[justinchuby]

Summary

Cherry-pick of commit dd6ab807f which missed the #762 merge window by 30 minutes and is stranded on the already-merged branch.

This fixes CURRENTLY-RED lanes on main:

• Rust Windows ARM64
• Rust coverage Windows x86_64
• Rust coverage macOS arm64

The root cause: 7 hardcoded libonnxruntime.so joins in plugin_ort_e2e.rs that fail on Windows (onnxruntime.dll) and macOS (libonnxruntime.dylib).

Fix: Route all ORT library path construction through ort_discovery::ort_lib_name() which returns the correct platform-specific filename. Also switches diag_ort_ep_api_nullcheck to use the skip_if_missing! macro.

Verification (Linux x86_64):

• cargo test -p onnx-runtime-ep-cpu-plugin --no-fail-fast: 24 passed, 0 failed, 0 ignored
• cargo clippy -p onnx-runtime-ep-cpu-plugin --all-targets -- -D warnings: clean
• No hardcoded lib names outside ort_discovery.rs

Windows/macOS correctness is BY CONSTRUCTION (via ort_lib_name() returning the right filename per cfg!(target_os)) and must be confirmed by CI — we have no such hardware in this environment.

Original commit: dd6ab807f on branch squad/ep-plugin-parity-cuda (PR #762, merged 2026-08-12T15:10:54Z; commit dated 30 min after merge).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.75%. Comparing base (ce5f14a) to head (cb2561b).
⚠️ Report is 10 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #823      +/-   ##
==========================================
+ Coverage   77.49%   77.75%   +0.26%     
==========================================
  Files         351      352       +1     
  Lines      141509   146507    +4998     
  Branches   141509   146507    +4998     
==========================================
+ Hits       109666   113923    +4257     
- Misses      27443    28120     +677     
- Partials     4400     4464      +64     

Flag	Coverage Δ
cli-ort-linux	83.50% <ø> (ø)
cli-ort-windows	82.99% <ø> (-0.10%)	⬇️
mlas	80.92% <ø> (-0.05%)	⬇️
offline	77.49% <ø> (+0.28%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 27 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

✅ Benchmarks — No Regression

Comparison of criterion micro-benchmarks: PR head vs merge-base, measured on the same runner in the same job (base first → PR second).

ℹ️ Absolute times are informational only — they vary with runner load. The % change column is the reliable signal because both sides ran under identical conditions.

Status	Scenario	Base	PR	Change
✅	qwen3_sampling_processors/top_k_partial_selection	131.60 µs	145.18 µs	+10.3%
✅	grammar_masking/llguidance_compute_mask/32	70.14 µs	73.40 µs	+4.7%
✅	qwen3_sampling_processors/top_k_top_p_full_sort_baseline	5.30 ms	5.55 ms	+4.6%
✅	matmul/medium_generic_bf16_threads=1/32x512x512	500.46 µs	520.63 µs	+4.0%
✅	matmul/large_generic_bf16_threads=1/32x1024x1024	1.87 ms	1.94 ms	+3.9%
✅	sampling_latency/min_p_per_token	191.33 µs	195.53 µs	+2.2%
✅	matmul/medium_generic_bf16_threads=8/32x512x512	377.02 µs	384.98 µs	+2.1%
✅	tokenization/encode_tokens_per_second	364.33 µs	370.43 µs	+1.7%
✅	qwen3_sampling_processors/top_k_full_sort_baseline	1.95 ms	1.97 ms	+1.4%
✅	matmul/large_generic_f32_threads=1/32x1024x1024	8.56 ms	8.66 ms	+1.3%
✅	matmul/large_generic_f16_threads=1/32x1024x1024	72.70 µs	73.39 µs	+1.0%
✅	sampling_latency/top_p_per_token	350.65 µs	353.15 µs	+0.7%
✅	block_quantized_matmul_cached_dense/mxfp4_uncached_dequant_each_call/1x1024x1024	532.60 µs	534.73 µs	+0.4%
✅	add/medium_f32_threads=1-internal/262144	2.52 ms	2.52 ms	+0.2%
✅	reduce_mean/small_f32_threads=1-internal/4096	14.54 µs	14.57 µs	+0.2%
✅	matmul/small_generic_bf16_threads=1/1x256x256	30.10 µs	30.09 µs	-0.0%
✅	tokenization/decode_tokens_per_second	5.73 ms	5.73 ms	-0.1%
✅	matmul/medium_generic_f32_threads=8/32x512x512	890.65 µs	889.31 µs	-0.1%
✅	matmul/small_generic_bf16_threads=8/1x256x256	30.69 µs	30.65 µs	-0.2%
✅	sampling_latency/greedy_per_token	3.02 µs	3.01 µs	-0.2%
✅	block_quantized_moe_cached_dense/mxfp4_cached_dense_expert_repeated_call/rows=1,H=256,I=256,E=4,top_k=1	72.33 µs	72.09 µs	-0.3%
✅	qwen3_sampling_processors/top_p_fast_after_top_k	484.73 µs	482.03 µs	-0.6%
✅	logit_processing/seven_processor_chain_per_step	300.79 µs	299.10 µs	-0.6%
✅	block_quantized_matmul_cached_dense/mxfp4_cached_dense_repeated_call/1x1024x1024	41.10 µs	40.80 µs	-0.7%
✅	qwen3_sampling_processors/top_p_full_sort_after_top_k_baseline	3.30 ms	3.28 ms	-0.8%
✅	matmul/medium_generic_f16_threads=8/32x512x512	28.04 µs	27.72 µs	-1.1%
✅	reduce_mean/medium_f32_threads=1-internal/65536	231.64 µs	228.91 µs	-1.2%
✅	matmul/large_generic_bf16_threads=8/32x1024x1024	1.34 ms	1.32 ms	-1.4%
✅	kv_cache/alloc_dealloc_pages	37.17 µs	36.61 µs	-1.5%
✅	sampling_latency/top_k_per_token	50.25 µs	48.76 µs	-3.0%
✅	block_quantized_moe_cached_dense/mxfp4_uncached_expert_dequant_each_call/rows=1,H=256,I=256,E=4,top_k=1	359.56 µs	348.24 µs	-3.1%
✅	matmul/large_generic_f32_threads=8/32x1024x1024	3.78 ms	3.66 ms	-3.2%
✅	add/large_f16_threads=1-internal/4194304	40.81 ms	39.08 ms	-4.3%
✅	gather/small_f16_threads=1-internal/4096	466.7 ns	443.1 ns	-5.1%
✅	reduce_mean/large_f32_threads=1-internal/262144	963.55 µs	911.64 µs	-5.4%
✅	add/large_f32_threads=1-internal/4194304	39.46 ms	37.24 ms	-5.6%
✅	gather/small_f32_threads=1-internal/4096	662.6 ns	613.2 ns	-7.5%
✅	add/medium_bf16_threads=1-internal/262144	2.62 ms	2.42 ms	-7.7%
✅	add/small_bf16_threads=1-internal/1024	12.83 µs	11.79 µs	-8.1%
✅	qwen3_sampling_processors/top_k_top_p_fast	664.64 µs	610.47 µs	-8.1%
✅	matmul/medium_generic_f16_threads=1/32x512x512	30.25 µs	27.76 µs	-8.2%
✅	matmul/medium_generic_f32_threads=1/32x512x512	2.38 ms	2.18 ms	-8.4%
✅	add/large_bf16_threads=1-internal/4194304	42.89 ms	39.22 ms	-8.6%
✅	matmul/small_generic_f16_threads=8/1x256x256	32.83 µs	29.43 µs	-10.4%
✅	add/medium_f16_threads=1-internal/262144	2.63 ms	2.35 ms	-10.4%
✅	add/small_f32_threads=1-internal/1024	202.6 ns	181.2 ns	-10.6%
✅	block_quantized_matmul_cached_dense/mxfp4_preexpanded_dense_oncelock_like_proxy/1x1024x1024	47.83 µs	42.76 µs	-10.6%
✅	add/small_f16_threads=1-internal/1024	13.66 µs	11.93 µs	-12.6%
🟢	matmul/small_generic_f16_threads=1/1x256x256	37.15 µs	28.52 µs	-23.2%
🟢	gather/small_bf16_threads=1-internal/4096	583.7 ns	440.1 ns	-24.6%
🟢	gather/large_bf16_threads=1-internal/131072	15.28 µs	10.79 µs	-29.3%
🟢	matmul/small_generic_f32_threads=8/1x256x256	49.24 µs	33.18 µs	-32.6%
🟢	gather/large_f16_threads=1-internal/131072	18.40 µs	12.14 µs	-34.0%
🟢	matmul/small_generic_f32_threads=1/1x256x256	51.48 µs	33.75 µs	-34.4%
🟢	gather/medium_bf16_threads=1-internal/32768	3.42 µs	2.19 µs	-35.9%
🟢	gather/medium_f32_threads=1-internal/32768	5.64 µs	3.39 µs	-39.9%
🟢	gather/large_f32_threads=1-internal/131072	42.71 µs	25.58 µs	-40.1%
🟢	gather/medium_f16_threads=1-internal/32768	3.86 µs	2.28 µs	-41.0%
🟢	matmul/large_generic_f16_threads=8/32x1024x1024	135.21 µs	77.72 µs	-42.5%

Visual flags: ⚠️ ≥ 15% slower, 🔴 ≥ 30% slower — calibrated against measured runner noise (~27% worst-case on multi-threaded matmul)

Host info

CPU: Apple M1 (Virtual)
Cores: 3
OS: Darwin 25.5.0 arm64
Rust: rustc 1.97.1 (8bab26f4f 2026-07-14)
Load avg: { 3.13 3.34 5.42 }

What this cannot catch

• Regressions in code paths not covered by these benchmarks (e.g., end-to-end decode with a real model)
• Sub-threshold regressions that compound over multiple PRs
• Performance changes that only manifest under GPU execution
• Latency changes in the ORT integration path (these benchmarks exercise the native Rust kernels)