What is Algomania?

Algomania is the fastest CPU-only vanity address generator for Algorand. It utilizes libsodium for ed25519 keypair generation and produces Algorand account mnemonics only when public key matches are found. For those wanting elite 8-char vanity addresses, using Algomania is essential. See the performance table for tested keygen speeds.

How to Build & Use

1. System Requirements (Cross-Platform supported)

   • .NET Core
   • libsodium - Windows / Mac / RHEL / Ubuntu

2. git clone https://github.com/kirse/algomania

3. Update the DllImport path in line 12 of Main.fs to point to your libsodium library (.dll, .so, etc.)

   • Recommended to put the libsodium .dll/.so (etc) in the same folder
   • If Windows, copy libsodium.dll next to this program and no edits will be needed

4. dotnet run -c Release

Optimizing Chances of Winning

Given the math, the best "bang for your buck" is an 8-char vanity prefix (if you have the hardware). A byte array of length mod 5 = 0 exactly specifies a desired Base32 string. What this means is a 7-char prefix (4 bytes of chance 1/256) * (1 "pinning" byte of chance 1/32) is about as hard to generate as an 8-char prefix (5 bytes of chance 1/256).

Any vanity prefixes that are 1-3 bytes are easily churnable on commodity hardware. Vanity prefixes of 4-5 bytes should use high performance many-core CPUs. Prefixes of 6+ bytes is uncharted space and not recommended. 6+ bytes is likely possible with an ed25519 GPU-focused solution (stay tuned).

Performance

To contribute figures, please PR with a runtime screenshot.

CPU	RAM	Keys/Sec	Test Notes	Platform
AMD EPYC 7443P - 2.85/4.0 GHz - 24C/48T	256GB	1.16M	11h:26m / 48b tries	Rocky8
Intel E-2288G - 4.0/4.9 GHz - 6C/12T	32GB	304K	10h:58m / 12b tries	Rocky8
Intel Core i7-7500U - 2.7/2.9 Ghz - 2C/4T	16GB	31K	8h:30m / 1b tries	Windows

Misc Links

• Base32 Converters
  • https://cryptii.com/pipes/base32-to-hex
  • https://tomeko.net/online_tools/base32.php?lang=en

Some Historical Notes

(On trying to generate keys faster)

• First cut is ~1:50/M addresses - 9k/s
• 2nd - remove .NET AlgoSDK, use Bouncy Castle direct = 1:32/M - 10.8k/s
• 3rd - separate fsharp = 1:12/M - 13.8k/s
• 4th - use Array and Array.Parallel.iter = 0:52/M (19.1k/s), 8h/500M (17.3k/s)
• 4th - fresh reboot - 7:35:00/500M (18.3k/s)
• 5th - used .NET TPL - 0:33/M - 30k/s (4 Tasks), 6:50/10M (24.4k/s), 9:21:00/800M (23.8k/s)
• 6th - refactor using TPL + libsodium for keygen - 0:24/M (41k/s), 2:18/5M (36k/s), 8:30/1B (33k/s) dotnet run -c Release
• 7th - tried span<byte> + stackalloc refactor but same times as round 6
• 8th - cloud 3cpu/9-core box - 0:07/M (143k/s)
• 9th - cloud Bare Metal 6cpu/12-core box -
  • 0:3.8s/1.2M (315k/s), 37.6s/12M (320k/s), 394s/120M (304k/s), 10h58m/12B (304k/s) = (26Bk/Day!!)
• 10th - cloud Bare Metal 24cpu/48-core box - 11:26:30/48B (1.16M k/s = 100Bk/Day)
  • Note: Successful generation of multiple 8-char vanity addresses on this config