[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: exception systems

|   Date: Thu, 18 Apr 1996 15:26:10 -0400
|   From: Olin Shivers <shivers@lambda.ai.mit.edu>
|   Cc: blume@cs.Princeton.EDU, will@ccs.neu.edu, rrrs-authors@mc.lcs.mit.edu
|   Reply-To: shivers@ai.mit.edu

|   What you are proposing certainly provides a way to get very cheap
|   interrupts -- it is the way that Henry Massalin gets low-overhead
|   interrupts in Synthesis, for example.
|   But I don't like the semantics of thinking of it as an unplanned-for
|   subroutine call. It's really just that the interrupt handler is borrowing
|   the current thread's resources to execute for a little while. But I
|   think of that handler as another entity; this resource-stealing trick
|   is just an efficiency hack.

That's only if you have a thread-oriented tasking model or program.
For some things it is the right thing.  For others it is not.

Since it is easy and cheap to build such a threading model on top of
the low-overhead interrupt delivery mechanism, while it does not force
you to use that model, I don't see how implementing it that way at the
bottom level makes you think in any particular way at the higher