> the JVM often competes in language performance closer to C++
I don't think it's even close. Benchmark game is one (biased) reference. And it's even worse because optimized Java is not canonical Java. Canonical C++ is pretty fast.
I don't think it's even close. Benchmark game is one (biased) reference. And it's even worse because optimized Java is not canonical Java. Canonical C++ is pretty fast.
I agree.
1. Benchmarks tend to do one thing, then repeat that one thing 25000 times. Non-server programs don't do that[1] so you don't get to amortise the time spent on JITting a code snippet across 25000 executions.
2. The runtime basically requires exclusive use of a core for short periods (GC, JIT). If you're running on a system that costs per core, you're using (worst case scenario) an extra core for each Java program. The benchmarks happily give multiple cores even to sequential (non-parallel) benchmarks.
[1] Even many server programs don't do that.
https://benchmarksgame-team.pages.debian.net/benchmarksgame/...
2. The benchmarks game happily gives multiple cores and shows that "cpu secs" measurement:
https://benchmarksgame-team.pages.debian.net/benchmarksgame/...
2) "one (biased) reference" is simply name calling.
3) What specifically are you saying is "biased" about these measurements?
https://benchmarksgame-team.pages.debian.net/benchmarksgame/...