OpenGL Game Rendering Tutorial: How FXAA works

In this video, we talk about how Fast Approximate Anti-Aliasing (FXAA) works, and discuss it's advantages/disadvantages over other anti-aliasing techniques

[@ChristopherGray00]

take a shot everytime this man says fundamental

[@dangi12012]

Watchable with speed set to 1.5

[@janmarucha9138]

you're ruining the asmr experience

[@FizzyhexDev]

informative, useful and broken down well. thanks!

[@ChenHuang]

awesome.

[@qqqq4915]

Benny , will you make SSDO || / && SSAO shadowind at some point?

[@DeveloperDesmond]

Even at double speed this goes by slowly

[@klikkolee]

do you remember which video you show that a program in c++ is not faster than one in java? I want to use it as a citation on that statement.

[@thebennybox]

I believe it was video 31 of the 3D Game Engine series. However, please don't take that statement out of context. The reason there is no significant performance difference is because that program's performance is bound by the GPU's speed. Any performance difference in the languages would only slow down the CPU, not the GPU, and that's why there was no performance difference in that case.

[@klikkolee]

***** do you know if there is any performance difference at all between the languages? they seem extraordinarily similar, so I wouldn't expect any at all. If I'm quite honest, the only difference that I see between the two is that c and c++ are very language independent and java wants nothing to do with pointers. java uses a virtual machine, but I don't believe that will have a significant difference on performance.

[@thebennybox]

klikkolee There is a difference. Static compilation and virtual machine systems are two fundamentally different ways to execute code, and both have their strengths and weaknesses. In general, a virtual machine system will be slower because you need to convert intermediate code into binary instructions before you can execute it. In statically compiled languages, like C/C++, that step will be performed beforehand, and the code can be immediately executed. There are, of course, exceptions to the rule. There are cases where a virtual machine can use information gathered during code execution to generate a more efficient instruction sequence than the statically compiled instructions, simply because the statically compiled instructions wouldn't have access to that information before the code was actually executed. However, in most cases, the cost of collecting that information, calculating how it can improve efficiency, and then generating an improved instruction sequence outweighs the benefits of doing so.

[@klikkolee]

***** about how significant do you think this difference is on performance? I'm not sure if it would be even worthwhile to turn a java program to c++ when it comes to performance.

[@thebennybox]

klikkolee As I said, it depends. If it's really that big of a deal, then set up an experiment that models the type of code you'll be running, and see what type of performance you get in both languages.