CUDA et OPENCL

[JacoTux]

Ta carte GPU semble compatible CUDA , c'est curieux que tu n'aies pas ses entrées dans les préférences.
Voilà ce que donne le retour de console à l'ouverture de dt chez moi.

Code :

$ darktable -d opencl -d perf
darktable 4.5.0+1014~g5f25f79812
Copyright (C) 2012-2023 Johannes Hanika and other contributors.

Compile options:
  Bit depth              -> 64 bit
  Debug                  -> DISABLED
  SSE2 optimizations    -> ENABLED
  OpenMP                -> ENABLED
  OpenCL                -> ENABLED
  Lua                    -> ENABLED  - API version 9.2.0
  Colord                -> ENABLED
  gPhoto2                -> ENABLED
  GMIC                  -> ENABLED  - Compressed LUTs supported
  GraphicsMagick        -> ENABLED
  ImageMagick            -> DISABLED
  libavif                -> ENABLED
  libheif                -> ENABLED
  libjxl                -> ENABLED
  OpenJPEG              -> ENABLED
  OpenEXR                -> ENABLED
  WebP                  -> ENABLED

See https://www.darktable.org/resources/ for detailed documentation.
See https://github.com/darktable-org/darktable/issues/new/choose to report bugs.

    0,4770 [dt_get_sysresource_level] switched to 2 as `large'
    0,4770  total mem:      15829MB
    0,4770  mipmap cache:    1978MB
    0,4770  available mem:  10821MB
    0,4770  singlebuff:      989MB
    0.4917 [opencl_init] opencl library 'libOpenCL' found on your system and loaded, preference 'default path'
    0.6357 [opencl_init] found 2 platforms
    0.6358 [check platform] platform 'Intel(R) OpenCL Graphics' with key 'clplatform_intelropenclgraphics' is NOT active
[opencl_init] found 1 device

[dt_opencl_device_init]
  DEVICE:                  0: 'Quadro P620'
  PLATFORM NAME & VENDOR:  NVIDIA CUDA, NVIDIA Corporation
  CANONICAL NAME:          nvidiacudaquadrop620
  DRIVER VERSION:          535.113.01
  DEVICE VERSION:          OpenCL 3.0 CUDA, SM_20 SUPPORT
  DEVICE_TYPE:              GPU, dedicated mem
  GLOBAL MEM SIZE:          1985 MB
  MAX MEM ALLOC:            496 MB
  MAX IMAGE SIZE:          16384 x 32768
  MAX WORK GROUP SIZE:      1024
  MAX WORK ITEM DIMENSIONS: 3
  MAX WORK ITEM SIZES:      [ 1024 1024 64 ]
  ASYNC PIXELPIPE:          NO
  PINNED MEMORY TRANSFER:  NO
  USE HEADROOM:            400Mb
  AVOID ATOMICS:            NO
  MICRO NAP:                250
  ROUNDUP WIDTH & HEIGHT    16x16
  CHECK EVENT HANDLES:      128
  TILING ADVANTAGE:        0.000
  DEFAULT DEVICE:          NO
  KERNEL BUILD DIRECTORY:  /usr/share/darktable/kernels
  KERNEL DIRECTORY:        /home/jacotux/.cache/darktable/cached_v2_kernels_for_NVIDIACUDAQuadroP620_53511301
  CL COMPILER OPTION:      -cl-fast-relaxed-math
  KERNEL LOADING TIME:      0.0868 sec
[opencl_init] OpenCL successfully initialized. internal numbers and names of available devices:
[opencl_init] 0 'NVIDIA CUDA Quadro P620'
    0.7800 [opencl_init] FINALLY: opencl is AVAILABLE and ENABLED.
[opencl_init] opencl_scheduling_profile: 'default'
[opencl_init] opencl_device_priority: '*/!0,*/*/*'
[opencl_init] opencl_mandatory_timeout: 200
[dt_opencl_update_priorities] these are your device priorities:
[dt_opencl_update_priorities] image preview export thumbs preview2
[dt_opencl_update_priorities] 0 -1 0 0 -1
[dt_opencl_update_priorities] show if opencl use is mandatory for a given pixelpipe:
[dt_opencl_update_priorities] image preview export thumbs preview2
[dt_opencl_update_priorities] 0 0 0 0 0
[opencl_synchronization_timeout] synchronization timeout set to 200
[dt_opencl_update_priorities] these are your device priorities:
[dt_opencl_update_priorities] image preview export thumbs preview2
[dt_opencl_update_priorities] 0 -1 0 0 -1
[dt_opencl_update_priorities] show if opencl use is mandatory for a given pixelpipe:
[dt_opencl_update_priorities] image preview export thumbs preview2
[dt_opencl_update_priorities] 0 0 0 0 0
[opencl_synchronization_timeout] synchronization timeout set to 200

Et la suite en passant une image dans la chambre noire
A préciser que 15 modules s'activent à ce moment, soit par défaut soit par préférence personnelle.

Code :

346,9090 [dt_dev_load_raw] loading the image. took 0,286 secs (1,782 CPU)
  347,5330 [histogram] took 0,000 secs (0,000 CPU) scope draw
  347,6953 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  347,7639 [dt_opencl_check_tuning] use 1584MB (headroom=ON, pinning=OFF) on device `NVIDIA CUDA Quadro P620' id=0
  347,7642 [dev_pixelpipe] took 0,000 secs (0,000 CPU) initing base buffer [full]
  347,7796 [histogram] took 0,000 secs (0,000 CPU) scope draw
  347,7823 [dev_pixelpipe] took 0,018 secs (0,033 CPU) [full] processed `rawprepare' on GPU, blended on GPU
  347,7896 [dev_pixelpipe] took 0,007 secs (0,008 CPU) [full] processed `temperature' on GPU, blended on GPU
  347,8422 [dev_pixelpipe] took 0,053 secs (0,057 CPU) [full] processed `highlights' on GPU, blended on GPU
  347,9113 [dev_pixelpipe] took 0,069 secs (0,232 CPU) [full] processed `hotpixels' on CPU, blended on CPU
  347,9314 [opencl copy_buffer_to_buffer] could not copy buffer on device 0: CL_MEM_OBJECT_ALLOCATION_FAILURE
  347,9647 [opencl_demosaic] markesteijn problem 'CL_MEM_OBJECT_ALLOCATION_FAILURE'
  347,9690 [default_process_tiling_opencl_roi] [full] couldn't run process_cl() for module 'demosaic' in tiling mode: CL_MEM_OBJECT_ALLOCATION_FAILURE
  347,9690 Error: process_tiling_CL  [full]        demosaic              (  0/  0) 6252x4162 scale=1,0000 --> (  0/  0) 1664x1108 scale=0,2662 device=0 (nvidiacudaquadrop620), CL_MEM_OBJECT_ALLOCATION_FAILURE
  347,9691 pixelpipe process CL      [full]        demosaic              (  0/  0) 6252x4162 scale=1,0000 --> (  0/  0) 1664x1108 scale=0,2662 couldn't run module on GPU, falling back to CPU
  349,4801 [resample_plain] plan 0,000 secs (0,000 CPU) resample 0,134 secs (1,528 CPU)
  349,5319 [dev_pixelpipe] took 1,621 secs (17,001 CPU) [full] processed `demosaic' on CPU, blended on CPU
  349,5896 [dev_pixelpipe] took 0,058 secs (0,052 CPU) [full] processed `denoiseprofile' on GPU, blended on GPU
  349,6145 [dev_pixelpipe] took 0,025 secs (0,025 CPU) [full] processed `lens' on GPU, blended on GPU
  349,6940 [histogram] took 0,000 secs (0,000 CPU) scope draw
  349,8014 [dev_pixelpipe] took 0,187 secs (1,050 CPU) [full] processed `cacorrectrgb' on CPU, blended on CPU
  349,8084 [dev_pixelpipe] took 0,007 secs (0,007 CPU) [full] processed `exposure' on GPU, blended on GPU
  349,8112 [dev_pixelpipe] took 0,003 secs (0,002 CPU) [full] processed `colorin' on GPU, blended on GPU
  349,8139 [dt_ioppr_transform_image_colorspace_cl] IOP_CS_LAB-->IOP_CS_RGB took 0,002 secs (0,002 GPU) [channelmixerrgb]
  349,8181 [dev_pixelpipe] took 0,007 secs (0,003 CPU) [full] processed `channelmixerrgb' on GPU, blended on GPU
  349,8228 [dev_pixelpipe] took 0,005 secs (0,004 CPU) [full] processed `colorbalancergb' on GPU, blended on GPU
  349,8284 [dev_pixelpipe] took 0,006 secs (0,005 CPU) [full] processed `filmicrgb' on GPU, blended on GPU
  349,8319 [dt_ioppr_transform_image_colorspace_cl] IOP_CS_RGB-->IOP_CS_LAB took 0,003 secs (0,003 GPU) [colorout]
  349,8468 [dev_pixelpipe] took 0,018 secs (0,015 CPU) [full] processed `colorout' on GPU, blended on GPU
  349,8613 [dev_pixelpipe] took 0,015 secs (0,066 CPU) [full] processed `gamma' on CPU, blended on CPU
  349,8614 [opencl_profiling] profiling device 0 ('NVIDIA CUDA Quadro P620'):
  349,8614 [opencl_profiling] spent  0,0373 seconds in [Write Image (from host to device)]
  349,8614 [opencl_profiling] spent  0,0045 seconds in rawprepare_1f
  349,8614 [opencl_profiling] spent  0,0049 seconds in whitebalance_1f_xtrans
  349,8614 [opencl_profiling] spent  0,0056 seconds in highlights_initmask
  349,8614 [opencl_profiling] spent  0,0076 seconds in highlights_dilatemask
  349,8614 [opencl_profiling] spent  0,0000 seconds in [Write Buffer (from host to device)]
  349,8614 [opencl_profiling] spent  0,0228 seconds in highlights_chroma
  349,8614 [opencl_profiling] spent  0,0000 seconds in [Read Buffer (from device to host)]
  349,8614 [opencl_profiling] spent  0,0107 seconds in highlights_opposed
  349,8614 [opencl_profiling] spent  0,0566 seconds in [Read Image (from device to host)]
  349,8614 [opencl_profiling] spent  0,0142 seconds in markesteijn_initial_copy
  349,8614 [opencl_profiling] spent  0,0130 seconds in [Copy Buffer to Buffer (on device)]
  349,8614 [opencl_profiling] spent  0,0010 seconds in denoiseprofile_precondition_Y0U0V0
  349,8614 [opencl_profiling] spent  0,0284 seconds in denoiseprofile_decompose
  349,8614 [opencl_profiling] spent  0,0037 seconds in denoiseprofile_reduce_first
  349,8614 [opencl_profiling] spent  0,0000 seconds in denoiseprofile_reduce_second
  349,8614 [opencl_profiling] spent  0,0069 seconds in denoiseprofile_synthesize
  349,8614 [opencl_profiling] spent  0,0030 seconds in [Copy Image (on device)]
  349,8614 [opencl_profiling] spent  0,0010 seconds in denoiseprofile_backtransform_Y0U0V0
  349,8615 [opencl_profiling] spent  0,0013 seconds in md_vignette
  349,8615 [opencl_profiling] spent  0,0209 seconds in md_lens_correction
  349,8615 [opencl_profiling] spent  0,0009 seconds in exposure
  349,8615 [opencl_profiling] spent  0,0010 seconds in colorin_unbound
  349,8615 [opencl_profiling] spent  0,0010 seconds in colorspaces_transform_lab_to_rgb_matrix
  349,8615 [opencl_profiling] spent  0,0013 seconds in channelmixerrgb_CAT16
  349,8615 [opencl_profiling] spent  0,0023 seconds in colorbalancergb
  349,8615 [opencl_profiling] spent  0,0006 seconds in filmic_mask_clipped_pixels
  349,8615 [opencl_profiling] spent  0,0020 seconds in filmicrgb_chroma
  349,8615 [opencl_profiling] spent  0,0009 seconds in colorspaces_transform_rgb_matrix_to_lab
  349,8615 [opencl_profiling] spent  0,0026 seconds in colorout
  349,8615 [opencl_profiling] spent  0,2559 seconds totally in command queue (with 1 event missing)
  349,8641 [dev_process_image] pixel pipeline took 2,100 secs (18,564 CPU) processing `DSCF0216.RAF'
  349,8665 [histogram] took 0,000 secs (0,000 CPU) scope draw
  349,8736 [dt_dev_process_image_job] loading image. took 0,000 secs (0,002 CPU)
  349,8736 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  349,8798 [dev_process_image] pixel pipeline took 0,006 secs (0,034 CPU) processing `DSCF0216.RAF'
  349,8940 [histogram] took 0,000 secs (0,000 CPU) scope draw
  349,8956 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  349,9053 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  349,9227 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  349,9405 [dt_dev_process_image_job] loading image. took 0,000 secs (0,000 CPU)
  349,9882 [dev_pixelpipe] took 0,000 secs (0,000 CPU) initing base buffer [preview]
  349,9897 [dev_pixelpipe] took 0,001 secs (0,000 CPU) [preview] processed `rawprepare' on CPU, blended on CPU
  349,9917 [dev_pixelpipe] took 0,002 secs (0,000 CPU) [preview] processed `temperature' on CPU, blended on CPU
  349,9933 [dev_pixelpipe] took 0,001 secs (0,003 CPU) [preview] processed `highlights' on CPU, blended on CPU
  349,9954 [dev_pixelpipe] took 0,002 secs (0,011 CPU) [preview] processed `hotpixels' on CPU, blended on CPU
  350,0969 [dev_pixelpipe] took 0,101 secs (0,897 CPU) [preview] processed `demosaic' on CPU, blended on CPU
  350,2305 [dev_pixelpipe] took 0,134 secs (1,295 CPU) [preview] processed `denoiseprofile' on CPU, blended on CPU
  350,2721 [dev_pixelpipe] took 0,042 secs (0,375 CPU) [preview] processed `lens' on CPU, blended on CPU
  350,4078 [dev_pixelpipe] took 0,136 secs (0,798 CPU) [preview] processed `cacorrectrgb' on CPU, blended on CPU
  350,4139 [dev_pixelpipe] took 0,006 secs (0,036 CPU) [preview] processed `exposure' on CPU, blended on CPU
  350,4198 [dev_pixelpipe] took 0,006 secs (0,027 CPU) [preview] processed `colorin' on CPU, blended on CPU
  350,4225 [dt_ioppr_transform_image_colorspace] IOP_CS_LAB-->IOP_CS_RGB took 0,002 secs (0,022 CPU) [channelmixerrgb]
  350,4381 [dev_pixelpipe] took 0,018 secs (0,168 CPU) [preview] processed `channelmixerrgb' on CPU, blended on CPU
  350,4926 [dev_pixelpipe] took 0,054 secs (0,585 CPU) [preview] processed `colorbalancergb' on CPU, blended on CPU
  350,5202 [dev_pixelpipe] took 0,028 secs (0,310 CPU) [preview] processed `filmicrgb' on CPU, blended on CPU
  350,5225 [dt_ioppr_transform_image_colorspace] IOP_CS_RGB-->IOP_CS_LAB took 0,002 secs (0,024 CPU) [colorout]
  350,5307 [dev_pixelpipe] took 0,010 secs (0,085 CPU) [preview] processed `colorout' on CPU, blended on CPU
  350,5342 [dev_pixelpipe] took 0,003 secs (0,025 CPU) [preview] processed `gamma' on CPU, blended on CPU
  350,5399 [dt_ioppr_transform_image_colorspace_rgb] RGB-->RGB took 0,006 secs (0,030 CPU) [final histogram]
  350,5411 histogram calculation 256 bins 2 -> -1 compensate 0 3 channels 1205120 pixels took 0,001 secs (0,011 CPU)
  350,5413 [histogram] took 0,007 secs (0,041 CPU) final histogram
  350,5427 [dev_process_image] pixel pipeline took 0,555 secs (4,658 CPU) processing `DSCF0216.RAF'
  350,5752 [histogram] took 0,004 secs (0,001 CPU) scope draw
  352,6869 [histogram] took 0,001 secs (0,001 CPU) scope draw
  352,8576 [histogram] took 0,001 secs (0,001 CPU) scope draw

On voit que certaines tâches sont dédiées au CPU d'autres au GPU