The last real graphic programming¹ I’ve done was in the late 80s targeting the Tektronics graphics mode of the Kermit terminal software on the PC.² It was all pretty much clear the screen and draw solid rectangles in 1-bit black and white. One of these days I’m going to join the twenty-first century.
⸻
1. I suppose some hand-written PostScript code might count as well, but I wouldn’t really count things like doing a simple function graph in python to explain something to my son as graphics programming.
2. This was for a DVI previewer running on an IBM mainframe running VM/CMS. As far as I know, this code is completely lost, which is probably a good thing.
If you want to play with software rendering, here's probably the shortest code that will get an ARGB8888 2D array from main memory to the screen efficiently for all platforms using SDL2 in C https://gist.github.com/CoryBloyd/6725bb78323bb1157ff8d4175d... you'll need to do the translation from a 320x200x8-bit palletized framebuffer to ARGB yourself ;)
At least with SDL3, you don't even need the renderer or the texture anymore. SDL_GetWindowSurface to get the surface and SDL_UpdateWindowSurface to present. That's the more software-graphics you can get from my understanding of the library. SDL still does the double-buffering for you.
Great article. I particularly enjoyed the approach to creating gibs.
Although it was a tech demo, I created something like this around the mid 90s. One thing I did that I don't see mentioned in this article was I used 8x8 (or 16x16) light maps on the textures, which allowed me to easily have things like flickering torches and rockets that lit up the hallways as they shot down them. Lightmaps can also be used to "bake in" lighting if desired.
Since the light map is "only" 8x8 you can afford to do some math on each luxel (each unit in the light map) to calculate distance and line of sight to light sources to determine a brightness value. When rendering the texture, the luxel was used with a lookup table to determine the actual color of the pixel being drawn. The light maps were updated 15 times a second if I recall correctly to help performance. Thanks to DJGPP, I was using inline assembly for the rendering. Since floating point math was slow at the time I used fixed point math which optimized well. The rendering was surprisingly performant on computers of the day.
I feel like the idea of fixed-point is under-utilized and very under appreciated. There are loads of applications where this is a better choice, let alone more performant.
This was the choice that Knuth made in TeX and Metafont although it’s also at least in part because floating point implementations in the late 70s/early 80s were so inconsistent from one platform to the next that using native floating point couldn’t satisfy his requirement of identical output on all platforms (likewise, the variations of Pascal implementations meant that he also used a highly restricted subset of Pascal features and no dynamic memory allocation).
This is taking a lot of inspiration from Doom, but the actual raycasting engine is more like Doom's predecessors, the most well-known of which is probably Wolfenstein 3D: perpendicular walls, constant floor and ceiling height. Wolf3D didn't have textured floors and ceilings because of performance reasons, but several other similar games had them. Doom and IIRC Duke Nukem as well used a BSP engine which was much more flexible (walls could intersect at any angle, variable floor and ceiling heights), although the levels were still "flat" (you couldn't have several "stories" inside a level, e.g. you couldn't design a bridge that you could walk over and under).
The Build engine didn't use BSP, it treated connections between sectors as portals and rasterized the walls as (90 degree rotated) trapezoids while performing clipping against those portals. This allowed it to have dynamic wall geometry (e.g. moving trains, rotating light fixtures, etc) as well as "room-over-room" setups as long as you couldn't see both rooms at the same time (in both Blood and Shadow Warrior they found a workaround for it allowing to create more "3D" spaces by making identically shaped sectors with the floor of one sector acting as a portal to the ceiling of the other sector - supposedly this wasn't "natively" supported by the engine, but it was flexible enough for the game studios who used it -without even having access to the source- to do it themselves).
The first level of Duke Nukem 3D does use a few Build tricks - e.g. another one is that sprites can be "axis aligned" instead of following the camera and they can also have collision - this can be used to create rudimentary 3D geometry by treating each sprite as an axis aligned quad and in the first level it is used to make a bridge between two buildings (right before the level exit button).
I always loved that the bridge you mentioned could take damage and fall down, screwing you over in the very first level, unless you knew where the Jetpack was stashed.
With regard to floors, afaik even DOOM didn't do them correctly. With vertical walls, the perspective divide needs to be done only once per column of pixels for a given wall segment.
For floors, unfortunately there's no such luxury, and if I remember correctly DOOM subdivided floors into patches, and only did proper perspective at the corners, and interpolated inbetween.
For floors the perspective divide is once per row, just like for walls it's once per column.
The BSP may have led to some floor subdivisions, especially as it needs convex sectors. I don't remember if the engine would coalesce adjacent floor spans into a single one, but I hope it did.
GLQuake introduced the r_wateralpha setting, which allowed transparent water, but the maps were still compiled with visibility calculations that assumed the water surfaces were opaque. You got visual artifacts unless you enabled r_novis to ignore the pre-calculated visibility calculations. Modern computers can handle it, but this was a heavy performance cost at the time.
To work around this, people used an unofficial tool to patch the maps to support transparent water:
I find the most interesting things are the internal tools -- like the Python script to generate the gib animation, or the other Python script to generate 2D spritesheet from Blender. OP is definitely a 10x engineer who can also do good arts. This is very rare IMO. I'm very surprised to find that OP has consistent art direction.
It seemed, as a fan of the genre in the 90s, like these Renaissance Engineers were behind every major hit. I remember some of their names, they are true artists.
I have no idea the names of nearly anyone but a CEO or lead Director in the games industry of the past 15 years.
Graphics programming in the early to mid 1990s was pretty fun: write pixel data into the memory-mapped video RAM and it appears on the screen! A pointer to 0xA0000 was all you needed - no API or anything. The reason for the non-square-pixel 320×200 VGA mode they mention was that the video buffer took 64000 bytes, which fit into a 16-bit segment, making addressing it easy in 16-bit code/CPUs.
Though your extender could make things a little more annoying on that front :-P
(DJGPP and Free Pascal -which use the same "go32" extender by DJ Delorie- do not do a full linear mapping so you need to do a bit more juggling to get stuff on screen there)
So refreshing to have this NOT read a few pageS IN; words to the effect of.. “and the I asked Claude to build most of it for me then set to tweaking a few parts at the edges before asking Claude to write up this blog post”
Step 0 is missing: having a great taste. One look at the video example is enough to figure out that the author keeps things in balance and in style. Explanations of why pixel grid mismatch looks wrong, or why mismatch between texture density and geometric complexity (in both ways) looks horrible, or why smoothing does not blend with pixel art are then made in retrospect.
Some details are a bit too cool for 1993, though, and assume high frame rate (won't work that well at low fps). Smooth weapon animations with a lot of frames, tiny per-pixel effects on bullet holes and flash sprites, smooth movement and object position calculations that use precise math instead of fast rough estimates resemble Chasm: The Rift or Quake (the concept of idle animations, e. g. objects moving in the starting view of difficulty selection room, assumes that there is some performance to waste on details that make the world less empty).
I just noticed that this might be one of the rare shooters with a female protagonist: the cat has a calico pattern, and those are almost always female (https://en.wikipedia.org/wiki/Calico_cat).
It's not that rare, is it? Off-hand, and very mainstream; Perfect Dark, Mirrors Edge, Dishonored (don't remember if it's the first or second one), Metroid and more are all kind of "shooters" with female protagonist, although maybe Mirror's Edge is more just "first-person" than "shooter" to be 100% accurate.
Not to mention the large selection of "RPG + FPS" where you can be either man or woman.
---------
Seems the author also realize the thing with the pattern and likely gender of the cat:
> After all, I do need to give the protagonist his fair share. [image] (Yes, I know it's a female, but call it convention rooted in dialect.)
Mirror's Edge has a female protagonist, but it's not an FPS (First Person Shooter). It's a parkour simulator which technically lets you shoot a gun in limited sections of the game, but the protagonist is a pacifist and you get a bonus for decommisioning guns rather than firing them.
So it looks like the two categorisations, for the most part, don't intersect.
Notable counterexamples would include Rise of the Triad, Ion Fury, No One Lives Forever, Wolfenstein: Youngblood and Far Cry 6, but definitely rare. You'd be clutching at straws to describe Portal or Alien: Isolation as FPS (they're a puzzle game and survival horror game respectively), likewise the Resident Evil / Clock Tower / Fatal Frame / etc. games with the novelty option of switching to first-person view, they're naturally third-person perspective. Left 4 Dead has one female character out of four you can play. You might count that one DLC for Bioshock: Infinite where Elizabeth gets a shot (https://www.youtube.com/watch?v=1E1lh-pb6Is). You might count the few FPS RPGs that there are with customisable characters (so yes Fallout, but not Mass Effect as it's third-person). But female protagonists are massively more prevalent in survival horror, metroidvania, third-person shooters (Tomb Raider, Monster Hunter, Horizon Zero Dawn, etc) and other genres besides FPS.
> Reaching back that far for an example just reinforces how rare it is.
Choosing one specific example when I also made more recent ones, isn't such a big dunk you think it is.
> If you tally all the FPS releases in a given year, a supermajority are going to have male protagonists.
Sure, I agree, I'm not saying it's more popular, just that I don't think it's that rare, but I guess ultimately I'm a bit nitpicky (sorry) and we're just disagreeing with the specific definition of "rare".
A lot of boomer shooters nowadays have a female protagonist, e.g. Selaco[0], Supplice[1], The Citadel[2] and its sequel[3], Zortch[4] (and its upcoming sequel[5]), Nighmare Reaper[6], COVEN[7], Viscerafest[8], Hedon[9], etc. If anything i'd say that nowadays there are way more boomer shooters with female protagonists than not :-P (combining the tags "boomer shooter" with "female protagonist" on Steam search gives 143 results, though that includes games where you can either choose your character's gender or you play as a woman for a part of the game even if you play as a man for most of it).
So a bit less than 1/10 of the games tagged with "boomer shooter". With your caveats above about being able to choose a gender, or a single brief segment where you're a lady in a game where you're mostly a dude. Is that a lot? I dunno, doesn't feel like a lot to me. Probably feels like a lot to the people who inevitably show up in the Steam discussions of any successful game that makes you be a lady for most of its length and complain about it being "woke", even one game with a female protagonist seems to be too many for them.
I doubt it was intentional, but in general, I am not impressed by that and don't find any value in that. Same with Hollywood's depiction of women knocking out guys twice their size. Unrealistic, ridiculous, and harmful.
This is so great. Another fun trick we used in the 90s was palette animation -- by swapping the palette you can create incredibly cool effects at a low runtime cost.
In the final video, it looks like the destructible vases take several shots to destroy. IMO, they should only take one. Real life vases only take one, so requiring more makes the gun feel weak. It seems to be cosmetic anyway, so there's no game balance reason to require more.
I am not even a noob with game development. I dabbled with Godot a couple years ago and was making a funny weight lifter game, some sorta stat management sim. It was actually pretty fun but I didn't get past some pretty early working versions.
I thought I could really level up with Claude and I started working on a boxing game. It's been a total disaster. .·°՞(˃ ᗜ ˂)՞°·.
We randomly chose magenta as our transparent pixel for shareware games too!
I consider 1993 the last "good" year of the pre-internet age. The web didn't go mainstream until around 95, and 94 felt like a liminal year (dunno why). In 93 one could still wrap a plaid shirt around one's waist without fear of ridicule. Grunge and alternative music hadn't quite landed in rural America yet, although we didn't know what we were missing. The Telecommunications Act, Digital Millennium Copyright Act (DMCA), Gramm-Leach-Bliley Act, USA PATRIOT Act, and so many other regressive/draconian laws hadn't passed yet to create the wealth inequality consuming the American Dream today. Although the Grand Upright Music vs Warner Bros decision had happened in 91 in an attempt to destroy hip-hop for racist reasons under the guise of protecting copyright. The Rodney King beating had happened the year before, but the OJ trial was still 2 years away. We were blissfully ignorant of the very ignorance and hate that would put us on this alternate timeline. It was like living in the Shire before the War of the Ring.
I can't stress enough how games like Wolfenstein 3D and DOOM completely blew our minds. They came out about 6-7 years before The Matrix, so the closest conceptual framework we had for it was probably The Lawnmower Man. Virtua Racing and Virtua Fighter came out about that time, but somehow couldn't compare. I remember using a drafting program on a 33 MHz PC with a 16 color monitor in drafting class, and DOOM revealed that even then, computers were running hundreds of times slower than they were capable of (millions of times slower today).
If I could go back to any time with what I know now, it would be spring of 93.
As a fellow 3d-engine-with-foolishly-unreasonable-constraints developer, I love the detail in the explanations here and seeing the process you went through.
This is terrific. I love reading about the creative process involved in a project like this, finding cool solutions to self-imposed boundaries.
I think the mix of highly rational reasoning and "it just feels right" is a killer combo too, it gives a rigorous basis for a lot of the decisions made, while also allowing for a strongly personal aesthetic to emerge. Very cool indeed.
I'm tinkering with a voxel space rendering tech demo as a PlayStation homebrew. After one weekend of work I'm getting decent results (like, 10-15 FPS) and I've yet to use the DMA, the GTE or even polylines primitives.
It's refreshing to dust up trigonometry and good old low-level optimization tricks. When the scratchbuffer has 1 KiB and the stack can only use a fraction of that, it makes me realize how spoiled I'm at work with the microcontrollers we have, with threads being allocated 8 KiB of stack and backtraces with over 50 functions of C++ templates on it.
Just a guess: if you want to scale a sprite at anything less than a whole ratio (e.g. 1.5, 0.7, etc), you have to choose pixels to drop out and pixels to repeat , on some pattern that looks good. There are going to be scaling ratios that look like a hot mess, especially at a low resolution like 320x240.
In context, it's talking about sprites that are going to get non-integer scaled anyway (in-game pickups), so it's just about maintaining a consistent detail level. If those specific sprites had their resolution increased, everything else would need its resolution increased to match them.
Inconsistent resolution isn't necessarily a bad thing, e.g. Elite for the BBC Micro changes video mode part way down the screen so it can display both high resolution monochrome wireframe 3D and a lower resolution color map/UI below, but it's not idiomatic to the MS-DOS style this game is going for.
>> What this actually means is, the constraints I have foolishly imposed upon myself are as follows....
Those kind of constraints can lead to increased creativity, and can also influence the overall style of a game. It's part of the reason early 80's arcade games had so much diversity.
I'm impressed by your sprite pipeline and gibs animations. Your attention to detail and navigation of constraints have really paid off, I can't wait to play this sometime
This is a great write-up of your process and behind-the-scenes peek at the making of what looks to be a really fun game! Can't wait to play and delve into the code once you release it.
Every time I think about graphics programming, I think about how we did it in the mid 90s when I was in high school messing around with exactly these things. XOR operations to drive animations, writing directly to memory, etc. (Clearly I do backend stuff now...)
I really loved that article. Creating games always seemed so daunting to me since I don't know a lot about how it's done, there are so many different processes involved. The solutions described here are so satisfyingly compact and so easy to understand given the simple constraints, and yet they produce an actual game that looks nice.
Much respect -- at first glance when I saw the animated gif I thought this was just a project making assets from scratch for an existing game engine (e.g. Doom or Wolf 3D) but then I realize it's creating all the game code from scratch too! (But using similar techniques from the old days). Amazing work.
This game looks great I really like the style it is inspiring.
The author seems to consider open-sourcing the engine, I would also be interested in the mentioned scripts for asset creation. Those scripts would make a great toolset for asset creation in this style.
one of my very first solo projects during high school was writing a wolfenstein-like raycaster from scratch. I still hold some very fond memories of programming it, arguably one of the moments I fell in love with the craft
I love this! I have been working on a similar project, recreating the originale BBC Elite but with multiplayer networking. Though I have not limited myself as much (I use SDL).
This is a wonderful deep dive into your project. I'm early days on creating pixel art style procedural art systems, and this gives me plenty to think about.
Nice, i've used similar approach for the lighting in Post Apocalyptic Petra[0] though i did use per-pixel LUT offset calculation[1] because it uses a generic 3D triangle rasterizer (the levels are based on grids like in Tomb Raider but they're rendered as triangles). Later i added sprite support for another gamejam but i never ended up finishing it and the sprite support is very rudimentary (and unoptimized - i just noticed i'm doing the LUT lookup for every pixel when drawing shaded sprites which isn't necessary).
I did write a tool for generating the sprites from 3D models though[2]. It uses plain old OpenGL 1.1 to draw the sprite and grabs the framebuffer directly. It is drawn fullbright so i can paint the lighting directly on the sprite's texture (using a Krita plugin i wrote[3][4] - the model is something i threw together with Blender's default generated UV since i didn't care for the details).
I wonder if doing some sort of postprocessing (after rendering with with shading) like you do with your game would help with the finer details since i also found that rendering from 3D models to sprites creates very "mushy" results most of the time because of all the details getting lost. I notice the colors also become more saturated after postprocessing in your examples, is this after it finds the closest color in the palette or the result of the postprocess? I'd like to keep the overall hue+saturation of the model so maybe doing post-processing on a grayscale render to shade the shadows/dark areas but keep highlights as-is and then multiplying that with the fullbright image would produce results that wont shift the saturation.
>I notice the colors also become more saturated after postprocessing in your examples, is this after it finds the closest color in the palette or the result of the postprocess?
It's the result of the Blender compositor postprocessing, just keep in mind it falls apart once you go low enough in resolution (it's an image space thing after all), so I'm not sure if that helps your case.
Wow, this takes me back of making my own software renderer and game engine as a teen in the 90s. Then OpenGL came out and fixed pipelines and some of the cool magic of doing anything with pixels disappeared (until pixel shaders came back). One cool rendering technique you don't see much today is voxel graphics.
The comments here are a cesspool unfortunately. People bickering about pronouns used for cats, how many shots it takes for a vase to explode, or whether or not some circa-1993 software was used or mentioned.
How so? Doing this with modern OpenGL would be much simpler than the software rasterizer solution.
I think I'm gonna have to do it anyway, because some players claim they get nausea when playing at such low resolution (320x240), and the only way to give them higher resolutions that perform reasonably is to have it hardware accelerated.
Renderer is abstracted away already, but the real difference would probably be occlusion culling... With raycasting, I get it for free, but if I'd go down the hardware accelerated path I'd have to pick something more clever.
Raycasting and software rendering in general tends to scale poorly with resolution, even with vectorization and all the bells and whistles of modern CPUs.
Unless you plan on rendering the level on some very retro hardware (think S3 Virge, maybe Voodoo 1) you can render the entire level in OpenGL with just zbuffer and alpha tested sprites and it'll run perfectly fine - if anything with such low polycount, chances are you're going to make the renderer slower by trying to do occlusion culling on any GPU released in the 21st century :-P. If you pack the geometry in a few vertex buffers (for each unique texture) even per-frame, you'll get four digit FPS in any relatively modern GPU.
As an example this[0] video shows the benchmark from Post Apocalyptic Petra running on my previous GPU (RX 5700 XT) which all it does is build a per-frame (client-side) vertex-buffer in OpenGL 1.1 (the engine was made for actual retro PCs running DOS and Win9x so it does some rudimentary occlusion culling but that mainly affects 90s hardware, not anything released since 2000 or so). If anything, the rendering has so little overhead that half of the framerate is "eaten" by the FPS counter overlay :-P.
Thinking about modern games, a single character model probably has more vertices than my entire level (and yours probably), so it's definitely reasonable to expect occlusion culling for such simple geometry might actually reduce performance rather than increase it.
Yeah, even this model[0] i made a few years ago for a game i wanted to make for the OG xbox (which has a GeForce3-like GPU) has ~2230 triangles and the entire first level of Post Apocalyptic Petra is ~5800 triangles, so you could say that even a turn of the century 3D character has more or less the same polycount as an entire 90s level (the game i wanted to make would have many characters on screen so i kept the polycount low but i've heard games having 5-6K character or more - e.g. Kingdom Under Fire had ~10K triangles for the main character and a game like Dead or Alive where there are few characters on screen had 15-20K triangles).
Meanwhile more current games have much higher polycounts, easily going above 100K triangles - e.g. Dante from DMC5, a ~7 year old game, apparently has ~190K triangles and that had to run on the more anemic PS4/XBone hardware :-P (though i'm not sure if it used the full 190K model there or some cut down version).
I rendered some Source engine levels on a shitty laptop in 2012ish and they still rendered at perfectly acceptable FPS (30+) just by rendering all the geometry in the level in one shot.
The synthesis technique would be to build the DOOM-style BSP tree and then construct a bunch of meshes for use depending on which portal space you're currently in, but .. as you say, you don't need to do that because it's at most a few hundred polygons.
Sure, but it becomes a question of how far you can push things. Maybe you raytrace the whole thing. Maybe there's some fractal geometry going on. Maybe you use a fisheye lens projection. Maybe your levels are dynamically tesselated. Maybe you have to do a few fancy tricks to achieve equivalent texturing etc.
But ignoring the GPU you have on your system is boring
I don’t know. Em-dashes are normal punctuation. The prose on the site doesn‘t strike me as particularly AI-y, but of course I might be wrong. Generally speaking, if the person wants to not use AI and tell people that, thats fine by me too.
We're getting to the point that building something with AI doesn't really indicate skill. So, for a prestige project, there is great value in avoiding AI use.
From the article they list a bunch of arbitrary constraints...
> If this sounds unreasonable to you, that is because it is.
Those listed, are tame. I don't understand this kind of faux modesty.
> My goal was to build a complete, shippable first-person shooter using techniques that were common in the early 90s
Goes on to explain how they used 3D blender...which wasn't available until 1998.
A vanity cat project being tailored and submitted for nostalgia clickbait. I don't think there's anything useful to take away from this other than some color shade selection ideas.
>Goes on to explain how they used 3D blender...which wasn't available until 1998.
In the early 90s, there was enough money in this kind of software that you could have hired a specialist 3D artist to use the software that was available at the time, e.g. LightWave 3D. When it's only a single-person project, I think it's reasonable to stick with what you know.
⸻
1. I suppose some hand-written PostScript code might count as well, but I wouldn’t really count things like doing a simple function graph in python to explain something to my son as graphics programming.
2. This was for a DVI previewer running on an IBM mainframe running VM/CMS. As far as I know, this code is completely lost, which is probably a good thing.
reply