Mario Rugiero
Mario Rugiero
I'm planning on implementing a feature to do this but I'd like early feedback on the design. The idea is to use a heap such as in the `roaring_bitmap_or_many_heap` implementation...
Our current hint handling has a few issues: - It exposes the whole VM state, limiting how much of the inner workings of the VM we can change further down...
We're facing a few limitations with running in GitHub-hosted workers, namely: - Benchmarks' results are less stable due to unpredictable environment; - Builds are slower due to inability to cache...
Cairo programs tend to rely heavily on recursion. A common optimization in languages with similar behavior is converting tail calls into jumps. It's not certain that it makes sense or...
Research whether there's a bound to the prime defining the field and use that to check whether we can use a smaller big integer implementation that can be allocated on...
We're currently using `BigInt` explicitly, which forces us to compute the `.mod_floor` of results after every operation, use costly divisions, etc. The first step to fix arithmetic performance issues is...
A significant part of runtime is spent in performing modular arithmetic. These operations can be greatly optimized by using [Montgomery form](https://en.wikipedia.org/wiki/Montgomery_modular_multiplication) to represent our fields or by using [Barrett reduction](https://en.wikipedia.org/wiki/Barrett_reduction)...
Currently our benchmark pipeline only posts comments when there's a significant performance regression. It would be useful for it to also post in all other cases, specially so we can...
## Summary Make the compilation of Python code to bytecode (or at least the parsing into AST) usable in constant contexts. ## Detailed Explanation The compilation to bytecode should be,...