Re: [PATCH] x86: Align jump targets to 1 byte boundaries

[Ingo Molnar]

* Denys Vlasenko <dvlasenk@xxxxxxxxxx> wrote:

> > What do you guys think about this? I think we should seriously 
> > consider relaxing our alignment defaults.
> 
> Looks like nobody objected. I think it's ok to submit
> this patch for real.

Yeah, so my plan is to apply the following three changes from that 
discussion:

--- tip.orig/arch/x86/Makefile
+++ tip/arch/x86/Makefile
@@ -77,6 +77,15 @@ else
         KBUILD_AFLAGS += -m64
         KBUILD_CFLAGS += -m64
 
+        # Pack jump targets tightly, don't align them to the default 16 bytes:
+        KBUILD_CFLAGS += -falign-jumps=1
+
+        # Pack functions tightly as well:
+        KBUILD_CFLAGS += -falign-functions=1
+
+        # Pack loops tightly as well:
+        KBUILD_CFLAGS += -falign-loops=1
+
         # Don't autogenerate traditional x87 instructions
         KBUILD_CFLAGS += $(call cc-option,-mno-80387)
         KBUILD_CFLAGS += $(call cc-option,-mno-fp-ret-in-387)

... and not do -fno-guess-branch-probability, because it destroys 
likely()/unlikely() annotations.

Which is a pity, considering the size effect on defconfig:

     text      data     bss      dec  filename
 12566383   1617840 1089536 15273759  vmlinux.expect=10 [==vanilla]
 11923529   1617840 1089536 14630905  vmlinux.-fno-guess-branch-probability
 11903663   1617840 1089536 14611039  vmlinux.align=1
 11646102   1617840 1089536 14353478  vmlinux.align=1+fno-guess-branch-probability

I.e. 2.6% of savings on top of the above three patches, while the 
effect of our hot/cold branch annotations is only around 0.4%, so if 
GCC preserved our annotations under -fno-guess-branch-probability we'd 
be good by at least 2%.

But GCC doesn't.

There were also these other changes I tested:

+        # Reduces vmlinux size by 0.25%:
+        KBUILD_CFLAGS += -fno-caller-saves
+
+        # Reduces vmlinux size by 1.10%:
+        KBUILD_CFLAGS += -fno-inline-small-functions
+
+        # Reduces vmlinux size by about 0.95%:
+        KBUILD_CFLAGS += -fno-tree-ch

We could maybe consider -fno-caller-saves. What do you think about 
that option?

-fno-inline-small-functions is probably a bad idea, and -fno-tree-ch 
is probably a bad idea as well and is a dangerously rare option in any 
case that could break in unexpected ways.

Thanks,

	Ingo
--
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/