Update TODO.txt

TODO.txt

@@ -1,6 +1,7 @@
 * Implement the contribution of a parallel implementation that speeds up 1.5 to 3x in pure python (but requires a rewriting of the module to be self-contained, no globals, to be compatible with the multiprocessing module)
 	* Maybe test and/or make a version compatible with the Mojo JIT compiler (a superset of Python), but it's only partially opensource so this version cannot replace the main one due to the constraint of being future-proof. https://www.macg.co/ailleurs/2024/04/le-code-source-de-mojo-bien-ete-rendu-public-comme-promis-143289 -- the huge speed-up is obtainable through parallelization, so it would be a step after the parallel implementation in pure python https://www.modular.com/blog/mojo-a-journey-to-68-000x-speedup-over-python-part-3 and https://github.com/modularml/mojo/discussions/843
 	* Test Photon JIT compiler, but not a drop-in replacement for CPython but tries to be as close as possible. Also requires to be parallelism-ready. https://github.com/exaloop/codon
+        * For reference, we can compare llama.mojo which is a port of llama.cpp by the same author as llama.py which is compatible with codon https://www.modular.com/blog/community-spotlight-how-i-built-llama2-by-aydyn-tairov and https://github.com/tairov/llama2.py and https://github.com/tairov/llama2.mojo
 * Test the new [galois](https://github.com/mhostetter/galois) jit-optimized extension module for numpy, which implements a reed-solomon codec and NTT transforms!
 * Implement my own code for probabilistic prime polynomials generator, because for higher fields it is necessary: https://en.wikiversity.org/wiki/Reed%E2%80%93Solomon_codes_for_coders/Additional_information#Probabilistic_prime_polynomial_generator
 * Cythonize more, read recent tutorials: