Hi Karl --

> It seems to be this that causes the problem when depth is 32 bits.

Yes, 32-bit depth forms are not meant to have a color map. See class comment in ColorForm.

> This doIt will cause same problem from a workspace:
> (1 to: (1 bitShift:32)) collect:[:i | i = 100]

Because it will try to allocate about 35 GiB of memory. See my previous answer.

> We should not use this for bit depth > 16 as it will cause problems

Maybe we should move the algorithm #copyWithColorsReducedTo: from Form (32-bit depth) down to ColorForm (up to 16-bit depth).

Best,
Marcel
Am 29.06.2022 20:19:06 schrieb karl ramberg <karlramberg at gmail.com>:
Hi,
It seems to be this that causes the problem when depth is 32 bits.


Form >> #copyWithColorsReducedTo:
...

colorMap := (1 to: (1 bitShift: depth)) collect:
....


This doIt will cause same problem from a workspace:

(1 to: (1 bitShift:32)) collect:[:i | i = 100]

We should not use this for bit depth > 16 as it will cause problems


Best,
Karl




On Wed, Jun 29, 2022 at 2:40 PM Marcel Taeumel <marcel.taeumel at hpi.de [mailto:marcel.taeumel at hpi.de]> wrote:

Hi Karl --

In a 64-bit image, that algorithm tries to allocate about 35 GiB of memory. CMD+Dot will not work because we are sitting in a primitive during that allocation.

I think that Form >> #copyWithColorsReducedTo: was originally (2001) used to reduce colors in 16-bit forms or maybe 8-bit forms, given a 32-bit image.

(1 << 32) * 8. "34 359 738 368 bytes"
(1 << 32) * 4. "17 179 869 184 bytes"
(1 << 16) * 4. "262 144 bytes"
(1 << 8 )* 4. "1024 bytes"

We cannot #collect: this:

(1 to: 1 << 32) collect: [:ea | ...].

:-)

In #copyWithColorsReducedTo: it looks like that we expect a color map. In 32-bit forms, there is no color map. Maybe we should move this algorithm down to ColorForm? See its class comment.

Best,
Marcel
Am 28.06.2022 22:49:43 schrieb karl ramberg <karlramberg at gmail.com [mailto:karlramberg at gmail.com]>:
Hi,

If I do Form>>copyWithColorsReducedTo:256 on any form the image stops responding.
No debug log.


This is on a updated Trunk image.

I'm on a Windows machine.

Best,
Karl



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20220630/06013d35/attachment.html>