[oexel]

I just purchased a new G5 iMac so I pulled an old project that I had been working on some years ago. A large portion of the cpu time in the code goes to blitting the screen. After experimenting a bit I discovered the performance to be VERY poor. I created a test application that simply creates an offscreen GWorld, paints it red and blits it to the screen a hundred times. I'm getting a throughput of around 140Mbytes/sec. What is going on? here is the snippet (copybits is even slower)

GWorldPtr offscreenGW = NULL;
Rect r = {0,0,900,1440};
Ptr screenBase = (*((*GetMainDevice())->gdPMap))->baseAddr;
Ptr offscreenBase;

//the screen resolution 1440x900x32bit
UInt32 bytes = 900*1440*4;

NewGWorld(&offscreenGW, 32, &r, 0,0,0);
SetGWorld(offscreenGW, NULL);
ForeColor(redColor);
PaintRect(&r);

//get base address of the offscreen GWorld
offscreenBase = GetPixBaseAddr(GetGWorldPixMap(offscreenGW));

for (int i=0;i<kNumBlits;i++){
memcpy(screenBase, offscreenBase, bytes);
}

any suggestions would be very much appreciated

[Detrius]

Well, it would be a *lot* faster if you used Core Graphics instead and drew into a CGLayer, blitting that to the screen instead. The CGLayer would be limited to 10.4, but a CGImage would work on any version of OS X. It wouldn't be as fast as a CGLayer, though.