Gert Svensson to retire soon from PDC/KTH

Background

In early 2025, thirty-five years after starting at PDC, Gert retired from full-time work at KTH and is now mainly working as part-time technical director at PDC and deputy technical director at the National Academic Infrastructure for Supercomputing in Sweden (NAISS), where he is heavily involved with the procurement and set-up of the Arrhenius EuroHPC Joint Undertaking mid-scale system that will be hosted in Sweden. Gert’s extensive expertise (as the person responsible for procuring the progressively faster systems hosted at PDC) has been invaluable to NAISS in the Arrhenius project, which is procuring an HPC system that will be the fastest ever in Sweden.

During Gert’s decades at PDC, there have been significant developments in the types and processing speed of the parallel supercomputers used for research, and PDC has evolved in step with these changes to continue offering advanced supercomputing resources to Swedish researchers. PDC has been an early adopter of new technologies and approaches, such as parallel and cluster computing and the use of graphics processing units (GPUs) for scientific computing (rather than gaming). PDC has also contributed to significant developments in infrastructure technologies, including water-cooled HPC systems and heat re-use.

Having been at PDC from the start, Gert has driven many of these changes. This is his story entwined with PDC’s evolution: How did PDC come into being? How did Gert become involved with HPC and PDC? And how did PDC evolve into a primary centre supporting Swedish research over Gert’s 36 years at PDC?

As the last of the original staff at PDC to leave and the longest serving staff member, Gert reflects back on the path that led him to PDC and how PDC evolved during his time there, culminating in PDC’s metamorphosis into the basis of the newly formed KTH Center for Scientific Computing (KCSC) as Gert fully retires from KTH. This retrospective article illustrates some of the vast changes and developments that have occurred in the field of computing in Sweden in the last half century.

Gert starts studying at KTH

Gert started studying engineering physics at KTH in 1975 as he wanted to “understand” physics and was keen to master the extremely challenging topics in that domain. Gert was also fascinated by computers, although they were quite different at that time. When he started at KTH, programming was mainly done using punch cards and batch processing. Using computers for calculations was a fascinating field, so Gert chose to specialise in applied mathematics and studied all the programming courses that existed at KTH at that time. Gert’s thesis work at the Department of Telecommunication and Computer Systems (TTDS) at KTH evaluated the real-time and parallel programming features of the new Ada language by implementing a real-time kernel. Gert graduated in 1981 and concluded that, although he still did not “understand” physics in the way he would have liked to, he could instruct computers to do all kinds of useful calculations and simulations.

Gert starts to work at KTH and discovers UNIX

Gert was then offered a position at TTDS and began teaching courses in real-time and parallel computing and operating systems. At TTDS, there was a Programmed Data Processor (PDP) 11/45 minicomputer from Digital Equipment Corporation (DEC) with one of the first Unix installations in Sweden. (It was probably the second in the country.) Gert found Unix to be simple and elegant compared to other operating systems (OSs). Back then, it was possible to understand most of the OS, unlike nowadays. Security was not a problem at that time since there were no connections to anywhere outside the department. Setting a password for your account was considered suspicious behaviour. People would wonder: Did you have something to hide? Almost everyone could be the root user (that is, have access to the whole system) and, if you found some bug or incorrect setting in the OS, you just had to recompile that part yourself. It was not until terminal switches (which, by the way, were developed at TTDS) were introduced at KTH that passwords were required. (Terminal switches were invented so computers could be used from all over KTH or even from remote locations by modems.)

Gert becomes an early open-source advocate

Early on, Gert became an advocate for open-source software (that is, software that is freely available without license fees for possible modification and redistribution) and became a board member of the Swedish UNIX Users Association (which was later called EurOpen.SE). In 1975, a UNIX Users Group had been formed in the USA to bring together the community of engineers, researchers, and practitioners working on the cutting edge of the computing world. The name of the organisation was changed to USENIX later as using the word UNIX in the name was not allowed. EurOpen was effectively USENIX in Europe and EurOpen.SE is an affiliate member of the original USENIX. EurOpen.SE had the role of promoting open-source software and also operated the Swedish part of an international electronic mail network that was based on modems in the early days. Gert invited many UNIX and open-source advocates to give presentations at EurOpen.SE and KTH. The most well-known of those today are probably Richard Stallman ( en.wikipedia.org/wiki/Richard_Stallman ) from the USA, a prime advocate for free software, and Linus Torvalds ( en.wikipedia.org/wiki/Linus_Torvalds ) from Finland, the lead developer of the Linux kernel and the Git distributed version control system (for keeping track of different versions of software as it is developed).

The UUCP mail network in Sweden

In the early days, the UNIX mail network in Sweden was operated by the consulting company ENEA, with Björn Eriksen ( sv.wikipedia.org/wiki/Björn_Eriksen ) as the guru. Connections between different computers were made using modems and diallers that called other modems in a hierarchical worldwide network using the Unix-to-Unix Copy (UUCP) tool. TTDS was the email hub at KTH (that is, it acted like the central post office) and, before domain addresses were introduced, the entire connection path for how the computers were connected was used. (For example, enea!ttds! meant send the email to enea then to ttds then to the specified department at KTH, so enea!ttds!nada!jespero would be used to send email to the the user jespero at the NADA department.) Important hubs like ENEA were hard-coded, and only the path from such a central hub needed to be given. Gert took part in operating the mail hub at KTH. ENEA soon introduced permanent connections to the emerging Internet (and electronic mail network). These connections relied on modems, which were used to let the computers communicate over rented phone lines.

Connecting KTH to the Internet

In the mid-1980s, Gert became fed up with the dial-up modems that were being used to connect KTH to the dial-up mail network, and he started investigating what would be required to connect the VAX 11/750 (which had replaced the PDP 11/45) to ENEA and consequently to the Internet. First of all, a 56 kbit/s connection with synchronous modems needed to be rented from Televerket (the Swedish Public Telephone and Telegraph Company). And then, an interface card needed to be purchased to handle the connection to the VAX. This would have been a significant investment, however, a cheaper card was found second-hand. The TTDS department board approved these expenditures, and Gert installed the card (which was also known as a board because of its size) and connected it to the modem. However, nothing worked! You can imagine how Gert felt after arguing for the funding! It turned out that the board was broken, and the company supplying them soon sent a new one, et voilà, KTH was connected to the Internet! This meant that domain names were required, and kth.se was introduced and set up by Gert’s colleague, Johan Widén. However, only basic Unix services, most importantly email, existed at this time. It was fantastic to be able to send emails to the US in just seconds, rather than waiting several hours or even days. The conferencing service Usenet (which used the readnews client) was also popular. All the selected conference groups were downloaded to KTH and could be read locally. The File Transfer Protocol (FTP), a protocol that is used to transfer files between computers, and remote login also existed, but an address and often a password were needed to connect. The World Wide Web had not been invented yet.

Early days of parallel computing at KTH

In 1985, a research lab for parallel computation was established by Prof. Lars-Erik Thorelli, professor of computer systems as part of TTDS at KTH. A Sequent Balance parallel computer system with shared memory was acquired and served as the basis for the new lab. Sequent was a system that pioneered symmetric multiprocessing (SMP) and introduced many novel ideas. The system at KTH initially used 10 MHz National Semiconductor NS32032 processors. Then the Sequent company decided to replace all Balance systems in the field with new systems based on Intel 80386 CPUs at no extra cost, maintaining the same total capacity. The only problem with this generous offer was that it would have meant there would be far fewer nodes in the upgraded KTH Sequent system, which would not have been ideal, as researchers were mostly interested in the degree of parallelism in order to study scalability. So, Gert approached the president of Sequent at a conference and briefly described the problem. To everyone’s surprise, the Sequent president just said, “We will solve the problem,” and he kept that promise by giving KTH a Sequent Symmetry with the same number of Intel processors.

The Connection Machine that became the basis of PDC

In 1988, a group of KTH researchers – under the lead of Prof. Anders Lansner from the Department of Numerical Analysis and Computer Science (NADA), Prof. Lars-Erik Thorelli from TTDS, and Jesper Oppelstrup from NADA – applied for and were awarded a grant to buy a massively parallel computer: a CM-2 Connection Machine from Thinking Machines Corporation (TMC). Gert handled the procurement of the system, which was installed at KTH at the end of 1989 and named “CM Bellman” in honour of the 18th-century Swedish poet, songwriter and entertainer Carl Michael Bellman. The Connection Machine used the data-parallel programming model, and each data element was treated by its own (physical or virtual) processor in a single instruction, multiple data (SIMD) fashion, meaning that all processors executed the same instruction (or no instruction depending on a data mask). The Connection Machine had three complete programming languages: CM Fortran, C* and *LISP.

PDC starts in 1990

After the Bellman system was installed, KTH came up with the idea of grouping together resources and activities that centred around the CM-2 and other parallel computers that already existed at KTH (the Sequent and some laboratory systems). As a result, the Parallelldatorcentrum (PDC) was established at KTH in 1990. Gert Svensson was the first person to be employed by PDC and was given the role of forming and setting up the new organisation. With the formation of PDC, Gert realised that HPC was the right area to combine his interests in computers and physics simulations. The second person hired at PDC was Fredrik Hedman, who was sent to TMC in Boston, USA, for several months of extensive training and became the first application expert (AE) at PDC. (An AE is a person with expertise in both programming and a research area, who can help other researchers to program and run their simulations.) Soon, PDC also hired a couple of system administrators (Lars-Johan Liman and Johan Ihrén) to manage and maintain the systems and infrastructure.

At this time, Lennart Johnsson (who was originally from Sweden but working overseas) was leading the development of the numerical library at TMC in the USA, and he served as a liaison between TMC and PDC when PDC purchased the CM-2 system and began using it for scientific calculations. Lennart met Lars Malinowsky, who was looking for thesis work for his degree in Sweden. This led to Lars helping implement the Fast Fourier Transform (FFT) routines for the CM-2 numerical library as part of his thesis work. After that, Lars joined PDC, where he still works as a system administrator.

PDC conference series

In 1992, Gert took the initiative to organise a groundbreaking conference on parallel computing at PDC which was held in December of that year. The event attracted distinguished speakers from across the globe, each of whom was contributing to important advancements in supercomputing (which is now known as HPC). This inaugural gathering gave rise to the highly popular PDC Annual December Conference series, which continued from 1992 through to 2000. Gert personally orchestrated these conferences, seeking out inspiring experts – often notable figures from the annual Superomputing Conference (SC) series – and personally inviting many of them.

PDC changes department at KTH

In 1993, PDC was transferred from the Computer Systems group of TTDS at KTH to the Department of Numerical Analysis and Computer Science (NADA). This change came about because Lars-Erik Thorelli, the head of the Computer Systems group at TTDS, wanted his group to relocate to the new KTH campus in Kista, north of Stockholm. However, most of the KTH researchers who were using the computer systems at PDC were based at the KTH main campus in central Stockholm. To continue serving the researchers most effectively, PDC remained in the same KTH building (at Lindstedtsvägen 5), but it became part of NADA. Prof. Björn Engquist from NADA took over Lars-Erik’s role as chairman of the board of PDC and later became its director.

PDC moves to new facilities, the first time

In 1993, KTH was planning a total renovation of the entire main building (where PDC and many other parts of KTH were housed). The idea was to divide the renovation of the building into four phases which would be spread over several years. Unfortunately PDC’s computer hall was located in a part of the building that would be renovated during the first phase. KTH proposed a plan for temporarily moving the computer hall to an area in the building that would be renovated during last phase in order to give KTH extra time to plan a more permanent location for PDC. PDC produced detailed specifications for the temporary computer hall, and KTH promised that everything would comply with those requirements: PDC would not need to take part in the building process. In fact, the PDC staff were not even allowed to enter the building site!

The renovation of the part of the building that included the new computer hall was eventually completed, and PDC organised for TMC to move the heavy and fragile Connection Machine to its new location. However, when the first weighty load was placed on the raised floor in the computer hall, the floor began to collapse, so the move had to be aborted. PDC checked and found that several of the important specifications for the computer hall had not been fulfilled, so PDC decided that the CM needed to be moved back to the old computer hall while the new computer hall was updated to meet the specifications. It took several months for KTH to correct the problems, after which TMC successfully moved the Connection Machine to PDC’s temporary computer hall.

PDC pioneers cluster systems

An important milestone for PDC in the 1990s was its application for and receipt of a grant for a large, multiple instruction, multiple data (MIMD) cluster. (A MIMD cluster is a type of supercomputer system that uses multiple processors to make parallel computing possible. This is still the approach that is used for most supercomputer clusters nowadays.) The procurement of the system was led by Gert and resulted in the installation of an IBM SP2 cluster in 1994. The system was named Strindberg after August Strindberg, a Swedish author and playwright who was also a painter.

The Strindberg SP2 supercomputer consisted of multiple individual advanced workstation nodes interconnected by a switch. (A switch in this context is a device that makes it possible to transfer information efficiently between many computer nodes.) The Strindberg system was significant because it demonstrated that supercomputers could be built by connecting many workstations. A single program could be run simultaneously across Stringberg’s multiple workstations to significantly speed up overall runtime. Programs that used this feature of Strindberg needed to be modified or written to take advantage of this kind of parallel processing, for example, by using the Message Passing Interface (MPI). The system could also run several serial programs at the same time without any of them needing to be rewritten, which is another form of parallel processing.

IBM had insisted on including an IBM tape robot (for long-term data storage) in its winning bid during the procurement process for Strindberg. That storage was extremely useful to researchers, and, although it has been upgraded many times, the current PDC tape robot still uses similar technology from IBM.

Many more cluster systems at PDC from new funding agencies

The supercomputer systems at PDC continued to be upgraded to provide top-notch HPC systems for research, thanks to KTH and the Swedish organisations that provide funding for the Swedish HPC research infrastructure, such as the current National Academic Infrastructure for Supercomputing in Sweden (NAISS) and previously the Swedish National Infrastructure for Computing (SNIC), the Swedish Council for Planning and Coordination of Research (FRN) and the Swedish High-Performance Computing Council (HPDR).

In 1995 PDC was awarded first place in a competition run by the HPDR, under the auspices of the Swedish Council for Technical Research (TFR), to form a Swedish national HPC centre. As a result, PDC became Sweden’s first national HPC centre and was granted several years of funding to support its operations, including maintaining the HPC systems as well as continuing a research agenda. In 1998, the FRN took over from the HPDR and managed the funding for HPC centres in Sweden until the role was handed over to the Swedish National Infrastructure for Computing (SNIC) in 2003. During that period, the role of the TFR was taken over by the Swedish Research Council (VR) in 2001.

PDC pioneers visualisation technology

PDC was involved in scientific visualisation (generating images to represent results from computer modelling and simulations) from the start of this discipline, since visualisation requires large computers. One major step in providing computer resources for generating visualisations was taken in 1996 when PDC acquired an ImmersaDesk, a large stereo screen with virtual reality (VR) capabilities driven by a system from Silicon Graphics International (SGI).

The experience with the ImmersaDesk was highly positive for researchers and led PDC to decide to build a unique 6-sided VR cube with the German company TAN. This was the world’s first Cave Automatic Virtual Environment (CAVE) with projection on all six sides. Gert was the project leader for the set-up process. The VR cube was inaugurated in 1998 by the Swedish king and soon became a big attraction for KTH. It was not only used for scientific visualisation but also attracted artists to create VR art installations.

Gert participates in European and Swedish research projects

The VR-cube project led to several European projects, for example, Gert participated in the VIRTUALFIRES project (coordinated by the Graz University of Technology, Austria), which simulated and visualised smoke and flame spread in fires in traffic tunnels (to improve fire prevention and firefighting methods), and the Uni-Verse project that developed novel tools for the Verse platform, which provided multi-user, interactive, distributed, high-quality 3D graphics and audio for public and personal use.

From 1998 to 2003, Gert was also involved in neuroscience research together with Prof. Per Roland from the Karolinska Institute in Stockholm. The project aimed to automate the processing and statistical analysis of diverse brain scan images, including positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). It also sought to develop a taxonomy of experimental conditions, streamlining the entry of both conditions and raw data into databases for supercomputer-driven analysis. This system enabled efficient reuse of analysis results, which were also stored in the database. The research began with the Swedish BINS project and advanced through the European NeuroGenerator project, where Gert served as technical manager.

PDC’s infrastructure needs more space so PDC moves for the second time

By 2004, PDC’s systems and infrastructure had grown too big for the old temporary computer hall in the main KTH building. Meanwhile KTH had had time to find a longer term site for PDC’s computer hall, so PDC moved to larger facilities at Teknikringen 14, which was nearby on the main KTH campus.

Gert Svensson led the project for moving PDC, which included designing the new computer facilities. This was a complex task, as the building's basement levels had to be modified to house the computer systems, storage, backup batteries, and fire suppression system, along with other infrastructure necessary to provide sufficient power to run the systems and adequate cooling.

This time around the collaboration with the KTH team worked much better, and the new specifications and successful collaboration led to a well-designed computer hall that has been able to house successively larger systems for more than 20 years. For example, at the time that PDC moved to Teknikringen 14, it was considered impossible to fit more than a few hundred kW of power infrastructure into the available space in the new computer hall. In comparison, now in 2026, the cooling capabilities and power distribution have been expanded to handle around 1.5 MW in the same facilities!

Organisational changes for PDC

In 2000, Björn Engquist stepped down as the director of PDC, and Lennart Johnsson was hired to take over that role and also became a professor at KTH. In 2005, there was a major reorganisation at KTH, where individual departments disappeared and were replaced by larger schools. NADA was dissolved and PDC was migrated (in organisational terms) into the KTH School of Electrical Engineering and Computer Science (EECS), while remaining in the Teknikringen 14 building at KTH.

In 2008, KTH established the service and infrastructure part of PDC as a separate entity reporting to the KTH administration and recruited Erwin Laure, previously the technical director of the EGEE (European Grid for E-SciencE), from the European Council for Nuclear Research (CERN) in Geneva, Switzerland, as the full-time director of the infrastructure and service part, which was called PDC-HPC. The research part of PDC continued as PDC under the directorship of Lennart Johnsson. Several years later, around 2012, both units were recombined into a single PDC organisation, and Lennart was appointed as chairman of the PDC board, while Erwin became the director of PDC.

In early 2020, Erwin was appointed as the director of the Max Planck Computing and Data Facility in Germany. Patrick Norman, a professor in the Division of Theoretical Chemistry at KTH and a previous director of the National Supercomputing Centre (NSC) at Linköping University, took on the role of acting director at PDC. At the start of 2021, Dirk Pleiter, a professor of theoretical physics at the University of Regensburg, Germany, and a research group leader at the Jülich Supercomputing Centre (JSC), Germany, was appointed as the new director of PDC. Then, in early 2023, KTH changed its approach to appointing PDC directors. Instead of having a permanent director, the idea was to appoint a KTH professor involved in HPC research to serve as a director for a four-year term, after which another KTH professor would take up the mantle. Patrick was appointed as the first such PDC director, and Dirk, who was already a professor at KTH, then moved on to become Professor of Innovative Computer Architectures at the University of Groningen in the Netherlands.

The first big Cray system at PDC

Late in 2009, PDC was granted funding for a new supercomputer funded by SNIC and KTH. Gert handled the procurement process, and the contract was awarded to Cray. The new system, a Cray XE6 known as Lindgren in honour of the Swedish author Astrid Lindgren, was installed in June 2010. The system was upgraded in late 2010 into a 305 teraflops XE6 system, which became one of the major systems in Europe at that time.

Gert’s environmental work on reusing heat from supercomputers

Gert has always been interested in environmental issues and was responsible for an early initiative to reuse heat from the supercomputers at PDC. This involved collecting the hot air from Lindgren, which was designed as an air-cooled system. The hot air was used to warm water in large heat exchangers, and that warm water was then sent to another building at KTH to be used for heating. The chemistry lab was selected for this purpose since it required a lot of fresh air (due to the chemicals that were used there). That meant a lot of heating was needed in the winter to warm up the fresh air used to ventilate the building. Fortunately, there were already water pipes (used for hydronic heating and cooling) running between the two buildings housing PDC and the chemistry lab, and it was possible to reuse those pipes in what was known as winter mode (that is, when there is no need for hydronic cooling so only heating is used).

The experiment was successful, however, when it came time to replace Lindgren with a bigger system several years later, the amount of heat that would have been produced by the new system would have been a significant increase compared to the amount Lindgren produced. This would have meant that there would have been excess heat being produced that could not have been re-used with the existing set-up. As the chemistry lab could only use a certain amount of heat, PDC would have needed more buildings at KTH to heat in order to use up the extra heat that would have been produced by the new system. However, that was not practical as most buildings at KTH needed higher temperature water due to the types of radiators that they had.

The success of the heat re-use experiment with Lindgren inspired Cray and PDC, and led to the next system at PDC, another Cray called Beskow, having heat exchangers built into it at the factory. Beskow was a Cray XC40 system that was installed at PDC in late 2015 and named in honour of the Swedish author Elsa Beskow. Gert coordinated the procurement process, as well as the modifications needed in the PDC computer hall to accommodate the new system and its pipes.

At this time, KTH also installed large heat pumps at the main campus in Stockholm. Their purpose was to increase the temperature of water heated by reused heat from various locations around KTH (primarily PDC) to a level suitable for use in the hydronic radiators in buildings on the KTH main campus.

Large investment in a general-purpose system at PDC

In 2019, PDC received a substantial grant from SNIC to procure a new SNIC-funded general-purpose HPC system for academic research, to be hosted at PDC. Unlike the previous Lindgren and Beskow systems at PDC, the idea was that the new system would have two partitions: one with only CPUs and the other also equipped with GPU accelerators. Gert led the procurement of the system. Hewlett Packard Enterprises (HPE), which acquired Cray in 2019, was one of the companies that put in tenders for the new system and was eventually awarded the contract. The proposal was the HPE would install the system in two phases: the first phase would consist mainly of the CPU partition, while the second phase (which included the GPU partition) would be installed later as the development of the cutting-edge GPU technology that would be used in the system was not complete.

In August 2021, the first phase (with the CPU partition) of the HPE Cray system was installed at PDC. The system was named Dardel in honour of the Swedish couple Nils and Tora Dardel, who were respectively a painter and a writer. Handling the logistics of the installation was particularly difficult due to pandemic-related restrictions at the time.

The second phase with the GPU partition of the Dardel system was mainly installed in mid-2022. However, there have been several upgrades and expansions to the system – partly thanks to delivery delays, which entitled KTH to compensation in the form of extra nodes and storage, and partly because of additional funding from SNIC. Dardel became by far the fastest academic system in Sweden.

In the years since Beskow was installed at PDC, HPE has improved the water cooling technology for its systems, and the Dardel system at PDC is directly water-cooled via a network of pipes that were installed as part of the supercomputer. Water flows directly to the components that need cooling; there is no intermediate step involving air. The heat obtained from cooling Dardel is reused at KTH via the KTH heat pumps. This approach is the way of the future, since direct water cooling is considerably more efficient and allows the water to be heated to a higher temperature.

Changes to HPC resource provision in Sweden

When the Swedish Research Council established NAISS to take over after SNIC at the start of 2023, it led to changes in the landscape of academic supercomputing in Sweden. The speed and complexity of supercomputer systems have increased exponentially since PDC was established, and so too has the cost of those systems. Funding multiple HPC systems in Sweden ceased to be efficient. Consequently, the Swedish Research Council and the universities and other organisations in Sweden that rely on supercomputers for academic research modified their approach. Instead of hosting several systems at different sites in Sweden, the focus is now on consolidating the sites so that all the top-level supercomputers are hosted at a single NAISS location. The new NAISS organisation made it possible for Sweden to obtain EuroHPC funding for a top-level system, known as Arrhenius, as well as setting up an “AI factory” in Sweden. While individual universities may still host smaller systems with local funding for their own researchers and partners, there is a move towards single extremely large systems that are shared nationally or on a European scale due to the high cost. As the Dardel system at PDC approaches the end of its life, the focus of PDC is shifting away from large systems hosted at PDC. Some PDC staff are already working as part of the NAISS distributed approach to user support, in which application experts from across Sweden are available to help researchers from anywhere in Sweden, rather than primarily providing help locally.

PDC metamorphoses into KCSC

PDC has always been a highly flexible organisation, adapting to and often front-running ongoing global changes and developments in high-performance computing. With the logical changes in the Swedish approach to providing supercomputing resources for research, PDC’s current director, Patrick Norman, has worked with the KTH administration to implement an inspiring change that will ensure PDC continues to provide first-class support for KTH researchers.

While the major European-level HPC system in Sweden will be hosted by NAISS, PDC is evolving into a new centre, known as the KTH Center for Scientific Computing (KCSC), providing support to researchers who need to adapt existing codes or develop new codes to take full advantage of the latest HPC systems. This is a vital step because the architecture of supercomputers has evolved and become orders of magnitude more complex (for example, including a mix of CPUs and GPUs of different types, as well as complex memory systems). The art and science of adapting code to run as efficiently as possible, to maximise the size of the problems that can be solved or simulated, or to minimise the runtime, have become ever more important. Domain experts in, for example, biological chemistry or fluid dynamics, need application experts who know how to optimise code for these systems.

The new KCSC centre was established on the 25th of October 2025, and PDC’s systems and staff have been incorporated into the new centre, which will particularly focus on further developing scientific codes such as GROMACS, Neko, and VeloxChem, as well as taking initiatives to develop new, modern scientific codes.

Au revoir, Gert!

In late March 2026, Gert will turn 70, and he will retire from KTH during the year, glad to have enjoyed over 50 years of interesting challenges at KTH and PDC. While Gert is certainly looking forward to spending more time in the Swedish mountains that he loves to explore, he may also lend a hand with future HPC challenges.