Fix/fix convert float to fp16 warning

[bluemiao3]

Fix: eliminate all float-to-fp16 compilation warnings

[tencent-adm]

All committers have signed the CLA.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 88.67925% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.94%. Comparing base (ce2c565) to head (2857988).
⚠️ Report is 9 commits behind head on master.

Files with missing lines	Patch %	Lines
src/layer/riscv/interp_bicubic_packn_fp16s.h	0.00%	10 Missing ⚠️
src/layer/riscv/interp_bilinear_packn_fp16s.h	0.00%	3 Missing ⚠️
src/layer/riscv/convolution1d_riscv_zfh.cpp	71.42%	2 Missing ⚠️
src/layer/riscv/interp_riscv_zfh.cpp	0.00%	2 Missing ⚠️
src/layer/riscv/deconvolution_packnto1_fp16s.h	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6525      +/-   ##
==========================================
- Coverage   93.08%   92.94%   -0.14%     
==========================================
  Files         808      809       +1     
  Lines      256136   257088     +952     
==========================================
+ Hits       238416   238963     +547     
- Misses      17720    18125     +405     

☔ 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.

[Copilot]

Pull request overview

This pull request aims to eliminate float-to-fp16 compilation warnings by adding explicit (__fp16) casts throughout RISC-V specific layer implementations. The changes target various operations including interpolation, convolution, deconvolution, instance normalization, and other neural network operations.

Changes:

• Added (__fp16) casts to float literals and variables when passing them to fp16 intrinsics
• Cast zero initializers for fp16 variables and vectors
• Cast alpha coefficients in interpolation operations
• Cast bias values and other scalar parameters in various operations

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 15 comments.

Show a summary per file

File	Description
src/layer/riscv/interp_riscv_zfh.cpp	Added casts for alphap coefficients in bilinear and bicubic interpolation
src/layer/riscv/interp_bilinear_packn_fp16s.h	Partially added casts for alphap coefficients (incomplete)
src/layer/riscv/interp_bicubic_packn_fp16s.h	Partially added casts for alphap coefficients (incomplete)
src/layer/riscv/instancenorm_riscv_zfh.cpp	Added casts for zero initializers and epsilon values
src/layer/riscv/innerproduct_riscv_zfh.cpp	Added casts for zero initializers in vector operations
src/layer/riscv/gru_riscv_zfh.cpp	Refactored sigmoid computation casts
src/layer/riscv/deconvolution_*_fp16s.h	Added casts for zero initializers
src/layer/riscv/convolutiondepthwise_3x3.h	Incorrectly added casts for float bias variables
src/layer/riscv/convolution_winograd_*.h	Added casts for various coefficients and initializers
src/layer/riscv/convolution_sgemm*.h	Incorrectly added casts for float bias variables and arrays
src/layer/riscv/convolution_*_fp16s.h	Added casts for zero initializers and coefficients
src/layer/riscv/convolution1d_riscv_zfh.cpp	Added casts for val parameters in widen-multiply operations
src/layer/riscv/celu_riscv_zfh.cpp	Added casts for alpha and scalar constants
src/layer/riscv/bias_riscv_zfh.cpp	Added cast for bias data access

---

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

[nihui]

see https://github.com/Tencent/ncnn/actions/runs/21436806039/job/61825161598?pr=6525

[nihui]

/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/deconvolution_packnto1_fp16s.h: In function 'void ncnn::deconvolution_packnto1_fp16sa_rvv(const Mat&, Mat&, const Mat&, const Mat&, int, int, int, int, int, int, int, const Mat&, const Option&)':
/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/deconvolution_packnto1_fp16s.h:186:36: warning: ISO C++ does not allow converting to '_Float16' from 'float' with greater conversion rank [-Wnarrowing]
  186 |                 sum = activation_ss(sum, activation_type, activation_params);
      |                       ~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/gru_riscv_zfh.cpp: In function 'int ncnn::gru_fp16sa(const Mat&, Mat&, int, const Mat&, const Mat&, const Mat&, Mat&, const Option&)':
/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/gru_riscv_zfh.cpp:586:30: warning: ISO C++ does not allow converting to '_Float16' from 'float' with greater conversion rank [-Wnarrowing]
  586 |             output_data[q] = H;
      |                              ^

/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/convolution1d_riscv_zfh.cpp: In member function 'int ncnn::Convolution1D_riscv::forward_fp16sa(const ncnn::Mat&, ncnn::Mat&, const ncnn::Option&) const':
/data/action/runner0/_work/ncnn/ncnn/src/layer/riscv/convolution1d_riscv_zfh.cpp:474:40: warning: ISO C++ does not allow converting to '_Float16' from 'float' with greater conversion rank [-Wnarrowing]
  474 |                     sum = activation_ss(sum, activation_type, activation_params);
      |                           ~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[nihui]

Thanks for your contribution !