On Wed, 2011-03-02 at 12:50 -0800, Brendan Conoboy wrote:
On 03/02/2011 07:31 AM, omalleys(a)msu.edu wrote:
> Is there a way to strip out What one you aren't using? IE i don't have
> an fpu, therefore I want to strip out the hardfpu code for a smaller
> binary.
Just to clarify, this script and its resulting binaries are to assist
bootstrapping armv7l. I don't think anybody is considering it for
anything beyond a one time rebuild of all the packages.
> What vfpu are the target? I thought the main reason why we didn't want
> hard fpu's was because of the differences, between them. Number of
> registers varied, etc. You can't really optimize for each of them and
> have any sanity.
This does bring up an interesting point- right now the flag difference
is simply '-mfloat-abi=hard'. Is this the right flag for all armv7
concerns?
I need to look closely at this, we'll need a 16-register VFP
configuration for maximum compatibility.
> If I have to run a script that changes the objects in the
binaries, it
> screws up checksums and can cause all sorts of issues.
It would definitely alter checksums. What other issues do you foresee?
The overall plan here is to try to short-circuit the armv7hl bootstrap
by:
- do a global rebuild of the packages as fat binaries
- flip all the switches to produce armv7hl versions (take apart the
RPMs, flip the switches, put the RPMs back together)
- do a global rebuild of the packages as optimized armv7hl versions
...instead of doing an ordered build from the ground up (similar in
scope to supporting a new arch because of the complete ABI break, and a
much longer process than the double-rebuild). There are numerous
challenges with this approach, but I think it's worth a thorough
investigation.
--
Chris