I believe Westmere does not have any explicit 1GB page TLB *entries* even
though
it ostensibly supports 1GB pages. Sandy Bridge is the first CPU with 1GB
TLB
entries, but it only has four. Not very helpful for big memory machines.
-tom
On Mon, Jul 9, 2012 at 5:02 PM, Sebastian Kuzminsky wrote:
Just wanted to share a new result.
Using huge pages and libhugetlbfs, I get a 48% speedup on my optimal
Rubik's cube solver.
That's a pretty dramatic improvement.
Thanks for all the hard work!
(Once transparent huge pages is available on ubuntu server, I'll see
how fast that is compared
to this
02:54:39PM -0700, Eric B Munson wrote:
>> > On Wed, 17 Nov 2010, Tom Rokicki wrote:
>> >
>> > > Why is it dangerous to use it in an inlined function? Inline functions
>> > > preserve lexical scoping.
>> > >
>> >
>> > I had it
, Nov 17, 2010 at 1:54 PM, Eric B Munson wrote:
> On Wed, 17 Nov 2010, Tom Rokicki wrote:
>
>> Why is it dangerous to use it in an inlined function? Inline functions
>> preserve lexical scoping.
>>
>
> I had it collide with a global variable in the test suite a
Why is it dangerous to use it in an inlined function? Inline functions
preserve lexical scoping.
On Wed, Nov 17, 2010 at 12:46 PM, Eric B Munson wrote:
> hpage_size is a fairly common variable name in this library so it
> is dangerous to use it in an inlined function. This patch renames it
> to
