On Wed, 25 Jul 2007, Trent Piepho wrote:
>
> Specifically, check test6_memasm.s. The C code looks like this:
>
> extern int a; /* keep asm from being elided for having no used output */
> static inline void bar(void) { asm("call bar" : "=m"(a) : : "memory"); }
> /* float x can't alias asm's output int a */
> void foo(float *x) { x[20] = 1; bar(); x[20] = 2; }
>
> The asm code ends up like this:
> foo:
> call bar
> movl 4(%esp), %eax # x, x
> movl $0x40000000, 80(%eax) #,
> ret
Hmm. I really think you should take this up with the gcc people. That
looks like a gcc bug - because there really is nothing that guarantees
that the asm doesn't change the array that "x" points to, and the asm
clearly talks about clobbering memory.
> Notice that the first write to x[20] was NOT done. It's also not done for a
> volatile asm without a memory clobber. But if you combine both volatile and a
> memory clobber, then it is! How to explain that?
I can't explain it. I do think you've found a gcc bug.
That said, the kernel mostly uses "asm volatile()" _together_ with a
memory clobber for these kinds of things, so it sounds like the kernel
wouldn't be impacted. But you're definitely right - the above report makes
me worry.
> The difference between test2_volasm.s and test2_normasm.s is hard to explain
> too. It seems like some times gcc forgets that imull is commutative. It will
> emit "imull %edx, %eax" in some cases, but change an asm slightly and it will
> decide it must do "imull %eax, %edx ; movl %edx, %eax" for no apparent reason.
Well, that's likely just a subtle register allocation issue, and
understandable. Generating perfect code is impossible, you want to
generate good code on average.
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
