Home » U++ Library support » Archive » you can download and compile AGG with Ultimate++
|
|
|
|
Re: you can download and compile AGG with Ultimate++ [message #1075 is a reply to message #1066] |
Sun, 19 February 2006 15:34 |
jadeite
Messages: 42 Registered: January 2006
|
Member |
|
|
fudadmin wrote on Sun, 19 February 2006 03:59 | I've have adapted AGG for U++ (only win32 for now but I think it will be not difficult for Linux). Try it.
|
I tried it. Installation and compile of GraphTest examples is smooth as silk.
Don't you just love the code of Maxim; it is sooo clean!
GraphTest example has a nice benchmarking feature. Are we able to render on a surface using AGG INSIDE A U++ APP yet (with new BGAR feature by luzr)? If so, you can build the GraphTest example in U++ framework, and run the benchmark to compare the 'pumping' of graphics to screen b/w the lightweight AGG widget set and U++ framework. They claim AGG is built for speed (not size), so it might be a good test case to test part of U++ system. Shouldn't be that much difference in 2 because AGG will be doing most of the dirty work, but could still be useful.
Cheers
[Updated on: Sun, 19 February 2006 15:35] Report message to a moderator
|
|
|
Re: you can download and compile AGG with Ultimate++ [message #1076 is a reply to message #1075] |
Sun, 19 February 2006 15:50 |
|
mirek
Messages: 14105 Registered: November 2005
|
Ultimate Member |
|
|
Considering pumping perfomance, there are my earlier results of "pumping" in "announcments" section.
The test was pumping 800x600 rectangle with various primitive painting (nothing, just cleaning with single color, text rendering, AA text rendering etc) to the screen.
Results varied, best machines (couriously, "best" in this case had little to do with actuall performance, we have seen Celeron 2.4Ghz with cheap VGA outperfoming 3.0Ghz Northwood with 6800GT) were pushing 600 rectangles / s with easy, however, some gone as low as 10 / s, which is clearly unacceptable.
There are also other issues to consider:
- printing
- terminal services in Win32, X11 over ethernet
BTW, during test, "pumping" was the most expensive operation, followed by "cleaning" (basically memset). Second one indicates that this is more or less memory bandwidth limited issue - it does not matter that much what operations are you performing to pixels, as the most expensive operation is simply the memory access itself.
Mirek
|
|
|
Goto Forum:
Current Time: Fri Nov 01 01:04:15 CET 2024
Total time taken to generate the page: 0.04575 seconds
|