Clinton Ingram

The sad thing is, the majority of devs will continue using System.Drawing because that's what they know. And a large percentage of them will take a Microsoft-authored package over literally...

That's the exact approach I use in MagicScaler. I make an effort to grab a SIMD-aligned section of a rented array here: https://github.com/saucecontrol/PhotoSauce/blob/master/src/MagicScaler/Core/PixelBuffer.cs#L179-L190 If the GC moves the array, which...

NT stores are one of those things where you really need to be sure you're smarter than the processor before you do it. Sort of the same as prefetch these...

I'm sure you would have heard about it by now if ImageSharp had any overruns 😆 It's more a time saver because it gives you instant feedback when you screw...

Yep, that'd be the one. I can't think of a way to SIMDfy huffman decoding since you're parsing an unknown number of bits per symbol, but I'll study up on...

Just a note in case anyone else wants to look at this before I get time. I forgot the values being encoded/decoded here are quantized DCT coefficients, so they're 16-bit...

I would add to that, FMA changes the equation as well, because in olden times you were looking at 4 cycles for float mul + 4 cycles for float add...

> So are we saying here that it's actually best to stick with `Avx` and `Vector256`? I hate to drop an "it depends" on you but... I'd say the general...

At the risk of muddying the issue a bit, although N-dimensional interpolation can be useful to have as a fallback, in practice it would rarely if ever be used. Real-world...

Do these help? https://github.com/mm2/Little-CMS/blob/master/src/cmsintrp.c#L623 https://github.com/mm2/Little-CMS/blob/master/src/cmsintrp.c#L855 First link is 3D tetrahedral, and second extends that to 4D (the float 4D one is macro-heavy so I linked the integer version). That file...