Hi all,
I've written a simple test application to measure gdiplus performace. Although currently it tests only GdipDrawImage() it could be easily extended to test anything else if desired.
I'm attaching the test source with results running 32-bit test binary with gdiplus=b and gdiplus=n (taken from win7). As could be observed, currently built-in gdiplus performs up to 17 times worse even in identity case in comparison with native gdiplus. With a scale/rotate graphics transform the results are up to 20 times slower.
Considerig that built-n gdiplus status regarding support for various complex things is pretty good, perhaps it's time to spend some time optimizing its performans.