Re: Extremly slow assembler on i686-pc-mingw32

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Sun, Mar 21, 2010 at 4:56 PM, Rainer Emrich
<rainer@emrich-ebersheim.de> wrote:
>>> This is for binutils cvs from today.
>>>
>>> The execution time for gas is several times higher than for gas from binutils-2.20.
>>> Discovered during gcc bootstrapping.
>>>
>>> Any ideas?
>>>
>>
>> Is that specific to mingw? Can you send the me .s file? I can run it with
>> cross binutils see if it also happens on Linux.
>>
> I don't know if that's specific to mingw. I first discovered it today and it's
> not specific to any source AFAIK. The time for bootstrapping gcc is more than
> doubled in my setup. I'm bootstrapping with make -j 8 on a quad core system.
>

Here are boostrap times on my quad core machine:

r143437: 4554.23user 535.67system 27:20.69elapsed 310%CPU
r143457: 8371.13user 520.69system 45:45.63elapsed 323%CPU
r145490: 8559.72user 522.42system 46:25.82elapsed 326%CPU
r145496: 7421.45user 522.26system 41:25.30elapsed 319%CPU
r145745: 7552.84user 514.96system 42:37.89elapsed 315%CPU
r145794: 7558.14user 1667.33system 56:04.96elapsed 274%CPU

We took several big hits and never really recovered.


-- 
H.J.