Hi, this boils down to:
i←63
t←? 1000 ⍴ i
t = 2÷⍨ {⍵×2} fasteach t
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
On Thu, Aug 06, 2020 at 08:50:27PM -0400, Christian Robert wrote:
∇z←(F fasteach) A;shp;vec;nub;res;idx
shp←⍴A ⍝ Shape and ...
vec←,A ⍝ ... ravel of arg.
nub←∪ vec⍝ Vector of unique elements.
nub←∪ vec⍝ Vector of unique elements.
res←F¨nub
Gentlemen,
thanks, fixed in SVN 1334.
Best Regards,
Jürgen
On 8/7/20 4:33 PM, Kacper Gutowski
wrote:
On Thu, Aug
06, 2020 at 08:50:27PM -0400, Christian Robert wrote:
∇z←(F fasteach) A;shp;vec;nub;res;idx
It fix the "Smaller reproducer:" but does not fix the whole problem:
$ apl -v
BUILDTAG:
-
Project:GNU APL
Version / SVN: 1.8 / 1334M
Build Date: 2020-08-07 19:19:29 UTC
Build OS: Linux 5.7.11-200.fc32.x86_64 x86_64
config.status: 'DYNAMIC_LOG_WANTED
This is some intermittent bug..
with less than 255 elements the all elements are always equal
t←? 255 ⍴ i
+/t = 2÷⍨ {⍵×2} fasteach t
255
t←? 512 ⍴ i
+/t = 2÷⍨ {⍵×2} fasteach t
50
t←? 300 ⍴ i
+/t = 2÷⍨ {⍵×2} fasteach t
9
256 elements random number of ele
Indeed, as Christian wrote, the fix is not sufficient. But I still
believe it's in the branch I previously mentioned as commenting out the
whole optimization by flipping #if 1 to 0 at PrimitiveFunction.cc:319
fixes it as far as I can see (at cost of performance of course).
Other examples in t
I've just discovered that you get a syntax error if you try to localise
a system variable in a named lambda:
lll←{⍳⍵⊣⎕io←0;⎕io}
lll 8
SYNTAX ERROR
lll[1] λ←⍳⍵⊣⎕IO←0;⎕IO
^ ^
Is it supposed to work that way? Or is that a bug?
