TMS320F2806: Why reading from pheripheral memory requires a lot of time?

Hello everybody,

i want to make 3 consecutive readings of registers from ePWM registers, and i noticed that it takes a lot of time. I have to do there operations:

1. read TBCTR register, from ePWM1
2. read TBSTS register, from ePWM1
3. read TBCTR register again, from ePWM1

performing these operation will require 16 ticks at 100MHz clock timing, and i would have expected 3 ticks only. Why? Here is my code. From my C software i call an asm function, which begins with:

_ePWMCouterRead:

  	;load periph. page
	MOVW	DP,#_Perif         ;periph value is 1a0h, address of ePWM1 regs

	;read counter, 1
	MOVZ 	AR0,@TBCTROff      ;TBCTROff is 0
	;read direction
	MOVZ	AR4,@TBSTSOff      ;TBSTSOff is 1
	;read counter 1
	MOVU	ACC,@TBCTROff
        
       ...

My ePWM counter is configured as 100MHz counter, no prescalers, 1tick is 10ns. Why my value in ACC is alaways greater than value in AR0 by 16? I would have expected greater by 2 only, not 16. I can't understand this and can't undestand if i have some settings in my register, that slow reading from pheripheral registers. Also writing the same code with C, produces the same result. Please give hints, i have do read these values very quickly.

Best Regards

Emanuele Peruzzi