tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: subr_kmem

On 07/10/2012 11:22 PM, Eric Haszlakiewicz wrote:
> On Tue, Jul 10, 2012 at 09:09:29PM +0200, Lars Heidieker wrote:
>> slight improvement on kmem(9):
>> splitting the lookup table into two parts
>> - this reduces the size of the lookup table -> touching less cache-lines
>>     (e.g. on 64bit down from 4096b to 1024b + 64b)
>> - expanding cache size up to PAGE_SIZE for larger then 4k PAGE_SIZE
>> machines (without blowing up the size of the lookup table further)
>> The size check under DIAGNOSTIC is moved back to the start of the
>> allocated as it makes it more useful and we have given up the PAGE_SIZE
>> alignment for kmem(9) on allocations >= PAGE_SIZE.
>> any objections?
> Have you done any performance tests?  How much of a difference does this make?
> eric

Some testing shows that system time goes down slightly
(taken from full release build on 4-core)

modififed       1478.03         1420.89
stock           1486.48         1436.89

the two runs are not comparable different checkout
all source pull to fs cache before run

So not statistically proven a slight win and better suited for PAGE_SIZE
> 4k



Mystische Erklärungen:
Die mystischen Erklärungen gelten für tief;
die Wahrheit ist, dass sie noch nicht einmal oberflächlich sind.

   -- Friedrich Nietzsche
   [ Die Fröhliche Wissenschaft Buch 3, 126 ]

Home | Main Index | Thread Index | Old Index