Re: [Sbcl-devel] [Sbcl-commits] master: Remove extra push/pop from alloc_tramp - no need to save r11 twice.

[Stas B. <sta...@gm...>]

Douglas Katzman <do...@go...> writes:

> argh, you're right. it should have been commented as such.
>
>
> On Tue, Apr 8, 2014 at 10:44 AM, Paul Khuong <pv...@pv...> wrote:
>
>> Douglas Katzman wrote:
>>
>>> - Log -----------------------------------------------------------------
>>> commit 7abfcccbfa5a65401ee6e69292318b83788254cc
>>> Author: Douglas Katzman<do...@go...>
>>> Date:   Tue Apr 8 10:25:55 2014 -0400
>>>
>>>      Remove extra push/pop from alloc_tramp - no need to save r11 twice.
>>> ---
>>>   src/runtime/x86-64-assem.S |    3 +--
>>>   1 files changed, 1 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/src/runtime/x86-64-assem.S b/src/runtime/x86-64-assem.S
>>> index 0547128..ab72bf0 100644
>>> --- a/src/runtime/x86-64-assem.S
>>> +++ b/src/runtime/x86-64-assem.S
>>> @@ -363,12 +363,11 @@ GNAME(alloc_tramp):
>>>         push    %r9
>>>         push    %r10
>>>         push    %r11
>>> -       push    %r11
>>>
>>
>> I'm pretty sure this extra push is for stack alignment to 16 bytes. I
>> expect this change makes windows (and probably clang) builds fail, but
>> perhaps not gcc.
But couldn't the stack be adjusted in the preceding:
	and	$-32,%rsp
	sub	$16*16,%rsp
used for xmms?

-- 
With best regards, Stas.