Re: State of Forth 200x
- From: "George Hubert" <georgeahubert@xxxxxxxxxxx>
- Date: 9 Jun 2006 01:01:58 -0700
Andreas Kochenburger wrote:
On 8 Jun 2006 06:32:12 -0700, "George Hubert"
<georgeahubert@xxxxxxxxxxx> wrote:
Doing cloning at run-time would add a lot of
run-time overhead
Overhead? It's just ALLOCATE and MOVE to copy a small memory region,
and perhaps adjustment of a few pointers, if any.
If the objected is ALLOTed in the dictionary then the address is known
at compile time so a smart compiler (not W32F; at least the current
version, though the STC version (not V4.2stc) that Alex McDonald and
co. are (slowly) working on might be less dumb) could use techniques
like constant folding to produce shorter/faster code.
I've nothing against ALLOCATEd objects (the NEON model supports them as
well) if an object only might be needed, or a variable number of the
same type are needed (though if they're the same type then IMO a class
makes perfect sense), however there is a noticeable overhead (even in
dumb old W32F) in using them. IMO objects that are definitely needed
should be ALLOTed for performance reasons, with ALLOCATEd objects used
only for objects that might be needed.
Andreas
-------
Et ceterum censeo, TV esse delendam.
George Hubert
.
- References:
- Re: State of Forth 200x
- From: John Passaniti
- Re: State of Forth 200x
- From: Doug Hoffman
- Re: State of Forth 200x
- From: John Passaniti
- Re: State of Forth 200x
- From: Doug Hoffman
- Re: State of Forth 200x
- From: John Passaniti
- Re: State of Forth 200x
- From: Doug Hoffman
- Re: State of Forth 200x
- From: Andrew Haley
- Re: State of Forth 200x
- From: Doug Hoffman
- Re: State of Forth 200x
- From: George Hubert
- Re: State of Forth 200x
- From: Andreas Kochenburger
- Re: State of Forth 200x
- Prev by Date: Re: New version of :REDa4
- Next by Date: Re: State of Forth 200x
- Previous by thread: Re: State of Forth 200x
- Next by thread: Re: State of Forth 200x
- Index(es):
Relevant Pages
|