COP8ACC5 Datasheet, PDF (23/41 Pages) National Semiconductor (TI) – 8-Bit CMOS ROM Based Microcontrollers with 4k Memory and High Resolution A/D

Interrupts (Continued)
and enable bits to determine the source(s) of the interrupt. If
more than one interrupt is active, the user’s program must
decide which interrupt to service.
Within a specific interrupt service routine, the associated
pending bit should be cleared. This is typically done as early
as possible in the service routine in order to avoid missing
the next occurrence of the same type of interrupt event.
Thus, if the same event occurs a second time, even while the
first occurrence is still being serviced, the second occur-
rence will be serviced immediately upon return from the cur-
rent interrupt routine.
An interrupt service routine typically ends with an RETI in-
struction. This instruction sets the GIE bit back to 1, pops the
address stored on the stack, and restores that address to the
program counter. Program execution then proceeds with the
next instruction that would have been executed had there
been no interrupt. If there are any valid interrupts pending,
the highest-priority interrupt is serviced immediately upon re-
turn from the previous interrupt.
VIS INSTRUCTION
The general interrupt service routine, which starts at address
00FF Hex, must be capable of handling all types of inter-
rupts. The VIS instruction, together with an interrupt vector
table, directs the device to the specific interrupt handling rou-
tine based on the cause of the interrupt.
VIS is a single-byte instruction, typically used at the very be-
ginning of the general interrupt service routine at address
00FF Hex, or shortly after that point, just after the code used
for context switching. The VIS instruction determines which
enabled and pending interrupt has the highest priority, and
causes an indirect jump to the address corresponding to that
interrupt source. The jump addresses (vectors) for all pos-
sible interrupts sources are stored in a vector table.
The vector table may be as long as 32 bytes (maximum of 16
vectors) and resides at the top of the 256-byte block contain-
ing the VIS instruction. However, if the VIS instruction is at
the very top of a 256-byte block (such as at 00FF Hex), the
vector table resides at the top of the next 256-byte block.
Thus, if the VIS instruction is located somewhere between
00FF and 01DF Hex (the usual case), the vector table is lo-
cated between addresses 01E0 and 01FF Hex. If the VIS in-
struction is located between 01FF and 02DF Hex, then the
vector table is located between addresses 02E0 and 02FF
Hex, and so on.
Each vector is 15 bits long and points to the beginning of a
specific interrupt service routine somewhere in the 32 kbyte
memory space. Each vector occupies two bytes of the vector
table, with the higher-order byte at the lower address. The
vectors are arranged in order of interrupt priority. The vector
of the maskable interrupt with the lowest rank is located to
0yE0 (higher-order byte) and 0yE1 (lower-order byte). The
next priority interrupt is located at 0yE2 and 0yE3, and so
forth in increasing rank. The Software Trap has the highest
rank and its vector is always located at 0yFE and 0yFF. The
number of interrupts which can become active defines the
size of the table.
Table 5 shows the types of interrupts, the interrupt arbitration
ranking, and the locations of the corresponding vectors in
the vector table.
The vector table should be filled by the user with the memory
locations of the specific interrupt service routines. For ex-
ample, if the Software Trap routine is located at 0310 Hex,
then the vector location 0yFE and -0yFF should contain the
data 03 and 10 Hex, respectively. When a Software Trap in-
terrupt occurs and the VIS instruction is executed, the pro-
gram jumps to the address specified in the vector table.
The interrupt sources in the vector table are listed in order of
rank, from highest to lowest priority. If two or more enabled
and pending interrupts are detected at the same time, the
one with the highest priority is serviced first. Upon return
from the interrupt service routine, the next highest-level
pending interrupt is serviced.
If the VIS instruction is executed, but no interrupts are en-
abled and pending, the lowest-priority interrupt vector is
used, and a jump is made to the corresponding address in
the vector table. This is an unusual occurrence, and may be
the result of an error. It can legitimately result from a change
in the enable bits or pending flags prior to the execution of
the VIS instruction, such as executing a single cycle instruc-
tion which clears an enable flag at the same time that the
pending flag is set. It can also result, however, from inadvert-
ent execution of the VIS command outside of the context of
an interrupt.
The default VIS interrupt vector can be useful for applica-
tions in which time critical interrupts can occur during the
servicing of another interrupt. Rather than restoring the pro-
gram context (A, B, X, etc.) and executing the RETI instruc-
tion, an interrupt service routine can be terminated by return-
ing to the VIS instruction. In this case, interrupts will be
serviced in turn until no further interrupts are pending and
the default VIS routine is started. After testing the GIE bit to
ensure that execution is not erroneous, the routine should
restore the program context and execute the RETI to return
to the interrupted program.
This technique can save up to fifty instruction cycles (t_c), or
more, (50µs at 10 MHz oscillator) of latency for pending in-
terrupts with a penalty of fewer than ten instruction cycles if
no further interrupts are pending.
To ensure reliable operation, the user should always use the
VIS instruction to determine the source of an interrupt. Al-
though it is possible to poll the pending bits to detect the
source of an interrupt, this practice is not recommended. The
use of polling allows the standard arbitration ranking to be al-
tered, but the reliability of the interrupt system is compro-
mised. The polling routine must individually test the enable
and pending bits of each maskable interrupt. If a Software
Trap interrupt should occur, it will be serviced last, even
though it should have the highest priority. Under certain con-
ditions, a Software Trap could be triggered but not serviced,
resulting in an inadvertent “locking out” of all maskable inter-
rupts by the Software Trap pending flag. Problems such as
this can be avoided by using VIS instruction.
23
www.national.com