Detlev, et al., On Aug 30, 2013, at 10:54 PM, Larry Baker wrote: > I have succeeded in building a functional gfortran cross compiler. I have > yet to actually run anything on my target. I'll add an update to confirm > everything works.
I tested my application and it runs -- C and Fortran! After some tweaks to the buffer settings (to allow the system to store more pending work until it could catch up), and allowing a lot more time for the floating-point calculations during my application's startup (~3 minutes on my 400 MHz 64MB no-FPU ARM V5TEJ), I was able to pass the validation test. Initially I used a USB drive for the rootfs. The I/O performance was abysmal -- the application wold not run properly. I have spent most of the past two days converting the JFFS2 rootfs to UBI. I already knew I wanted to do that -- I built my own kernel partially to add UBI support. The Atmel tools for loading UBI images (.ubi) did not work right, even with the option to enable trimmffs. I gave up trying to figure out what they did to make Linux unbootable. (I also did not try to use the UBI support built into U-Boot.) I ended up creating a minimal Yocto/ELDK with mtd-utils on a USB drive, which I then used to flash the UBI roots (.ubifs) from Linux. Now I am on to Yocto/ELDK on a BeagleBone Black -- an ARM processor with FPU this time! One thing I discovered is that when I log off and log back on to do a Yocto/ELDK build, it is not sufficient to cd to the build directory and source set-path. I had to also run the TEMPLATE=... command from the ELDK documentation, and source set-path. That should be described somewhere in the documentation, like where it talks about modifying set-path. Thanks for your help. Larry Baker US Geological Survey 650-329-5608 [email protected]
_______________________________________________ eldk mailing list [email protected] http://lists.denx.de/mailman/listinfo/eldk
