[Top][All Lists]

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

[avr-gcc-list] Re: an idea: adding serial sram to avr to extendheapstora

From: David Brown
Subject: [avr-gcc-list] Re: an idea: adding serial sram to avr to extendheapstorage
Date: Tue, 15 Mar 2011 22:45:06 +0100
User-agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv: Gecko/20101209 Fedora/3.1.7-0.35.b3pre.fc14 Thunderbird/3.1.7

On 15/03/11 22:30, Weddington, Eric wrote:

-----Original Message----- From:
On Behalf Of David Brown Sent: Tuesday, March 15, 2011 2:47 PM To:
address@hidden Subject: [avr-gcc-list] Re: an idea: adding
serial sram to avr to extendheapstorage

My thought was that the user-specific code should be in the user
program - I certainly was not suggesting that there should be code
to access a serial SRAM within the standard gcc builds.  What I was
hoping for is that the standard builds could contain generic
extendible address spaces, whose actual implementation could be in
user code.  The idea is to get some way to allow a user to
implement their own "__serialsram" address space with only AVR code
- no changes to gcc, and no gcc plugins.

I don't know if it is possible to do this with gcc multiple
address space support as it stands today.

That's my concern too. It's a whole other kettle of fish.

Even if it is possible, it may not be possible to do efficiently.
But if it /were/ possible, then it would open up a range of
features for the AVR and many other gcc targets.  It would mean for
example that implementing the __flash and __eeprom address spaces
would become an AVR library issue rather than a gcc development

Well, I certainly don't mind that __flash and __eeprom would turn
into compiler issues. That's been desired since at least 2001/2002,
if not earlier.

I agree, user-defined address spaces would be a cool feature as you
describe it. But unless there's a volunteer to help with that, it's
going to be yet another cool feature that will have to wait until we
get the other cool features on the list done.

If it is possible to make efficient user-defined address spaces, then it would be a cool feature for all of gcc - not just the AVR. In all likelihood, some of the other mainline gcc developers will already have thought of the possibilities of this cool feature. Either they will have decided that it can't realistically be done at the moment, or there is a good chance that it will gradually appear as the support for user-defined address spaces matures and improves.

reply via email to

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