From Newsgroup: comp.lang.lisp


I was trying to wrap my head the >>> operator

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Unsigned_right_shift

The unsigned right shift (>>>) operator returns a number whose
binary representation is the first operand shifted by the
specified number of bits to the right. Excess bits shifted off
to the right are discarded, and zero bits are shifted in from
the left. This operation is also called "zero-filling right
shift", because the sign bit becomes 0, so the resulting number
is always positive.

This is what I understood but it looks too complicated

(defun signed->unsigned (num bits)
(if (< (- (1+ (expt 2 (1- bits)))) num 0)
(+ (expt 2 bits) num)
(if (< num (expt 2 bits))
num
(logand (1- (expt 2 bits)) num))))

(defun fixed-width-ash (width x n)
(if (> x 0)
(logand (1- (ash 1 width)) (ash x n))
(fixed-width-ash width
(signed->unsigned x 32) n)))

-9 >>> 2
1073741821

(fixed-width-ash 32 -9 -2)
1073741821

I'm looking for any insights on this.

PS. Anyone know what happened to PJB, his online presence seems to have
been discontinues.

https://stackoverflow.com/questions/1691292/how-to-do-bit-wise-zero-filling-right-shift-in-scheme
https://stackoverflow.com/questions/26151644/why-is-there-no-unsigned-left-shift-operator-in-java
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp


Ar an c||igi|| l|i d|-ag de m|! Deireadh F||mhair, scr|!obh Madhu:

I was trying to wrap my head the >>> operator

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Unsigned_right_shift

The unsigned right shift (>>>) operator returns a number whose
binary representation is the first operand shifted by the
specified number of bits to the right. Excess bits shifted off
to the right are discarded, and zero bits are shifted in from
the left. This operation is also called "zero-filling right
shift", because the sign bit becomes 0, so the resulting number
is always positive.

This is what I understood but it looks too complicated

Something like:

(defun fixed-width-ash (width value count)
(ash (logand value (lognot (ash -1 width))) count))

Construct an integer of the desired width that is all 1s, logand it with VALUE, then shift by COUNT.

PS. Anyone know what happened to PJB, his online presence seems to have
been discontinues.

I hadnrCOt followed him, worth reading?

https://stackoverflow.com/questions/1691292/how-to-do-bit-wise-zero-filling-right-shift-in-scheme
https://stackoverflow.com/questions/26151644/why-is-there-no-unsigned-left-shift-operator-in-java

--
rCyAs I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after fourteen pints of stoutrCO
(C. Moore)
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

Aidan Kehoe <kehoea@parhasard.net> writes:

Ar an c||igi|| l|i d|-ag de m|! Deireadh F||mhair, scr|!obh Madhu:

Something like:

(defun fixed-width-ash (width value count)
(ash (logand value (lognot (ash -1 width))) count))

Construct an integer of the desired width that is all 1s, logand it
with VALUE, then shift by COUNT.

Your compiler will probably recognize that this is equivalent to:

(defun fixed-width-ash (width value count)
(ash (logandc2 value (ash -1 width)) count))

But I would write it that way in the first place because I like to
remember that the PDP-10 had an ANDCM instruction...
--
Alan Bawden
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp


Ar an s|-|| l|i d|-ag de m|! Deireadh F||mhair, scr|!obh Alan Bawden:

Aidan Kehoe <kehoea@parhasard.net> writes:

Ar an c||igi|| l|i d|-ag de m|! Deireadh F||mhair, scr|!obh Madhu:

Something like:

(defun fixed-width-ash (width value count)
(ash (logand value (lognot (ash -1 width))) count))

To comment on my own code; this wonrCOt do the desired fixed-width treatment of COUNT is positive, the logand would need to be done on the result of the #'ash. But the discussion was about implementing >>>, where COUNT is negative.

Construct an integer of the desired width that is all 1s, logand it
with VALUE, then shift by COUNT.

Your compiler will probably recognize that this is equivalent to:

(defun fixed-width-ash (width value count)
(ash (logandc2 value (ash -1 width)) count))

But I would write it that way in the first place because I like to
remember that the PDP-10 had an ANDCM instruction...

I donrCOt expect to remember any other PDP-10 instructions, but I might remember
this one. Thanks!
--
rCyAs I sat looking up at the Guinness ad, I could never figure out /
How your man stayed up on the surfboard after fourteen pints of stoutrCO
(C. Moore)
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp


* Aidan Kehoe <87jz0utvlt.fsf@parhasard.net> :
Wrote on Thu, 16 Oct 2025 21:55:26 +0100:

Alan Bawden:

Aidan Kehoe <kehoea@parhasard.net> writes:
Something like:

(defun fixed-width-ash (width value count)
(ash (logand value (lognot (ash -1 width))) count))

To comment on my own code; this won't do the desired fixed-width treatment of COUNT is positive, the logand would need to be done on the result of the #'ash.
But the discussion was about implementing >>>, where COUNT is negative.

Construct an integer of the desired width that is all 1s, logand it
with VALUE, then shift by COUNT.

Your compiler will probably recognize that this is equivalent to:

(defun fixed-width-ash (width value count)
(ash (logandc2 value (ash -1 width)) count))

something like this then, does it miss out any case?

(defun fixed-width-ash (width value count)
(if (< count 0)
(ash (logand value (lognot (ash -1 width))) count)
(ash (logandc2 value (ash -1 width)) count)))

But I would write it that way in the first place because I like to remember that the PDP-10 had an ANDCM instruction...

I don't expect to remember any other PDP-10 instructions, but I might remember
this one. Thanks!

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

E R R A T A ERRATA jftr

* In <m3ms5s3atl.fsf@pison.robolove.meer.net> :
I Wrote on Wed, 15 Oct 2025 12:34:22 +0530:
[...]

This is what I understood but it looks too complicated

(defun signed->unsigned (num bits)
(if (< (- (1+ (expt 2 (1- bits)))) num 0)
(+ (expt 2 bits) num)
(if (< num (expt 2 bits))

^^^^^
this condition should be (-1 < num (expt 2 bits))

num
(logand (1- (expt 2 bits)) num))))

(defun fixed-width-ash (width x n)
(if (> x 0)
(logand (1- (ash 1 width)) (ash x n))
(fixed-width-ash width
(signed->unsigned x 32) n)))

^^^^ 32 here should be `width'

-9 >>> 2
1073741821

(fixed-width-ash 32 -9 -2)
1073741821

[...]

PS. Anyone know what happened to PJB, his online presence seems to have
been discontinues(d)

(PJB pascal j was of course a one time fairly prolific former poster CLL, who could be usually counted on for a review)
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

Madhu <enometh@meer.net> writes:

something like this then, does it miss out any case?

(defun fixed-width-ash (width value count)
(if (< count 0)
(ash (logand value (lognot (ash -1 width))) count)
(ash (logandc2 value (ash -1 width)) count)))

Uh, guys, my point was that (logandc2 x y) and (logand x (lognot y))
compute _exactly_ the same thing...
--
Alan Bawden
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp


* Alan Bawden <86plac3j8e.fsf@williamsburg.bawden.org> :
Wrote on Fri, 24 Oct 2025 02:29:53 -0400:

Madhu <enometh@meer.net> writes:

something like this then, does it miss out any case?

(defun fixed-width-ash (width value count)
(if (< count 0)
(ash (logand value (lognot (ash -1 width))) count)
(ash (logandc2 value (ash -1 width)) count)))

Uh, guys, my point was that (logandc2 x y) and (logand x (lognot y))
compute _exactly_ the same thing...

Oops. cut-paste typo on my part, sorry

Aidan had written

* <87wm4uu3j6.fsf@parhasard.net>
|(defun fixed-width-ash (width value count)
| (ash (logand value (lognot (ash -1 width))) count))
|
|Construct an integer of the desired width that is all 1s, logand it
|with VALUE, then shift by COUNT.

* <87jz0utvlt.fsf@parhasard.net>
| To comment on my own code; this wonrCOt do the desired fixed-width
| treatment of COUNT is positive, the logand would need to be done on
| the result of the #'ash. But the discussion was about implementing
| >>>, where COUNT is negative.

So I wanted to add a alternate branch for (< count 0) which would call
regular ash and do the logand, while retaining the proposed form (either version,) for the consequent branch

if i hadn't C-yed the wrong kill I would have pasted

(if (< count 0)
(ash (logand value (lognot (ash -1 width))) count)
(logand (1- (ash 1 width)) (ash value count)))

but it would have still been wrong, given modular arithmetic

leaving me hopelessly confused again. I have to find a reliable text.

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

On 2025-10-24, Alan Bawden <alan@csail.mit.edu> wrote:

Madhu <enometh@meer.net> writes:

something like this then, does it miss out any case?

(defun fixed-width-ash (width value count)
(if (< count 0)
(ash (logand value (lognot (ash -1 width))) count)
(ash (logandc2 value (ash -1 width)) count)))

Uh, guys, my point was that (logandc2 x y) and (logand x (lognot y))
compute _exactly_ the same thing...

So I'm looking for the logtrunc function and it's not there!

Right, it's something I added in TXR Lisp, which is not in Common Lisp.

(defun lsh (width value count) ;; ash -> arithmetic shift; lsh -> logical
(ash (logtrunc value width) count))

logtrunc truncates an integer to the specified number of binary
digits, producing a nonnegative value. Negative integers are taken to be in the familiar "infinite two's complement", where we have a mantissa followed by an infinite number of sign bits that are 1, so that (logtrunc -1 8) produces 255, etc.
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

So I'm looking for the logtrunc function and it's not there!

Do you mean something like:

(defun logtrunc (n bits)
(mod n (ash 1 bits)))

aka

(defun logtrunc (n bits)
(logand n (1- (ash 1 bits))))

?


=== Stefan
--- Synchronet 3.21d-Linux NewsLink 1.2

From Newsgroup: comp.lang.lisp

On 2026-03-12, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

So I'm looking for the logtrunc function and it's not there!

Do you mean something like:

(defun logtrunc (n bits)
(mod n (ash 1 bits)))

aka

(defun logtrunc (n bits)
(logand n (1- (ash 1 bits))))

Yes, these look equivalent to me. A compiler could recognize
these patterns and contract to a "logtrunc" operation that
operates on the representation to do the truncation directly.
--
TXR Programming Language: http://nongnu.org/txr
Cygnal: Cygwin Native Application Library: http://kylheku.com/cygnal
Mastodon: @Kazinator@mstdn.ca
--- Synchronet 3.21d-Linux NewsLink 1.2