Alright, where do I sign these forms…
Forms are after you send an email to GHI telling them why you need RLP… Again RLP may not be the right choice for your design.
Well, I am doing a project soon that requires several filters and I am a professional coder, so I might look into RLP on my Cobra instead of using a coprocessor for filters and things. 
Assuming that the precision of the ADC is sufficient, with RLP could a Cobra do 6000 signal samples per second and a FFT and stay real-time?
If not, ChipX?
This is exactly why RLP is left out. A user must study this very well.
RLP is native code…you need FFT, then sure you can do it. But, what kind of FFT can be done on 72Mhz ARM processor? GHI doesn’t have experience with FFT. Maybe an assembly-ARM-targeted FFT would be VERY fast, maybe not!
Also, can I do 6000 samples? Maybe, this info is in the processor datasheet/manual, LPC2478.
When you use RLP, you are running directly on the processor and doping things directly the way you like them. GHI doesn’t know or have control over what you do. I hope I was able to explain it without scaring you from using RLP. This is a great feature but you must know exactly what you are getting into.
I did a similar system on a PC with .NET and C# at a higher sampling frequency and in a feedback loop.
The LPC2478 datasheet talks about a 10-bit conversion time of >= 2.44 microseconds. I think 6000 samples a second looks possible.
I have to read up more on RLP. I have no immediate needs for it, but it sounds like something to know about.
My feelings are the same.
As for “can it do samples at X hz?”: I think it depends on the quality of your code.
We will write up a document to cover this in details.
I understand that there may be things that are outside of GHI’s experience; this includes Fast Fourier Transforms, and that’s ok. I think that we understand the gravity of doing c++ applications that run “directly” on the processor as this could cause physical damage to the cpu if done incorrectly (yes peeps you can damage the cpu using this). Anyone using these functions should understand that this is “out of the scope” of the device and that if you run these type of applications that they are “at your own risk”; i.e. we may not be able to support you.
Most people who do this type of development will understand this completely. Most people who are beginners won’t. GHI’s stance in this matter is valid, but they should make ppl aware of how they can get “permission” to do this kind of work.
And who knows, if someone writes something kewl like Native FFT’s it could even become part of the GHI managed namespace; extending the usefulness of the EMX even further 