[coreboot] Using only one romcc copy
c-d.hailfinger.devel.2006 at gmx.net
Wed Sep 3 18:24:10 CEST 2008
Hi Mats Erik,
if you want to sign off on my patch version of your utility, please do
so. Having an Ack from you would be cool as well.
On 03.09.2008 03:17, Carl-Daniel Hailfinger wrote:
> Hi Mats Erik,
> I'm sorry that I don't know the exact convention for writing your first
> name. Do you prefer to be called Mats Erik or just Mats?
> On 02.09.2008 22:13, Mats Erik Andersson wrote:
>> I did one round of timing the build processes:
>> Original make command: 0m42.323s
>> My reduction and a clean start: 0m30.955s
>> After that a 'make clean' and renewed make: 0m23.081s
>> The timings might not be typical, but the time saving seems
>> to be substantial in the 'middle case' where only one romcc
>> needs to be built, as well as in the last case were romcc
>> is fully reused for both rom-images.
I changed this a bit to trigger romcc rebuild on "make clean" in the
main target directory, but not on "make clean" in each normal/fallback
> I measured an abuild run and the time savings on my machine are between
> 0% and 30% (average 12%), so there is no downside if you look at it from
> a compile time standpoint.
> Your work is really helpful for anybody working on v2 and I took your
> suggestion to fix the make rules at the source.
>> Since I only have used coreboot-v2, I have no knowledge
>> whether my utility could be applied also to coreboot-v3.
> Fortunately v3 uses only the system gcc, so this doesn't apply.
> This patch is the equivalent of running Mats' reduce.sh before every
> compilation. Thanks to him for the initial idea and realization.
> The only thing left out is the make rule for cleaning the romcc binary
> shared between each fallback/normal image. That change should be
> discussed separately.
> Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006 at gmx.net>
> Patch attached because it is >100k.
More information about the coreboot