Oleg,
I've used similar techniques at times. Recently, one of my designs I found one task was taking more than it's alotted time so I thought that talking to the lcd was taking most of the time, so I altered my lcd library to talk to a ram buffer and the timer tick would update the lcd display over many ticks. The task still took too much time and I have yet to find the time to look further - could be all the printfs. The side benefit from the changes was that I could access the lcd buffer via the comms subsystem (Modbus) so that I could write a utility that would display what was on the lcd remotely. Feature creep has started to occur!

Most of my code is based around a very simple co-operative task dispatcher so the technique of splitting a task into little chunks done regularly gets used a lot. Most of my code also features finite state machines - my collegues expect this!!!

Good Days!