Replacing a Rust Enum with a 64-bit Word Made My Interpreter 17% Faster
<...> That being said, sha256_unfixed is still running 12% faster with the new value representation than with the old version, even though it's allocating a total of 3,149,052 boxed integer values in a span of about 490 ms. That seemed a bit crazy to me, but having a copying GC with a bump allocator means that we can allocate objects really fast.