Re: Unboxed structures (yes, again, only this time for real)

>>>>> On Fri, 16 Jan 2026 19:32:21 +0100, Yuri Davidovsky said:
> 
> > On 16 Jan 2026, at 19:24, Martin Simmons <[email protected]> wrote:
> > 
> > Your code uses single-float where the 32-bits of data are hidden in the 64-bit
> > pointer, not allocated on the heap.  You should see a difference if you make
> > it use double-float.
> 
> Yep, we have figured it out already. There appears to be confusion sometimes
> in relation to immediate/boxed in discussions where it is not always clear
> if “boxed” means located on the heap, or simply a tagged pointer.

Yes, I sometimes use the term "boxed" to mean something that is encoded in a
tagged pointer.  That is possibly a misuse of the term.

> > I wouldn't trust any timings that are less than a few seconds.  This is
> > especially true on hardware where the clock speed varies according to CPU
> > usage (hence the same code runs more slowly until the speed has ramped up).
> 
> Hm, I was actually having some issue with this testing code where the
> results suddenly started fluctuating wildly, sometime 3x per run. Could not
> figure it out, I wonder if it could be the clock speed indeed.

We have also seen strange performance after reloading code, but it is usually
resolved in LispWorks by calling (clean-down t).

-- 
Martin Simmons
LispWorks Ltd
http://www.lispworks.com/

_______________________________________________
Lisp Hug - the mailing list for LispWorks users
[email protected]
http://www.lispworks.com/support/lisp-hug.html