Simulation delegate overhead?

[dsharlet]

@Federerer mentions he found the simulation delegate to have significant overhead.

[dsharlet]

I calculated the overhead by varying the block size of the benchmark on a medium size circuit:

Samples/Block	Rate (kHz)
2^6	210
2^10	215
2^12	215

Seems like the overhead is 2-3% when using 64 samples/block (1.4ms at 44.1 kHz). 64 samples/block is pretty small (though I know people do use this setting) and 2-3% overhead isn't bad, so I think this overhead is negligible.

[Federerer]

I haven't said that the overhead is significant, just that there is some and it's related to the usage of DynamicInvoke() ;) Also, creating array for parameters adds another memory allocation to the audio signal path. It looks like it's unnecessary in this case. I've created a pull request with my changes.

[dsharlet]

Oh, I see now, what you are proposing is also better for type safety anyways even if the overhead isn't huge. Thanks for the PR, I'll take a look now.