68000 CE 8x speed problem ?!

[amilo3438]

Below are some tests that explain/show influence of MHz and Fast RAM to A500 speed:

Case 1) A500 7.09MHz 512 Chip + 512 Slow RAM = 553 Dhrystones (pic.1)

Case 2) A500 7.09Mhz 512 Chip + 1Mb Fast RAM = 729 Dhrystones (pic.2) ... ~1.32 X Case1

Case 3) A500 14.18MHz 512 Chip + 512 Slow RAM = 602 Dhrystones (pic.3) ... ~1.09 x Case1

Case 4) A500 14.18MHz 512 Chip + 1MB Fast RAM = 1520 Dhrystones (pic.4) ... ~2.75 x Case1 (OR ~2.08 x Case2)

EDIT: In addition...

Case 5) A500 28.4MHz 512 Chip + 512 Slow RAM = 640 Dhrystones (no pic.) ... ~ 1.16 x Case1

Case 6) A500 28.4MHz 512 Chip + 1MB Fast RAM = ... In SySinfo this test goes crazy (over 3x A4000@25) ?!? ... see EDIT3 below!

The test Case 6 was set to CE 8x speed (same as Case 5), so why it goes crazy, the reason cant be only because of Fast RAM ? ... test done with WinUAE2810b6.

EDIT2: If same test as Case 6 but with CPU 68020 instead of 68000 (or 68010) and it works fine !!! (SySinfo doesnt go crazy )


EDIT3: With a lot of reluctance the WinUAE problem finally fixed (see post 11), so now is able to complete the test:

Case 3) A500 14.18MHz 512 Chip + 512 Slow RAM = 593 Dhrystones (no pic.) ... ~1.07 x Case1

Case 4) A500 14.18MHz 512 Chip + 1MB Fast RAM = 1471 Dhrystones (no pic.) ... ~2.66 x Case1 (OR ~2.02 x Case2)

Case 5) A500 28.4MHz 512 Chip + 512 Slow RAM = 620 Dhrystones (no pic.) ... ~1.12 x Case1

Case 6) A500 28.4MHz 512 Chip + 1MB Fast RAM = ... 2956 Dhrystones (pic.5) ... ~5.35 x Case1 (OR ~4.05 x Case2 AND ~2.01 x Case4)

[Toni Wilen]

Sysinfo with config that does not exist in real world does not really mean anything at all. (and sysinfo even with real config does not really show anything useful either )

[amilo3438]

Originally Posted by Toni Wilen

Sysinfo with config that does not exist in real world does not really mean anything at all. (and sysinfo even with real config does not really show anything useful either )

How strange, testing the same config. (A500 68K 28MHz 512 Chip 1Mb Fast) on older WinUAE 2.0.1 and Sysinfo doesnt go crazy (see pic.) ?!

[Toni Wilen]

So? It still means nothing at all.

It could be overflow in sysinfo. Anything is possible. No way to test: don't care.

[amilo3438]

same A500 CE with custom 28.375160 setting and 512Chip + 1MB Fast...

on 2.6.1 test works fine (pic.1)

on 2.7.0 test goes crazy (pic.2) ?! (something is obviously broken and you still dont care ?!)

[Toni Wilen]

Because I don't bother until it can be confirmed and if it is interesting. This can't be confirmed and is not interesting. You are also assuming that if something looks wrong, it is wrong. Never ever do that!

Maybe it becomes more interesting if you find the exact beta that started working differently.

[amilo3438]

I found it...

the problem starts from WinUAE 2.7.0 beta4 !!!

in 2.7.0 beta 3 it works fine with 8x or custom setting ! (lazybones )

EDIT:
2.7.0.beta3 -> http://eab.abime.net/showpost.php?p=901068&postcount=74
2.7.0.beta4 -> http://eab.abime.net/showpost.php?p=902679&postcount=88

p.s. CIA tod bug ON/OFF doesnt help !

[Toni Wilen]

Don't really see anything useful in change logs but try toggling CIA bug option in Advanced chipset to see if it has any effect.

[amilo3438]

Originally Posted by Toni Wilen

Don't really see anything useful in change logs

Therefore it is also important to keep the test versions between the two beta ... which of course you probably do not do.
(unfortunately, I started to collect a test betas after 2700 beta8)

Originally Posted by Toni Wilen

You are also assuming that if something looks wrong, it is wrong. Never ever do that!

Do what !? Spending my free time to help finding a possible problem ?! (if you wish, no problem)

p.s.
I must say that it is not an interesting job to seek other's mistakes and if you find them, nobody will say thanks.
Yeah, thus one year of volunteering on this topic would be quite enough.

Cheers

1 year volunteering = cumulatively with 20 days of unused annual leave (just in time before the World Cup)


EDIT:

Originally Posted by Toni Wilen

Sysinfo with config that does not exist in real world does not really mean anything at all.

According to BigFan (see below) there really exist a similar config:

GVP Supra turbo 28Mhz, A500 Accelerator
http://eab.abime.net/showthread.php?t=28503
http://www.bigbookofamigahardware.co...uct.aspx?id=87


Yeah, also my mistake was I have not tried another example instead of Sysinfo. (BigFan, many thanks)

[BigFan]

I don't give a shit on Sysinfo, but amilo is right.
I can confirm emulation is way too fast with this setting. Lha packs/unpacks archives more than 10x faster.
Workbench is loading and running as if JIT is enabled. It is odd to complain about additional speed , but it's definitely to much if one wants to emulate a Supra28 board f.e.

[Toni Wilen]

Originally Posted by amilo3438

Therefore it is also important to keep the test versions between the two beta ... which of course you probably do not do.
(unfortunately, I started to collect a test betas after 2700 beta8)

No need, source tree shows all changes in detail. I was simply not interested enough to check it yet, because of using sysinfo results

Originally Posted by BigFan

I don't give a shit on Sysinfo, but amilo is right.
I can confirm emulation is way too fast with this setting. Lha packs/unpacks archives more than 10x faster.

Thanks. This is much better report, I always prefer real time speed tests because they can't show incorrect times.

about additional speed , but it's definitely to much if one wants to emulate a Supra28 board f.e.

It does not emulate Supraturbo28, it has 64k cache which allows CPU to run really fast if data is already in cache. But without fast ram results should be similar (supraturbo only caches fast mem data because caching chip ram would break most programs)

http://www.winuae.net/files/b/winuae.zip should fix it.