To go on with shift register. If you don't need UART, configure it in mode 0 as an 8 bit shift register, and you can get a transfer rate of 1 Mbit/sec. So you can toggle an output of the shift register every 10 us theoretically (xtal = 12MHz), and you have no code overhead of shifting, masking etc. When TI flag set, just toggle the shift register strobe input twice.