Which free software could acquire 48 bits color depth pictures from a scanner ?

On Sun, 16 Nov 2008 16:55:35 -0500, nospam wrote:

In article , mark raif
wrote:

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran operating systems.
Not early mainframes that have nothing to do with the topic. "The topic" which
you so quickly try to avoid at every turn. Trolls are like that. That's all they
can bring to a discussion, to get that attention that they so desperately crave,
from anyone or anything possible.


If someone hands me a complimentary copy of
Photoshop,
I merely thank them, then after they have left I throw it in the waste-basket,
where it's always belonged. I wouldn't dare even give it to a friend, I wouldn't
want them to have to put up with that Adobe nonsense.

so ebay it to a stranger.

I have more respect than that for someone I don't know. But to you? I'd sell it
to you in a heartbeat for $1 less than the going rate.

In article , mark raifr
wrote:

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran operating systems.

mainframes certainly did, but if you want to restrict it to desktop
computers, that's fine too. the macintosh was a 32 bit machine since
its introduction in january 1984 and photoshop debuted on the mac in
1990, appearing on windows a couple of years later with 32 bit math
internally on both platforms.

and cpu bus width isn't the determining factor either. even on an 8
bit computer, one can do higher precision math, it just takes more
instructions.

mark raifr wrote:
On Sun, 16 Nov 2008 16:55:35 -0500, nospam wrote:

In article , mark raif
wrote:

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran operating systems.
Not early mainframes that have nothing to do with the topic.

I wasn't aware that the 80386 (introduced 1985, produced in significant
numbers from 1986) )was used in mainframes instead of in desktop
computer.
Windows 3.1 came in 1992, which at least to me is 6 years _AFTER_ the
introduction of 32-bit math.

jue

On Sun, 16 Nov 2008 17:40:34 -0500, nospam wrote:

In article , mark raifr
wrote:

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran operating systems.

mainframes certainly did, but if you want to restrict it to desktop
computers, that's fine too. the macintosh was a 32 bit machine since
its introduction in january 1984 and photoshop debuted on the mac in
1990, appearing on windows a couple of years later with 32 bit math
internally on both platforms.

Too bad that never applied to the majority of Photoshop users that are on PCs.
They've had to deal with a 16-bit math package in Photoshop all these years, but
never realizing it.

btw: The operating system does not define the math bit-depth of the software.
You could very well have been running the same 16-bit Photoshop on your Mac all
those years without even realizing it. Mac users never have been too bright,
they want some corporation to wipe their ass for them and pay out the ass for
having them do so.

mark raif added these comments in the current discussion du jour
....

While all operating systems can load and view 32 and 64 bit
color-depths (usually by averaging to lower bit-depths if your
display and software can't handle it), that is where it ends.
The moment you use an editor with only 16-bit math you are
starting to truncate valuable data during each process due to
rounding errors. Most will never notice it due to the
limitations of their display, but this loss does exist. This
is why they put up with it in PhotoShop for so long, what you
can't see won't hurt you. For those that paid dearly to get
every bit-depth in their images they would like to have all
that information retained.

Let's say for example that you only have a 4-bit depth math
package. 4-bit floating-point math means that only FOUR
significant binary characters can be used to do the math. Any
operation done on any color values will be truncated by
averaging mathematically. A decimal (not binary) example: If
you have an RGB pixel's red value of 38756 on a 16-bit color
depth (values of 0 to 65535), and you want to reduce that by
33%, you end up with a value of 12789.48. In a 4-bit math
platform that is rounded to 12790.00 Only 4-significant digits
may be used, the 89.48 is rounded to a value of 90. (Keeping
in mind that in a 4-bit math depth which is all performed in
binary, using values 0-7, so the significant digits in decimal
(values 0-9) as presented here for examples, becomes much less
than this, even greater rounding is done sooner. This is only
provided as a quick example of what happens.) As more editing
operations are performed those approximated color values will
carry over their errors and be duplicated. Each time an
operation is performed on that data then more math errors will
be introduced, always being rounded-off (lost) in the math
functions. This is why it's a good rule-of-thumb to always use
at least the same or higher floating-point math depth as your
image color-depth. Perform 2 or more operations on any set of
pixels with a lower bit-depth math platform and you may have
drastically changed your color values by the time you are
done. This is precisely why PhotoShop could never incorporate
the more advanced Lanczos-8 resampling algorithms. Its math
platform was/is just incapable of doing the calculations
necessary to retain the image details during rotations and
resizings. The best that PhotoShop could ever offer was (and
is) simplistic bicubic interpolations, always resulting in
muddy images and soft edges due to lost details every time
those operations are performed.

32 bits of binary floating point are only good for about 8 or
maybe nine significant digits hence prior to the 60-bit CDC
mainframes contracted by DoD back in the 1960s and 1970s, IBM
mainframes, mini-computers and such needed double precision to
get to 11-12 digits after the decimal point. But, there was never
any rounding, it was really simply binary truncation, a far worse
problem.

There are also issues involving whether the floating point is in
hardware or software wrt rounding or truncating.

None of this had diddly to do with color depth unless you are
saying that any advantage is destroyed by a processor, data bus,
O/S or something else throw away what color there is.

But, on an ordinary PC even with a 32 bit O/S, 64 bit arithmetic
can still be simulated IF the data is grabbed in two chunks,
processed, then stored back. However, that would be deathly slow.

This is the first I've ever heard of Photoline, but then I'm
hardly a pro. But, 15 years ago? What motherboard or CPU was
even remotely capable of floating point math at 32
bits/channel? Or, am I again misunderstanding you?

32-bit math has been around since Windows 3.1 if you installed
the System32 math package, and more commonly found in the very
first versions of Windows 95, the very first version of Win95
happening in 1993 (if you were one of the alpha testers for
MS, as I was). Windows 3.1 was primarily 16-bit math but
allowed you to use 32-bit software if your CPU allowed for it
and if you installed some accessory files (the origin of the
"system32" Windows folder, there used to be just a "system"
folder in Win3.1). Photoline, originally called Photoline32,
recently renamed to just Photoline, was named that just
because it was the ONLY graphic editor that fully supported
the new 32-bit math platform that started during the Windows
3.1 to Window 95 bridge years.

Can't really speak to that as I was neither a system programmer
nor an application one for Win 3.1 or any other version. But,
what good would any of this be in the first place with the
motherboards and memory of that era? Win 3.1 could barely stay up
for more than an hour without crashing, so I'm sorry, but I can't
see how this might work. I clearly do not know, I just wonder,
that's all.

A slight error, checking online I see now that Photoline32 was
released in 1995. I only know I was using it on the very first
versions of Windows 95 when Win95 was released to the public
(as the Windows-95 beta version due to a marketing lawsuit
that Gates didn't want to deal with, in-house memos that I was
privy to). The 2 year discrepancy about Photoline due to my
alpha-testing phase of Windows 95 which started in 1993. I
only recall that I latched onto Photoline as soon as I
discovered its existence. It's been the backbone of my graphic
editing platform all these years due to how much more it can
do and how much more accurately it can do it. If someone hands
me a complimentary copy of Photoshop, I merely thank them,
then after they have left I throw it in the waste-basket,
where it's always belonged. I wouldn't dare even give it to a
friend, I wouldn't want them to have to put up with that Adobe
nonsense.

If I am following this at all, I'm still confused - for purely
curiosity reasons - why anyone would want to use 32 bit math in a
64-bit world (now).

This has been an excellent theoretical discussion and I thank you
for taking the time to write all of this. Now, could you please
return to Earth and explain how I might be affected or the OP,
who simply wants - for some reason - to do some free 48 bit work.
I stopped doing anything in FORTRAN more than 20 years ago, and
about the time that Photoline was released, I'd come to a mind-
boggling conclusion: that my PC was to do useful work for me, not
be a hobbyist play toy.

I assume you have some professional need for all of this
technocracy or maybe you just love pushing a PC to it's limits. I
am envious of your ability, so again, I thank you and I think
I'll go back to watching TV. grin

--
HP, aka Jerry

"Never attribute to malice that which can be adequately explained
by stupidity!" - Hanlon's Razor

nospam added these comments in the current discussion du jour
....

In article , mark
raif wrote:

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

In the world of computer science or PCs? I programmed on a 60-bit
word CDC mainframe for a number of years from the late 1970s to
1985 when I went over to the Dark Side and became a
supervisor.Those things were fantastic with at least 11 decimal
digits of precision after the decimal point. CDC originally was
formed to build computers for DoD that could do floating point far
more quickly than the IBMs of the day that needed double-precision
in SW that was inherently very slow.

If someone hands me a complimentary copy of
Photoshop,
I merely thank them, then after they have left I throw it in
the waste-basket, where it's always belonged. I wouldn't dare
even give it to a friend, I wouldn't want them to have to put
up with that Adobe nonsense.

so ebay it to a stranger.

--
HP, aka Jerry

"Never attribute to malice that which can be adequately explained
by stupidity!" - Hanlon's Razor

mark raifr added these comments in the current discussion du
jour ...

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran
operating systems. Not early mainframes that have nothing to
do with the topic. "The topic" which you so quickly try to
avoid at every turn. Trolls are like that. That's all they can
bring to a discussion, to get that attention that they so
desperately crave, from anyone or anything possible.

this is precisely the confusion I have, mark. even after reading
your excellent treatise, I admit to being very confused, but in my
case, it really isn't necessary that I be conversant here.



--
HP, aka Jerry

"Never attribute to malice that which can be adequately explained
by stupidity!" - Hanlon's Razor

nospam wrote:
HEMI-Powered wrote:

mark,

He's not 'mark', he's the P&S troll posting so many crazy rants in here
recently with dozens of different names. Don't take him seriously.


my computer math is pretty rusty but what does 16 bit
floating point or whatever vs. 32 bit or even 64 bit have anything
at all to do with color depth?

it doesn't. photoshop uses 32 bit math internally (or 64 bit in cs4)
when making calculations on an 8 bit per channel image.

nospam added these comments in the current discussion du jour
....

32-bit math has been around since Windows 3.1

32 bit math was available *long* before that.

I believe we are discussing desktop computers and home ran
operating systems.

mainframes certainly did, but if you want to restrict it to
desktop computers, that's fine too. the macintosh was a 32
bit machine since its introduction in january 1984 and
photoshop debuted on the mac in 1990, appearing on windows a
couple of years later with 32 bit math internally on both
platforms.

and cpu bus width isn't the determining factor either. even
on an 8 bit computer, one can do higher precision math, it
just takes more instructions.

one more time, please, isn't what is really necessary is 64-bit
math? and, yes, you are obviously correct that any bit-length math
can be done IF one is willing to spend enough instructions and
enough CPU cycles to fetch the data, compute a result of some
algorithm, then write it back out across the short bus. my point in
an earlier reply was that this is a fairly academic debate since
this isn't either 1984 or 1995, we now have vastly superior HW
architectures, but we are also plagued with O/S and app bloatware
so severe that even Moore's Law is overwhelmed.

At one time, talented programmers hand optimizing assembly code
could wring fantastic performance out of even 8bit CPUs yet now, we
have reached the point where heat buildup prevents any increase in
clock speeds now requiring parallel processing. And, so the saga
continues.

--
HP, aka Jerry

"Never attribute to malice that which can be adequately explained
by stupidity!" - Hanlon's Razor

nospam wrote:
In article , Ray Fischer
wrote:

32-bit pixels is different from being able to handle a 64-bit address
space.

right.

Whining that PS cn't do 32-bit pixels for all functions is
silly given that there's no output device that can handle even a
16-bit range.

16 bit printing is supported on cs4/mac.

Non sequitur. I referred to devices, not support for drivers.

--
Ray Fischer