Re: [PATCH 5/7] x86/jump labels: Use etiher 5 byte or 2 byte jumps

[Steven Rostedt]

On Mon, 2012-03-12 at 12:17 -0400, Jason Baron wrote:
>  
> > By allowing the jump labels to be 2 bytes, it speeds up the
> > nops, not only 2 byte nops are faster than 5 byte nops, but also
> > because it saves on cache foot print.
> > 
> >    text    data     bss     dec     hex filename
> > 13403667 3666856 2998272 20068795 13239bb ../nobackup/mxtest/vmlinux-old
> > 13398536 3666856 2998272 20063664 13225b0 ../nobackup/mxtest/vmlinux-new
> > 
> > Converting the current v3.2 trace points saved 5,131 bytes.
> > As more places use jump labels, this will have a bigger savings.
> > 
> 
> Hi Steven,
> 
> Strange. I'm not seeing the text size savings with this patch, relative
> to the 'old' jump label compiled code. Is your comparison against jump
> labels disabled?

Note, the code went through several changes since I first pushed these.
I did not redo the size tests in my rebase.

> 
> Here's the size without your patch 'CONFIG_JUMP_LABEL' set:
> 
>    text    data     bss     dec     hex filename
> 10809465        1023976 1159168 12992609         c64061 vmlinux
> 
> And with your patches and 'CONFIG_JUMP_LABEL' set:
> 
>    text    data     bss     dec     hex filename
> 10812613        1023976 1163264 12999853         c65cad vmlinux

Interesting that your test had a 3k increase?? but the jump label code
increased by only 720bytes. This looks very fishy.  And yes, I did have
jump label enabled for my changes. I didn't change the config in the two
checks. IIRC, I did a localyesconfig so that I would get the updates of
the modules that would be used it my code, against a distro config.

> 
> So an increase in text of 3148. Which is not completely explained by the
> increase in arch/x86/kernel/jump_label.o:
> 
> without your patch 'CONFIG_JUMP_LABEL' set:
> 
>    text    data     bss     dec     hex filename
>     229       0       0     229      e5 arch/x86/kernel/jump_label.o
> 
> 
> with your patch 'CONFIG_JUMP_LABEL' set:
> 
>    text    data     bss     dec     hex filename
>     943       0       8     951     3b7 arch/x86/kernel/jump_label.o
> 
> So jump_label.o is 714 bytes larger, which is not enough to explain the
> 3148 byte increase.

Right this looks strange.

> 
> I'm using gcc (GCC) 4.6.2 20111027 (Red Hat 4.6.2-1).

I'm still using 4.6.0.

> 
> Can you please double check the savings.

I'll check again.

-- Steve


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/