1 |
On Sunday 08 May 2005 21:47, Miguel Filipe wrote: |
2 |
> Are there any numbers (benchmarks) about the performance penalty of |
3 |
> pageexec and/or segmexec on intel x86 machines? |
4 |
> |
5 |
> The idea that I have is that page-exec on x86 involves a page-fault |
6 |
> for every (execute) access to a new page that will be treated by |
7 |
> pax... and that is performance-wise .. bad.. |
8 |
|
9 |
It's a page-fault for every access, whether it is write, read or execute. |
10 |
Access type is verified later. Am I right? |
11 |
|
12 |
> And that segmexec is a diferent approach that involves, mirroring the |
13 |
> process address space on two segments with diferent "write" |
14 |
> permissions, and compairing those two, to check if there was any |
15 |
> overwrite of the code segment. |
16 |
> This would mean doubling the mem-usage, at least for the code-segment |
17 |
> in segmexec mode. |
18 |
> |
19 |
> I have the idea that segmexec is advised for being faster (on x86), |
20 |
> but I don't have any numbers, and I was trying to understand the |
21 |
> performance-wise consequences of each implementation. |
22 |
|
23 |
Someone told me the same thing :-p |
24 |
|
25 |
> And in arches that suport no-exec pages (has sparc or amd64), what are |
26 |
> the performance penalties? Anyone can give me some pointers? |
27 |
|
28 |
Using PAGEEXEC, supposedly none. right? |
29 |
|
30 |
> stuff like: kernel compiles, mysql benches, or... any other benchmark |
31 |
> is good for me.. just to "grasp" a idea... |
32 |
|
33 |
On a simple SMP pentium III I can make such benchmarks :-) |
34 |
|
35 |
regards, |
36 |
pedro venda. |
37 |
-- |
38 |
|
39 |
Pedro João Lopes Venda |
40 |
email: pjvenda < at > arrakis.dhis.org |
41 |
http://arrakis.dhis.org |