mirror of
https://codeberg.org/ziglang/zig
synced 2026-11-18 07:05:05 +00:00
Tests heavy on comptime arithmetic show a ~7% improvement in wall-clock time, and a ~11% improvement in peak RSS. There are two optimizations: 1. Avoid interning the overflow bit Previously, the overflow bit was being interned and immediately read back. Instead, a new OverflowArithmeticResultInt type is introduced that contains the overflow result as a boolean directly. When converted, that boolean points to the already existing zero_u1 and one_u1 interned values. 2. Try arithmetic operations without BigInt first. This did not show significant performance improvements, but it did show a significant memory usage improvement, almost going 0.5x! Instead of resorting to BigInt first, arithmetic operations are tried in a u64. Reviewed-on: https://codeberg.org/ziglang/zig/pulls/36495 Reviewed-by: mlugg <mlugg@mlugg.co.uk>
2580 lines
105 KiB
2580 lines
105 KiB