"faster than QEMU" isn't a particularly high bar to clear, though. In particular if you're only emulating a single process rather than a full guest OS then you can reduce the overhead quite a bit. Plus if you design something from the start for the single-process emulation case, and only need to worry about one guest and one host architecture, and focus on emulation performance as a key goal, you can likely do better than QEMU without too much trouble.