@zorg thanks for the suggestion, this was the closest test! For a single static image the result is identical to when using the shader with a filter size of 1.0 (one pixel). The problem is when you add moving things, and pretty much all games would have those. This is the exact same context (sprites...

Someone on LÖVE's Discord channel suggested some tests. If you draw your sprites scaled/rotated/subpixel-translated with 'linear' filtering onto a Canvas, and then draw that Canvas (scaled up or not) with 'nearest' filtering, the result you get looks blurry like with the standard 'linear' filter: ht...

This is a pixel shader for use with modern pixel-art-style games that make use of smooth scrolling and scaling. It filters the edges of pixels, if needed, so they have an antialiased look. The size of the antialising is adjustable as a constant in the pixel shader in case you want it sharper or smoo...

That looks dope AuahDark. I consider myself good at googling and couldn't find any Lua video encoding libraries, I wonder how I missed your ls2x. Very useful.

Hi, I think the closest thing is the Canvas type, used for off-screen rendering. See these: - https://love2d.org/wiki/Canvas - https://love2d.org/wiki/Canvas:newImageData (a copy of the pixels in the canvas as an ImageData type) - https://love2d.org/wiki/ImageData:encode (saving that ImageData as a ...

Furthermore, I'm pretty sure that the big branching factor adds a performance penalty with respect to a pure atan2 + lookup implementation. See for example the benchmarks at https://love2d.org/forums/viewtopic.php?p=218409#p218409 Totally, if i had to pick a method I'd also go with ivan's atan2, it...

Hm, if it were some arbitrary float number then I would agree, but it's zero. I think it should be safe, even if either dx or dy are +0.0 or -0.0. If you remove that check, which you can do, then whenever the source and destination are at the same position the result will be whatever passes the test...

For the challenge I did it a different way (comparing the tangents inside a quadrant to pick the "slice"/octant, then picking the quadrant based on the sign of dx and dy). I also changed the order of parameters in the direction function, LÖVE usually puts the position parameter first and t...

I still don't see why the events processing time should not be included in the total measurement. I don't understand the confusion, timerStep() (alias for love.timer.step()) returns the time that's elapsed since the last call to it. So in that first code that I posted that we're talking about, when...

The goal is getting a main loop that's as generous as possible to the CPU so you have low usage/no busy-spinning, it should do a game tick and then sleep to give all the remaining time back to the OS. The loop I posted I based on the first one in the deWiTTERS game loop article . To get my head arou...