RVV1.0 Dequantize Layer

[Deepdive543443]

Road to int8 optimization Episode 2: Dequantize Layer

[Deepdive543443]

All "elempack == vlm2" checks are added to bypass the "FORCE_INPUT_PACK8" unit test. It seems work, but I wonder, in RVV, when we will have pack8 input for dequantize layer? Isn't Net Module always packing elements in single vector length?

[nihui]

All "elempack == vlm2" checks are added to bypass the "FORCE_INPUT_PACK8" unit test. It seems work, but I wonder, in RVV, when we will have pack8 input for dequantize layer? Isn't Net Module always packing elements in single vector length?

yes, always use the native packn. packn can be pack4 pack8 pack16 etc .

[codecov-commenter]

Codecov Report

❌ Patch coverage is 97.45223% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.96%. Comparing base (71b1a61) to head (ca3e8c2).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
src/layer/riscv/dequantize_riscv_zfh.cpp	96.20%	3 Missing ⚠️
src/layer/riscv/dequantize_riscv.cpp	98.71%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           master    #6658    +/-   ##
========================================
  Coverage   93.96%   93.96%            
========================================
  Files         932      934     +2     
  Lines      299059   299395   +336     
========================================
+ Hits       280998   281321   +323     
- Misses      18061    18074    +13     

☔ 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.