3/17

FIELD UPDATES FOR FLASH BASED ST7 APPLICATIONS USING A PC COMM PORT

2 ABSOLUTE ADRESSES AND VECTORS

The example program was written for the ST72264 but, so long as the reset and interrupt

vector locations are the same, it can be executed without change on any ST7 regardless of the

size of its flash memory. This is because whatever the size of the flash, it will occupy the top

of the 0000 to $FFFF 64k possible addressing space of an ST7. We would thus always want

sector0 to occupy the top 0.5k, the space from $FE00 to $FFFF. We would also want the skip-

jump table (which is technically part of sector1) to always be located in the area just below

sector0 (also the top of sector1). In the example program this table occupies from $FDD0 to

$FDFF. Looking at the vector section at the end the program, the segment directive is used to

place the program pointer at each of the absolute memory locations for the various interrupt

vectors (and the reset vector) as described in the data sheet for the MCU. At each of these lo-

cations, a dc.w directive is used to place a 16 bit absolute address vector. In most micro con-

troller software we would be more accustomed to seeing a symbolic (and relocatable) address

reference used here and this points out an important concept that must be understood. After

we make our first version of the software application and send it “out into the world” by pro-

gramming it into the MCU with a standard programmer (sector0 and sector1), sector0 will

probably (if all goes according to plan) never be changed again, although sector1 can have

many revisions. When sector1 code is revised, the addresses of any or all of the interrupt

service routines associated with the interrupt vectors will move. Since sector0 and the vector

table will not change we need the address references in the vector table to be fixed. This lim-

itation would cause a severe loss of flexibility in the program if it were not for the use of the

skip-jump table located at $FDD0. This table has one entry for each interrupt vector. The table

consists of long jump (to cover the entire 64k space) instructions spaced every four bytes. Ac-

tually a long jump only takes up three bytes (one for the jp op code and two for the address)

but it’s easier to count by four.

Please note that, in the example program, all of the jumps are to the symbolic (and re-locat-

able) address “dummyisr” except for the one associated with the reset vector, which jumps to

“start”. This is because the example program does not make use of any interrupts. The pro-

gram must provide a mechanism to use any of the interrupts since we cannot “get at” the

vector table once we send our program out into the world. If we send out a version 1.0 of our

application which perhaps uses only the SCI interrupt and later decide to revise the program

and make use of the TIMERA interrupt we can do that by simply replacing the appropriate “jp

dummyisr” instruction in the skip-jump table with a jump to our TIMERA interrupt service rou-

tine. Since the skip-jump table resides in sector1 we are free to change the destination of the

jumps whenever we please although the absolute location of the jump instruction itself is fixed

so that it always agrees with sector0.

A careful examination of the reset vector reveals that it is treated specially. This vector does

use a symbolic (and re-locatable) address reference, “BANK0”. This is permissible because