Re: [PATCH v2] slof/engine.in: refine +COMP and -COMP by not using COMPILE

Hi Thomas,

On 2024-02-26 10:02:55, Kautuk Consul wrote:
> On 2024-02-23 15:04:56, Segher Boessenkool wrote:
> > On Fri, Feb 23, 2024 at 11:29:23AM +0530, Kautuk Consul wrote:
> > > > > difference (e.g. by running the command in a tight loop many times)?
> > > Running a single loop many times will not expose much because that loop
> > > (which is NOT within a Forth colon subroutine) will compile only once.
> > > In my performance benchmarking with tb@ I have put 45 IF-THEN and
> > > IF2-THEN2 control statements that will each compile once and reveal the
> > > difference in compilation speeds.
> > 
> > All of this is only for things compiled in interpretation mode anyway.
> > Even how you get the source code in (read it from a slow flash rom in
> > the best case!) dominates performance.
> > 
> > You do not write things in Forth because it is perfect speed.  Write
> > things directly in machine code if you want that, or in another high-
> > level language that emphasises optimal execution speed.  The good things
> > about Forth are rapid prototyping, immediate testing of all code you
> > write, simple compact code, that kind of goodness.  Ideal for (system)
> > firmware!
> > 
> > 
> > Segher
> 
> Yes, but SLOF will be there on the product we sell to our customers.
> Considering that there is a noticeable improvement in performance I just
> thought maybe IBM management would be interested in it.

On this note, what did you also try to understand the performance
implications of my patch ? What improvements did you observer on your
set up?