avr-gcc-list
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [avr-gcc-list] avr-libc-1.7.0 fails to compile


From: Georg-Johann Lay
Subject: Re: [avr-gcc-list] avr-libc-1.7.0 fails to compile
Date: Tue, 02 Nov 2010 22:56:28 +0100
User-agent: Mozilla Thunderbird 1.0.7 (Windows/20050923)

Weddington, Eric schrieb:

currently working on a set of changes to avr-libc that should get
 it to build without the -k switch to make. - Please don't use
gcc 4.5.1 for any real-world work. I've heard that there are bug reports against AVR GCC 4.5.x that need to get fixed. However, if you're using 4.5.1 to do some testing on the AVR port then that would be very helpful.

Never actually used avr-gcc 4.x on my projects. Just doing some benchmarks with 4.x then and when and therefore still stuck at
3.4.6.

Understood. AVR GCC 4.4.3 seems stable so far. But, depending on your
benchmarks you may, or may not, want to stick with 3.4.6, also
depending on your criteria.

Looking at the numbers I first thought 4.5 code is too poor to be used. But skimming its output it might be the first 4.x that has the potential to catch up with 3.4.6. There are minor issues that blow code size like the X-reg reloads and some unlucky inline heuristics, but these are local in some sense and the "quality gap" is not scattered all over the code. However, I don't know anything about its stability.

Was just being curious and tried to build latest avr tools and to catch up with current state of avr-gcc. Seems to me as if
development of avr-gcc is dead somehow? So the development had not
been stopped because avr-gcc is mature, though :o)
>
I wouldn't say that development of avr-gcc is dead. It is the usual
culprit: never enough people to test and fix. Hopefully that should
be fixed soon-ish, but I'm always hesitant to promise anything.

Ya, who has the power and endurance to bring avr-gcc forward if not Atmel...?

Would be a pity if avr-gcc was discontinued because the backend is defunct because no one (or hardly anyone) cares...

Are there plans for new features? I'm thinking of taking advantage of named address spaces or implementing for unexploited optimization opportunies.

Georg





reply via email to

[Prev in Thread] Current Thread [Next in Thread]