adds engine benchmark to integration tests

[Prathik Rao (prathikr)]

By default, the pipeline runs the
smoke test config which uses qwen2.5-0.5b-instruct.

Select Run Full Engine Benchmark to run the
full benchmark config with qwen2.5-1.5b-instruct + qwen2.5-7b-instruct.

Also contains some bug fixes to get the benchmarking pipeline running w/ H100 CI machine.

[Copilot]

Pull request overview

Adds a dedicated Azure Pipelines integration stage to build and run the native Engine benchmark on Linux (H100 pool), defaulting to the smoke-test.json benchmark config and optionally running the full config.json suite via a new pipeline parameter. This integrates benchmark execution into the existing integration-tests pipeline flow.

Changes:

• Link engine_benchmark against Threads::Threads in the benchmark CMake target.
• Introduce a new integration job template that builds the benchmark binary, exports required HF models via the in-repo builder, runs the benchmark, and publishes results.
• Wire the new “Engine Benchmark” stage into the integration pipeline templates and add a run_full_engine_benchmark boolean parameter to control config selection.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
benchmark/engine/CMakeLists.txt	Ensures the benchmark executable links thread support explicitly.
.pipelines/stages/jobs/integration-engine-benchmark-job.yml	New job template to build models + run the engine benchmark in a CUDA container and publish outputs.
.pipelines/stages/integration-stage.yml	Adds an Engine Benchmark stage to the shared integration stage template.
.pipelines/integration-tests.yml	Adds run_full_engine_benchmark pipeline parameter and passes it into the stage template.

---

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