Re: [PATCH 0/7] preempt_count rework -v2

["H. Peter Anvin" <hpa@zytor.com>]

On 09/10/2013 03:02 PM, Linus Torvalds wrote:
> On Tue, Sep 10, 2013 at 2:51 PM, H. Peter Anvin <hpa@zytor.com> wrote:
>> On 09/10/2013 02:43 PM, Linus Torvalds wrote:
>>
>> Actually, the right thing here really is "er" (which I think you meant,
>> but just to make it clear.)
> 
> Yes, I was just answering the i-vs-e confusion.
> 
>> "e" doesn't work on versions of gcc older than the first x86-64 release,
>> but we don't care about that anymore.
> 
> Indeed.
> 
>> A final good question is if we should encapsulate the add/inc and
>> sub/dec into a single function; one could easily do somethin glike:
> 
> Yes. However, I would do that at a higher level than the one that
> builds the actual functions.
> 
> That said, there's a few cases where you might want to specify
> add-vs-sub explicitly, but they are rather odd, namely the fact that
> "-128" fits in a byte, but "128" does not.
> 
> So it can be better to add 128 by doing a "subl $-128" than by doing
> an "add $128".
> 
> But we probably don't have any situation where we care about that
> special value of "128". I've seen the trick, though.
> 

Yes, and if __builtin_constant_p() we could even do it explicitly.
Unfortunately I don't think gcc allows alternatives in asm() statements,
unlike in its own pattern tables.

	-hpa