On 07.16, J.A. Magallon wrote:
>
> On 07.16, Gwenole Beauchesne wrote:
> > Hi,
> >
> > > After some binary search and a 'couple' kernel builds, I narrowed the
> > > problem to mm/memory.c. With that file built at -O1:
> > >
> > > CFLAGS_memory.o = -O1
> >
> > Very good! There are also other new options at -O2 related to profiling
> > (in 3.3-hammer) but I don't exactly remember their names. If you could
> > also try with them disabled.
> >
>
> This is even better: -Os _WORKS_
> This narrows search space. From info gcc:
>
> `-Os'
> Optimize for size. `-Os' enables all `-O2' optimizations that do
> not typically increase code size. It also performs further
> optimizations designed to reduce code size.
>
> `-Os' disables the following optimization flags:
> -falign-functions -falign-jumps -falign-loops
> -falign-labels -freorder-blocks -fprefetch-loop-arrays
>
> I will try to get the assembler listings at -O2 and -Os and compare...
>
As expected, -Os does more that disabling those things. With
CFLAGS_memory.o = -Os -falign-functions -falign-jumps -falign-loops -falign-labels
-fprefetch-loop-arrays -freorder-blocks
it still WORKS. And the assembler is very different from -O2:
(just an example)
@@ -53,13 +45,11 @@
.globl check_pgt_cache
.type check_pgt_cache, @function
check_pgt_cache:
- subl $8, %esp
- movl pgt_cache_water+4, %eax
- movl %eax, 4(%esp)
- movl pgt_cache_water, %eax
- movl %eax, (%esp)
+ pushl pgt_cache_water+4
+ pushl pgt_cache_water
call do_check_pgt_cache
- addl $8, %esp
+ popl %edx
+ popl %ecx
ret
.size check_pgt_cache, .-check_pgt_cache
.section .rodata.str1.1,"aMS",@progbits,1
But this is enough to fill a bug report. But where ? gcc.gnu.org does
not know anything about hammer. Any SuSE developer site ?
TIA
--
J.A. Magallon <[EMAIL PROTECTED]> \ Software is like sex:
werewolf.able.es \ It's better when it's free
Mandrake Linux release 9.2 (Cooker) for i586
Linux 2.4.22-pre5-jam1m (gcc 3.3.1 (Mandrake Linux 9.2 3.3.1-0.2mdk))
