Re: [darktable-user] Optimising GPU memory usage in dt 4.4.1 on Windows ?

[Remco Viëtor]

On mercredi 12 juillet 2023 22:35:10 CEST marc.ca...@gmail.com wrote:
>  MAX MEM ALLOC:2048 MB

That's not the maximum darktable can use, it's the maximum opencl can get in 
one allocation request (a quarter of the total GPU memory). See e.g.
https://registry.khronos.org/OpenCL/sdk/3.0/docs/man/html/clGetDeviceInfo.html

And of course, there usually are other programs also wanting GPU memory 
(desktop manager, Firefox, ...)

Remco




darktable user mailing list
to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org

[darktable-user] Optimising GPU memory usage in dt 4.4.1 on Windows ?

[marc.cabuy]

In a test to optimize GPU memory for Opencl execution of dt, I have tried
different updates in "preferences > processing > cpu / gpu / memory > tune
OpenCL performance" and in "preferences > processing > cpu / gpu / memory >
darktable resources". 

But when checking in "darktable-log.txt", I seem to get all the time the
same memory allocation, only 2 GB from the 8 GB available GPU memory as
shown hereunder. 

Now I am fine with the response time of dt execution on my computer. But
this seems not in line with what I read in the dt 4.4 documentation.

Does anybody have further advice about Opencl GPU memory optimisation for dt
4.4.1 on Windows on a NVIDIA GeForce RTX?

Thank you,

Marc.

EXTRACT from darktable-log.txt

--

version: darktable 4.4.1

start: 2023:07:12 20:56:44

 

 0,4468 [dt_get_sysresource_level] switched to 3 as `unrestricted'

 0,4468   total mem:   32607MB

 0,4469   mipmap cache:4075MB

 0,4469   available mem:   28531MB

 0,4469   singlebuff:  32607MB

 0,4469   OpenCL tune mem: WANTED

 0,4469   OpenCL pinned:   WANTED

[opencl_init] opencl related configuration options:

[opencl_init] opencl: ON

[opencl_init] opencl_scheduling_profile: 'default'

[opencl_init] opencl_library: 'default path'

[opencl_init] opencl_device_priority: '+0/+0/*/+0/+0'

[opencl_init] opencl_mandatory_timeout: 725

[opencl_init] opencl library 'OpenCL.dll' found on your system and loaded

[opencl_init] found 2 platforms

[opencl_init] found 2 devices

 

[dt_opencl_device_init]

   DEVICE:   0: 'NVIDIA GeForce RTX 2080 with Max-Q Design'

   PLATFORM NAME & VENDOR:   NVIDIA CUDA, NVIDIA Corporation

   CANONICAL NAME:   nvidiacudanvidiageforcertx2080withmaxqdesign

   DRIVER VERSION:   535.98

   DEVICE VERSION:   OpenCL 3.0 CUDA, SM_20 SUPPORT

   DEVICE_TYPE:  GPU

   GLOBAL MEM SIZE:  8192 MB
<

   MAX MEM ALLOC:2048 MB
<

   MAX IMAGE SIZE:   32768 x 32768

   MAX WORK GROUP SIZE:  1024

   MAX WORK ITEM DIMENSIONS: 3

   MAX WORK ITEM SIZES:  [ 1024 1024 64 ]

   ASYNC PIXELPIPE:  YES

   PINNED MEMORY TRANSFER:   WANTED

   MEMORY TUNING:WANTED

   FORCED HEADROOM:  400

   AVOID ATOMICS:NO

   MICRO NAP:0

   ROUNDUP WIDTH:16

   ROUNDUP HEIGHT:   16

   CHECK EVENT HANDLES:  1024

   PERFORMANCE:  5.938

   TILING ADVANTAGE: 0.000

   DEFAULT DEVICE:   NO

   KERNEL BUILD DIRECTORY:   C:\Program
Files\darktable\share\darktable\kernels

   KERNEL DIRECTORY: C:\Users\Marc
Cabuy\AppData\Local\Microsoft\Windows\INetCache\darktable\cached_v1_kernels_
for_NVIDIACUDANVIDIAGeForceRTX2080withMaxQDesign_53598

   CL COMPILER OPTION:   -cl-fast-relaxed-math

   KERNEL LOADING TIME:   11.4726 sec

 



darktable user mailing list
to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org