I recently watched fellow particle physicist talks about a calculation he has taken to new heights of precision. His instrument? A 1980s computer program called FORM.
Particle physicists use some of the longest equations in all of science. To look for signs of new elementary particles in collisions at the Large Hadron Collider, for example, they draw thousands of pictures, called Feynman diagrams, that depict the possible outcomes of collisions, each encoding a complex formula that could have millions of terms. It is impossible to summarize such formulas with pen and paper; even adding them with computers is a problem. The rules of algebra we learn in school are fast enough for homework, but for particle physics they are terribly inefficient.
Programs called computer algebra systems seek to cope with these tasks. And if you want to solve the biggest equations in the world, one program stands out in 33 years: FORM.
Developed by a Dutch particle physicist. Hos Vermaseren, FORM is a key part of the particle physics infrastructure required for the most complex computations. However, as is the case with many important elements of the digital infrastructure, the maintenance of FORM is largely the responsibility of one person: Vermaseren himself. And at 73, he began to move away from the development of FORM. Due to the incentive structure of academia, which values published papers over software tools, no successor has emerged. If the situation does not change, particle physics may be forced to slow down dramatically.
FORM was born in the mid-1980s, when the role of computers was rapidly changing. Its predecessor, Martinus Veltman’s Schoonschip, was released as a custom chip that you plugged into an Atari computer. Vermaseren wanted to make the program more accessible and available for download by universities around the world. He started programming it in the FORTRAN computer language, which means translation of formulas. The name FORM was a reference to this. (He later switched to a programming language called C.) Vermaseren released his software in 1989. By the early 1990s, more than 200 organizations around the world had downloaded it, and the number continued to grow.
Since 2000, a particle physics article with a link to FORM has been published on average every few days. “Most of [high-precision] the results our group has obtained over the past 20 years have been largely based on the FORM code,” said Thomas Hermannprofessor at the University of Zurich.
Part of the popularity of FORM has come from specialized algorithms that have been developed over the years, such as a trick for quickly multiplying certain parts of a Feynman diagram, and a procedure for rearranging equations to have as few multiplications and additions as possible. But the oldest and most powerful advantage of FORM is how it handles memory.
Just as humans have two types of memory, short-term and long-term, computers have two types: primary and external. Main memory – your computer’s RAM – is easily accessible on the fly, but is limited in size. External storage devices such as hard drives and solid state drives contain much more information but are slower. To solve a long equation, you need to store it in main memory so you can work with it easily.
In the 80s, both types of memory were limited. “FORM was created at a time when there was almost no memory, and also no disk space – in fact, there was nothing,” he said. Ben Ruyle, a former student of Vermaseren and developer of FORM, who is now a PhD researcher at the Swiss Federal Institute of Technology in Zurich. This created a problem: the equations were too long for main memory. To calculate it, your operating system had to treat your hard drive as if it were also main memory. The operating system, not knowing how big your equation needs to be, will store data in a set of “pages” on the hard drive, switching between them frequently as needed – an inefficient process called paging.
FORM bypasses paging and uses its own technique. When you work with an equation in FORM, the program allocates a fixed amount of hard disk space to each term. This method allows the software to more easily keep track of where the parts of the equation are. It also makes it easy to bring these fragments back to main memory when they are needed without access to the rest.
Memory has grown since the early days of FORM, from 128 kilobytes of RAM in an Atari 130XE in 1985 to 128 gigabytes of RAM in my advanced desktop computer—a million-fold improvement. But the techniques developed by Vermaseren are still decisive. As particle physicists sift through petabytes of data from the Large Hadron Collider for evidence of new particles, their need for precision, and thus the length of their equations, increases.