Why C?
:date: 2026-07-23 07:56 :tags:
For the last couple of months I've been deep in the software mines laboring away at a big project that I will be revealing and writing more about very soon. Today I wanted to get things going with a preemptive justifcation of the primary tool I used for the project.
Here is my list of best reasons to use the C programming language.
- C is the unambiguous high performance GOAT. Other languages measure their speed by how much slower they are than C. When you write software in another language you will always wonder how fast it could have been had you written it in C. BTW, whenever a language is reported to outperform C in some weird edge case and that language itself was written in C, I'm sharing the win with team C.
- Generally better ecologically. This interesting paper shows that fastest execution speed does not always equal lowest calories. But generally it does and except for some unusual cases, C is the champion of both. "By observing the data in Table 4, we can see that the C language is, overall, the fastest and most energy efficient."
- Used by very serious, very well respected projects. It was also heavily used by the people who created it. To say that the Bell Labs people made better software than what most people use today is like saying that the artwork commissioned by the Medici popes is better than the typical Powerpoint presentation.
- It's quite free. GCC actually stands for the
GNU Compiler
Collection (because it supports
C++, Fortran, Go, Rust, et al) but of course it was originally the
GNU C compiler just as the original
cccommand stood simply for "C compiler". The GNU project literally invented explicit software freedoms and GCC is its flagship software. Obviously GCC is primarily written in C. If you like your software freedom to have a rich robust aroma, yet be slightly less jittery, and with fruity notes of apple, you can also use LLVM/Clang.
- It is not (naturally) object oriented. 25 years ago I went all in with OOP. I developed several of my big projects in C++ and Python. Despite having faith in the hope and hype of OOP, over the years I have had to honestly assess its value. I have discovered that in retrospect it has almost never been helpful to me. It has definitely never been more beneficial than its overhead in extra complexity, ambiguity, and loss of control. (See my xedNN project where I finally am able to see how a neural network really functions because I unravel it from obfuscated OOP.) I now suspect that OOP is like flowcharts or UML in that it seems like it must be a reasonable step toward progress but in reality it is an obfuscatory muddle with mostly imagined benefits. Note that OOP is often a win for "end user" type programmers who are just deploying a library; I feel like OOP makes the job of library developers much harder but can make using the library easier. If you're writing your own software properly from scratch, the benefits are questionable.
- Like object oriented? C can do that! Like garbage collection? C can do that too! Yes, a good rule of thumb is that C can do what computers can do.
- Properly portable. From microcontrollers to the emulated CPUs of long forgotten systems to the latest chips running all modern hardware. A processor chip is still in development until it has the support of a C compiler.
- Porting software from one language to another very often goes with C libraries as a common format that pretty much all sensible languages can link to.
- Low level memory and resource management. Do you like shooting yourself in the foot more than being shot in the foot by other people? Then you'll appreciate C memory management.
- Massive ecosystem of libraries which are often the reference standard and highest performing implementation. For example, one is unlikely to find a vector math library or a regular expression library that beats the most popular C one.
- While there are some goofy (and possibly useful) "modern" versions of C (e.g. C23), the development of C is extremely stable. Even C23's new features are pretty well aligned with not fixing something that isn't broken. Ancient original K&R is seldom used today but it is completely comprehensible by any C programmer and it does still compile. Reading the original K&R C book is still probably one of the best ways to understand the language. In contrast to C, C++ is constantly metastasizing new and baroque features and syntax. I even created this little meme for my C++ notes to highlight the absurdity of it.
- It's deterministic and has low run time overhead. C has no hidden overhead, no hidden garbage collection pauses, no virtual machine startup time, and no runtime checks (unless explicitly enabled). This makes it ideal for critical real time hardware control and, really, any software if you can pull it off.
- Closest to the hardware. What's important to me is that this means it is closest to The Truth. Sometimes the truth is not pleasant and you may wish something else were true, but I never wish to be deceived about The Truth.
Here's Linus selling it with this exact point.
Excerpt...
I like interacting with hardware from a software perspective and I have yet to see a language that comes even close to C in that respect. It's not just that you can use C to generate good code for hardware. It's that if you think like a computer, writing C actually makes sense.
Why not C?
Are there ever any good reasons not to use C? Absolutely. And they help illustrate when you should.
- When someone offers to pay a lot of money to use something else, reasonable people will sometimes do that. Similarly contract software generally has little incentive to be lightweight, transparent, and easy to hand off to one's competitors.
- C asks you to do things yourself, to program the computer yourself. If you are a professional computer programmer, C is perfectly compatible with your skills. If you do not have the skills to program in C — if you can not think like a computer — you many not have what it takes to be a professional computer programmer.
- C is the correct choice if the goal is to create your own software. If you are trying to use your own software to produce something else (a geotechnical report, a gearbox design, a web site, an AI model, etc), then it is often much easier to use tools that are already much closer to the true goal.
- Some people love Python's semantic whitespace syntax. Such people can struggle with C because it forces you to use syntax that is either redundant or an abomination. What do I mean by that? Stay tuned, for I have cured this problem!