Re: [darktable-dev] error during building current git on 14.04

[Josep V. Moragues]

El 28/07/16 a les 11:20, Roman Lebedev ha escrit:

>> I need the AMD closed source graphics driver because, as far as I know,
>> the open source radeon driver does not allow using the 12 bit LUT
>> present on AMD cards. And this is needed to acurately profile a monitor
>> (using ArgyllCMS or DispcalGUI), without banding artifacts.
>
> Hm, where did you read that?
> Also, if it really is the case, is this issue already reported against
> open source radeon driver?
>
> (and, does argyll actually produce > 8bit vcgt? because if it does not,
> the issue is bogus)
>
> TBN i don't really believe there is any banding even with 8-bit bit
depth..
>


I did not read it, I have seen it myself. Using the "dispcal" command
from ArgyllCMS with these params:

dispcal -y l -R

the programs shows the value for the apparent videoLUT's bit depth. With
an Intel integrated graphics, or a Nvidia Geforce gamer card, or an AMD
using the opensource radeon driver, the value shown is 8 bit.

But with an AMD card using the Catalyst binary driver for Linux, the
returned value is 10 bit.

Old ATI cards have 10 bit LUTs, current AMD cards have 12 bit LUTs. The
test used by dispcal only detects the difference between ordinary 8 bit
LUTs, and LUTs with higher bitdepths. So a returned value > 8 bit means
that the AMD 12 bit LUT is working.

And, yes, profiling the monitor with Argyll and the Catalyst driver
takes advantage of this, I have evidence of it. Profiling with and
ordinary 8 bit LUT does not show much banding i an sRGB monitor with a
good "from factory" response, with good neutral grays, and a white point
defined in OSD that does not differ too much from the target white point
you chose in Argyll, and also the gamma setting needs to be the same.

But with widegamut monitors, or monitors with a less than ideal
response, o when you can't adjust the desired white point or the gamma
on the OSD ... a 8 bit LUT shows banding that a 12 bit LUT does not
show. And when you validate the profile with DispcalGUI you can see how
many of the originally available 256 levels per channel get lost using a
8 bit LUT.

I use a Dell UP2516D (wide gamut). And, even being a hardware calibrated
monitor, I need my AMD card with 12 bit LUT to get the final icc
profile, because Dell's internal calibration and also precalibrated
modes have several flaws. These "cheap" wide gamut monitors suffer badly
from lack of uniformity. You can get rid of it enabling "Uniformity
compensation" in Dell's OSD, but this value has no effect in custom
callibrated modes (CAL1 and CAL2), only in a generic wide gamut "Custom
color mode", (quite similar to the ECI RGB v.2 colorspace) but the
uniformity compensation disables the individual color channels gain
controls ... so I can't chose the white point in OSD, so I need to let
it in native value and set te desired white point as a target in
DispalGUI when profiling ... I could'nt do that with good enough results
with a 8 bit only VideoLUT.


I have learned all this stuff thanks to the expertise of a contributor
in a spanish Canon's users forum, "Canonistas.com". His nickname is
"ColorConsultant". He has written a series of FAQ's about monitor
profiling, and buyers recommendations on monitors for photography, and
callibrating devices, that are worth reading, if you know a little of
spanish:

http://www.canonistas.com/foros/blogs/colorconsultant


Josep V. Moragues

___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] error during building current git on 14.04

[Roman Lebedev]

Yes, we understand that version bumps are troublemakers at times...

On Thu, Jul 28, 2016 at 10:20 AM, Josep V. Moragues  wrote:
>
>
> El 27/07/16 a les 13:02, Roman Lebedev ha escrit:
>
>> Yes, darktable git master now requires gtk3 3.14+ to build.
>> The 3.14 was chosen because even debian stable has that version.
>>
>> That is not a bug.
>> This was intentional.
>> Therefore this will not be fixed.
>
> I understand the reasons to match Darktable's build requirements with
> current stable distributions (Debian stable, Ubuntu 16.04) and I am not
> trying to convince the developers to change that.
>
> But currently there is a reason that makes me stick with Ubuntu 14.04
> for photo work: the AMD Catalyst binary driver. This driver is disabled
> in 16.04. And the AMDGPU Pro that will replace it (only for GCN based
> AMD cards) is not still available.
I'm also in this situation (no more fglrx and no amdgpu yet, Radeon 7970HD)

> I need the AMD closed source graphics driver because, as far as I know,
> the open source radeon driver does not allow using the 12 bit LUT
> present on AMD cards. And this is needed to acurately profile a monitor
> (using ArgyllCMS or DispcalGUI), without banding artifacts.
Hm, where did you read that?
Also, if it really is the case, is this issue already reported against
open source radeon driver?

(and, does argyll actually produce > 8bit vcgt? because if it does not,
the issue is bogus)

TBN i don't really believe there is any banding even with 8-bit bit depth..

> No AMD propietary driver for Ubuntu 16.04 means also no OpenCL
> acceleration for Darktable ... but my main concern is the LUT color depth.
>
> I expect this situation will get solved by the time Darktable 2.2 will
> be released. My current card is a Radeon R7 250, that is a low-end card,
> but I like it because is a fanless design, it matches well my silent
> self built computer. It is GCN 1.0, and AMD has announced AMDGPU support
> for it, but only with the 4.8 Linux kernel, i.e. Ubuntu 16.10. We will
> see. If this do not allow me to use 12 bit LUT, i will have to stick
> with Ubuntu 14.04 and Darktable 2.0.x. Or to buy a new card.
>
> Josep V. Moragues
Roman.

> ___
> darktable developer mailing list
> to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
>
___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] error during building current git on 14.04

[Josep V. Moragues]

El 27/07/16 a les 13:02, Roman Lebedev ha escrit:

> Yes, darktable git master now requires gtk3 3.14+ to build.
> The 3.14 was chosen because even debian stable has that version.
>
> That is not a bug.
> This was intentional.
> Therefore this will not be fixed.

I understand the reasons to match Darktable's build requirements with
current stable distributions (Debian stable, Ubuntu 16.04) and I am not
trying to convince the developers to change that.

But currently there is a reason that makes me stick with Ubuntu 14.04
for photo work: the AMD Catalyst binary driver. This driver is disabled
in 16.04. And the AMDGPU Pro that will replace it (only for GCN based
AMD cards) is not still available.

I need the AMD closed source graphics driver because, as far as I know,
the open source radeon driver does not allow using the 12 bit LUT
present on AMD cards. And this is needed to acurately profile a monitor
(using ArgyllCMS or DispcalGUI), without banding artifacts.

No AMD propietary driver for Ubuntu 16.04 means also no OpenCL
acceleration for Darktable ... but my main concern is the LUT color depth.

I expect this situation will get solved by the time Darktable 2.2 will
be released. My current card is a Radeon R7 250, that is a low-end card,
but I like it because is a fanless design, it matches well my silent
self built computer. It is GCN 1.0, and AMD has announced AMDGPU support
for it, but only with the 4.8 Linux kernel, i.e. Ubuntu 16.10. We will
see. If this do not allow me to use 12 bit LUT, i will have to stick
with Ubuntu 14.04 and Darktable 2.0.x. Or to buy a new card.

Josep V. Moragues
___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org

Re: [darktable-dev] error during building current git on 14.04

[Roman Lebedev]

On Wed, Jul 27, 2016 at 1:47 PM, Nazarii Vitak  wrote:
> Hi everyone,
>
> I tried to update the darktable through git. The command ./build.sh ended up
> in error.
> Last lines from terminal:
> "CMake Error at
> /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108
> (message):
>   Could NOT find GTK3 (missing: VERSION_OK) (Required is at least version
>   "3.14")
> Call Stack (most recent call first):
>   /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:315
> (_FPHSA_FAILURE_MESSAGE)
>   cmake/modules/FindGTK3.cmake:58 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
>   src/CMakeLists.txt:146 (find_package)
>
>
> -- Configuring incomplete, errors occurred!
> "
> As I understand it requires gtk3 >3.14, but I have found that installation
> of gtk3 3.14 brokes unity. The only way is to upgrade the system, but I
> usually do this when xx.04.2 LTS appears. Could it possibly be fixed to
> build darktable at 14.04.
> I have attached ~/darktable/build/CMakeFiles/CMakeOutput.log and
> ~/darktable/build/CMakeFiles/CMakeError.log.
Yes, darktable git master now requires gtk3 3.14+ to build.
The 3.14 was chosen because even debian stable has that version.

That is not a bug.
This was intentional.
Therefore this will not be fixed.

The solution is:
*). Either use stable darktable version (2.0.x), which still requires 3.10+
*) or keep using the git master at the moment just before gtk version
requirement was bumped
   Do note that it will no longer be considered a build of "git
master", and is perhaps more
   riskier than using the actual "git master".
*) Or update gtk, possibly by updating the system to something newer.

LTS is only about the packages present in said LTS release, it has
nothing to do whatsoever
with the choices about version requirement made by the developers of
the software...

Also see "[darktable-user] Compiling last git master failed because of
GTK-3 outdated" thread.

> If you need more information let me know how I can gather this information
> and I try to obtain it.
>
Roman.

> ___
> darktable developer mailing list to unsubscribe send a mail to
> darktable-dev+unsubscr...@lists.darktable.org
___
darktable developer mailing list
to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org