TOS RAM test routine ?

[exxos]

I'd been hacking that file already for my own nefarious ends.

I'm afraid a debug Linux version is probably not much use to you.

dammit

Can you try these and see if any are faster ?

viewtopic.php?p=95016#p95016

[Badwolf]

Can you try these and see if any are faster ?
viewtopic.php?p=95016#p95016

Just eyeballing these as don't have time to measure them properly, but TP32 doesn't seem to make any odds, but the inverse video looks corrected.

80000 is noticeably quicker, but something goes wrong with the ST-RAM test -- it aborts a quarter of the way through.

I noticed the tests run with both caches disabled. Whilst this makes sense for the data cache, I wonder if enabling the instruction cache would have an effect? It's slowing to 8MHz to read each instruction ATM. That said, just hacking it on in Hatari (r cacr=$1) didn't seem to make much difference, but it doesn't have the differential slowness of the real thing.

BW

[exxos]

Just eyeballing these as don't have time to measure them properly, but TP32 doesn't seem to make any odds, but the inverse video looks corrected.

80000 is noticeably quicker, but something goes wrong with the ST-RAM test -- it aborts a quarter of the way through.

I noticed the tests run with both caches disabled. Whilst this makes sense for the data cache, I wonder if enabling the instruction cache would have an effect? It's slowing to 8MHz to read each instruction ATM. That said, just hacking it on in Hatari (r cacr=$1) didn't seem to make much difference, but it doesn't have the differential slowness of the real thing.

Thanks,On the right track then basically.

TP32 is supposed to enable the caches. I could probably figure out how to enable it in the memory test routines directly. And possibly patch the STram test to run at the original $20000 value. Maybe chuck out part of the memory test as well is not really doing anything anyway.

Difficult for me to test unless some kind soul can do me a Hatari Windows build

[Badwolf]

Difficult for me to test unless some kind soul can do me a Hatari Windows build

Changing your TOS version to 306 doesn't work? (Maybe 307 to be safe)

BW

[exxos]

Changing your TOS version to 306 doesn't work? (Maybe 307 to be safe)

Already mentioned all that previously.. It just completely skips the memory tests.

[Badwolf]

Changing your TOS version to 306 doesn't work? (Maybe 307 to be safe)

Already mentioned all that previously.. It just completely skips the memory tests.

That's probably the memvalid check. I'm not sure if that can be bypassed with the patch systemvariables to boot faster switch in the system options screen.

I'm afraid I can do you Linux and possibly MacOS, but not Windows.

BW

[exxos]

I'm afraid I can do you Linux and possibly MacOS, but not Windows.

Don't worry about it.

I really need to get Ubuntu onto my laptop at some point. Just never really on my to do list and don't really fancy that world of pain right now

Saying that someone donated me a MacBook so I could actually run a Mac version but would have to steal it off the girlfriend first

The fallback option is just erasing programming the ROM 100 times over like I had to do last time. But don't really have the enthusiasm to go through all that again at the moment.

[Badwolf]

I really need to get Ubuntu onto my laptop at some point. Just never really on my to do list and don't really fancy that world of pain right now

All my linux boxes (bar the Pis) are virtual machines. Something goes wonky? Meh. Do another.

Debian with XFCE is my weapon of choice. Umbungo's constantly updating.

BW

[exxos]

@Badwolf When you get a moment can you please give this one a try. I tried only altering the number on the TTram test..

If it doesn't work I will just build @frank.lukas a German one from the original build files and just leave it like that until I get chance to go back into it all again.

tos206uk.zip

(127.23 KiB) Downloaded 31 times

[dml]

I noticed the tests run with both caches disabled. Whilst this makes sense for the data cache, I wonder if enabling the instruction cache would have an effect? It's slowing to 8MHz to read each instruction ATM. That said, just hacking it on in Hatari (r cacr=$1) didn't seem to make much difference, but it doesn't have the differential slowness of the real thing.

While Hatari emulates the correctness of the caches (at least, for a working program), it does not emulate the time aspects of cache behaviour. So doing performance comparisons with caches on/off won't give you anything very sensible IMO