Fix run-level profiling for subgraph operators

[adrianlizarraga]

Description

Run-level profiling (introduced in PR #26846) does not currently capture profiling events for operators inside subgraphs. This PR fixes that by threading the run_profiler pointer through OpKernelContextInternal to subgraph execution, following the same pattern as terminate_flag.

Root Cause

utils::ExecuteSubgraph() had no run_profiler parameter and always passed nullptr to ExecuteGraphImpl, so nested operators (inside If, Loop, Scan, BeamSearch, GreedySearch) were never profiled at the run level.

Fix

1. OpKernelContextInternal — Added run_profiler_ member and GetRunProfiler() accessor.
2. SessionScope / ExecuteKernel() — Pass the run profiler into OpKernelContextInternal.
3. ExecuteSubgraph() — Added profiling::Profiler* run_profiler = nullptr parameter, forwarded to ExecuteGraphImpl().
4. Control flow ops (if.cc, loop.cc, scan_utils.cc) — Pass context_.GetRunProfiler() to ExecuteSubgraph().
5. Contrib transformer ops (beam_search_impl_gpt.h, beam_search_impl_t5.h, beam_search_impl_whisper.h, greedy_search_impl_gpt.h) — All 8 ExecuteSubgraph() call sites updated to pass this->context_.GetRunProfiler().

Plugin EP control flow kernels (PluginEpIfKernelImpl, etc.) delegate to the same internal kernels, so the fix propagates automatically.

Tests

• CheckRunProfilerWithSubgraph (inference_session_test.cc) — Runs if_mul.onnx, enables run profiling, asserts mul_0 (inside If's then-branch) appears in the profile JSON.
• CheckRunProfilerWithBeamSearch (beam_search_test.cc) — Runs tiny_gpt2_beamsearch.onnx, enables run profiling, asserts decoder subgraph Node entries (beyond the top-level BeamSearch op) appear in the profile JSON.

Files Changed (12 files)

File	Change
core/framework/op_kernel_context_internal.h	Added run_profiler_ member, GetRunProfiler(), constructor param
core/framework/sequential_executor.cc	SessionScope::GetRunProfiler(), pass to OpKernelContextInternal
core/framework/utils.h / utils.cc	run_profiler param on ExecuteSubgraph()
core/providers/cpu/controlflow/if.cc	Forward GetRunProfiler()
core/providers/cpu/controlflow/loop.cc	Forward GetRunProfiler()
core/providers/cpu/controlflow/scan_utils.cc	Forward GetRunProfiler()
contrib_ops/cpu/transformers/beam_search_impl_gpt.h	2 call sites
contrib_ops/cpu/transformers/beam_search_impl_t5.h	2 call sites
contrib_ops/cpu/transformers/beam_search_impl_whisper.h	2 call sites
contrib_ops/cpu/transformers/greedy_search_impl_gpt.h	2 call sites
test/framework/inference_session_test.cc	CheckRunProfilerWithSubgraph test
test/contrib_ops/beam_search_test.cc	CheckRunProfilerWithBeamSearch test

[Copilot]

Pull request overview

Fixes run-level profiling so it also records operator events that occur inside subgraphs executed by control-flow operators (e.g., If/Loop/Scan bodies), by propagating the per-run profiler pointer through subgraph execution and kernel contexts.

Changes:

• Thread profiling::Profiler* run_profiler through utils::ExecuteSubgraph into ExecuteGraphImpl so subgraph execution can record run-level events.
• Plumb the run profiler into OpKernelContextInternal (from SessionScope) and expose it via GetRunProfiler().
• Update CPU control-flow kernels (If, Loop, Scan) to pass context.GetRunProfiler() when executing subgraphs; add a regression test that validates profiling includes a nested subgraph op.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
onnxruntime/test/framework/inference_session_test.cc	Adds a regression test that run-level profiling includes events for ops inside an If subgraph, and parses the generated JSON profile.
onnxruntime/core/providers/cpu/controlflow/scan_utils.cc	Passes the run profiler through to subgraph execution in Scan iteration.
onnxruntime/core/providers/cpu/controlflow/loop.cc	Passes the run profiler through to subgraph execution in Loop iterations.
onnxruntime/core/providers/cpu/controlflow/if.cc	Passes the run profiler through to subgraph execution for the selected If branch.
onnxruntime/core/framework/utils.h	Extends ExecuteSubgraph API to accept an optional run_profiler pointer.
onnxruntime/core/framework/utils.cc	Forwards run_profiler from ExecuteSubgraph into ExecuteGraphImpl.
onnxruntime/core/framework/sequential_executor.cc	Exposes the run profiler from SessionScope and passes it into OpKernelContextInternal during kernel execution.
onnxruntime/core/framework/op_kernel_context_internal.h	Stores the run profiler pointer in the kernel context and adds a GetRunProfiler() accessor for kernels.

---

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

[xiaofeihan1]

Thanks @adrianlizarraga for looking into this. I have two questions

1. How do you find this issue?
2. I found there are some other codes to use ExecuteSubgraph. Do we need to handle that?

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated 1 comment.

---

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

[adrianlizarraga]

Hi @xiaofeihan1, Thanks for taking a look.

How do you find this issue?

I found it while working on a separate PR that ports the C++ EP profiling interface to the public C API for plugin EPs. I noticed that the run profilers did not profile nodes within subgraphs like Loop or If.

I found there are some other codes to use ExecuteSubgraph. Do we need to handle that?

Yep, thanks for catching those. I have now updated the PR to capture all call sites.