Enable Floating Point Operations on ARC from Quark SE


Michael Rosen
 

Zephyr Devs,

I am working on the project that requires a number of floating point operations to take place on the ARC core in the Intel Quark SE microcontroller. As such, it would be great to fully utilize the FP HW that the core contains. So far, I haven't been able to get the compiler to use the FP instructions in all but the most basic of cases by adding the following flags to my compilation (from my prj.conf file): CONFIG_COMPILER_OPT="-mspfp -mdpfp"

Without these flags, it seems we get no floating point instructions. With them, we get some fsmul and fssub instructions and what not, but some sections of code do not use the floating point operations (these seems to be b.c theres an integer to float conversion or constant float in the equation). Ive heard the latest version of gcc for ARC supports more floating point, is it planned to be included newer versions of the SDK? If not, how difficult is it to replace the compiler with a newer one (not sure how your build system is set up)?

Thanks,
Mike


-------------------------------------
Michael R Rosen
Engineer
New Devices Group | Wearables
RNB2-H4
Santa Clara, CA | 95054

Join devel@lists.zephyrproject.org to automatically receive all group messages.