCE_PLATFORM={gnulinux|wsl|windows} as override. The gnulinux body is moved verbatim from linux_kfd_provider.cpp so existing KFD behavior stays byte-identical. The WSL body is moved from dxg_provider and gains an RAII DxcoreHandle wrapper for the dlopen'd library; review comments on get_agent_offset rationale and the WSL num_xcc=1 limitation are addressed inline. update_agent_runtime_visibility moves to agent.hpp so the now-redundant agent_internal.hpp is removed. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

Three D3DKMTQueryAdapterInfo calls (ADAPTERADDRESS, ADAPTERREGISTRYINFO, GETSEGMENTSIZE) previously dropped their status and continued, so a failure would publish a rocprofiler_agent_t with a zeroed BDF, missing adapter name, or zero VRAM size. Check each status; on failure log a WARNING with the adapter index and vendor/device IDs and skip the adapter. Adapter cleanup now uses common::scope_destructor instead of duplicating DxcCloseAdapter at every exit path, matching how the rest of the codebase handles scope-bound resources and also closing the leak window if the rocprofiler_agent_t allocation throws. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

Adds tests/platform_dispatch.cpp covering the runtime selection in agent.cpp and the per-platform entry points: - wsl::is_available() / wsl::enumerate() degrade safely off-WSL - gnulinux::is_available() honors the local-topology fixture - ROCPROFILER_FORCE_PLATFORM={gnulinux,wsl,unknown} dispatch behavior Tests with /dev/dxg dependencies self-skip when not in WSL so they are useful on both bare-metal Linux CI and on a WSL workstation. wsl/agent.cpp: cache the random agent offset as const and drop the redundant gpu_count counter (was always equal to the logical id since every DXCore-enumerated adapter is a GPU). Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

…ws/ subdirs Relocate the platform-abstraction implementations into per-OS subdirectories under details/platform/, matching the layout introduced in rocprofiler-sdk PR #6399: loader_posix.cpp -> gnulinux/loader.cpp pe_parser_posix.cpp -> gnulinux/pe_parser.cpp loader_win.cpp -> windows/loader.cpp pe_parser_win.cpp -> windows/pe_parser.cpp encoding_win.hpp -> windows/encoding.hpp loader.hpp and pe_parser.hpp stay in details/platform/ since they are the platform-neutral interface. Includes in the moved .cpp files now reach those headers via the rooted "details/platform/..." form (resolved against the rocprofiler-register target's include path on source/lib/rocprofiler-register), and the Windows .cpp files include the relocated encoding helper via "details/platform/windows/encoding.hpp". Layout-only change: no behavioral edits to the .cpp content. Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>

…yout Replace the suffix-based file selection in details/platform/CMakeLists.txt with the add_subdirectory pattern used by rocprofiler-sdk PR #6399. Each per-OS subdirectory now owns its own sources via a self-contained CMakeLists.txt: * details/platform/CMakeLists.txt: dispatches to windows/ on WIN32 and gnulinux/ otherwise. * details/platform/gnulinux/CMakeLists.txt: adds loader.cpp and pe_parser.cpp to rocprofiler-register. * details/platform/windows/CMakeLists.txt: adds loader.cpp, pe_parser.cpp, and encoding.hpp, and links psapi (used by EnumProcessModulesEx / GetModuleInformation in loader.cpp and pe_parser.cpp). CONTRIBUTING.md compliance: only target-scoped commands, and each subdirectory's CMakeLists.txt owns its own sources. Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>

…7016) ## Motivation Enable `rocprofv3` counter collection and profiling on WSL2 for gfx11-class GPUs. On WSL, the complete KMT topology is read up front through `librocdxg`, using the same thunk interface as the HSA runtime. An agent is published only after all required topology information has been collected, so the published record remains stable for its lifetime. As a result, `rocprofiler_query_available_agents()` returns consistent agent information regardless of when it is called, including before the HSA runtime has initialized. This is required by `rocprofv3-avail` and counter metric lookup. Native Linux continues to use the existing `gnulinux` path. `select_platform()` (`agent.cpp:355`) checks `gnulinux::is_available()` first, which succeeds whenever the KFD sysfs nodes directory is present, so the WSL path is not selected on a normal Linux system. There are two ungated changes that also affect native Linux. Both are fixes and are called out under **Native-Linux impact** below. This work builds on [[#6399](#6399)], which introduced the platform dispatch structure used here. `platform_dispatch.cpp` and its six tests came from that PR (`d2f0b27ad5`) and are therefore listed as pre-existing under **Tests**. This PR adds WSL as another implementation behind the existing `select_platform()` interface. ## Dependency on [#10034](#10034) [#10034](#10034) contains the libhsakmt and ROCr prerequisites for this work. There is no CI gate, submodule pin, or build-time dependency that enforces merge order, so the two PRs can technically merge in either order. Their changes to `core/runtime/amd_topology.cpp` also do not overlap. However, two parts of this PR depend on #10034 before they are fully functional. ### 1. WSL enumeration is disabled in a stock build until the required hsakmt headers are installed WSL enumeration is controlled by the CMake probe `rocprofiler_sdk_dxg_topology_supported()` (`cmake/Modules/rocprofiler-sdk-utilities.cmake:120`). The probe requires the hsakmt headers used by the build to provide: * `HsaStructureSizes` * `HsaNodeProperties::LuidLowPart` * `HsaNodeProperties::LuidHighPart` * `OverrideEngineId` Released ROCm headers, including 7.2.4, do not provide all of these. When the probe fails, `agent_unsupported.cpp` is built instead of the WSL enumerator, `platform::wsl::is_available()` returns false, and `enumerate()` returns an empty result. The copies of `hsakmttypes.h` in this repository already contain the required definitions (lines 359-360 and 1578), but they are not currently installed. #10034 adds the WSL header installation rules in `libhsakmt/CMakeLists_wsl.txt` through `ROCDXG_PUBLIC_HEADERS`. Until that lands, a stock build of this PR will not enable WSL enumeration. ### 2. The DXG topology snapshot is not reference-counted on current `develop` Current `develop` unconditionally calls `topology_drop_snapshot()` from `hsaKmtReleaseSystemProperties()` (`libhsakmt/src/dxg/topology.cpp:417`). This frees the global topology snapshot and destroys its `WDDMDevice` objects even if another consumer still has the topology acquired. #10034 adds `src/dxg/snapshot_refcount.hpp` and makes snapshot destruction conditional on the reference count. The normal `rocprofv3` path is not affected because the SDK topology read happens from `ROCPROFILER_LIBRARY_CTOR` and completes before `main()`, and therefore before the application's `hsa_init()`. `rocprof-attach` is different because it injects into an already-running process. In that case there is no ordering guarantee between the SDK and the existing HSA runtime consumer, so late attach can fail until the reference-counting change from #10034 is present. ## Technical Details ### Agent enumeration (WSL) * `platform/wsl/dxg_topology.cpp` and `dxg_thunk.hpp` read KMT topology through `librocdxg` using: * `hsaKmtOpenKFD()` * `hsaKmtAcquireSystemProperties()` * `hsaKmtGetNodeProperties()` * `hsaKmtReleaseSystemProperties()` * `hsaKmtCloseKFD()` The symbols are resolved with `dlsym`. `DxgThunk::complete()` verifies that all five are available before any are called. * `HsaNodeProperties` has no size parameter, so the destination buffer includes `kNodePropsSlack = 128` bytes of additional space. * `platform/wsl/agent.cpp` matches DXG adapters to HSA nodes by LUID, with device ID as a fallback. Enumeration is allowed to be partial: an unmatched adapter is skipped with a diagnostic while successfully matched GPUs are still published. * Memory banks, caches, and IO links are not currently published by the WSL path. * If the thunk reports no GPU node, no GPU agent is published. A warning is emitted and the application continues. * `agent_mapping.{hpp,cpp}` adds a WSL-specific mapping policy. Native Linux keeps the existing strict behavior (`agent.cpp:755`), while WSL logs mismatches, keeps successfully paired GPUs, and continues. ### Counters * `counters/controller.cpp` emits a one-time warning that WSL counters are collected through the DXG vendor-packet path rather than the KFD profiler ioctl. * `WSLKMT_VENDOR_PACKET=1` is now set from `wsl::enumerate()` instead of `is_available()`. This matters when using `ROCPROFILER_FORCE_PLATFORM=wsl`, where the old placement was bypassed. Without the environment variable, the DXG path drops aqlprofile's vendor-specific PM4 IB and counter values are returned as zero. * `basic_counters.xml` and `derived_counters.xml` each gain an empty `<gfx1150 base="gfx11">` section to keep the installed metric files consistent. These XML files are not used by the runtime metric loader. Runtime metrics come from `config.yaml` (`counters/metrics.cpp:348`), which already supports gfx1150 and is unchanged by this PR. ### KFD * `kfd::init()` now probes for a usable KFD device. If `/dev/kfd` is unavailable or cannot be opened, initialization continues with a warning instead of aborting in `kfd_device_fd()`. ## Native-Linux impact Two changes are not WSL-gated and also apply to native Linux: * `aql/helpers.cpp:51`: `get_query_info()` returns an empty ID query for metrics that do not have a hardware block. Constant and topology metrics such as `simd_count` have an empty block name. Passing that empty name to `aqlprofile_get_pmc_info()` can dereference a null pointer and abort. * `kfd/kfd.cpp:1586`: the KFD availability check is unconditional. A system without a usable `/dev/kfd`, for example because the driver is not loaded or access is denied, now receives a warning and continues instead of aborting. Other rocprofiler services are unaffected. All other WSL-specific behavior remains isolated from native Linux. The DXG implementation requires `/dev/dxg` and `libdxcore.so`, and the PM4 command-buffer bounds change is limited to libhsakmt's DXG/WDDM submission path. ## Tests * **44 new tests** * 14 `wsl_dxg_topology` tests * 16 `wsl_dxg_thunk` tests * 10 `agent_mapping` tests * 4 `parse_gfx_target_version` tests The first two groups are in new executables. The mapping and target-version tests are added to the existing `rocprofiler-sdk-lib-tests`. None of these tests requires a GPU, HSA runtime, or real DXG thunk. The DXG tests use a fake thunk and fake dynamic loader. The 30 DXG-specific tests use the same CMake capability probe as the WSL enumerator, so they are only built when the required hsakmt interface is available. In a stock configuration, 14 of the 44 new tests are built. * `platform-dispatch-tests` and its six tests are pre-existing from #6399. * `rocprofv3-avail` * Three tests are moved out of the PC-sampling gate and are now enabled on Navi. * The L2 hit counter is selected by GPU family instead of being hardcoded. These tests have not previously run in Navi CI, so they may expose existing Navi-specific issues. * `hsa_barrier` * Skip when no GPU agent is available instead of dereferencing an empty agent map. * `agent` * `agent_visibility_multigpu` performs its `substr(1)` calls after the no-GPU skip, avoiding `std::out_of_range` when no agents are present. * `grid_max_size` tolerates the difference between rocprofiler's hardcoded `INT32_MAX` value and the value reported by the installed ROCr runtime. ## JIRA ID JIRA ID: ROCM-23912 ## Test Plan ## Test Result ## Submission Checklist - [ ] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. --------- Co-authored-by: Mika Laitio <mika.laitio@amd.com>

Copilot AI review requested due to automatic review settings May 25, 2026 18:10

bgopesh requested review from a team as code owners May 25, 2026 18:10

Copilot started reviewing on behalf of bgopesh May 25, 2026 18:10

View session

github-actions Bot added the project: rocprofiler-sdk label May 25, 2026

bgopesh mentioned this pull request May 25, 2026

[rocprofiler-sdk] Introduce TopologyProvider abstraction #6352

Closed

1 task

bgopesh changed the title ~~Users/bgopesh/topology provider~~ [rocprofiler-sdk] Introduce TopologyProvider abstraction May 25, 2026

This comment was marked as outdated.

Sign in to view

systems-assistant Bot added the organization: ROCm label May 25, 2026

meserve-amd suggested changes May 26, 2026

View reviewed changes

ApoKalipse-V reviewed May 26, 2026

View reviewed changes

Comment thread projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/platform/wsl/agent.cpp Resolved

Comment thread projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/topology/dxg_provider.cpp Resolved

bgopesh requested review from ApoKalipse-V and meserve-amd May 27, 2026 07:12

bgopesh changed the title ~~[rocprofiler-sdk] Introduce TopologyProvider abstraction~~ [rocprofiler-sdk] Platform specific topology information May 27, 2026

meserve-amd approved these changes May 28, 2026

View reviewed changes

bgopesh assigned jrmadsen May 29, 2026

ApoKalipse-V requested changes Jun 1, 2026

View reviewed changes

bgopesh force-pushed the users/bgopesh/topology-provider branch 3 times, most recently from 0b4794f to 8796818 Compare June 2, 2026 05:04

bgopesh requested a review from ApoKalipse-V June 2, 2026 05:28

ApoKalipse-V approved these changes Jun 2, 2026

View reviewed changes

bgopesh force-pushed the users/bgopesh/topology-provider branch 3 times, most recently from 7aab19d to 7342b2f Compare June 2, 2026 11:12

bgopesh and others added 9 commits June 2, 2026 20:06

addressing review comments

bc9d5b3

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

bgopesh force-pushed the users/bgopesh/topology-provider branch from 7342b2f to 4882a01 Compare June 2, 2026 14:36

jrmadsen approved these changes Jun 2, 2026

View reviewed changes

bgopesh merged commit d2f0b27 into develop Jun 2, 2026
40 checks passed

bgopesh deleted the users/bgopesh/topology-provider branch June 2, 2026 18:47

jrmadsen mentioned this pull request Jun 9, 2026

[rocprofiler-sdk] Fix cmake and build issues #7008

Merged

1 task done

ihhethan mentioned this pull request Jun 10, 2026

feat(rocprofiler-sdk): WSL2 compute and profiling support for RDNA 3.5 #7016

Merged

1 task

Conversation

bgopesh commented May 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Directory structure

Key design points

Technical Details

JIRA ID

Test Plan

Test Result

Submission Checklist

Uh oh!

This comment was marked as outdated.

Uh oh!

github-actions Bot commented May 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Coverage Report

Code Coverage Report

Tests Only

Samples Only

Tests + Samples

Uh oh!

jrmadsen commented May 26, 2026

Uh oh!

bgopesh commented May 26, 2026

Uh oh!

bgopesh commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

meserve-amd left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

bgopesh commented May 27, 2026

Uh oh!

bgopesh commented May 27, 2026

Uh oh!

bgopesh commented May 28, 2026

Uh oh!

meserve-amd left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

bgopesh commented May 29, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ApoKalipse-V left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bgopesh commented Jun 2, 2026

Uh oh!

Uh oh!

Reviewers

bgopesh commented May 25, 2026 •

edited

Loading

github-actions Bot commented May 25, 2026 •

edited

Loading

bgopesh commented May 26, 2026 •

edited

Loading

meserve-amd left a comment •

edited

Loading

ApoKalipse-V left a comment •

edited

Loading