8 seconds only for IPL and startup? Can’t be. Maybe you include the image copy/decompress time? Then the first thing you should do is reduce your image size. Put only kernel and flash driver in image, then start flash driver first and you can start anything else from flash. This will keep the image copy/decompress time to a minimum. Be warned though, that starting stuff from flash takes a little longer (but this will be when kernel is already running).
Next is to compare if uncompressed maybe is faster than compressed; often it is. Also, check if your IPL is ‘scanning’ the flash for the image. If it does, re-program it to get the image from a fixed address. Also get rid of checksumming if you don’t need it. Disable debug output of startup and remove any -v from procnto aswell.