ORBit speed measure.

[Xavier Ordoquy <xordoquy aurora-linux com>]

Someone said:
> There is a large perception of CORBA being slow, but for the most part it
> is unjustified.  

Well, I've measured using function calls through ORBit is 300 times slower
than using dynamic loading.

I've made three sample programms that calls an empty function 1000000
times. It tooks:

  - For the ORBit programme: 99s 990ms

            client      server
    user   25s 090ms   25s 570ms
    sys    25s 930ms   23s 400ms

  - For the ORBit initialisation (to see the start/end of ORBit
time): 50ms

            client    server
    user     20ms     20ms
    sys       0ms     10ms

 - for the dynamic loading: 290ms
    user  280ms, sys 10ms

That gives me about 0.1  ms par function called throught ORBit
                    0.29 ns ---------------------------- dl

Which gives me the dl is about 333 times faster than ORBit.

BUT those tests are falshed because of the CPU memory cache.
I did not took the time to make random calls to many functions
so dl, since it's less code, could be handled in CPU caches...

However, the number of 10,000 calls through ORBit seems realistic.
I've just noticed that they were lots of disk access during the test.
I've check that it wasn't the cache. So what could it be ?
May be it is what slowed the process down ?

Xavier Ordoquy