philihp / pcg / 26122724051

99%

main: 100%

Pull #336

github

perf: add BigInt-free fast variant (SETSEQ + XSH_RR)

Adds a parallel `fast*` API that runs the PCG32 hot path on plain
number arithmetic instead of BigInt. State is carried as two uint32
halves and 64-bit multiply uses a 16-bit-limb split so partial
products stay inside 2^53.

The fast variant commits to the default scheme (SETSEQ) and output
function (XSH_RR), letting it skip all per-step dispatch. Output
matches the BigInt path for that configuration; tests assert
equivalence across many seeds.

Microbench (1M iterations, node 22): ~27-29x faster than the BigInt
hot path for both nextState and randomUint32.

51 of 54 branches covered (94.44%)

Branch coverage included in aggregate %.

143 of 143 new or added lines in 3 files covered. (100.0%)

256 of 256 relevant lines covered (100.0%)

6593.11 hits per line