lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]


> bertie wrote:
> [...]
> 
>>Exactly. I started to compare them. Couldn't do more, question of time.
> 
> 
> Extremely interesting, but I'd recommend checking out The Great Language
> Shootout --- http://shootout.alioth.debian.org. Benchmarking is really a hard.

Some of the Shootout benchmarks scripts are available in the
neko/src/benchs. In general, when meaningful computation is performed
(i.e. not "hello world") then Neko tends to be one of the fastest
dynamicly typed VM.

> From looking at your Fibonacci code, for example, I suspect that what it's
> actually measuring is how long it takes to allocate stack frames; it's calling
> fib() a *lot of times*. This means that the results don't necessarily mean
> anything in the real world.

Calling functions is something that you often do in a program. You tend
to call them a lot of times. Fib benchmark is a benchmark, so it doesn't
cover all the runtime features, but it measures the following :
  * function call overhead
  * integer operations + - <=
That gives you an idea about your program performances if you want to do
a lot of these.

But to be exactly fair, one should run Neko in interpreted mode (using
neko -interp <script>).

> (Incidentally, your bar charts look awfully strange --- how come the high
> numbers have shorter bars than the low numbers?)

I think it's 1/T ratio.

Nicolas