otland
Optimize XTEA encryption
Pushing the same code present in TFS. Tired of seeing this being sold in plain sight :stuck_out_tongue_closed_eyes:
"This patch enables auto-vectorization for XTEA encryption and decryption, exploring vector (SSE, AVX) instructions to be auto-enabled by the compiler. Should render up to 16x faster encryption/decryption throughput on AVX2, 8x on AVX and 4x on SSE/Neon."
"Out with all that optimal block size width, let the compiler figure it out. Reduces the amount of code by half and increases performance by ~20%."
Let's hold this, there is a possible optimization in otland/forgottenserver#3406
@ranisalt since #3406 was merged, do you want to close this or there's still something useful here?
This was already merged upstream, so I suggest we rebase and then just add the precomputed key changes