Optimize vote distribution between PRs and non-PRs

[gr0vity-dev]

Currently, PRs receive votes both from flood_vote and flood_vote_pr, leading to duplicate voting, while fewer non-PR receive votes. This is particularly noticeable in small test networks.

This change ensures votes are distributed more efficiently:

• PRs receive votes only through flood_vote_pr
• Non-PRs receive votes through flood_vote_non_pr
• Overall outgoing vote load remains unchanged

Here is the difference seen in a local test network:

I included a small optimization to nano::network::list_non_pr since it's called often.

Test Configuration & Benchmark Results:
PR Count: 100
Non-PR Count: 200
Iterations: 10000

--------------------------------------------------
Current method   - Duration: -- 521 ms
Partition method - Duration: -- 354 ms
--------------------------------------------------

Target count per iteration: 29 [ceil( 2* sqrt(200))]
Actual Current method:      29
Actual Partition method:    29

[gr0vity-dev-bot]

Test Results for Commit 5bc04e8

Pull Request 4766: Results
Overall Status: ✅

Test Case Results

• ✅ 5n4pr_conf_10k_bintree: PASS (Duration: 110s)
• ✅ 5n4pr_conf_10k_change: PASS (Duration: 212s)
• ✅ 5n4pr_conf_change_dependant: PASS (Duration: 149s)
• ✅ 5n4pr_conf_change_independant: PASS (Duration: 158s)
• ✅ 5n4pr_conf_send_dependant: PASS (Duration: 136s)
• ✅ 5n4pr_conf_send_independant: PASS (Duration: 140s)
• ✅ 5n4pr_rocks_10k_bintree: PASS (Duration: 109s)
• ✅ 5n4pr_rocks_10k_change: PASS (Duration: 219s)

Last updated: 2024-10-24 20:55:44 UTC

[pwojcikdev]

Looks good. Do I understand correctly that before votes were being sent twice to PRs?

[gr0vity-dev]

Looks good. Do I understand correctly that before votes were being sent twice to PRs?

Not systematically but they were sent:

• to all PRs in flood_vote_pr
• to 2 * sqrt of all peers (including PRs) in flood_vote

This means that some PRs might get the same votes twice