Hi, <snip>
> I know a similar question was asked in one of your earlier patch > versions, but, is there a reason why full hardware floating point > operation is not enabled? There's no reason this hasn't been enabled other than I hadn't got round to trying it yet, and I wanted to avoid changing too much at once. > The performance of the device can almost double in certain situations > if compiled with -mfloat-abi=hard rather than -mfloat-abi=softfp. [1] > > Jon > > [1] http://elinux.org/RPi_Performance#Results I would expect only a modest floating point performance improvement. Most the improvement seen in [1] is probably due to -mfpu=vfp, and this is already in the patch. However it would still be worthwhile making this change -mfpu=vfp - enables use of hardware floating point instructions. (in patch already) -mfloat-abi=hard - makes calling hardware floating point more efficient. Ian p.s. Let me know how you get on if you try this. _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/mailman/listinfo/openwrt-devel
