Early versions of PureBasic started on Amiga, with the support of the Motorola 680×0 CPU series. PureBasic has been designed from start to generate raw assembly code, and the first working CPU assembly back-end was the one for 680×0. As the audience on Amiga was quickly falling, we decided to move on PC. In 1998, there was only one PC CPU architecture, the x86, and it was decided to add a new assembly back-end for it. It is a very time consuming task and requires a lot of learning to be able to handle a new assembly. 680×0 and x86 architectures are quite opposite but it fitted in the PureBasic architecture and PureBasic for x86 was a reality.

Shortly after, we decided to go really cross-platform and support OS X, which was running on PowerPC processors. A new assembly back-end was needed and we did it again, during almost a full year to get something working. Shortly after, Apple announced they will drop support for PowerPC and choose x86 instead ! That was a huge blow for us, and a lot of time wasted. In the meantime, x86-64 was released and slowly gaining traction and we knew we should support it as well. Unlike the name suggests it, x86 and x86-64 are two different beasts, and a lot of work was needed to make it happen. Once done, we though we could finally focus on functionalities for a while, but computer science is evolving at quick rate and a new CPU architecture appeared and took the world by storm, first on mobiles, then on tiny computers and now on desktop: ARM and ARM-64.

So here we are, to support these new platforms, we face the same issue again: we need new assembly back-ends. As a small company, we can’t dedicate a whole year every time a new assembly emerge, so we need to find a future proof solution. Meanwhile, a fork of the PureBasic compiler has been created to generate high-level language instead of raw assembly. This allowed to create SpiderBasic, a PureBasic-like language which can run on the browser by generating JavaScript code.

When Apple announced their new computers will be running on ARM, we knew we needed to change our way to handle new processors architecture and we decided to leverage the work already done of SpiderBasic and port it back for PureBasic. We experimented a lot with LLVM for quite a time now, but it is an headache to find it on all architectures. Also it’s a kind of mix between assembly code and high-level language so it didn’t fit either PureBasic or SpiderBasic back-end architecture. For the past 8 months, we decided to focus on a new high-level back-end for PureBasic: the C language.

The good news is the whole PureBasic compiler test suit is successfully running on the C back-end, so we are very confident we could have a compiler ready to test quickly. Stay tuned, more details to come soon !