[Top][All Lists]

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

Re: Register VM WIP

From: Stefan Israelsson Tampe
Subject: Re: Register VM WIP
Date: Fri, 11 May 2012 22:29:11 +0200


This looks very good. i like the hole approach and this approach has the
potential to address most of the issues I have seen when disassembling
guile-2.0 output. A few notes.

1. What about growing stacks any coments if they will be easier to manage
for this setup. Can one copy the C stack logic?

2. Is there an instruction that does what call does but can be used for tail call's
when it needs it e.g. the code
 for (n = 0; n < nargs; n++)
        LOCAL_SET (n, old_fp[ip[4 + n]]);
that is missing for the tail code

3. I would appriciate if the frame is always below say 256 SCM:s of the fp stack limit
that way when preparing tail calling one doesn't usally need to check if the argument fit's
when issuing a tail call. If you compile a function that tail call more then 254 (?) arguments
then you can as well check because then be free relative the argument handling.

4. I think the logic code hook I recently investigated could easily fit into this VM engine with
using similar techniques as I described in previous mails.

Thanks for your work on this

On Fri, May 11, 2012 at 6:19 PM, Andy Wingo <address@hidden> wrote:
Hi all,

This mail announces some very early work on a register VM.  The code is
in wip-rtl ("work in progress, register transfer language".  The latter
bit is something of a misnomer.).  There is not much there yet:
basically just the VM, an assembler, and a disassembler.  Still, it's
interesting, and I thought people might want to hear more about it.

So, the deal: why is it interesting to switch from a stack VM, which is
what we have, to a register VM?  There are three overriding
disadvantages to the current VM.

 1) With our stack VM, instructions are smaller.  They do less, so you
    need more of them.  This increases dispatch cost, which is the
    largest cost of a VM.

 2) On a stack VM, there is a penalty to naming values.  Since the only
    values that are accessible to an instruction are the ones on the
    top of the stack, whenever you want to use more names, you end up
    doing a lot of local-ref / local-set operations.  In contrast an
    instruction for a register VM can address many more operands, so
    there is much less penalty to storing something on the stack.  (The
    penalty is not so much in the storage, but in the pushing and
    popping needed to access it.)

 3) Our stack VM has variable-sized stack frames, so we need to check
    for overflow every time we push a value on the stack.  This is
    quite costly.

The WIP register VM fixes all of these issues.

The basic design of the VM is: 32-bit instruction words, 8-bit opcodes,
variable-length instructions, maximum of 24-bit register addressing, and
static, relocatable allocation of constants.

Also, with the wip-rtl VM there is no stack pointer: locals are
addressed directly via the frame pointer, and the call frame for a
function is of a fixed size.  Indeed the IP and FP are the only state
variables of the VM, which makes it much easier to think about native
compilation, given the scarcity of CPU registers on some architectures.

See vm-engine.c from around line 1000 for a commented set of
instructions.  It's messy in many ways now, but hey.

As far as performance goes, we won't know yet.  But at least for a
simple loop, counting down from a billion, the register VM is a few
times faster than the stack VM.  Still, I would be happy if the general
speedup were on the order of 40%.  We'll see.

Here's that loop in rtl VM:

  (use-modules (system vm rtl))

    '((assert-nargs-ee/locals 1 2)
      (br fix-body)
      (make-short-immediate 2 0)
      (br-if-= 1 2 out)
      (sub1 1 1)
      (br loop-head)
      (mov 1 0)
      (br loop-head)
      (make-short-immediate 0 #t)
      (return 0)))

There are various ways to improve this, but its structure is like what
the stack VM produces.

Compare to the current opcode:

   scheme@(guile-user)> (define (countdown n) (let lp ((n n)) (or (zero? n) (lp (1- n)))))
   scheme@(guile-user)> ,x countdown
   Disassembly of #<procedure countdown (n)>:

      0    (assert-nargs-ee/locals 17)     ;; 1 arg, 2 locals
      2    (br :L186)                      ;; -> 30
      6    (local-ref 1)                   ;; `n'
      8    (make-int8:0)                   ;; 0
      9    (ee?)
     10    (local-set 2)                   ;; `t'
     12    (local-ref 2)                   ;; `t'
     14    (br-if-not :L187)               ;; -> 21
     18    (local-ref 2)                   ;; `t'
     20    (return)
     21    (local-ref 1)                   ;; `n'
     23    (sub1)
     24    (local-set 1)                   ;; `n'
     26    (br :L188)                      ;; -> 6
     30    (local-ref 0)                   ;; `n'
     32    (local-set 1)
     34    (br :L188)                      ;; -> 6

OK, time to set down the keyboard; been working far too much on this in
recent days.  I still need to adapt the compiler to produce RTL
bytecode.  I am going to let it sit for a week or two before touching it
again.  Comments welcome.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]