<div dir="auto">Because I tend to track this issue<div dir="auto"><br></div><div dir="auto"><a href="https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19232" rel="noreferrer noreferrer noreferrer noreferrer noreferrer" target="_blank">https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19232</a><br></div><div dir="auto"><br></div><div dir="auto">see latest notes there (you need llvm 15.0.4 - yeah, in this case 0.0.1 version matters) and mesa git + some merge requests manually merged, but results relatively promising - darktable starts and enjoy some 2x speedup (on Power CPU, no less!), Lux render also starts and works.</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">Ffmpeg I guess not tested yet :)</div><div dir="auto"><br></div><div dir="auto">for me ffmpeg -filters | grep opencl shows</div><div dir="auto"><br></div><div dir="auto">... avgblur_opencl V->V Apply average blur filter ... boxblur_opencl V->V Apply boxblur filter to input video </div><div dir="auto">... colorkey_opencl V->V Turns a certain color into transparency. Operates on RGB colors. ... convolution_opencl V->V Apply convolution mask to input video</div><div dir="auto"> ... deshake_opencl V->V Feature-point based video stabilization filter ... dilation_opencl V->V Apply dilation effect ... erosion_opencl V->V Apply erosion effect ... nlmeans_opencl V->V Non-local means denoiser through OpenCL ... overlay_opencl VV->V Overlay one video on top of another </div><div dir="auto">... pad_opencl V->V Pad the input video. ... prewitt_opencl V->V Apply prewitt operator ... program_opencl |->V Filter video using an OpenCL program </div><div dir="auto"> ... roberts_opencl V->V Apply roberts operator ... sobel_opencl V->V Apply sobel operator ... tonemap_opencl V->V Perform HDR to SDR conversion with tonemapping.</div><div dir="auto"> ... transpose_opencl V->V Transpose input video </div><div dir="auto"> ... unsharp_opencl V->V Apply unsharp mask to input video </div><div dir="auto">... xfade_opencl VV->V Cross fade one video with another video. ... openclsrc |->V Generate video using an OpenCL program<br></div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">for tonemapping on OpenCL this line was suggested (on Nvidia) - see below.</div><div dir="auto"><br></div><div dir="auto"><a href="https://superuser.com/questions/1549479/ffmpeg-4k-hdr-to-1080p-sdr-hardware-tonemap-conversion-opencl" rel="noreferrer noreferrer" target="_blank">https://superuser.com/questions/1549479/ffmpeg-4k-hdr-to-1080p-sdr-hardware-tonemap-conversion-opencl</a><br></div><div dir="auto"><br></div><div dir="auto">so I guess we can use it as example for trying on Radeons (comment about OpenCL 1.2 needed still apply, but rusticl exposes 3.0 :-) )</div><div dir="auto"><br></div><div dir="auto">there is also AUR package, but not sure if it tracks right branch from Karol Herbst (developer) repo ..</div><div dir="auto"><br></div><div dir="auto"><a href="https://aur.archlinux.org/packages/mesa-rusticl-git" target="_blank" rel="noreferrer">https://aur.archlinux.org/packages/mesa-rusticl-git</a><br></div><div dir="auto"><br></div><div dir="auto">I personally look at </div><div dir="auto"><a href="https://gitlab.freedesktop.org/karolherbst/mesa/-/commits/rusticl/wip_nv/">https://gitlab.freedesktop.org/karolherbst/mesa/-/commits/rusticl/wip_nv/</a><br></div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">note enabling this also require some environment flags to be set.</div><div dir="auto"><br></div><div dir="auto"><pre><code>ffmpeg -vsync 0 -hwaccel cuda \</code></pre><pre><code> -init_hw_device opencl=ocl \</code></pre><pre><code>-filter_hw_device ocl -extra_hw_frames 3 \</code></pre><pre><code>-threads 16 -c:v hevc_cuvid \</code></pre><pre><code>-resize 1920x1080 -i INPUT \</code></pre><pre><code>-vf "format=p010,hwupload,tonemap_opencl=tonemap=mobius:param=0.01:desat=0:r=tv:p=bt709:t=bt709:m=bt709:format=nv12,hwdownload,format=nv12" \</code></pre><pre><code> -c:a copy -c:s copy \</code></pre><pre><code>-c:v libx264 -max_muxing_queue_size 9999 OUTPUT
</code></pre><pre><code><br></code></pre><pre><code><br></code></pre><pre><code><br></code></pre><pre><code><br></code></pre><pre><code><br></code></pre></div></div>