The generated assembly looks more optimized. In addition, this avoids widened load, which would cause a TSan-detected data race with parallel --gc-sections (#189321).
64 KiB
64 KiB