| 1 |
On Thu, May 2, 2013 at 3:33 PM, Frank Peters <frank.peters@×××××××.net> wrote: |
| 2 |
> During kernel configuration there is an option called CONFIG_TRANSPARENT_HUGEPAGE |
| 3 |
> which allows the use of Transparent Hugepages or application memory pages larger |
| 4 |
> than the traditional 4K size. This option has been available for a while but |
| 5 |
> I've never enabled it. However, it would seem like a good idea to use. |
| 6 |
> |
| 7 |
> Searching for more information I can find no comments on actual Hugepage |
| 8 |
> performance for a simple desktop Linux system. Can anyone verify or refute |
| 9 |
> the idea that Transparent Hugepages will lead to improvements on a desktop |
| 10 |
> system? Does an application have to specifically request Hugepages or |
| 11 |
> does the allocation occur for all applications? |
| 12 |
|
| 13 |
AFAIR there were claims of up to 10% performance improvement in |
| 14 |
certain test scenarios, and that you don't need to do anything special |
| 15 |
to exploit THP in existing software (but applications can be rewritten |
| 16 |
in such a way as to specifically make better use of it). |
| 17 |
|
| 18 |
There is a setting that lets you expose THP _only_ to those programs |
| 19 |
which specifically request it, rather than backwards-compatible to all |
| 20 |
programs if you are worried about that. |
| 21 |
|
| 22 |
Check out the in-kernel documentation (vm/transhuge.txt) for details. |
| 23 |
|
| 24 |
Anecdotal: I have turned it off and on and don't "feel" any difference |
| 25 |
in the operation of my computer either way... |
Archives