nope, I’m actually trying to open a file: /home/alb/Downloads/DSC03922.ARW
[alb@redspot:~]$ ls -lh /home/alb/Downloads/DSC03922.ARW
-rw-r--r-- 1 alb alb 25M dic 20 2016 /home/alb/Downloads/DSC03922.ARW
EDIT also, if I try to open a directory with raw files, I get a screen full of bombs…:
hm this second listing looks like it fails to create the device, possibly caused by another one of these missing features. since you’re saying you reinstalled your drivers i very much suspect the cause there. unfortunately i haven’t used an amd card since i think 2010, so i have absolutely no clue about installing drivers there.
not sure this mobile radeon would be powerful enough to make working with it fun. still i suppose several seconds would be faster than current darktable (because it always processes the full image, usually with heavy things like wavelet denoise and local laplacians in the default pipeline). i’m guessing this is another candidate for the (probably broken) LOD switch in the esoteric tab.
haha looks great. i see in your log it does not find bin/modules/denoise/noop.comp.spv. do you really not have this file? if you make, does it give you an error about this particular compilation unit?
silly me… I was compiling without a proper config.mk and didn’t notice that…
now it works! It’s not super smooth but considering it’s running at full scale, it’s actually pretty darn impressive! hats off!
I will definitely try to play with this now that I got it working 
Thanks a lot for your help!
…and yes, if it’s too slow, the simple stuff seems to still work with LOD=2 or higher, if you need to work lowres. quite sure the guided filter drawing is not correct then.
Thanks for reply. I’ll continue to search. Last week it was not as fast as with a graphic card but still really fast (less sucks !). The hardware seems ok. Just remains to find out yo make it work.
iirc @asn had some success trying it on an amd card. maybe he’ll remember which driver works?
A while back I ran it just fine on the open driver (amdgpu) on my RX 480.
I just have the Linux distribution defaults running.
Linux Kernel 5.9.1
Mesa 20.2.1 which provides libvulkan_radeon
Hi !
So, I tried again … could not stop after my last failure to compile.
Note : I’m know my way thru computers & OS, but I’m not developers … compile & code are mostly mysteries to me.
I’m running Kubuntu 20.10
uname -a : Linux ultra 5.8.0-25-generic #26-Ubuntu SMP x86_64 GNU/Linux
gcc --v : gcc version 10.2.0 (Ubuntu 10.2.0-13ubuntu1)
make --v : GNU Make 4.3 Built for x86_64-pc-linux-gnu
So, with this, I did
git clone --recursive https://github.com/hanatos/vkdt.git
cd bin && cp config.mk.example config.mk
make all
and then the difficulties arise
I get a long message, which includes a mention that
CMake Warning at cmake/Modules/LibFindMacros.cmake:263 (message):
WARNING: MISSING PACKAGEWe could not find development headers for Pugixml. Do you have the
necessary dev package installed? This package is NOT REQUIRED and you may
ignore this warning but by doing so you may miss some functionality of
rawspeed.
As its optional, I ignore the pugixml (that I did not manage to compile+install last time).
ending with
-- Configuring incomplete, errors occurred!
See also "/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeOutput.log".
See also "/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeError.log".
make[2]: *** [Makefile:10: ../built/ext/rawspeed/Makefile] Error 1
make[2]: Leaving directory '/home/gllm/vkdt/ext'
make[1]: *** [Makefile:29: ext] Error 2
make[1]: Leaving directory '/home/gllm/vkdt'
make: *** [Makefile:6: all] Error 2
So i do
more /home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeError.log :
Performing C++ SOURCE FILE Test RAWSPEED_HAVE_CXX_FLAG_WOLD_STYLE_CASTS failed with the following output:
Change Dir: /home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeTmp
Run Build Command(s):/usr/bin/gmake cmTC_44bde/fast && gmake[2]: Entering directory '/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeTmp'
/usr/bin/gmake -f CMakeFiles/cmTC_44bde.dir/build.make CMakeFiles/cmTC_44bde.dir/build
gmake[3]: Entering directory '/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeTmp'
Building CXX object CMakeFiles/cmTC_44bde.dir/src.cxx.o
/usr/bin/g++ -std=c++14 -Wall -Wextra -Wcast-qual -Wextra -Wextra-semi -Wformat=2 -DRAWSPEED_HAVE_CXX_FLAG_WOLD_STYLE_CASTS -std=c++14 -Wall -Wext
ra -Wcast-qual -Wextra -Wextra-semi -Wformat=2 -fPIE -Wold-style-casts -std=c++14 -o CMakeFiles/cmTC_44bde.dir/src.cxx.o -c /home/gllm/vkdt/bui
lt/ext/rawspeed/CMakeFiles/CMakeTmp/src.cxx
g++: error: unrecognized command-line option '-Wold-style-casts'; did you mean '-Wold-style-cast'?
gmake[3]: *** [CMakeFiles/cmTC_44bde.dir/build.make:66: CMakeFiles/cmTC_44bde.dir/src.cxx.o] Error 1
gmake[3]: Leaving directory '/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeTmp'
gmake[2]: *** [Makefile:121: cmTC_44bde/fast] Error 2
gmake[2]: Leaving directory '/home/gllm/vkdt/built/ext/rawspeed/CMakeFiles/CMakeTmp'
And here I am, stuck with this
g++: error: unrecognized command-line option ‘-Wold-style-casts’; did you mean ‘-Wold-style-cast’?
Not sure where to change this 
Many thanks for any help !!
Gllm
hmm actually i never tried to compile without pugixml, thanks for the reminder. that’ll probably mean to convert cameras.xml from rawspeed and put into something other, like rawtherapee’s camconst.json or so. anyways at this point to read useful white points etc for the raws it’ll be quite useful to have pugixml for now (need to work a bit to remove it).
at least in debian it would be as simple as apt install libpugixml-dev to install it. i’m told it’s one of the less bloaty xml libraries, not that it’s any help.
the other thing: i’m usually building with clang 9 or 10. whic version gcc is that you’re using? and is that also what you declared in config.mk? because the default config.mk.example will also use clang. but is this test actually something that should succeed or is it just some configure test case that tests for said compiler option?
I can confirm this is also the case for debian sid.
I can even run vkdt without debug, not as fast I believed however, not fun but usable.
With debug I get this error:
[qvk] validation layer: Validation Error: [ VUID-VkPushConstantRange-size-00298 ] Object 0: handle = 0x560f51b91810, type = VK_OBJECT_TYPE_DEVICE; | MessageID = 0xed1e4a95 | vkCreatePipelineLayout() call has push constants index 0 with offset 0 and size 156 that exceeds this device’s maxPushConstantSize of 128. The Vulkan spec states: size must be less than or equal to VkPhysicalDeviceLimits::maxPushConstantsSize minus offset (Vulkan® 1.2.277 - A Specification (with all registered extensions))
vkdt: qvk/qvk.c:95: vk_debug_callback: Assertion `0’ failed.
Aborted
I’ve also tried different ways to install amdgpu-pro drivers, but no success there.
oh interesting. can you post me a backtrace which one module has these offending large push constants? maybe i can reduce it a bit.
I get this error as soon as I launch vkdt. I don’t even the thumbnails screen.
Here is the console output (not sure if that’s what you request).
RESEND (more complete) log.txt (191.4 KB)
The gdb output says nothing I’m afraid.
okay i’m guessing here but i think i got this. could you git pull and git submodule update -f please and try again?
good guess ! I can run vkdt with debug and colors are better too
Thanks Hanatos for your reply.
I apt install libpugixml-dev and it unblocked the compilation !!
I was not sure why I did not try this (shame on me) …
When compiling, I get this at some point
make[1]: Entering directory '/home/glllm/vkdt/src'
ls: cannot access 'pipe/modules/*/*.tesc': No such file or directory
Compilation seems to go on and succeed, but when I attempt to start vkdt, I get
[gui] monitor [0] DP-0 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
INTEL-MESA: warning: Performance support disabled, consider sysctl dev.i915.perf_stream_paranoid=0
vkdt: qvk/qvk.c:95: vk_debug_callback: Assertion `0' failed.
Aborted (core dumped)
I looked for this dev.i915.perf_stream_paranoid on the internetz … not sure what to do.
uname -r → 5.8.0-25-generic
nvidia-settings → Linux-x86_64 450.80.02 (running a GeForce GTX 1070 w/ 8GB NVRAM)
Sudo changed dev.i915.perf_stream_paranoid to 0
tried again → yet another core dumped
Where is the core dumped ?
Edit : compile with debug argument
[gui] monitor [0] DP-0 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain
Thanks
Gllm
so is this a laptop? you’ve got an intel GPU as well as an nvidia GPU in your setup? how did you configure them? are you using the modesetting driver? the nvidia GPU needs to be able to draw on your screen.
can you run ./vkdt -d all ~/Pictures or similar? this assertion probably means some vk error has been caught, but it’s not verbose enough to tell you which. also running make clean && make debug might help tracking down things (it enables the validation layers).
HI @hanatos !
Thanks for your reply ! Much appreciated
I’m working on a desktop (I7 7700k + NVidia GFX 1070 + 32GB RAM) but your question made me remember some setting I changed in my BIOS long ago to try leverage on both my real GPU and my iGPU.
Maybe that’s a use case not managed at the moment 
So, I changed the setting in the BIOS, make clean and make all again
When compiling, there is a fair number of warnings.
Done with gcc version 10.2.0 (Ubuntu 10.2.0-13ubuntu1)
I now see vkdt start, with an empty window :
:~/vkdt$ ./bin/vkdt /media/gllm/mypath/myphoto.NEF
[gui] monitor [0] DP-0 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[gui] no display profile file display.DP-0, using sRGB!
[gui] no display profile file display.DP-0, using sRGB!
[ERR] individual config /media/gllm/mypath/myphoto.NEF.cfg not found, loading default!
[rawspeed] load /media/gllm/my_path/myphoto.NEF in 64ms
[ERR] failed to read shader '/home/gllm/vkdt/bin/modules/denoise/noop.comp.spv': No such file or directory!
[ERR] running the graph failed (VK_ERROR_INVALID_EXTERNAL_HANDLE)!
I tried to pass a folder instead
[gui] monitor [0] DP-0 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[gui] no display profile file display.DP-0, using sRGB!
[gui] no display profile file display.DP-0, using sRGB!
[rawspeed] load /media/gllm/my_path/myphoto1.NEF in 69ms
[ERR] failed to read shader '/home/gllm/vkdt/bin/modules/denoise/noop.comp.spv': No such file or directory!
[ERR] [thm] running the thumbnail graph failed on image '/media/gllm/my_path/myphoto2.NEF'!
[rawspeed] load /media/gllm/my_path/myphoto2.NEF in 202ms
[ERR] failed to read shader '/home/gllm/vkdt/bin/modules/denoise/noop.comp.spv': No such file or directory!
[ERR] [thm] running the thumbnail graph failed on image '/media/gllm/my_path/myphoto2.NEF'!
[rawspeed] load /media/gllm/my_path/myphoto3.NEF in 187ms
[ERR] failed to read shader '/home/gllm/vkdt/bin/modules/denoise/noop.comp.spv': No such file or directory!
[ERR] [thm] running the thumbnail graph failed on image '/media/gllm/my_path/myphoto3.NEF'!
Finally, I launched vkdt with an additional argument -d all
No visible error … not more luck
What else can i do ?
One side note : I’m working on a 3840*2160 screen, the interface is quite … visible 
Thanks
Gllm
…isn’t that exactly the same issue as we had in this thread above? likely your bin/config.mk is outdated and does not define the GLSL compiler? if you copy over bin/config.mk.example again and edit to taste? at least it contains a glsl compiler. probably during make all your .spv files did not get built.
well yes it scales with height of the window, doesn’t care about pixels. maybe not the most clever way of doing it, but at least i can work on it on all my screens (1200p, 1440p, 4k).