This looks like a fun project, but claiming that it “beats luajit” is very premature based on the benchmarks listed.
The README lists only an allocation stress test (fast allocs are good, to be sure) and recursive Fibonacci (always good fun to try but doesn’t resemble real usage).
Looks like it doesn’t have a JIT compiler, and without that I’d put a lot of money against it ever being as fast as LuaJIT, V8, the JVM, etc etc in the general case.
Edit to add: on reading more closely, there’s more benchmark-gaming going on here than I realised. On the second benchmark, it says:
As you can see, minivm (no JIT) is a hair slower than Node JS (JIT) but beats luajit --joff by a fair margin (no JIT).
They already noted that Node has a longer startup time, so that obviously dominates in a benchmark that’s only a few milliseconds long; and disabling the JIT in LuaJIT doesn’t tell us much.
The README lists only an allocation stress test (fast allocs are good, to be sure) and recursive Fibonacci (always good fun to try but doesn’t resemble real usage).
Looks like it doesn’t have a JIT compiler, and without that I’d put a lot of money against it ever being as fast as LuaJIT, V8, the JVM, etc etc in the general case.
Edit to add: on reading more closely, there’s more benchmark-gaming going on here than I realised. On the second benchmark, it says:
As you can see, minivm (no JIT) is a hair slower than Node JS (JIT) but beats luajit --joff by a fair margin (no JIT).
They already noted that Node has a longer startup time, so that obviously dominates in a benchmark that’s only a few milliseconds long; and disabling the JIT in LuaJIT doesn’t tell us much.