GPU Next vs GPU

This page documents the known differences between vo_gpu_next and vo_gpu (that is, between --vo=gpu-next and --vo=gpu).

There is currently no support for the libmpv / render API in vo_gpu_next.

https://github.com/mpv-player/mpv/labels/vo%3Agpu-next
--dscale-antiring is not applied currently when using an orthogonal filter for quality reasons

--sharpen: Very low quality and easily replaced by (better) custom shaders (#9387, #9845)
ewa_lanczossoft, haasnsoft: These were arbitrary filters to begin with and are easily replicated by tuning ewa_lanczos
--vf=format:light=...: OOTFs are tied to the EOTF in gpu-next, and can no longer be adjusted independently
--gamma-factor and --gamma-auto: considered deprecated by design, the new BPC logic handles this much better
--fbo-format: libplacebo requires the use of floating point buffers (to store unnormalized linear light data), and there is no quality gain from going above rgba16f
--blend-subtitles=video: libplacebo always renders subtitles during the final output pass, and never blends them directly onto video frames.
--tone-mapping-max-boost: crutch for bad tone-mapping functions and lack of black point handling

gpu-next should be generally faster across the board, owing to refactors and rewrites of many key algorithms (EWA scaling, frame interpolation, color management, PRNG/debanding, etc.)
gpu-next aggressively merges shader passes wherever possible, leading to a significant reduction in the number of full-screen draw calls required per frame; in addition to this it also more aggressively reuses internal frame caches
gpu-next, unlike gpu, performs black point compensation during color management, leading to subtly different results for many of the tone mapping-related options
gpu-next generally defaults to bt.1886 instead of gamma2.2 for the display output, due to the better/different black point handling
gpu-next doesn't rotate/flip frames until the final output pass, whereas gpu applied this during the main scaling pass. This generally makes it more robust when combining custom user shaders with rotated videos

ewa_lanczos defaults to using radius 3.2383... in gpu-next, rather than the radius 3.0 of gpu.
--tone-mapping=auto: picks spline (instead of bt.2390)
--tone-mapping=gamma: scaling of the parameter has been changed slightly
--tone-mapping=bt.2390: knee point is now tunable, defaulting to 1.0 instead of the hard-coded 0.50 from gpu
--tone-mapping=linear: is now perceptually linear instead of linear light, use linearlight to get the old behavior back
--gamut-mapping-mode=auto: picks perceptual instead of clip
--hdr-compute-peak: gpu-next generally uses frame-exact results, rather than delaying detected peaks by a frame
--scale can now be used directly on window functions, removing the need for --scale-window + --scale=box combo; in addition, a few new functions and presets are available while others have been removed

TEXTURE_rot is no longer needed (it still exists for compatibility, but it's defined as an identity matrix)
The distinction between MAIN and MAINPRESUB no longer exists - either name can be used to refer to the same texture

Support for Dolby Vision content (reshaping only)
Support for GPU film grain application (AV1, H.264), see --vd-lavc-film-grain
Significantly better dynamic tone mapping, and colorimetrically accurate gamut mapping

--interpolation-preserve
--lut, --image-lut, --target-lut
--gamut-mapping-mode: most gamut mapping modes
--target-contrast
--target-colorspace-hint: cross-platform HDR passthrough, with metadata forwarding and automatic tone-mapping
--tone-mapping=bt.2446a
--tone-mapping=spline
--tone-mapping=st2094-40, --tone-mapping=st2094-10: HDR10+ and DV tone mapping curves
--tone-mapping-visualize
--tone-mapping-param for --tone-mapping=bt.2390
--inverse-tone-mapping
--allow-delayed-peak-detect
--hdr-peak-percentile
--hdr-contrast-recovery, --hdr-contrast-smoothness
--vd-lavc-film-grain=gpu/auto
--glsl-shader-opts
--corner-rounding
--icc-use-luma
--libplacebo-opts
--hwdec=vulkan

Support for custom, runtime-tunable shader parameters has been added
Two new macros linearize and delinearize are available in all shader stages, resolving to GLSL functions for going back and forth between the video's native color space and linear light

Provide feedback