low cost open science instrumentation was: beagleBone clone, scaling down features to decrease cost

So to reiterate, I want to design a line of hackable yet robust open
scientific equipment. I think the power and flexibility afforded by
linux is awesome and opens my equipment to a large market, but I
wonder if Angstrom linux can be dependable for controlling low level
control signals for analog peripherals (through GPIO, PWM, or hardware
Timers) as well as 30Mbps of data coming through SPI from fast ADC.
I've been recommended to look into a co-processor to so the signalling
and data collection, and I was just informed about PRUs in the A8
chips. In either case I would still need a fast way to get the data
into linux, so I can serve it over Wifi, ethernet, USB, bluetooth,

I'm currently working with a C2000 Picollo controlStick (TMS320F28069)
and a Stellaris Cortex M4 EKS-LM4F232

So could the beagleBone PRU to do the signalling for this CCD I have
(and similar chips I could potentially expand into using)... 1 Mhz
master clock, another line (SH, shutter) with 30% duty cycle and a
period of 10uS to seconds, and another line that is basically a gate
clear line that is called once to begin a read (all of these have
100ns to 1000ns timing requirements, the SH line continues to cycle as
long as the pixel data is being clocked out). Analog data is clocked
out at 250khz to a 16bit ADC to SPI which I've calculated will need
about 30Mbps throughput (so 30Mhz SPI clock signal)

How would a user space program get the 30Mbps in from SPI, would I
have to setup some kind of DMA (sounds complicated), or would it just
be a memcpy from the SPI register address?

Alternatively to using the PRUs, I could just have an SPI link to
something like the C2000, which is cheap and has tons of hardware PWM
for doing all sorts of cool motion control (beagleBone CNC or 3D
printer???), GPIO, 2 SPI (so I could easily pass the sensor data
through the device via SPI to linux, if I could figure the linux side
out of course), etc...