SystemPatch3 - WHDLoad preload incompatible?

[Shoonay]

Just found a few secs of free time, between posting more cat pics...

Copied and installed it on my real Amy's HD in WinUAE, restarted, and already felt the difference.
There was an even bigger speed, wbscreen refresh rate and overall improvment when I've reconnected the HD to my real Amy!

But, unfortunately... now the WHDLoad's preload function doesn't work
Tried it with Ultima 6 for the first time and thought... maybe it's a buggy slave? Then tried Prehistorik - the same = not preloaded...

After that, tried it with the older quicklayers.lib and SystemPatch2 and WHDLoad preloaded just right.

My specs are: Amiga 1200 with Blizzard IV 030 50mhz 32MB, KS3.1 blizkicked, OS3.9BB2 on 60gb hd

Thankyou very much for your hard work ziosante, it is a *REALLY* helpfull patch, and I hope you fix that problem!

---=== EDITED ===---
Further investigation revealed that the problem is with the new SystemPatch (3) in the C drawer.
The old SystemPatch (2) and the new quicklayers.library (3) don't cause that problem.

[alexh]

The new version of System Patch includes Chris "Platon42" Hodges new TLSFMem memory allocator, perhaps there are a few issues there?

Perhaps try the no MMU type?

[Shoonay]

Thanks alexh, but neither nommu, nocache nor both, do anything.
It's still preloading while in game

[alexh]

The TLSFMem patches Deallocate(), FreeMem(), AvailMem(), Allocate(), AllocMem(), AllocAbs(), AddMemList()

I would imagine that there might be a problem between WHDload and the new AvailMem, perhaps it thinks there is less RAM available than there actually is.

Only Wepl will know for sure. Bert are you reading??

[Wepl]

I'm reading
But have no good idea, the memory allocs in WHDLoad should be 100% os conform. You may try option NOMEMREVERSE with WHDLoad 16.9 but I doubt that it will help.

For debugging use option TRACE and post the .whdl_trace file here.

[Shoonay]

Alrighty, thanks guys. I'll try that Wepl, but it might take time...

[alexh]

It does say in the TSLFMem docs:

Currently does not support MEMF_REVERSE (could be added, but there is not much sense in this).

So maybe the NOMEMREVERSE may work?

[Wepl]

Originally Posted by alexh

It does say in the TSLFMem docs:

So maybe the NOMEMREVERSE may work?

my interpretation of this is, that the flag MEMF_REVERSE is ignored, if so it will make no difference.
but the tslfmem docs are not detailed enough, so the option is worth a try

[Shoonay]

Nope, with the whdload *.slave preload NOMEMREVERSE I can't even load a game anymore.

Did try the trace function with some games, here's the results:

[Wepl]

It seems that TLSF returns 2,018,011,950 for AvailMem(MEMF_LARGEST).
Which causes that nothing is preloaded because WHDLoad uses this to calculate the puddle size for the memory pools used for the PreLoad memory.

I would recommend to request a fix for TLSF and for now disable it to run WHDLoad.

I don't understand why it should not work at all with NOMEMREVERSE. What happens in that case?

[Shoonay]

Originally Posted by Wepl

I don't understand why it should not work at all with NOMEMREVERSE. What happens in that case?

Well, uhm, nothing.
Let's say I'm in FileMaster, I go into the prefered games directory, double-click on a slave, type whdload *.slave preload nomemreverse and press [enter], and... nothing happens = the game does not get loaded, I'm still in FileMaster

[Wepl]

what whdload version?

[Shoonay]

The newest I could find, 4474, also tested on the one before that: 4412

[Shoonay]

*BUMP*

Any news?
Ziosante? Wepl?
Wepl? Ziosante?

[Shoonay]

*pwease*

[alexh]

The new author of MemTLSF (as it is now known) hasn't released any bugfixes AFAIK.

[Wepl]

This must be fixed in TLSF.

[Cosmos]

I have a raw source of TLSFMem, maybe I'm able to fix...

First, WHDLoad don't run with TLSFMemPool, but ok with TLSFMem here : please Wepl and Shoonay, confirm on your Amiga setup (WinUAE is used for me)

So, the problem is somewhere in TLSF_CreatePool, TLSF_DeletePool, TLSF_AllocPooled, or TLSF_FreePooled


>It seems that TLSF returns 2,018,011,950 for AvailMem(MEMF_LARGEST).
>Which causes that nothing is preloaded because WHDLoad uses this to
>calculate the puddle size for the memory pools used for the PreLoad memory

I tried a jsr -$D8(a6) with d1=MEMF_LARGEST and I get the right available memory, with TLSFMemPool and TLSFMem

Wepl : tell me more about your 2,018,011,950-test !

[Wepl]

the AvailMem bug has been fixed in later releases of tlfsmem, it's also mentioned in the changelog. my assumption was that it works since them without problems together.
since them I never tried it again (I cannot use it because the incompatibility with warpos).
if it doesn't work with TLSFMemPool this is probably a different problem. so what doesn't work???

[Cosmos]

Give me ALL R_CreatePool, R_DeletePool, R_AllocPooled, and R_FreePooled args from your WHDLoad great proggy, and I'll see !