I tested engine a bit, noted that speed increase is about 10x, going over 900fps in blending sample.
Fullscreen mode in your base project gives access violation if i before or after alt+enter press space to use hardware.

I'm still unable to use hardware acceleration with old flexbattle game. It renders images using normal, rotate and blending functions in custom TDirectDrawSurface. Then clips player splitscreen parts from it. All graphics, except the ones using rotate and blendings show ok. Player itself uses rotate, so main character is invisible now :?