ST536 STE EDITION

[exxos]

Won't that be slower than the original test?

[stephen_usher]

Yes, but more thorough.

The original test won't pick up addressing errors

[exxos]

I'm looking to speed it up by like 50% without sacrificing to much testing wise. There's YAARTT for in-depth testing.

[dml]

If you know at boot time where and how much ram should exist, the march test is the best & simplest way to verify the RAM is good. The mapping thing I suggested above is useful when the RAM config may depend on other factors like the user, jumpers, board revision, IC / SIMM sizes etc. etc. as it will find the RAM for you.

For this board though, maybe that config is always fixed (?) so going straight to a march test without mapping might be good enough.

[stephen_usher]

Off topic: I wish YAART would add a March test. It won't detect addressing errors with the tests it's doing, well, only a few with "own address".

[exxos]

The OS should only be used on the STE536/ST536 and its always going to have 64MB RAM. I just find the power up TTram check very useful, it just takes way to long.

But the original test doesn't test own address which checks bad address lines, so I think that is a must , rather than writing more "fixed" values. Then the walking 1s would test every bit. I think that's better and quicker than the original tests and should be a bit faster as well.

Technically bht walking 1s would kinda detect bad address lines as well as the bits should be wrong. But I think address test is a good one to have ?

[exxos]

Off topic: I wish YAART would add a March test. It won't detect addressing errors with the tests it's doing, well, only a few with "own address".

Yeah. Maybe I will get AI to write a new TTram tester one day ;)

[dml]

The OS should only be used on the STE536/ST536 and its always going to have 64MB RAM. I just find the power up TTram check very useful, it just takes way to long.
But the original test doesn't test own address which checks bad address lines, so I think that is a must , rather than writing more "fixed" values. Then the walking 1s would test every bit. I think that's better and quicker than the original tests and should be a bit faster as well.

Right. So for a 'quick' power-on ram test, a page-based own-address test with a very limited bit-level test per page seems like a good option vs a full march test on everything.

Having an option to properly test the whole lot would be nice though, maybe with some held keys or whatever. But as stephen_usher wrote, diagnostic tests like YAART should really be using march for certainty.

[stephen_usher]

Technically bht walking 1s would kinda detect bad address lines as well as the bits should be wrong. But I think address test is a good one to have ?

"Own address" will only catch address errors which (a) are out of cadence with the address rotation, (b) affect all bits equally.

e.g. address error only affects the lower 8 bits and repeats memory every 1K. Own address won't spot this at all.

Because a March test uses in-RAM data and progressively modifies it by XORing any faulty addressing will flip the bits multiple times and amplify the resulting change. This is why it's so good at spotting addressing errors.

[exxos]

Having an option to properly test the whole lot would be nice though, maybe with some held keys or whatever. But as stephen_usher wrote, diagnostic tests like YAART should really be using march for certainty.

Yeah having a extended test could be a good feature on a keypress or something. Not sure what key as TOS uses a few already for various things. Though possible something like the control key before the RAM test wouldn't effect anything at that point anyway I guess..

I suppose TOS could have its own built in version of something similar like YAARTTT. Having a bit more diagnosis built in would be useful.