Re: Avoid treating the stack as an array [Re: "Back & Forth" is back!]

mhx@iae.nl (mhx) writes:

The experimental PARAMS| a | construct does not support 'OF and tries
to keep integer locals in a register.

Great.  And using the same order as {: ... :} is also great.  Now if
only (LOCAL) (which is used by the reference implementation of {:
... :}) used the same mechanism.

I just tried VICHECK1 with PARAMS| ... | instead of {: ... :}

401       336   gforth-fast (AMD64)
179       132   lxf 1.6-982-823 (IA-32)
182       119   VFX FX Forth for Linux IA32 Version: 4.72 (IA-32)
241       159   VFX Forth 64 5.43 (AMD64)
163       175   iforth-5.1 mini (AMD64)
182             iforth-5.1 mini using PARAMS|

Looking at the code, you store these registered locals on the locals
stack before the IF, and then load them into registers again after the
IF, and then reload them after every call (so apparently the registers
you use for them are caller-saved in iforth).  And the problem in this
code is that ever local is used at most once between calls, so storing
it in a caller-saved register results in no better code than storing
it in memory.

Let's see how the 3DUP.3 example fares:

   instr. bytes  system
    28    103    Gforth AMD64
    16     44    iforth 5.0.27 (plus 20 bytes entry and return code)
     8     11    iforth 5.0.27 PARAMS| (plus 20 bytes entry and return code)
     7     19    lxf 1.6-982-823 32-bit
    32    127    SwiftForth 4.0.0-RC89 (calls LSPACE)
    26     92    VFX Forth 64 5.11 RC2

Yes, in the right setting PARAMS| is very nice, too bad it's not used
for (LOCAL) (or directly for {:).

- anton
--
M. Anton Ertl  http://www.complang.tuwien.ac.at/anton/home.html
comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
     New standard: https://forth-standard.org/
   EuroForth 2024: https://euro.theforth.net