@vkrasnov This is the tool I used to review #1702. The tool existed in a different form, and I modified it to produce (almost) identical output to that PR. My plan is to use this tool to help with the P-521 work and to bulk rewrite all of src/ec/suite_b.

Codecov Report

Merging #1756 (3094b9d) into main (3565fe8) will increase coverage by 0.11%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1756      +/-   ##
==========================================
+ Coverage   95.90%   96.01%   +0.11%     
==========================================
  Files         132      138       +6     
  Lines       15102    20788    +5686     
  Branches      169      226      +57     
==========================================
+ Hits        14483    19959    +5476     
- Misses        598      792     +194     
- Partials       21       37      +16     

... and 28 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

Yes, I generated the params with python as well, they are the same, not sure why my code doesn't pass the 32bit tests, since code is reused from p384. I was unable to get a 32bit dev enviroment unfortunately.

I updated this PR substantially so that it can generate everything in in p384.rs except the addition chains.

I also extended the P-521 support so that it does the right thing for the 32-bit vs. 64-bit Montgomery stuff; it does the #[cfg(...)] stuff slightly differently than your P-521 PR does it. See the second commit in this PR. I verified the numbers match your numbers for both 64-bit and 32-bit.

Yes, I generated the params with python as well,

I didn't extend the code generator to generate gfp_p{384,521}.c. If you have Python code that can generate the constants in the C code, then could you please submit it as a change to this Python script?