| 1 |
Am 19.01.2013 12:36, schrieb Daniel Troeder: |
| 2 |
> Am 19.01.2013 11:49, schrieb Florian Philipp: |
| 3 |
>> Hi list! |
| 4 |
>> |
| 5 |
>> I've noticed that there is probably a lot of performance to be gained |
| 6 |
>> from activating backend acceleration in cairo. It's just that I have no |
| 7 |
>> clue about the details: |
| 8 |
>> |
| 9 |
>> drm, gallium and opengl are all described as accelerating backends. Does |
| 10 |
>> that mean they are mutually exclusive or can I activate them all and |
| 11 |
>> cairo will sort it out? |
| 12 |
>> |
| 13 |
>> What does the qt4 use flag do? Is it just a language binding? |
| 14 |
>> |
| 15 |
|
| 16 |
To answer myself on this: Yes it is. |
| 17 |
|
| 18 |
>> Does anyone know a simple benchmark for cairo? |
| 19 |
>> |
| 20 |
>> Thanks in advance! |
| 21 |
>> Florian Philipp |
| 22 |
>> |
| 23 |
> I found your question very interesting, and so I used... google :D |
| 24 |
> |
| 25 |
> -> Cairo seems to bring a benchmark with the source in cairo\perf. |
| 26 |
> -> http://cworth.org/intel/performance_measurement/ |
| 27 |
> |
| 28 |
|
| 29 |
I stumbled across this, too, while looking at the source tar-ball. I've |
| 30 |
done some benchmarks with the "macro" tests (i.e. traces) on mylaptop |
| 31 |
with Intel graphics and a desktop with the free radeon driver. |
| 32 |
|
| 33 |
In both cases, opengl was usually faster (sometimes considerably) than |
| 34 |
xlib. xcb never finished for some reason. |
| 35 |
|
| 36 |
> -> The Phoronix test suit ofc has a performance test for cairo :) |
| 37 |
> -> I looked into some examples, and there are BIG differences between |
| 38 |
> gfx cards: |
| 39 |
> http://openbenchmarking.org/prospect/1204129-SU-NVIDIAGEF19/3653c29c52e5da38d1879d193dc660515f97d242 |
| 40 |
> |
| 41 |
> -> This posting is most interesting: |
| 42 |
> http://lists.cairographics.org/archives/cairo/2012-October/023609.html |
| 43 |
> |
| 44 |
> As a nvidia blob user it seems I should be using the XRender backend. |
| 45 |
> But how? I can USE=-opengl to have that out of the way, but what's with |
| 46 |
> the other USE flags? |
| 47 |
[...] |
| 48 |
> Now I wonder if cairo doesn't simply choose the best one automatically? |
| 49 |
> Or should I "X -xcb -opengl -openvg"? |
| 50 |
> (Keeping "glib qt4 svg" for apps on.) |
| 51 |
> |
| 52 |
> [..] |
| 53 |
> |
| 54 |
> I used cairo-trace to start some programs and it seems like programs can |
| 55 |
> set their desired backend. There is a "/type" setting: |
| 56 |
> |
| 57 |
> /type 42 set (firefox, soffice, clementine, evince, opera, |
| 58 |
> crack-attack, digikam, inkscape) |
| 59 |
> /type /xlib set (firefox, soffice, clementine, evince, |
| 60 |
> crack-attack, digikam, inkscape) |
| 61 |
> /type /xrender set (firefox, soffice, clementine) |
| 62 |
> |
| 63 |
> |
| 64 |
> I wonder what "42" is... |
| 65 |
> |
| 66 |
|
| 67 |
I think that's related to a font format. |
| 68 |
|
| 69 |
Regards, |
| 70 |
Florian Philipp |
Archives