Fixed a hardware interrupt signalling check in MAINDEC Test #15
This commit is contained in:
parent
d0911c49e2
commit
d533ae69aa
10 changed files with 342 additions and 334 deletions
|
|
@ -745,7 +745,7 @@ CPUStatePDP11.prototype.removeTrigger = function(trigger)
|
|||
triggerPrev = triggerNext;
|
||||
}
|
||||
}
|
||||
// We could also set trigger.next to null now, but strictly speaking, that shouldn't be necessary
|
||||
// We could also set trigger.next to null now, but strictly speaking, that shouldn't be necessary.
|
||||
};
|
||||
|
||||
/**
|
||||
|
|
@ -757,16 +757,12 @@ CPUStatePDP11.prototype.removeTrigger = function(trigger)
|
|||
*/
|
||||
CPUStatePDP11.prototype.setTrigger = function(trigger)
|
||||
{
|
||||
/*
|
||||
* We COULD dispatch interrupts immediately, but there are compatibility reasons for ONLY doing so
|
||||
* inside the stepCPU() loop; see that function for details.
|
||||
*
|
||||
* if (this.dispatchInterrupt(trigger.vector, trigger.priority)) {
|
||||
* return true;
|
||||
* }
|
||||
*/
|
||||
this.insertTrigger(trigger);
|
||||
this.opFlags |= PDP11.OPFLAG.INTQ;
|
||||
/*
|
||||
* See the writeXCSR() function for an explanation of why signalling an INTQ hardware interrupt condition
|
||||
* should be done using INTQ_DELAY rather than setting INTQ directly.
|
||||
*/
|
||||
this.opFlags |= PDP11.OPFLAG.INTQ_DELAY;
|
||||
return false;
|
||||
};
|
||||
|
||||
|
|
|
|||
Loading…
Reference in a new issue