vulkan int8 kernels

[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 81.27871% with 489 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.78%. Comparing base (b16501a) to head (cd08386).

Files with missing lines	Patch %	Lines
src/layer/vulkan/convolution_vulkan.cpp	77.21%	206 Missing ⚠️
src/layer/vulkan/gemm_vulkan.cpp	62.01%	158 Missing ⚠️
src/layer/vulkan/padding_vulkan.cpp	84.75%	50 Missing ⚠️
src/gpu.cpp	70.58%	25 Missing ⚠️
src/layer/vulkan/convolutiondepthwise_vulkan.cpp	96.44%	16 Missing ⚠️
src/layer/vulkan/innerproduct_vulkan.cpp	97.11%	9 Missing ⚠️
src/layer/convolution.cpp	40.00%	6 Missing ⚠️
src/layer/x86/convolution_x86.cpp	50.00%	5 Missing ⚠️
src/layer/arm/convolution_arm.cpp	50.00%	3 Missing ⚠️
src/layer/loongarch/convolution_loongarch.cpp	62.50%	3 Missing ⚠️
... and 3 more

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6751      +/-   ##
==========================================
- Coverage   95.78%   94.78%   -1.00%     
==========================================
  Files         946      946              
  Lines      410528   412347    +1819     
==========================================
- Hits       393237   390861    -2376     
- Misses      17291    21486    +4195     

☔ View full report in Codecov by Harness.
📢 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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 46154093ca

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".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fd9a06adcb

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".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ee43f88d03

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Fall back before rejecting dynamic int8 Gemm inputs

When an int8 Gemm has a non-constant A input that is already an 8-bit VkMat (for example fed by a previous requantized int8 layer), this new Vulkan path is selected because the old support_vulkan = false guard was removed, but the first thing it does is return -1. There is no CPU fallback at this point in execution, so extraction fails for a graph shape that the loader now advertises as Vulkan-capable; either keep this case off Vulkan during loading or handle the supplied int8 scale path here.

Useful? React with 👍 / 👎.