Re: [PATCH 00/10] jump label: introduce very_[un]likely + cleanups+ docs

[Steven Rostedt]

On Wed, 2012-02-22 at 15:56 +0100, Ingo Molnar wrote:

> > Because it really just looks like a stronger "unlikely()" and 
> > fundamentally it really isn't. [...]
> 
> Well, the fact is that right now it *is* a stronger unlikely() 
> on architectures that have jump-labels and it's mapped to 
> unlikely() on others.
> 

Has gcc been fix to make it truly an unlikely case and remove the "jmp;
jmp" problem of before? I'm still using gcc 4.6.0 which has the
following code for a tracepoint (example is the
trace_sched_migrate_task().

    5b4a:       e9 00 00 00 00          jmpq   5b4f <set_task_cpu+0x5e>

The above is the jump label that turns into a nop at boot up.

    5b4f:       eb 19                   jmp    5b6a <set_task_cpu+0x79>

Here we jump over some of the trace code (this is the fast path)

    5b51:       49 8b 7d 08             mov    0x8(%r13),%rdi
    5b55:       44 89 e2                mov    %r12d,%edx
    5b58:       48 89 de                mov    %rbx,%rsi
    5b5b:       41 ff 55 00             callq  *0x0(%r13)
    5b5f:       49 83 c5 10             add    $0x10,%r13
    5b63:       49 83 7d 00 00          cmpq   $0x0,0x0(%r13)
    5b68:       eb 41                   jmp    5bab <set_task_cpu+0xba>

Below is the continuation of the fast path.

    5b6a:       48 8b 43 08             mov    0x8(%rbx),%rax
    5b6e:       44 39 60 18             cmp    %r12d,0x18(%rax)
    5b72:       74 0c                   je     5b80 <set_task_cpu+0x8f>


Again, I'm using gcc 4.6.0 and maybe it has been fixed.

-- 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/