Somehow, this makes sense, yes. We don't know how to divide by anything
but a matrix, so the right rank is 2, but we know how to multiply two
things that are more than matrices, so it's ok to have a left right _.
It took me some time to understand why you were mentioning A%.A, as
matrix multip
This thing is turning into fascinating reading, thanks.
There is nevertheless a small tidbit of information concerning the
creation of numbered locales I would like to add.
As pointed out:
locs=: cocreate @''@>i.8
I would add that special code is accepted as well:
locs =: cocreate &'' i.8
On Sat, May 21, 2022 at 9:55 AM Robert Cyr wrote:
> I would add that special code is accepted as well:
> locs =: cocreate &'' i.8
> The first line will work for any verb, the second is specific to cocreate
> and, as far as I can tell, not documented.
It could be better documented, but this actu
On Sat, May 21, 2022 at 3:39 AM Adrien Mathieu
wrote:
> It took me some time to understand why you were mentioning A%.A, as
> matrix multiplication is not commutative. Turns out, by reading the
> documentation, that x%.y is actually (%.y) +/ .* x, not x +/ .* %.y as I
> implicitly assumed. Btw, wh
That's very cool!
A =: 3000 4000?.@$0
B =: 4000 2000?.@$0
10 timespacex 'C =: A +/ .* B'
4.15287 6.71109e7
0&T.@''^:4 ''
4
10 timespacex 'C =: A +/ .* B'
2.30104 6.71109e7
JVERSION
Engine: j904/j64avx/linux
Beta-d: commercial/2022-05-19T20:41:10
Library: 9.04.01
Qt IDE: 2.0.3/5.12.8(5.
No, but the code to divide the operation into tasks erroneously splits the
job even when there's only one thread, and also makes too many tasks.
We'll tune that for the next beta.
Henry Rich
On Sat, May 21, 2022, 5:28 PM Jan-Pieter Jacobs
wrote:
> That's very cool!
>
> A =: 3000 4000?.@$0
>
> B
I am confused, for the dimension of A and B, the product */3000 4000 2000
should exceed the threshold and BLAS is used so that 0 T. should have no
effect on it.
can you check what is the result of (9!:58)"0 i. 3
and also set _1(9!:58)"0 i.3 to disable BLAS before testing
If the J binaries is co
904-beta-d is available for android and raspberry pi.
Questions/bug reports/etc. should go to the beta forum.
On Fri, May 20, 2022 at 11:07 PM Eric Iverson
wrote:
> J904-beta-d is available for windows/linux/macos.
>
> If you already run J904-beta, then upgrade is easy:
>load'pacman'
>'
I changed to use j mm instead of blas by default (except on openmp+nonavx).
It was faster when using multiple threads, and it should definitely be faster
for a multithreaded j app, since it uses the unified job queue.
On Sun, 22 May 2022, bill lam wrote:
I am confused, for the dimension of A
Currently J is on version 903, and soon 904. On the jsoftware.com website,
versions all the way back to J602 are installable.
There is no backward compatibility, so old scripts often only run on specific
versions of J.
It would be good to have an environment manager, which can manage the
insta
I have installed J903 on my Ubuntu 18.06 LTS machine. I have also installed all
the relevant QT libraries. Note that every version of J80X jQT seems to work
fine on my machine.
However, running jqt in J903 gives the following error:
./jqt: /usr/lib/x86_64-linux-gnu/libQt5Core.so.5: version `Qt_
I don't have the expertise you seek Jon, but I have cross posted this to the
general forum, which I think is the place that this sort of a problem is
addressed.
Cheers, bob
> On May 21, 2022, at 22:16, 'Jon Hough' via Programming
> wrote:
>
> I have installed J903 on my Ubuntu 18.06 LTS mach
12 matches
Mail list logo
