Hi Tom,
On Mon, Feb 18, 2013 at 2:45 PM, Tom Rini trini@ti.com wrote:
On Mon, Feb 18, 2013 at 09:06:01AM -0800, Simon Glass wrote:
Hi Wolfgang,
On Mon, Feb 18, 2013 at 3:35 AM, Wolfgang Denk wd@denx.de wrote:
Dear Tom,
In message 51216721.1010603@ti.com you wrote:
There's another thread I don't have yet (and I don't have this one in gmail yet even). But, I am OK with custodians using their repos, but not the master branch, for unrelated but otherwise good patches. I'm also fine with patchwork bundles. I suppose we could use the staging repository for these changes instead.
What I mostly object about there is that these patches would go into mainline basicly unreviewed, as patch submission and pull request is all done from a single person, with no other feedback on the patches at all. And this affects a lot of common code...
Actually, I see this change when pulling u-boot-x86.git/master:
-> bloat-o-meter u-boot-before u-boot
What board is this please?
Some specific notes here - I think it boils down to moving crc32 into the hash framework. This adds some overhead, but has a few benefits.
So you're going to v2 this part?
I have just sent 3 v2 patches to the mailing list. There is a new patch 16 to be inserted before this patch, an updated version of this patch, and an updated version of the final patch in the series (which was number 20 and is now number 21).
I have ended up just using #ifdef to remove most of the hash code if only crc32 is in use. It's not particularly elegant but I was not able to significantly reduce the code size impact any other way. Please take a look at it and see what you think.
Just in case it is useful, I have pushed a new 'us-mem3' branch to x86, but this is not for pulling (it has not been through patchwork), just for testing / comparison.
I will be away until Monday so probably no further response from me until then.
Regards, Simon
-- Tom