style: rustfmt repair after #1100

[justinchuby]

What

$ git checkout origin/main && cargo fmt --all -- --check
Diff in crates/onnx-runtime-ep-cpu/src/kernels/governed_weight_cache.rs:263
Diff in crates/onnx-runtime-ep-cpu/src/kernels/matmul.rs:354, 2127, 2148, 2163, 3206
Diff in crates/onnx-runtime-ep-cpu/src/lib.rs:92

main is fmt-broken at a0ed99a (#1100). main is unprotected and the merge queue does not re-run CI on the merge result, so this fails the Check formatting step of both Fast (Linux x86_64) and Rust quality on every branch cut from it — including #1101, which is how I found it.

What this is

The literal output of cargo fmt --all, nothing else. No logic change, no test change. 17 insertions, 20 deletions, all whitespace and line-wrapping.

Verification

• cargo fmt --all -- --check — clean
• cargo test -p onnx-runtime-ep-cpu --lib — unchanged

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.40%. Comparing base (a0ed99a) to head (0f00684).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1102      +/-   ##
==========================================
+ Coverage   79.76%   80.40%   +0.64%     
==========================================
  Files         367      367              
  Lines      156416   156410       -6     
  Branches   156416   156410       -6     
==========================================
+ Hits       124765   125764     +999     
+ Misses      26947    25940    -1007     
- Partials     4704     4706       +2     

Flag	Coverage Δ
cli-ort-linux	83.79% <ø> (ø)
cli-ort-windows	83.31% <ø> (-0.10%)	⬇️
offline	80.28% <100.00%> (+0.66%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...untime-ep-cpu/src/kernels/governed_weight_cache.rs	89.55% <100.00%> (+0.07%)	⬆️
crates/onnx-runtime-ep-cpu/src/kernels/matmul.rs	86.34% <100.00%> (+5.05%)	⬆️

... and 5 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]

🔴 Benchmark Regression Detected

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
🔴	block_quantized_matmul_cached_dense/mxfp4_preexpanded_dense_oncelock_like_proxy/1x1024x1024	40.56 µs	88.88 µs	+119.1%
🔴	gather/large_f32_threads=1-internal/131072	28.94 µs	38.65 µs	+33.6%
🔴	block_quantized_matmul_cached_dense/mxfp4_uncached_dequant_each_call/1x1024x1024	828.73 µs	1.09 ms	+32.1%
🔴	matmul/large_generic_bf16_threads=8/32x1024x1024	1.65 ms	2.14 ms	+30.0%
⚠️	matmul/large_generic_f32_threads=8/32x1024x1024	5.52 ms	6.81 ms	+23.4%
⚠️	gather/medium_f32_threads=1-internal/32768	3.49 µs	4.21 µs	+20.7%
✅	matmul/small_generic_f16_threads=8/1x256x256	32.88 µs	37.52 µs	+14.1%
✅	gather/medium_f16_threads=1-internal/32768	2.28 µs	2.53 µs	+11.3%
✅	matmul/small_generic_bf16_threads=8/1x256x256	33.72 µs	36.70 µs	+8.8%
✅	reduce_mean/medium_f32_threads=1-internal/65536	234.09 µs	254.21 µs	+8.6%
✅	add/large_f32_threads=1-internal/4194304	665.89 µs	721.49 µs	+8.3%
✅	add/medium_f32_threads=1-internal/262144	24.61 µs	26.25 µs	+6.6%
✅	matmul/medium_generic_bf16_threads=8/32x512x512	598.75 µs	635.39 µs	+6.1%
✅	matmul/large_generic_bf16_threads=1/32x1024x1024	2.03 ms	2.11 ms	+3.8%
✅	matmul/small_generic_f16_threads=1/1x256x256	35.61 µs	36.77 µs	+3.3%
✅	add/medium_f16_threads=1-internal/262144	103.35 µs	106.68 µs	+3.2%
✅	add/small_f16_threads=1-internal/1024	459.6 ns	474.3 ns	+3.2%
✅	reduce_mean/small_f32_threads=1-internal/4096	14.50 µs	14.96 µs	+3.2%
✅	gather/small_f16_threads=1-internal/4096	492.5 ns	501.5 ns	+1.8%
✅	sampling_latency/top_k_per_token	54.51 µs	55.36 µs	+1.6%
✅	add/small_bf16_threads=1-internal/1024	434.2 ns	440.9 ns	+1.5%
✅	add/large_bf16_threads=1-internal/4194304	1.67 ms	1.67 ms	+0.2%
✅	add/medium_bf16_threads=1-internal/262144	104.50 µs	104.50 µs	+0.0%
✅	tokenization/decode_tokens_per_second	6.92 ms	6.84 ms	-1.2%
✅	add/small_f32_threads=1-internal/1024	204.5 ns	200.5 ns	-2.0%
✅	logit_processing/seven_processor_chain_per_step	335.30 µs	327.79 µs	-2.2%
✅	gather/small_f32_threads=1-internal/4096	675.8 ns	658.5 ns	-2.6%
✅	matmul/medium_generic_f16_threads=1/32x512x512	36.04 µs	35.09 µs	-2.6%
✅	add/large_f16_threads=1-internal/4194304	1.73 ms	1.69 ms	-2.7%
✅	matmul/small_generic_bf16_threads=1/1x256x256	38.04 µs	36.97 µs	-2.8%
✅	reduce_mean/large_f32_threads=1-internal/262144	1.02 ms	985.42 µs	-3.1%
✅	grammar_masking/llguidance_compute_mask/32	82.07 µs	79.49 µs	-3.1%
✅	matmul/small_generic_f32_threads=1/1x256x256	40.32 µs	38.94 µs	-3.4%
✅	matmul/medium_generic_f16_threads=8/32x512x512	43.72 µs	41.81 µs	-4.4%
✅	matmul/medium_generic_f32_threads=1/32x512x512	2.58 ms	2.46 ms	-4.7%
✅	matmul/large_generic_f16_threads=8/32x1024x1024	109.15 µs	103.28 µs	-5.4%
✅	qwen3_sampling_processors/top_p_fast_after_top_k	566.74 µs	525.73 µs	-7.2%
✅	matmul/medium_generic_bf16_threads=1/32x512x512	604.47 µs	555.38 µs	-8.1%
✅	matmul/large_generic_f32_threads=1/32x1024x1024	10.55 ms	9.68 ms	-8.3%
✅	matmul/large_generic_f16_threads=1/32x1024x1024	90.67 µs	82.61 µs	-8.9%
✅	sampling_latency/greedy_per_token	3.63 µs	3.25 µs	-10.3%
✅	kv_cache/alloc_dealloc_pages	45.12 µs	40.26 µs	-10.8%
✅	qwen3_sampling_processors/top_p_full_sort_after_top_k_baseline	4.13 ms	3.67 ms	-11.1%
✅	qwen3_sampling_processors/top_k_partial_selection	168.54 µs	148.63 µs	-11.8%
✅	sampling_latency/top_p_per_token	461.18 µs	398.33 µs	-13.6%
🟢	sampling_latency/min_p_per_token	266.12 µs	226.03 µs	-15.1%
🟢	gather/large_bf16_threads=1-internal/131072	16.64 µs	13.68 µs	-17.8%
🟢	gather/small_bf16_threads=1-internal/4096	600.4 ns	491.8 ns	-18.1%
🟢	tokenization/encode_tokens_per_second	480.29 µs	388.23 µs	-19.2%
🟢	qwen3_sampling_processors/top_k_top_p_full_sort_baseline	7.25 ms	5.85 ms	-19.3%
🟢	block_quantized_moe_cached_dense/mxfp4_cached_dense_expert_repeated_call/rows=1,H=256,I=256,E=4,top_k=1	227.78 µs	176.27 µs	-22.6%
🟢	matmul/small_generic_f32_threads=8/1x256x256	45.16 µs	34.59 µs	-23.4%
🟢	qwen3_sampling_processors/top_k_full_sort_baseline	2.89 ms	2.19 ms	-24.2%
🟢	gather/large_f16_threads=1-internal/131072	17.45 µs	13.02 µs	-25.4%
🟢	qwen3_sampling_processors/top_k_top_p_fast	920.03 µs	665.82 µs	-27.6%
🟢	matmul/medium_generic_f32_threads=8/32x512x512	1.66 ms	1.15 ms	-30.5%
🟢	block_quantized_moe_cached_dense/mxfp4_uncached_expert_dequant_each_call/rows=1,H=256,I=256,E=4,top_k=1	785.41 µs	506.34 µs	-35.5%
🟢	gather/medium_bf16_threads=1-internal/32768	4.36 µs	2.41 µs	-44.6%
🟢	block_quantized_matmul_cached_dense/mxfp4_cached_dense_repeated_call/1x1024x1024	129.26 µs	68.39 µs	-47.1%

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: { 3.01 3.46 5.18 }

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)