Mask does not scale properly on HiDPI or Retina display

[civet]

Is there any API for getting devicePixelRatio, or how to fix this issue in the C++ part?

I'm using an old retina MacbookPro. a simple snippet is here:

HiseSnippet 1003.3ocsV01aaaCDlxNJHVaEXEX+.HxmjKxRk655JVwPayKtyaMIFQccCnnnfghxhHRjBTzIQan+m6+f1ijxQxodYcFX5CB5d+4Nd2QMUIorpJoB4M3U0kLj2W6GWKzY6mQ3BzjC.99wBdYISi1qtjTUwRPdd8egQr2fMP1mO7z8H4DAk0xBgdsjSYujWv0sbm9reimmOljvdEuni1O7YSnRw9xb4b.J88iPkD54jYriIF054i717vDtVph0DMqB4swdxj53L4kBm9ulWwOKmYHFghAG4XOVlmXPrgKZ+LddxzEobEB7xz1BPeWA3a8Ohmvuleag3arBvsVzsd30aY30eI3M5eBdq.RdcfzFNHcW+XphWpakXvyW4OQnYpTBT16BEmtnd0882WBZHz6VPNmMVADWaQ3ihh1ACuF9jf.nzWowWPTXJQbAoB+y3EVRRRlRDr7vsch1dGLXmwpEZTwL3pjozbVEDjheM9ji6n9eGLX6q19mvif.BeV294k7DcVKYFiOKS2RelTkvTw7+hA75v4TRBedkgWv6uA54JZNqC5oJFbbLknyBAMchMozyUzv2DsqISLuF075s1j6HP8cmNAeO7Crt2lH1zD5DzmJmq4BVX5bAUykhvYCCfTzDdBDMH3uIx5FcFuZ2YL8eXxxvgsL9EadFN7sOIXPvf6eebEzlfujqXoJnaOXvLSvbyBgQWMFdhhrGTfjDE4xSYTcnIXVPObI2nyX3RHAvxzlxwp7m4o0e15iS4cveteSgIVLAqfnBnDbsIFMk5BR04F2bFaFW7RRMSElRxqXNm2MrONpaXM975zvwhTVlWeD3ua.kN9iIRbw.Hgi9EGAXo3XolchHzdTD79.7MEkltRYl9DkLO23xUH1riRcaFFJlWbFSsCz8kOmcshvf6xaC17Kaa.0011QQoXhfqOoj0Pe6qyPM88vW+9jCHZhYcRCu1YTf2ArKfcytkKC7OfUctVVZ0snTJLdvaKsU5cVr5wrD.wgPukuah.cEX6lIMK6q6RXGr6xvMZ2kS6vsgKZIttA7t7Qe9NQXarLYdNQu7JZy8RMBfSpk1KZ18Ip355t2a8eXuczst29KEh20eJWSyVMF6sBLBmg+efwla6ti+goovfXK.2ve7ettWs8uDd2tyYvBVEG5d7OddQLrcfxfnKf9Kv49d8L8wN5HCsoBDCS9VhOBOMBGYn8ZDNZgPTAgpjui5lOM2mtkkCfIg8WIF3ejgFOBYmYA67g6.PEvU6uiRMo+2AiVq1lGrF178qgMObMr4GVCadzZXyOtF173a0FyeU874ZYgab.XL8P6xKOuCEDnyx1Eh9DHh84iA

[ustk]

@civet This issue is here for a long time and I've opened an issue:
https://github.com/christophhart/HISE/issues/269
https://forum.hise.audio/topic/4851/zoom-factor-offsets-graphics-effects/10

As soon as you use a layer, interface zoom factors other than 100% are messing up the scale

[d.healey]

@ustk Does Content.setUseHighResolutionForPanels(true); make any difference?

[Natan]

@d-healey said in Mask does not scale properly on HiDPI or Retina display:

@ustk Does Content.setUseHighResolutionForPanels(true); make any difference?

Is this Unique to Develop Branch?

[ustk]

@d-healey It does not seem to be related, or at least it does not solve the problem

[d.healey]

@Natan I don't know, I don't use Master branch :p

[ustk]

@d-healey I think the paint routine does take into account the zoom factor, but the layer part might miss this rescaling somehow. I'll have another look at the source but I'm not confident I can find anything...

[whoopsydoodle]

Is there an update to this in 2024?

It seems to still be an issue. I'm using a mask and even just zooming the view in and out inside hise, the mask keeps it's absolute pixel dimensions, completely moving around within my interface relative to everything else. This bug makes using masks completely pointless.

[aaronventure]

@whoopsydoodle nope, still a bug. You could refer to the scaling ratio as a dimensions modifier when drawing your mask.

Though that's only a bandaid.

[prehm]

does this mean masks are basically unusable?
i really need this feature..

[ustk]

@prehm Unfortunately yes, like blur and maybe some other graphic functions...

[prehm]

@ustk
That’s a bummer.
What’s wrong with blur? Seemed fine the one time I tried it

[ustk]

@prehm Very CPU heavy like shadows, except text shadow because it uses a different algorithm that we wait to be applied to other functions. I think the blur has a strange behaviour when scaling too...

[aaronventure]

@prehm It uses the old JUCE routine for drawing blur, which is about 100x slower than the melatonin version that HISE is still missing, and it's still done on the CPU.

If you want fancy graphics and visual effects, use a WebView and enjoy all the modern frontend tech.

[prehm]

@aaronventure
Cool, thanks for the hint.
I am new to programming so I‘ll have to read into that..

[Christoph Hart]

It uses the old JUCE routine for drawing blur, which is about 100x slower than the melatonin version that HISE is still missing, and it's still done on the CPU.

FYI, I have replaced the shadow functions in the scripting Graphics object with the melatonin routines in the upcoming batch of commits.

But if you want to be 100% accurate, the melatonin shadow is also rendered on the CPU but it's highly optimized withs lots of dirty tricks. The default JUCE implementation (or my failed attempt of replacing it with another algorithm) is just unoptimized code.

[ustk]

@Christoph-Hart So the shadow is optimised, but not yet the blur, right?