furs

abs_addresses
Login

abs_addresses

Help

Compute the Register Address (absolute) and add it to the new ABS_ADDR_SVD as register.address

$(UNFOLDED_SVD) $(XSL_PROC) $(ABS_ADDR_SVD) $(ABS_ADDR_STY) 
C:circle rad 40px "$(UNFOLDED_SVD)" fit
arrow right 100% 
E:box rad 10px "$(XSL_PROC)" fit
arrow right 100%
G: circle rad 40px "$(ABS_ADDR_SVD)" fit
F:circle rad 40px "$(ABS_ADDR_STY)" fit at 1.5 s of C
arrow from F.e \
then right until even with E \
then to E.s

The CMSIS design contains a register base and offset which are summed by Forth using the on-mcu compiler to derive the register absolute address.

As FURS was designed for small Cortex-M mcu's such as the Cortex-M0 series, I decided to move this function to the PC which has thousands of times more computing power and memory than a slow embedded mcu. I also believe this saves time compiling Forth source code on the same slow mcu.

But is this method efficient compared to the MCU doing the computations ?

Furthermore, this computation need only be done once (on the PC) as the result is then saved in a file or database for any further use.

Makefile

$(ABS_ADDR_SVD):  $(UNFOLDED_SVD)
		$(XSL_PROC) -o $(ABS_ADDR_SVD) $(ABS_ADDR_STY) $(UNFOLDED_SVD) 
		$(XMLLINT) --format $(ABS_ADDR_SVD) > temp1.tmp && mv temp1.tmp $(ABS_ADDR_SVD) || rm -f temp1.tmp

Example: 


<register>
          <name>ARR</name>
          <displayName>ARR</displayName>
          <description>auto-reload register</description>
          <addressOffset>$2C</addressOffset>
          <size>$20</size>
          <access>read-write</access>
          <resetValue>$00000000</resetValue>
          <address>$4001482C</address>  <---- ADDITION