GVP A4060DT 68060.library troubles

[mousehouse]

About 3 months ago I purchased a A4060DT through Ebay... I got lucky I think because I didn't put in a very high bid but still ended up winning the auction. Ah well...

Ever since I've occasionally tried to get the accelerator working, just never put any real time in it until yesterday. I took off al jumpers and used Greg Donner's page to set them all as I thought they should be. Now finally SCSI worked, auto booting worked and I could actually hear the SCSI disk (2GB Seagate USCSI drive) being accessed. But the system always freezed. First I needed to discover that it won't work with my IDE-CF installed, so I took it out. Now there is a long delay on boot up as the Amiga frantically tries to do something silly with IDE... I need to get a terminator for that.

After playing around with the "boot without startup-sequence" I found out that the system would freeze on SetPatch, after the first reboot (e. the second run). This had me puzzled. I took out SetPatch and was left with a b0rked 3.9 setup. Now "WhichAmiga" would also crash, which led me to believe something else was wrong.

I was using the Phase4 set of 4 libraries (68060.library, 68040old, -new, etc.). which turned out to be the problem. After putting the GVP supplied libraries from the install disk in LIBS: everything runs fine and speedy.

Now I'm wondering what the differences are between those library sets... the Phase5 ones are much larger (130K vs. 60K for the 68060.library if I recall correctly).

Anyway... this just for anybody who manages to get their hands on a A4060DT / Trex2 and is wondering why it is all messed up :hammer:

[mousehouse]

And attached tech p0rn...

[matthey]

After playing around with the "boot without startup-sequence" I found out that the system would freeze on SetPatch, after the first reboot (e. the second run). This had me puzzled. I took out SetPatch and was left with a b0rked 3.9 setup. Now "WhichAmiga" would also crash, which led me to believe something else was wrong.

The 68060.library is loaded at SetPatch. The AmigaOS 3.9 SetPatch is the first to load 68060.library directly and doesn't require the small "stub" 68040.library with the 68060. WhichAmiga does have some bugs also.

I was using the Phase4 set of 4 libraries (68060.library, 68040old, -new, etc.). which turned out to be the problem. After putting the GVP supplied libraries from the install disk in LIBS: everything runs fine and speedy.

Now I'm wondering what the differences are between those library sets... the Phase5 ones are much larger (130K vs. 60K for the 68060.library if I recall correctly).

The Phase5 CPU libs have extra support code for P5 accelerators. They do run on most accelerators with minimal problems and good speed. I don't know why the P5 68060.library would not run on your 4060. I wouldn't think Jeff Boyer would do any dirty hardware tricks. You could try ThoR's Mu Libs (including 68060.library) and see if they run. It's a little bit more difficult to setup (read the docs) but is very stable and fast on my P5 CSMK3...

http://aminet.net/util/libs/MMULib.lha

ThoR is still at the same e-mail and may provide support if there is a problem with this accelerator. Also, you might check out this web site about the GVP 4060 which has some similarities and some possible alternative CPU libs...

http://www.gregdonner.org/gvp4060/downloads.html

[mousehouse]

Thanks for the great tips. I'll check out the MMUlib install, looks OK to me. What would be a good stress-test that hits these libs extensively?

[itix]

68040 and 68060 libraries are always vendor specific.

[matthey]

Thanks for the great tips. I'll check out the MMUlib install, looks OK to me. What would be a good stress-test that hits these libs extensively?

The most important thing the 68060.library does is trap missing instructions. Some are integer (e.g. multiply 32x32=64 bits) while others are floating point emulating the instructions in the 68881/68882 but not the 68060. These instructions are pretty common in older software that is not compiled for the 68060 and are "tested" when trapped all the time. The 68060.library also configures the 68060 and works around bugs in the 68060. It sets up and maps the MMU in the case of the Mu libs which are used by various expansion hardware like gfx cards. You shouldn't notice that the 68060.library is there when it's working correctly. Reading the Mu docs, even if you don't use it, is very informative.

68040 and 68060 libraries are always vendor specific.

They shouldn't be if the accelerator is designed according to C= hardware specs. Not all accelerators are. The original C=, P5 and Mu 68060.library all work on my P5 CSMK3 and it's not completely compliant. ThoR wrote a little fix for P5 hardware to deal with the lack of compliance. He complains about their "hack" job in his docs.

[stefcep2]

So what should I use on my CSII 68060?

[matthey]

So what should I use on my CSII 68060?

ThoR's 68060.library should work on the CSMK2. I believe it was less hackish than most P5 accelerators. Some of the 1200 P5 accelerators were the worst hack jobs. ThoR's 68060.library is among the fastest out there when set up correctly. I have this line in my S:Startup-Sequence...

MuFastZero MOVESSP ON

I know the trapped 32x32=64 bit multiply is up to 50% faster than some other 68060.libraries including the P5 68060.library. These commonly used multiplication functions are also patched in the utility.library. I believe ThoR's 68060.library patches more potential bugs and does a better job of setting up the MMU. ThoR's MMU knowledge is second to none. He also still supports his library. It is used as the Natami 68060.library for their 060 card.

On the other hand, the P5 68060.library was fast and reliable for me. Using a 68060.library created by the same company as your accelerator should insure excellent compatibility. The only down side I found was the bloated size of the P5 libs.