When last I worked on ARM7, it took somewhat less than half a day to put together a native environment based on one of the available low-cost Samsung ARM7 chip EVB's, ATAPI disk drives, power supply, cabling, and even a reasonable "enclosure" (not really, but a subchassis, sufficient to support everything yet allow probing the target as needed.) The software was free, worked perfectly "out of the box" and was running with a PC as a terminal during the first half-hour. Adding mass storage was what took most of the time, though software was provided with the OS files.

This provided a low-cost development environment that allowed us to run the code on the target, fiddle with it as desired, and save the work product all on the simple yet powerful development environment. The target-required hardware was on-chip and only the development effort required addition of ATAPI hardware. When we were done, we removed our EVB and HDD and CDROM, and what remained was the $18 target PCB, and the desired code, all of which required a bit under 12 working days, beginning with negotiation of system requirements and testing. It took 6 weeks (1000 hours) to complete physical/environmental testing (shake-and-bake), but that required no changes.

So, I conclude you'd prefer to use a PC running a cross-compiler, with all the PC quirks and foibles, rather than to run LINUX on the target hardware and perform trial/debugging there. If you're happy with that, then, well, fine. I doubt you'll get to the heart of an SDRAM channel bug in the target very quickly if you go that route.

RE