Why not latest version of Quartus Prime?

[Solskogen]

I'm pretty sure that there's a very good reason for why not using the latest (and greatest?) version of Quartus Prime, but I've failed to figure out why. Perhaps something to put into a FAQ?

[ExCyber]

I guess only Sorgelig can give the authoritative answer to this, but my understanding is that most of the development work on newer Quartus versions is targeted at newer product families (10 series, Agilex), so the benefit of upgrading is minimal for designs targeting Cyclone V. Upgrading also changes the project files for the cores to fit the new version. The effect is that moving to the latest version would basically just create pointless churn for MiSTer.

Also, it's standard practice in embedded systems to freeze some concept of the build environment/configuration when a product ships. An upgrade of the toolchains in particular is usually considered to have significant risks and might call for a full round of regression testing. It's not unheard of to unplug the whole build machine and store it somewhere to preserve the build environment as fully as possible (some tools care about host MAC addresses, serial port numbering, drivers, etc., so this isn't quite as silly as it might sound). Compared to that, Sorgelig's approach is pretty flexible.

[dshadoff]

Also, please read the other topics in the "Development for MiSTer" forum here.
This is a topic which has come up and been answered many times.

[Solskogen]

Even the latest version of Quartus supports Cyclone V. I managed to compile both Minimig/Amiga core and ao486 with it, although the compile times were hideous (over 1 hour for ao486 on a Ryzen 3900x) with some very small changes. That's why I asked
I've looked at the forum and can't find a reasonable thread about it, @dshadoff. Perhaps it was on the old atari-forum? I've only seen discussions regarding 17.0 vs 17.1.

[rhester72]

There are a few cores that won't compile cleanly on 19 and no real updates to Cyclone V support (specifically) beyond 17.0.2, hence the recommendation. Nothing mystical, just unnecessarily compiler breakage and headaches without benefit.

17 won't compile any faster, tho.

[dshadoff]

Well, perhaps not so much of an explanation, but rather a demonstration of why...

viewtopic.php?f=28&t=370

viewtopic.php?f=28&t=376

viewtopic.php?f=28&t=692

Basically, lots of the internals changed, and require lots of changes to be made to things like build files.
Such changes come at a cost (and on a semi-annual basis), but with zero benefit.

And if no changes were required, they would be compatible and there would be no reason to discuss which version is necessary (only which version is recommended).

[RedskullDC]

I managed to compile both Minimig/Amiga core and ao486 with it, although the compile times were hideous (over 1 hour for ao486 on a Ryzen 3900x) with some very small changes. ....

If you have enough memory, I recommend you create a >=1GB ramdisk, and compile your project there.
Compile times will come down significantly.
I use "Primo Ramdisk", but there are plenty of other similar products.

Also turn of any virus checker software, which might slow down disk read/writes.

Regards,
Red

[Solskogen]

One of the problems with version 17 is that it's not possible to use it on newer Linux distributions