On 1/23/07, Ralf Baechle <ralf@xxxxxxxxxxxxxx> wrote:
On Tue, Jan 23, 2007 at 03:18:17PM +0100, Franck Bui-Huu wrote:
> From: Franck Bui-Huu <fbuihuu@xxxxxxxxx>
>
> This trivial change reduces considerably code size of these
> 2 functions callers. For instance, here is the figures for
> arch/kernel/signal.o objects:
>
> text data bss dec hex filename
> 11972 0 0 11972 2ec4 arch/mips/kernel/signal.o~old
> 5380 0 0 5380 1504 arch/mips/kernel/signal.o~new
Have you ran any benchmarks on this? Unrolling the loops used to make
a noticable difference.
OK, I ran a micro benchmark on setup_frame() which calls
setup_sigcontext(). I got the execution time of setup_frame() by
counting the number of cycles spent in it (by using cp0_count
register).
Without this patchset applied it takes about 14600 cycles for
setup_frame() execution whereas with this patchset applied it takes
10300 cycles. These figures are averages.
So it appears that this patchset has a positive impact on both size and speed.
thanks
--
Franck
