[QUOTE=Antartica
- (optional) if the processor has floating point unit or not, and if it hasn't, what type of emulation is used (in-kernel fp emulation or soft-float compiled binaries)[/quote]
From previous ARM experience, emulated float point is very slow. This is important when using the pnmtools. Fortunately the netpbm package provides integer alternatives for the more usual manipulations.
Quote:
|
- what display "driver" they are using; I doubt they use an X server, it makes more sense for them to use something along the lines of FBDev.
|
Well fbdev >(or was it vfbdev?) is an xserver, only that it does not incorporate input devices. My strategy on another e-ink device was to use it, and then to use a simple script to capture the virtual framebuffer screen, reprocess into e-ink format, and dump into the real screen framebuffer. In this way I was able to run xclock.
Quote:
|
The main limitaion right now is how to inject a new executable into the iliad and try it (if there is a security hole as in the librie and we can exec scripts in the SD/CF).
|
In the librie, the solution was to made a new firmware update incorporating an exec of script during init. It worked, but nobody become involved in the development. Perhaps as you say, the lack of a real SDK. I hope that iRex will deploy his [SDK] soon, but I am not sure if it is going to be free, so I keep considering the hacking alternative.
Quote:
|
BTW: I'm eager to start programming for the iliad. I've plenty of experience programming for small embedded linux platforms (for in-house use, but that includes gui and gui-less programs for the zaurus SL-5500 w/ SharpROM, ipaq w/ familiar, Zaurus SL-6000L w/ SharpROM, WRT54G w/OpenWRT, Asus WL-550g w/ Oleg Firmware), and it boils down to having a good crosscompiler toolchain and some patience to iron the bugs that appear when executing your app in the resource-starved embedded linux ;-).
|
I agree. Myself I have crosscompiled for ARM but not really developed, just deployed packages. One of my students were assigned to the zaurus SL-5500, I asked him to write a new kernel module to simulate a usbmouse in the zaurus (so that the machine can be attached to the usb port of the host and start moving the cursor in the host screen without installing any software on it, just having the usbmouse driver loaded). He did it well, so even kernel hacking is a feasible possibility when one has got, as you say, a decent toolchain.