It has nothing to do with pymalloc's performance per se. Rather, the performance issue only occurs when using `rep movsb` on AMD CPUs with certain page/data al

It has nothing to do with pymalloc's performance per se. Rather, the performance issue only occurs when using `rep movsb` on AMD CPUs with certain page/data alignment. Pymalloc just happens to be using page/data alignment that makes `rep movsb` happy while Rust's default allocator is using alignments that just happen to make `rep movsb` sad.