[wxpython-users] On performance

[wxpython-users] On performance	30 of 31
Pablo Antonio

On Wed, Apr 22, 2009 at 02:50:01PM +0200, Nitro wrote:

> Am 22.04.2009, 14:28 Uhr, schrieb Nitro <nitro@dr-code.org>:

>> Am 22.04.2009, 04:38 Uhr, schrieb Robin Dunn <robin@alldunn.com>:

>>> Christopher Barker wrote:

>>>> if self.USE_NUMPY:

>>>> bmp = wx.BitmapFromBuffer(width, height, self.__bytes)

>>>>

>>>

>>>> the first should be a simple pointer transfer, though it may make a

>>>> copy of the data.

>>>

>>> The wx.Bitmap's internal pixel buffer format is platform dependent

>>> and is not likely to be the same format as the buffer expected by

>>> wx.BitmapFromBuffer. However the copy that is made inside

>>> wx.BitmapFromBuffer (from the buffer to the wxBitmap) is using the wx

>>> raw bitmap APIs that, while abstracting out the platform differences,

>>> is still doing essentially a C pointer to pointer copy of the bytes.

>>> So given the circumstances it should be as fast as possible. It's

>>> still a copy though, so takes O(N) time to be done.

> Just as a little follow-up to my last post, you might try to use RGBA

> instead of RGB. In some cases I have found RGBA to be faster, probably

> because a) the driver uses rgba internally and has to convert rgb to rgba

> before it can be used and b) because rgba is 32-bit aligned in memory

> which processors somehow like. Not sure if all of this also applies to

> RGB/RGBA in wxPython/wxWidgets.

Nice tip! I'll give that a try. Thanks :)

Pablo Antonio (AKA crazy2k)

http://www.pablo-a.com.ar/

_______________________________________________

wxpython-users mailing list

wxpython-users@lists.wxwidgets.org

http://lists.wxwidgets.org/mailman/listinfo/wxpython-users

-->

[wxpython-users] On performance	31 of 31
Pablo Antonio

On Wed, Apr 22, 2009 at 12:56:18PM +0200, Werner F. Bruhin wrote:

> Hi Pablo,

> Pablo Antonio wrote:

>> On Tue, Apr 21, 2009 at 7:08 PM, Pablo Antonio <pabloa@gmail.com

>> <mailto:pabloa@gmail.com>> wrote:

>> Now this is fixed. However, since I'm working on the windowing

>> algorithm the code could break any moment :) So be aware of that if

>> you happen to checkout now from svn.

> Just FYI, now it works find on Windows Vista .

> Played around with the timer setting (Intel Core Duo 6300 1.86 Ghz - 4

> GB Ram).

> 1. initial value 300 = CPU usage: 30 - 45 %

> 2. value 100 = CPU usage: 50 - 55 %

> 3. value 1 = CPU usage: 50 - 67 %

> Animation is a bit slower in case 1, no significant difference between

> case 2 and 3 - but this is difficult to judge just be looking at one

> after the other.

Thanks for taking the time to test this :) I really appreciate it.

> Might be interesting to have an option where you can say e.g. do "n"

> animation iterations and measure start and end.

Yes. I'll make a better profiling of this when I get the time.

Thanks again,