View Single Post
Old 05-10-2013, 07:09 PM   #135
kaito
Connoisseur
kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.kaito understands the mechanisms of the catecholamine pathways.
 
Posts: 52
Karma: 76234
Join Date: May 2013
Device: k3w, gloHD
Quote:
Originally Posted by baf View Post
The difference between binaries suggests that the faster one uses hardware fpu. Try to recompile with "--with-float=softfp"
Thanks for the help baf & knc1. Soft FP was definitely the problem with CPU use.

I wasn't able to compile with --enable-thumb (some register error), but adding --enable-armvfp was fine. My readelf tag output now exactly matches the fbdev binary and cpu use is the same lower number with Mplayer v1.1.1.

The build.sh script could be updated to use softfp:

export CFLAGS="-O4 -I$ALSAPREFIX/include/ -march=armv6j -mtune=arm1136jf-s -mfloat-abi=softfp -mcpu=arm1136j-s -U_FORTIFY_SOURCE"

Last edited by kaito; 05-11-2013 at 02:57 PM. Reason: quote
kaito is offline   Reply With Quote