Nicholas Carlini Crams a Working Intel 4004 Calculator Into 4004 Bytes of Obfuscated C Code

Research scientist Nicholas Carlini has celebrated the birth of the microprocessor with a gate-level emulation of Intel's classic 4004 chip — distributed, for good measure, as 4004 bytes of obfuscated C code.

"The International Obfuscated C Code Contest results were announced this last weekend, and I was thrilled to see that my entry was among the winners," Carlini writes. "[It's] a feature-complete gate-level emulator for the Intel 4004 that's capable of emulating the original Busicom 141-PF calculator ROM — the application for which the 4004 was originally designed."

Intel's 4004 was launched back in 1971, and is generally recognized as the first commercially-successful microprocessor built using large-scale integration (LSI) — though its success came outside the project for which it was designed, a printing calculator from Busicom. While a four-bit design, it would lay the groundwork for what would become the x86 architecture and Intel's rise to one of the biggest names in computing.

Carlini's recreation works as a simulation at the level of individual gates — capable of running, he says, at a few hundreds of instructions per second on modern hardware or enough to complete a calculation in around three to 10 minutes, "which is actually fast enough to be usable, just as long as you're not in a hurry." How it's made, though, is the interesting part: the emulation comes courtesy of a design written in miniHDL, a Python-based domain-specific language (DSL), which is then compressed down into a 4004-byte C program of seeming gibberish deliberately designed to hide its intended purpose to serve as an entry to the IOCCC.

"If one were to rank programs by human-hours-per-byte, this program would be certainly top the ranking for the all the code I've ever written," Carlini guesses. "I didn't keep good track of this, but I'd estimate I spent probably a few hundred hours over the last four years working on it. (Honestly I felt kind of empty after submitting the code back in June. I've been working on this since almost the start of the pandemic, and whenever I felt like I had to take a break from my large (mostly Python) actual work this was the perfect place to come and fiddle with the bytes of a program that fits comfortably on a single piece of paper.)"

The project is documented in detail on Carlini's website; the source code is available on GitHub under the reciprocal GNU General Public License 3.

Main article image courtesy of Thomas Nguyen, CC-BY-SA 4.0 International.