brianr@tekig5.PEN.TEK.COM (Brian E Rhodefer) (06/29/90)
My program uses a Vertical Blanking Interrupt server whose only
function is to signal a task every N frames. To call Exec's
Signal() function, of course, the server loads A6 with &SysBase,
but I have a few misgivings about this.
My copy of the Exec volume of the RKM, in the chapter on Interrupts
(Ch 7, I believe), says that Interrupt Server routines can't count
on the contents of A6, and that only regs D0-D1/A1/A5 are scratch.
I disassembled the Vertical Blank interrupt (for my system), and found
that D0-D1/A1/A5-A6 are all saved & restored by the Kernel. With the
explicit warning against depending on A6's contents, it seems to me
that either
a) if everyone indeed preserves A6, as the RKMs appear to mandate, then
they should be able to depend on the Kernel's loading of A6 with SysBase
or
b) if nobody depends on A6, as the RKMs suggest, and the kernel
saves&restores it, then A6 can be regarded as scratch
Is the RKM inaccurate/incomplete here? I don't have to push&pop A6
in my Server routine, do I?
Brian Rhodefervalentin@cbmvax.commodore.com (Valentin Pepelea) (07/04/90)
In article <6493@tekig5.PEN.TEK.COM> brianr@tekig5.PEN.TEK.COM (Brian E Rhodefer) writes: > > My copy of the Exec volume of the RKM, in the chapter on Interrupts > (Ch 7, I believe), says that Interrupt Server routines can't count > on the contents of A6, and that only regs D0-D1/A1/A5 are scratch. To be precise, what the RKM says is that the registers d0,d1,a0,a1 are scratch registers that are automatically saved by the kernel, and so are registers a5 and a6, because they are often used by servers. > a) if everyone indeed preserves A6, as the RKMs appear to mandate, then > they should be able to depend on the Kernel's loading of A6 with SysBase The kernel pre-saves a6 for you so that *you* may load it with whatever you want. The a6 register is indeterminate at the beginning. > b) if nobody depends on A6, as the RKMs suggest, and the kernel > saves&restores it, then A6 can be regarded as scratch Correct. > Is the RKM inaccurate/incomplete here? I don't have to push&pop A6 > in my Server routine, do I? You don't have to, the kernel does it for you. Just don't make any assumptions as to the contents of the a6 register when your interrupt server gets started. (interrupt handlers have ExecBase in a6 though) The new RKM are accurate and complete, but I suggest you also buy the "Guru's Guide, Meditation #1 - Interrupts" by Carl Sassenrath, the creator of the Exec. >Brian Rhodefer Valentin -- The Goddess of democracy? "The tyrants Name: Valentin Pepelea may distroy a statue, but they cannot Phone: (215) 431-9327 kill a god." UseNet: cbmvax!valentin@uunet.uu.net - Ancient Chinese Proverb Claimer: I not Commodore spokesman be