Hacker News new | past | comments | ask | show | jobs | submit login

Few things consistently blow my mind as insane graphics demos

https://www.shadertoy.com/view/4dfGzS (or basically anything on that site)

How is that 400 lines of code.

Or this one which even generates the sound on the GPU

https://www.shadertoy.com/view/4ts3z2

With the wide adoption of WebGL, it's a good time to get involved in graphics. Furthermore, GPUs are taking over esp. with the advent of machine learning (nvidia stock grew ~3x, amd ~5x last year). The stuff nvidia has been recently doing is kinda crazy. I wouldn't be surprised if in 15 years, instead of AWS, we are using geforce cloud or smth, just because nvidia will have an easier time building a cloud offering than amazon will have building a gpu.

These are some good resources to get started with graphics/games

# WebGL Programming Guide: Interactive 3D Graphics Programming with WebGL

https://www.amazon.com/WebGL-Programming-Guide-Interactive-G...

Historically, C++ has definitely been THE language for doing graphics but if you are starting these these, you would have to have really compelling reasons to start with C++ and not JavaScript and WebGL. And that's coming from someone who actually likes C++ and used to write it professionally.

# Book of Shaders

https://thebookofshaders.com/

# Game Programming Patterns

http://gameprogrammingpatterns.com/contents.html

https://www.amazon.com/Game-Programming-Patterns-Robert-Nyst...

HN's own @munificent wrote a book discussing the most important design patterns in game design. Good book applicable beyond games.

# Game engine architecture

https://www.amazon.com/Engine-Architecture-Second-Jason-Greg...

# Computer graphics: Principles and Practice

https://www.amazon.com/Computer-Graphics-Principles-Practice...

This is more of college textbook if you'd prefer that but the WebGL one is more accessible and less dry.

# Physically Based Rendering & Real-Time Rendering

These discuss some state of the art techniques in computer graphics. I'm not going to claim to have really read them but from what I've seen they are very solid.

https://www.amazon.com/Computer-Graphics-Principles-Practice...

https://www.amazon.com/Physically-Based-Rendering-Third-Impl...




Pardon me asking, Is there a way to save comments for later? This is gold, and thank you for the resources. I have been very interested in graphics since I took Computer Graphics in College (It felt like an applied linear algebra course, but I loved it. It was subsequently the only course I felt like I was challenged beyond my abilities -- I had to take the course twice to get the credit, but I loved that class)

Also just kind of asking for curiosity, do you think a language like go or rust will become popular for developing game engines? I realize game programmers are anti-GC but what if GC technology advances that the performance drop is negligible I wonder.


Lol, I'm flattered that you think it's worthy of saving. If you upvote it will show up in your profile in upvoted comments. Also, if you click on the date in the header of the comment, a favorite button should appear. When you favorite a comment it will show up in your profile in favorite comments.

> Also just kind of asking for curiosity, do you think a language like go or rust will become popular for developing game engines? I realize game programmers are anti-GC but what if GC technology advances that the performance drop is negligible I wonder.

I think it will but on some level, the needs of a game engine are different from the needs of say a DNS server. Jonathan Blow, the developer behind Braid (http://braid-game.com/) and Witness (http://store.steampowered.com/app/210970/) has been working on a language called Jai https://github.com/BSVino/JaiPrimer/blob/master/JaiPrimer.md even though he's aware of Rust and Go. He talks about some of his reasons in this video https://www.youtube.com/watch?v=TH9VCN6UkyQ

One of the things he mentions is that the game industry doesn't care about security that much (which I didn't realize until then but it makes sense) compared with a DNS server or something so his ideal language might have different design considerations than Rust.


> Is there a way to save comments for later?

Click on its timestamp to go to its page, then click 'favorite'. Favorite stories and comments are visible from your profile page. Note that these are public, so users can browse each others' favorites.


> do you think a language like go or rust will become popular for developing game engines?

That depends on the type of engine :)

Many games will be perfectly fine in a GC language. (People do write JS games all the time :). But as you approach the limits o

f a devices performance, GC has an overhead that is visible.

You can work around it - there are patterns that essentially work around the collector by recycling objects - but it's quite a bit of engineering effort, and it's a very different style.

But the big issue in GC'ed environments is that you give up control over heap growth. Working with fixed memory budgets becomes a very difficult thing to achieve. And there are few things game developers hate more than unpredictability :)

If you want to look at ongoing efforts, Amethyst is trying in Rust: https://github.com/amethyst/amethyst


Tap the time (x mins/hours ago) of the comment you want to favourite. That should take you to a page where you can favourite it


> Is there a way to save comments for later?

I usually just upvote. You can view your list of upvoted posts/comments in your profile page.


That first shader is sampling from three textures, so the image is not entirely generated by an algorithm - see iChannel0, iChannel1 and iChannel2 at the bottom of the page. (With iChannel3 being the music.)

The shaders on ShaderToy aren't quite the same as straight OpenGL fragment shaders - there is quite a bit of boilerplate code built into the site that allows ShaderToy shaders to have access to stuff like mouse coordinates and audio signal info. (Expand the 'Shader inputs' area above the code block - none of those are available in OpenGL.)

But OpenGL also has its own library[1] of functions: `texture2D`, `smoothstep`, `mix` and all the built-in vector math - and you can see all of these in action in the shader you linked. The ShaderToy boilerplate - in partnership with these libraries - is the reason the code is so concise.

[1] http://www.shaderific.com/glsl-functions/


You are right, I didn't pick the best example. But I believe everything is procedurally generated on the GPU in this demo https://www.youtube.com/watch?v=pnn6QC8a5g8. Note that all of this had to be fit into a binary of 4k or smaller.


Oh it's a fine example - I'm sorry - it didn't occur to me that my reply would sound critical! I just wanted to add some explanation, for anyone who was curious. The original post was good!


for the maximum wow effect, I suggest one of the latest shader by iq : https://www.shadertoy.com/view/4ttSWf

This is almost Pixar level graphics running realtime in your browser.


That guy's whole production is insane https://shadertoy.com/user/iq


I don't want to dis shaders on shadertoy, they are amazing toys (which is why it's called shadertoy). And, that shader you linked to is amazing as well. But, for example on my 2015 MBP it runs at 8fps in a tiny sub portion of a window. If I go fullscreen in runs in 0.3fps. Yet this same machine can run The Witness at 60fps or GTA5 at 30fps and see the entire world.

My point is the shaders on shadertoy are exercises in doing something for fun with limits (one shader and a few inputs) and not about speed or optimization or the anything to do with the way any game engine would go about to get performance.


I'd like to just point out that iq used to work for Pixar[0] :)

[0] http://www.iquilezles.org/personal/curri/curri.htm


iq has a ton more examples of these on his site (including tiny executables which generate such images and animations):

http://www.iquilezles.org/www/articles/raymarchingdf/raymarc...

http://www.iquilezles.org/prods/index.htm


> you would have to have really compelling reasons to start with C++ and not JavaScript and WebGL. And that's coming from someone who actually likes C++ and used to write it professionally.

As someone working in game programming a long time and now also spending more time in machine learning, I must disagree. I suppose I can see your implied point about getting started quickly or something like that, but there's really few compelling reasons to do graphics programming in JavaScript professionally. As a hobby or for a web-based project like a tutorial site, sure, JavaScript and WebGL make perfect sense. For most professional uses though, I can't agree.

Writing a decent performing and well-architected graphics engine is much more than just using an API like OpenGL, WebGL, or any of the successors or predecessors. Further, writing a graphics engine is not about creating an efficient island, but rather a piece of a larger eco-system. Consider that quite often some of the biggest hinderances to graphics performance are creating the data to actually be rendered (ex: via the game update), sending the data to the GPU efficiently, and still getting all kinds of other program logic to run before you even have a frame or want to display a frame. It would be a long post to go into all the details, so forgive me for oversimplifying. Among many things vital to graphics programming that C and C++ provide that JavaScript doesn't or either fails to do well:

- Control of memory allocations and memory shape/characteristics

- Maximizing cache lines (due to architecture, memory footprint, etc.)

- Multi-threading, concurrency, parallel processing

- Data pipelining

- Predictable execution

- Fast, efficient, correct math (or possibility thereof)

- Compilation tweaking/assembler output tweaking

- Efficient IO

- Fast compression/decompression algorithms

- Integration with existing toolchains for graphics including middleware, apps, and shader languages

JavaScript has made great strides in some of these areas and certainly things like web asm help. Nevertheless, most JavaScript runtimes people are using are wholly unsuitable for professional graphics programming. This is akin to people who used to say that you can write a professional game in BASIC or later, Visual Basic. Yes, it's true you can, but should you do it or start there? Probably not if your goal is to learn professional graphics programming. Have people taken such routes and become awesome programmers? Sure, but there are just as many people who never learn, grow, and develop nasty habits that are harder to unlearn than learning the difficult stuff.

If your goal is having fun and learning a few things before moving on to C++, dabble in some JS graphics programming, but since you called our professional programming as no longer being compelling to start, I have to point out that this is both wrong and bad advice to someone new. If you are struggling with the concepts, you will struggle in any language. If you personally need instant gratification and a self-esteem boost, from this point of view JS might not be bad for learning, but it won't teach you a lot of vital things and will encourage some horrific practices. The presence of GC and the lack of a proper parallel and concurrent programming model poison the experience quite a bit.

Ironically, other languages that are more friendly to parallelism and concurrency tend to be better than C, C++, and JS. A lot of the realities of graphics programming are simply caused by historical reasons, market dominance of things like x86 and companies such as nVidia, and existing knowledge, code, and toolchains. Functional languages for instance in many ways have the potential to be better suited to take advantage of the direction we're headed in with regard to numbers of cores, Moore's Law (or diminishing of), GPU design, memory architecture, and so on, but this is just hypothetical mostly. Given JavaScript is hardly known for its ability to deal with concurrency and parallelism in most contexts (no, nodejs doesn't count), again I have to find the above dubious considering you mention the future WRT nVidia.

Of course in the end, you can use things like transpiling, engines that may or might compile to JS as a runtime target, and so on and getting a running product that might be decent in the eye of the beholder. Personally, I've dabbled in JS enough to write a high-performance Voxel engine many years ago when I thought about making a game in JS as well as a skeleton of a 3D adventure game engine for a contract. I was productive enough but ran up against walls that weren't worth working around or just made things feel so kneecapped that I wondered why I was even using JS at all in its current state. The old adage holds that in the end, you can write anything in any language, but there is indeed value of selecting the best tool for the job. Even just making something run and at a reduced frame rate typically is something fundamentally different than being THE language for graphics programming.

In truth, when I first saw JS in the 90s, I thought there was no way it would reasonably do anything in 3D. Things are getting better, but I think you are dramatically underselling what people use and exists today in a professional environment. Your comment about 400 lines of code being that impressive and small seems to also hint you haven't dabbled that much in the area professionally. While your links are just fine and the above shader toy link is indeed impressive in the context of now, personally I was much more blown away when I dabbled in the demo scene and saw what people were doing with an Atari ST, Amiga, and PCs with no GPUs and on systems with less power than today's fitness trackers.

Anyway, I can't think of many compelling reasons to start in JS when IMO, it's better to learn to do things the proper way even if they can be a bit rough and punishing. JS will surely be more productive at the beginning, but also skip teaching you vital things you need to know as well as some of the fundamental primitives of graphics programming. This is the real world and no one in a professional environment cares if you are awesome in JS but don't understand the tools and best practices people actually use industry-wide. Moreover, you won't get far if you can't be productive from your first day because you tried to take the easy way out. If anything, just comparing the amount of resources for JS vs. C/C++, especially from professional vendors like nVidia, Microsoft, AMD, and so on makes me think at best, JS as a starting language for graphics programming only holds for web programmers. If anything this seems to makes things an extra layer of difficult.


Oh yeah I totally meant to get started. There's no way we'll see AAA games written in JS anytime soon. I just remember how long it takes me to setup my environment when I want to do some graphics programming.

However WebGL will outlive JS, when webassembly is introduced.


Fair enough. Definitely there need to be better resources for getting started in 3D programming, particularly in C/C++, but the barriers to entry are also related to what makes things powerful at a AAA level.

In some ways, just grabbing an engine like Unreal or Unity is a decent alternative to something like JS to learn (some even let you use JS or use other languages that also have traps). Big game engines leave a lot to be desired and at times abstracts too much or makes things like handling shader code annoying. Still, most larger game engines like these two are the closest thing to having LOGO for 3D games programming.

You can at least get stuff on the screen relatively quickly, learn a few things, and then start replacing it from there. I personally learned quite a bit way back when just decompiling or reverse engineering stuff from people much smarter than me.

Sometimes I feel like there's a lack of things like what the C64 provided for younger kids and adults today. I suppose as expectations have risen, so has complexity of getting going.


> Sometimes I feel like there's a lack of things like what the C64 provided for younger kids and adults today. I suppose as expectations have risen, so has complexity of getting going.

Although I understand the gist of this statement, I wish I had all the amazing (and cheap!) powerful stuff we have around now to play with when I was young. Also the effect of easy access to information on the internet now cannot be overstated. You can just pull up a youtube video on any subject you might be interested in immediately. It's amazing.

I remember manually typing out pages of C64 code from a magazine to generate a fractal. After typing for literally hours, the actual single fractal picture took hours to generate... Still satisfying in the end to witness it being generated pixel by pixel, but it was surely a lot of work and needed a lot of patience for a kid. Plus if I had mistyped any of that code, it would have been a big disappointment for sure. Kids nowadays have no idea how tough using computers was back then.


Yes, very true, obviously I just mean conceptually poking around. In many ways we definitely have it better now. I am infinitely more productive today. Simultaneously have gone further from teaching the low-level stuff properly.

I spent hours, days, and weeks of my life on things related to graphics/games programming such as:

- Building mouse drivers from scratch or implementing them from alternative vendors

- Implementing/Working with DOS protected mode

- Manually compressing memory/implementing swapping

- Implementing blitting from scratch

- Implementing z-buffers from scratch

- Reverse engineering consoles to steal processing power from idiotic sources just to render a tiny bit more data or later, a few more polygons

- Spending a huge chunk of cash to throw in a math co-processor into my machine at home

- Debugging code for hours only to realize things like the problems are caused by seemingly unrelated problems such as tape media is at fault, the floppy is corrupt, or the file system doesn't work the way the vendor said it does in the specs

- Rendering a scene and then going home, only to come in the next day and see it is still not done

- Converting from 72 billion formats and interpreting, finding, and/or correcting corrupt data from each one

- Rewriting entire pieces of code bases to squeeze out several more bytes of EMS and XMS

- Implementing 2 or more graphics APIs for the same game. Thank you 3dfx, S3, and many others that pained me, not to mention at a higher-level, OpenGL, DirectX, and so on.

- Doing all my work on 1 platform, then loading it on another. Thank you SGI for taking years off my life.

- Writing matrix operations in pure assembler for the simplest of operations

- Having multiple workstations for reasons such as "this one has the Matrox card in it."

The list goes on. Yeah, I don't miss those days. But I learned a lot, we all did. And the barriers to entry definitely reduced the signal to noise IMO.


There always has been and always will be problems, there are just different problems now...

We can't even imagine what kinds of crazy stuff the next generation will come up with, with all the resources they have available now. The barriers to entry are still there, but the goalposts have moved significantly.


You sound like you've had some interesting experiences in the trenches, do you have a blog or something where you share your thoughts about the craft?


Funny that, I've used that exact expression before, with Matrox framegrabbers ;-)


> # Game Programming Patterns

This is also available on the authors website: http://gameprogrammingpatterns.com/contents.html


Added.


Another one that blows my mind, for a completely different reason, is https://www.shadertoy.com/view/MddGzf . That's a simple breakout clone that stores all game data in a separate render buffer. I love that the author managed to implement a game loop in a system that was never designed to support it.


I'm going to shamelessly push

http://webglfundamentals.org


> ...I wouldn't be surprised if in 15 years, instead of AWS, we are using geforce cloud...

NVidia's already there with the hardware:

http://www.nvidia.com/object/nvidia-grid.html


And they are building that thing that will let you game in the cloud and stream it to your computer (is there a name for this type of product). There were startups that tried that before that generally failed but fundamentally if you manufacture the platform, sky is the limit. This will let them build out an infrastructure while still selling to the other cloud providers but eventually, they can just start keeping the secret sauce for themselves and their products.


LiquidSky is doing it now, and I'm using it as my sole gaming PC. It's amazing.


Thanks for the mention of this, I'm quite interested, looking at their FAQ right now, curious to see how they handle latency. Do you notice any lag? Also, what FPS do you achieve?


My ping in Overwatch is 1ms, with 40ms latency to the server from my computer at home. Your bandwidth will determine what your FPS will be, of course!


OK thanks again for the info, now very much worth a serious look.


Marking this for later. I did some graphics in the 00s.

Your JS plus WebGL comments are compelling.

Thanks for a golden comment.


I would say C more than C++, with the latter only emerging into popularity due to major game engines within the last couple of decades.


Now mostly retired, but I've been doing game dev for a couple of decades. I would say based on my career and that of my peers, it is more correct to say C or even assembler is the traditional graphics programming language with regard to games. It just depends on your definition. For that matter, if we consider things like CAD, 3D modeling/animation (using APIs/scripting), movie production, etc. to be graphics programming, again there are some differing results as to which language you use.

It also really depends on what time period we're counting as the "start" and if we value older games or periods of higher but simpler productions vs. now. For instance, I wrote quite a lot of code against NES, SNES, Genesis, Atari, Arcade machines, C64, and so on in assembler. 68000-based systems alone are such a huge chunk of sheer volume of games and most of us wrote quite a bit in pure assembler with regard to graphics.

On other systems that were faster or for different game requirements, we definitely used a ton of C, only moving to C++ later. In the case of C++, it's almost hard to even call things C++ at times. To be honest, most decent game engines I've worked on essentially use C++ to practically rewrite the language and go to great lengths to avoid some of the primary selling points of C++. It's sort of pick and choose some good things about C++ or at least powerful things while avoiding some of the burdens of C with regard to games. Obviously general C++ programming is different, so keep in mind I'm only referring to professional games programming. As things have progressed, people are indeed using more of what the newer C++ standards offer, but most of the same things still hold true. Where one has a tough argument with regard to graphics is the fact that so much now happens on the GPU, which one can argue is its own thing given stuff like shader languages.


"I would say Latin more than English, with the latter only emerging into popularity due to major nations within the last couple of centuries"


Hmmm. The first official C++ standard was published in 1998, so 'the last couple of decades' covers pretty much the entirety of languages standardised life.




Applications are open for YC Winter 2022

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: