adding SIMD/TensorPrimitives to dotent-diag/analyzing-dotnet-performance

[jeffschwMSFT]

Incorporated feedback, trained on 100+ scalar loop examples, moved to analyzing-dotnet-performance reference

[Copilot]

Pull request overview

Adds SIMD/TensorPrimitives guidance and evaluation coverage to the dotnet-diag/analyzing-dotnet-performance skill, expanding it to recognize when scalar loops are good candidates for vectorization (or when SIMD is not applicable).

Changes:

• Adds new SIMD-focused fixtures and corresponding eval scenarios (TensorPrimitives reductions, SIMD-friendly loops, and a “no SIMD opportunity” case).
• Introduces a new simd-vectorization.md reference with decision gating (TensorPrimitives-first vs manual intrinsics).
• Updates the skill description and detection signals to include SIMD/vectorization.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/dotnet-diag/analyzing-dotnet-performance/fixtures/simd-tensor-primitives-product.cs	New fixture for product reduction intended for TensorPrimitives optimization.
tests/dotnet-diag/analyzing-dotnet-performance/fixtures/simd-tensor-primitives-minmax.cs	New fixture for min/max reduction intended for TensorPrimitives optimization.
tests/dotnet-diag/analyzing-dotnet-performance/fixtures/simd-no-opportunity-catalog.cs	New fixture representing a case where SIMD is not a meaningful optimization.
tests/dotnet-diag/analyzing-dotnet-performance/fixtures/simd-conditional-increment.cs	New SIMD-friendly loop fixture (conditional increment).
tests/dotnet-diag/analyzing-dotnet-performance/fixtures/simd-bit-reverser.cs	New SIMD-friendly byte processing fixture (bit reversal).
tests/dotnet-diag/analyzing-dotnet-performance/eval.yaml	Adds five new scenarios covering TensorPrimitives, SIMD intrinsics, and “no opportunity” detection.
plugins/dotnet-diag/skills/analyzing-dotnet-performance/references/simd-vectorization.md	New reference doc defining the SIMD/TensorPrimitives decision gate and implementation patterns.
plugins/dotnet-diag/skills/analyzing-dotnet-performance/SKILL.md	Updates description and signal detection to include SIMD vectorization and reference loading.

---

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

[Copilot]

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

---

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

[Copilot]

Pull request overview

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

---

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

[tannergooding]

Few nits/feedback, but this looks overall good!

[Copilot]

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 7 comments.

---

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

[jeffschwMSFT]

/evaluate

[tannergooding]

Couple more nits, but LGTM

[jeffschwMSFT]

we are adding a new experimental branch to this repo and this may be one of the first skills to give it a try. @artl93 and I discussed and although this seems useful, we are not sure how many people it will be most helpful for. though willing to discuss. (fwiw, I am on the list of people that have found it helpful for numeric libraries)

[github-actions]

Skill Validation Results

Skill	Scenario	Quality (Isolated)	Quality (Plugin)	Skills Loaded	Overfit	Verdict
directory-build-organization	Organize build infrastructure for a multi-project repo	3.0/5 → 5.0/5 🟢	3.0/5 → 4.7/5 🟢	✅ directory-build-organization; tools: skill, create, edit, bash, task / ✅ msbuild-antipatterns; directory-build-organization; tools: task, bash, skill	✅ 0.15	✅
dotnet-trace-collect	High CPU in Kubernetes on Linux (.NET 8)	3.7/5 → 4.7/5 🟢	3.7/5 → 4.3/5 🟢	✅ dotnet-trace-collect; tools: report_intent, skill, view, glob, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	.NET Framework on Windows without admin privileges	2.0/5 → 5.0/5 🟢	2.0/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill / ✅ dotnet-trace-collect; tools: skill	✅ 0.16	✅
dotnet-trace-collect	.NET 10 on Linux with root access and native call stacks	1.7/5 → 4.0/5 🟢	1.7/5 → 4.0/5 🟢	✅ dotnet-trace-collect; tools: skill / ✅ dotnet-trace-collect; tools: skill	✅ 0.16	✅
dotnet-trace-collect	Memory leak on Linux (.NET 8)	2.3/5 → 3.0/5 🟢	2.3/5 → 3.0/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	Slow requests on Windows with PerfView	3.7/5 → 5.0/5 🟢	3.7/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, glob, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	Excessive GC on Linux (.NET 8)	3.3/5 → 5.0/5 🟢	3.3/5 → 4.7/5 🟢	✅ dotnet-trace-collect; tools: skill, glob / ✅ dotnet-trace-collect; tools: skill	✅ 0.16	✅
dotnet-trace-collect	Hang or deadlock diagnosis on Linux	2.7/5 → 3.7/5 🟢	2.7/5 → 3.0/5 🟢	✅ dotnet-trace-collect; tools: skill / ✅ dotnet-trace-collect; dump-collect; tools: skill, report_intent, view	✅ 0.16	❌ [1]
dotnet-trace-collect	Windows container high CPU with PerfView	1.7/5 → 4.3/5 🟢	1.7/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill, glob / ✅ dotnet-trace-collect; tools: skill	✅ 0.16	✅
dotnet-trace-collect	Long-running intermittent issue with PerfView triggers	2.3/5 → 5.0/5 🟢	2.3/5 → 4.3/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash, glob / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	Linux pre-.NET 10 needing native call stacks	2.7/5 → 4.7/5 🟢	2.7/5 → 4.3/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	Windows modern .NET with admin high CPU	2.0/5 → 4.7/5 🟢	2.0/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash, glob / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
dotnet-trace-collect	Memory leak on .NET Framework Windows	3.3/5 → 5.0/5 🟢	3.3/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: report_intent, skill, view, glob, bash / ✅ dotnet-trace-collect; tools: report_intent, skill, view	✅ 0.16	✅
dotnet-trace-collect	Kubernetes with console access prefers console tools	4.3/5 → 4.7/5 🟢	4.3/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	❌ [2]
dotnet-trace-collect	Container installation without .NET SDK	3.0/5 → 3.3/5 🟢	3.0/5 → 4.7/5 🟢	✅ dotnet-trace-collect; tools: skill / ✅ dotnet-trace-collect; tools: skill	✅ 0.16	❌ [3]
dotnet-trace-collect	HTTP 500s from downstream service on Linux (.NET 8)	4.3/5 → 5.0/5 🟢	4.3/5 → 5.0/5 🟢	✅ dotnet-trace-collect; tools: skill, report_intent, view, bash, glob / ✅ dotnet-trace-collect; tools: report_intent, skill, view	✅ 0.16	✅
dotnet-trace-collect	Networking timeouts on Windows with admin (.NET 8)	2.0/5 → 5.0/5 🟢	2.0/5 → 4.7/5 🟢	✅ dotnet-trace-collect; tools: report_intent, skill, view, bash / ✅ dotnet-trace-collect; tools: skill, report_intent, view	✅ 0.16	✅
analyzing-dotnet-performance	Detects compiled regex startup budget and regex chain allocations	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ✅ analyzing-dotnet-performance; tools: skill	✅ 0.14	❌ [4]
analyzing-dotnet-performance	Detects CurrentCulture comparer and compiled regex budget in inflection rules	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [5]
analyzing-dotnet-performance	Finds per-call Dictionary allocation not hoisted to static	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [6]
analyzing-dotnet-performance	Catches compound allocations in recursive number converter with ToLower	1.0/5 → 1.3/5 🟢	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [7]
analyzing-dotnet-performance	Finds StringComparison.Ordinal missing and FrozenDictionary opportunities	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [8]
analyzing-dotnet-performance	Detects Aggregate+Replace chain and struct missing IEquatable	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [9]
analyzing-dotnet-performance	Finds branched Replace chain in format string manipulation	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [10]
analyzing-dotnet-performance	Catches LINQ on hot-path string processing and All(char.IsUpper)	1.0/5 → 1.0/5	1.0/5 → 1.0/5	✅ analyzing-dotnet-performance; tools: glob, skill / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [11]
analyzing-dotnet-performance	Detects LINQ pipeline in TimeSpan formatting and collection processing	1.0/5 → 1.0/5	1.0/5 → 1.0/5	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [12]
analyzing-dotnet-performance	Flags Span inconsistencies and compound method chains in truncation library	1.3/5 → 1.0/5 🔴	1.3/5 → 1.0/5 🔴	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌
analyzing-dotnet-performance	Identifies unsealed leaf classes and locale hierarchy patterns	1.0/5 → 1.0/5	1.0/5 → 1.3/5 🟢	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [13]
analyzing-dotnet-performance	Optimize manual min/max with TensorPrimitives	1.0/5 ⏰ → 1.0/5	1.0/5 ⏰ → 1.0/5 ⏰	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌ [14]
analyzing-dotnet-performance	Optimize manual product with TensorPrimitives	1.0/5 ⏰ → 1.0/5 ⏰	1.0/5 ⏰ → 2.0/5 ⏰ 🟢	✅ analyzing-dotnet-performance; tools: skill / ✅ analyzing-dotnet-performance; tools: skill, edit	✅ 0.14	✅
analyzing-dotnet-performance	No optimization opportunity — dictionary-based lookup service	1.0/5 → 1.0/5 ⏰	1.0/5 → 1.0/5 ⏰	✅ analyzing-dotnet-performance; tools: edit, create, skill / ✅ analyzing-dotnet-performance; tools: create, skill	✅ 0.14	❌ [15]
analyzing-dotnet-performance	Optimize int array conditional increment with SIMD	4.7/5 ⏰ → 3.3/5 ⏰ 🔴	4.7/5 ⏰ → 3.7/5 ⏰ 🔴	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌
analyzing-dotnet-performance	Optimize byte buffer bit reversal with SIMD	3.7/5 ⏰ → 1.0/5 ⏰ 🔴	3.7/5 ⏰ → 1.0/5 ⏰ 🔴	⚠️ NOT ACTIVATED / ⚠️ NOT ACTIVATED	✅ 0.14	❌

[1] (Plugin) Quality improved but weighted score is -8.8% due to: judgment, quality
[2] (Isolated) Quality improved but weighted score is -10.0% due to: tokens (11654 → 105134), tool calls (0 → 6), time (8.8s → 39.4s)
[3] (Isolated) Quality improved but weighted score is -46.7% due to: judgment, quality
[4] (Isolated) Quality unchanged but weighted score is -13.2% due to: judgment, tokens (35194 → 40879)
[5] (Plugin) Quality unchanged but weighted score is -1.3% due to: tokens (34971 → 43307)
[6] (Plugin) Quality unchanged but weighted score is -0.8% due to: tokens (34933 → 38931)
[7] (Plugin) Quality unchanged but weighted score is -5.7% due to: tokens (23092 → 39001), tool calls (2 → 3), time (12.2s → 17.1s)
[8] (Isolated) Quality unchanged but weighted score is -0.8% due to: tokens (34917 → 40485)
[9] (Plugin) Quality unchanged but weighted score is -0.1% due to: tokens (34957 → 38958)
[10] (Isolated) Quality unchanged but weighted score is -3.3% due to: tokens (38898 → 52988), tool calls (3 → 4), time (13.4s → 16.7s)
[11] (Isolated) Quality unchanged but weighted score is -7.7% due to: tokens (34909 → 60848), tool calls (3 → 6), time (11.9s → 19.0s)
[12] (Isolated) Quality unchanged but weighted score is -1.4% due to: tokens (42920 → 48986), time (17.5s → 22.0s)
[13] (Isolated) Quality unchanged but weighted score is -0.3% due to: efficiency metrics
[14] (Plugin) Quality unchanged but weighted score is -14.1% due to: errors (0 → 1), tokens (109364 → 204707), time (71.7s → 180.1s), tool calls (9 → 17)
[15] (Isolated) Quality unchanged but weighted score is -15.0% due to: tokens (47249 → 196950), errors (0 → 1), tool calls (4 → 16), time (20.2s → 127.0s)

⏰ timeout — run hit the scenario timeout limit; scoring may be impacted by aborting model execution before it could produce its full output

Model: claude-opus-4.6 | Judge: claude-opus-4.6

Full results

[Copilot]

Pull request overview

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

---

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

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.

---

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

[jeffschwMSFT]

/evaluate

[github-actions]

Skill Validation Results

Skill	Scenario	Quality (Isolated)	Quality (Plugin)	Skills Loaded	Overfit	Verdict
exp-simd-vectorization	Optimize manual min/max with TensorPrimitives	1.0/5 ⏰ → 5.0/5 🟢	1.0/5 ⏰ → 5.0/5 🟢	✅ exp-simd-vectorization; tools: skill, glob, create, bash, edit / ✅ exp-simd-vectorization; tools: skill, glob, create, edit, bash	🟡 0.23	✅
exp-simd-vectorization	Optimize manual product with TensorPrimitives	1.0/5 → 5.0/5 🟢	1.0/5 → 5.0/5 🟢	✅ exp-simd-vectorization; tools: skill, glob, create, bash / ✅ exp-simd-vectorization; tools: skill, glob, create, bash	🟡 0.23	✅
exp-simd-vectorization	No optimization opportunity — dictionary-based lookup service	4.3/5 → 4.7/5 🟢	4.3/5 → 5.0/5 🟢	✅ exp-simd-vectorization; tools: skill / ✅ exp-simd-vectorization; tools: skill	🟡 0.23	✅
exp-simd-vectorization	Optimize int array conditional increment with SIMD	4.0/5 → 3.7/5 🔴	4.0/5 → 4.0/5	✅ exp-simd-vectorization; tools: skill / ✅ exp-simd-vectorization; tools: skill, glob	🟡 0.23	❌ [1]
exp-simd-vectorization	Optimize byte buffer bit reversal with SIMD	2.0/5 ⏰ → 4.7/5 🟢	2.0/5 ⏰ → 4.0/5 🟢	✅ exp-simd-vectorization; tools: skill, edit, glob, bash / ✅ exp-simd-vectorization; tools: skill, edit, bash	🟡 0.23	✅
analyzing-dotnet-performance	Detects compiled regex startup budget and regex chain allocations	2.7/5 ⏰ → 3.7/5 ⏰ 🟢	2.7/5 ⏰ → 3.0/5 ⏰ 🟢	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	❌ [2]
analyzing-dotnet-performance	Detects CurrentCulture comparer and compiled regex budget in inflection rules	5.0/5 → 4.0/5 ⏰ 🔴	5.0/5 → 5.0/5	✅ analyzing-dotnet-performance; tools: skill, bash, grep / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	❌ [3]
analyzing-dotnet-performance	Finds per-call Dictionary allocation not hoisted to static	5.0/5 → 5.0/5	5.0/5 → 3.7/5 ⏰ 🔴	✅ analyzing-dotnet-performance; tools: skill, bash, grep / ✅ analyzing-dotnet-performance; tools: skill, bash, write_bash, grep	✅ 0.16	❌
analyzing-dotnet-performance	Catches compound allocations in recursive number converter with ToLower	3.7/5 → 4.3/5 🟢	3.7/5 → 4.7/5 🟢	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	✅
analyzing-dotnet-performance	Finds StringComparison.Ordinal missing and FrozenDictionary opportunities	4.7/5 → 5.0/5 🟢	4.7/5 → 5.0/5 ⏰ 🟢	✅ analyzing-dotnet-performance; tools: skill, grep, bash / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	❌ [4]
analyzing-dotnet-performance	Detects Aggregate+Replace chain and struct missing IEquatable	4.0/5 → 4.3/5 ⏰ 🟢	4.0/5 → 2.7/5 ⏰ 🔴	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash, write_bash, stop_bash	✅ 0.16	❌
analyzing-dotnet-performance	Finds branched Replace chain in format string manipulation	1.3/5 ⏰ → 3.3/5 🟢	1.3/5 ⏰ → 3.7/5 🟢	✅ analyzing-dotnet-performance; tools: skill, bash, grep / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	✅
analyzing-dotnet-performance	Catches LINQ on hot-path string processing and All(char.IsUpper)	4.3/5 → 4.7/5 🟢	4.3/5 → 5.0/5 🟢	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	✅
analyzing-dotnet-performance	Detects LINQ pipeline in TimeSpan formatting and collection processing	5.0/5 → 4.0/5 🔴	5.0/5 → 3.7/5 🔴	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash	✅ 0.16	❌
analyzing-dotnet-performance	Flags Span inconsistencies and compound method chains in truncation library	4.3/5 → 4.0/5 🔴	4.3/5 → 4.7/5 🟢	✅ analyzing-dotnet-performance; tools: skill, bash / ✅ analyzing-dotnet-performance; tools: skill, bash, write_bash, stop_bash, read_bash	✅ 0.16	✅
analyzing-dotnet-performance	Identifies unsealed leaf classes and locale hierarchy patterns	3.3/5 ⏰ → 4.3/5 ⏰ 🟢	3.3/5 ⏰ → 3.0/5 ⏰ 🔴	✅ analyzing-dotnet-performance; tools: skill, bash, read_bash, stop_bash, grep / ✅ analyzing-dotnet-performance; tools: skill, bash, read_bash, stop_bash	✅ 0.16	❌ [5]

[1] (Plugin) Quality unchanged but weighted score is -7.4% due to: quality, tokens (57025 → 83232), tool calls (5 → 6)
[2] (Plugin) Quality improved but weighted score is -62.5% due to: judgment, quality, tool calls (6 → 11), tokens (87694 → 103331)
[3] (Plugin) Quality unchanged but weighted score is -11.2% due to: tokens (28705 → 139554), tool calls (2 → 11), time (40.1s → 91.1s), quality
[4] (Plugin) Quality improved but weighted score is -5.1% due to: tokens (28967 → 132919), tool calls (2 → 11), time (39.7s → 104.5s)
[5] (Isolated) Quality improved but weighted score is -48.4% due to: judgment, quality, errors (0 → 1), tool calls (8 → 14), time (73.3s → 112.1s)

⏰ timeout — run(s) hit the (120s, 180s) scenario timeout limit; scoring may be impacted by aborting model execution before it could produce its full output (increase via timeout in eval.yaml)

Model: claude-opus-4.6 | Judge: claude-opus-4.6

📖 See InvestigatingResults.md for how to diagnose failures. Additional debugging guidance may be provided by your workflow.

Full results

To investigate failures, paste this to your AI coding agent:

Download eval artifacts with gh run download 23552325478 --repo dotnet/skills --dir /tmp/eval-results, then fetch https://raw.githubusercontent.com/dotnet/skills/4ec893eae1c36a42f48b2aa8f61138a4e91041c7/eng/skill-validator/InvestigatingResults.md and follow it to analyze the results.json files. Diagnose each failure, suggest fixes to the eval.yaml and skill content, and tell me what to fix first.