Re: [PATCH] Disable fast softints on OMAP (was Re: Status of current on ARM? )

[Chris Gilbert <chris%dokein.co.uk@localhost>]

Mikko Rapeli wrote:
> On Thu, Nov 13, 2008 at 08:56:43PM -0500, Rafal Boni wrote:
>> On Wed, Nov 12, 2008 at 08:46:55AM +0200, Mikko Rapeli wrote:
>>> On Tue, Nov 04, 2008 at 11:51:34AM +0200, Mikko Rapeli wrote:
>>>> The current kernel is not very well tuned for OMAP 2420 since it takes 
>>>> 2 minutes and 44 seconds to boot it to the "Enter pathname of shell..."
>>>> point. In comparison our kernel branch takes 19 seconds. Userspace was the 
>>>> same yesterdays current on both.
>>> An update to the slow boot: current from Aug 5th 2008 boots in 23
>>> seconds. Changes since then have slowed the kernel considerably on OMAP
>>> 2420.
>> Just as a datapoint, this isn't generic to all the evbarm targets; my PXA250
>> boots an INSTALL kernel (gzip'ed, with ramdisk) from WinCE in 35-40 sec, and
>> from the point of the 2nd-stage loader handing control to the kernel it's ~ 5
>> sec to sysinst running.
> 
> This is propably because __HAVE_FAST_SOFTINTS isn't enabled on your target
> platform.
> 
> Current kernel from today, some old userspace:
> 
> # time cc -Wall hello.c
>       361.09 real         1.24 user         1.23 sys
> 
> Current kernel from today, some old userspace and __HAVE_FAST_SOFTINTS
> disabled with patch below:
> 
> # time cc -Wall hello.c
>        10.33 real         0.67 user         1.69 sys
> 

I'd say we'd be better finding out why FAST_SOFTINTS isn't fast :)  In
theory it should be.  The fact it isn't suggests something is broken
somewhere (perhaps not clearing a flag somewhere...)

Thanks,
Chris