unroll 16 for convolution x86 bf16s avx512

[nihui]

No description provided.

[tencent-adm]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov-commenter]

Codecov Report

❌ Patch coverage is 93.76580% with 74 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.93%. Comparing base (10cee2a) to head (24aefe3).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
src/layer/x86/convolution_im2col_gemm.h	94.62%	42 Missing ⚠️
src/layer/x86/convolution_3x3_winograd.h	92.11%	32 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6680      +/-   ##
==========================================
+ Coverage   93.92%   93.93%   +0.01%     
==========================================
  Files         933      933              
  Lines      300879   303335    +2456     
==========================================
+ Hits       282599   284942    +2343     
- Misses      18280    18393     +113     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[nihui]

@codex review

[Copilot]

Pull request overview

This PR updates x86 AVX512 convolution GEMM kernels to use a wider N-dimension unroll of 16 (vs 12), and adjusts tiling/packing paths accordingly to better match AVX512 register width and improve throughput in im2col+GEMM and Winograd codepaths.

Changes:

• Update AVX512F GEMM microkernels to compute/store 16 output columns per inner loop (new sum registers, updated pointer increments, transpose/store paths).
• Add AVX512F-specific jj+=16 packing/compute loops in several remaining kernels, keeping 12/8/4 fallbacks where applicable.
• Update optimal tile selection to prefer TILE_N = 16 (AVX512F) and TILE_N = 12 (AVX).

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/layer/x86/convolution_im2col_gemm.h	Expands AVX512F im2col+GEMM N-tile unroll to 16 and updates tiling heuristics accordingly.
src/layer/x86/convolution_3x3_winograd.h	Updates Winograd BT packing and GEMM kernels to use 16-wide AVX512F N-tiles and adjusts related tiling.

---

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

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Hooray!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[nihui]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Chef's kiss.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

Copilot reviewed 3 out of 4 changed files in this pull request and generated 1 comment.