|
Rockbox mail archiveSubject: gcc 4.4.1 & binutils 2.20gcc 4.4.1 & binutils 2.20
From: Thomas Martitz <thomas.martitz_at_student.htw-berlin.de>
Date: Thu, 10 Sep 2009 20:20:24 +0200 So, I today decided to experiment with a newer gcc/binutils version. My results are very much better than the previous tests. For reference, the last tests were discussed here: http://www.rockbox.org/mail/archive/rockbox-dev-archive-2009-02/0069.shtml http://www.rockbox.org/mail/archive/rockbox-dev-archive-2009-03/0160.shtml In details: My sytem: gcc: arm-elf-gcc (GCC) 4.4.1 (<latest release) ld: GNU ar (GNU Binutils) 2.20.51.20090910 (< today's snapshot) Host gcc: gcc (Debian 4.4.1-1) 4.4.1 Host system: Linux In contrast to the previous test, this time everything seems to run. AAC now runs also, but I didn't test WMA (both didn't work in the last test). The changes needed to build rockbox are very minimal, removing the memcpy wrapper in rockdoom.c is enough. The rest builds fine, although there are a few (I think minor) warnings. On my e200: Binsize and RAM usage are *down* by 10k each. Codecs: [codec] -- [4.0.3: %rt/MHz needed for rt] -- [4.4.1: %rt/MHz needed for rt] mp3 -- 381.27%/20.98 -- 382.84%/20.89 aac -- 223.06/35.86 -- 219.01%/36.52 vorbis -- 281.06%/28.46 -- 275.01%/29.07 flac 632.50%/12.64 -- 554.01%/14.44 While the binsize goes down quite nicely and my tested codecs (except flac) don't differ much in their performance, flac is for some reason noticeably slower. At least rockbox builds and runs (apparently) fine with this combo again which is a major improvement over older combos. But what it gives us is not worth changing our standard toolchain I think. Let's hope for gcc 4.4.2 or binutils 2.20(.1) :) Best regards Received on 2009-09-10 Page template was last modified "Tue Sep 7 00:00:02 2021" The Rockbox Crew -- Privacy Policy |