VAX port needs help

June 05, 2020 posted by Martin Husemann

The VAX is the oldest machine architecture still supported by NetBSD.

The support for it sometimes causes heated discussions, but it also has benefits:

  • It uses a pre-IEEE 754 version of floating point numbers, while all other (later) architectures use some variant of IEEE 754 floating point in hardware or emulate it.
  • By today's standards all machines are slow and have little memory.

This are severe challenges for a general purpose operating system like NetBSD, but also provides reality checks for our claim to be a very portable operating system.

Unfortunately there is another challenge, totally outside of NetBSD, but affecting the VAX port big time: the compiler support for VAX is ... let's say sub-optimal. It is also risking to be dropped completely by gcc upstream.

Now here is where people can help: there is a bounty campaign to finance a gcc hacker to fix the hardest and most immediate issue with gcc for VAX. Without this being resolved, gcc will drop support for VAX in a near future version.




How is LLVMs support for VAX? Can it be used as an alternative? Is adding support to LLVM a thing to consider?

Posted by Reinoud on June 06, 2020 at 01:20 PM UTC #

According to the LLVM webpage current supported hardware includes AArch64 & ARM, Itanium (ia64), Lanai, MIPS, PowerPC, AMDGPU, RISC-V, SPARC, SystemZ, X86, XCore and Hexagon. This is far less than what gcc supports and never included VAX. Rather than pushing in different directions, focusing on creating one bounty large enough to fund development is likely the only way to preserve the support for VAX. For some people VAX is most interesting as a platform for running DEC clusters under VMS. At the same time, aside from the no longer supported PDP-11, the VAX it is one of the earliest architectures to run Unix. From this point of view, it is not clear if trying to keep up with current versions of NetBSD is as important as carefully preserving historic versions of Unix with only the minimal changes necessary to fix security and bugs.

Posted by Eric Olson on June 07, 2020 at 01:42 AM UTC #

Is there any value in allocating development resources to support a system that can no longer be bought? This is an essential question as NetBSD's resources are scarce and supporting antiques is done to the detriment of contemporary platforms - whose the "Of course, it runs NetBSD" should apply first and foremost!

Posted by Vincent on June 09, 2020 at 01:09 PM UTC #

Wow! Here we have an OS that could still keep a VAX running depending on the support of a compiler maintained by a large community, but that community needs the help of this one. Could LLV come to rescue instead? And kudos to DEC! Their HW still runs!

Posted by on June 11, 2020 at 09:32 PM UTC #

Why not port the original BSD compiler forward?

Posted by Greywolf on August 15, 2020 at 09:27 AM UTC #

Post a Comment:
Comments are closed for this entry.