Sonnet 7200 PCI programming

[grelbfarlk]

It has extra mediatioring. Don't be like asking what things do, gawds.

[DarrenHD]

Originally Posted by grelbfarlk

it has extra mediatioring. Don't be like asking what things do, gawds.

lol

[jkdsteve]

You’d think this would be fun but my latest point of torture is trying to run this on a firestorm backplane. I’ve been trying for two days now and nothing PPC works right :-/

Anyone try this config?

[grelbfarlk]

Yes they did. A lot of the good stuff was in a Firestorm thread. Maybe this was it, or not, whatever.
https://eab.abime.net/showthread.php?p=1532713

[Hedeon]

[ Show youtube player ]

[Hedeon]

The bus timeout on the firestorm seems to be different than the mediator. Leading to much more issues (i.e. bus errors)

So what is your config?

[jkdsteve]

Originally Posted by Hedeon

The bus timeout on the firestorm seems to be different than the mediator. Leading to much more issues (i.e. bus errors)

So what is your config?

It’s an AA3000+ with the PCI board….hardly standard :-/

I haven’t had much look, it kind of stutters (timeouts? Exceptions?) to the point where the 060 can beat the Ragnarok (?) for decoding data types

[jkdsteve]

Originally Posted by Hedeon

[ Show youtube player ]

That’s the video….I knew he had it working. Maybe I try again without Zorro cards:

It makes no difference.

CyberPi works as does Bogomips.

Voxelspace works sort of but will lock up. There seems to be a diff response with p96 version compared to CGX. Looks like p96 is locked at 30fps, CGX is not and will hit 100fps before crash/lock.

Library versions:

PowerPC. library 17.14
Prometheus.library 4.3
Prometheus.card 7.66
3dfxVoodoo.chip 7.16

[trixster]

Originally Posted by jkdsteve

You’d think this would be fun but my latest point of torture is trying to run this on a firestorm backplane. I’ve been trying for two days now and nothing PPC works right :-/

Anyone try this config?

There are plenty of people running hedeon’s powerpc.library on Firestorm Prometheus machines (either aa3000+ or A4000) with a variety of ppc pci cards, so it’s not an uncommon ‘config’

[trixster]

@jkdsteve wrangler and I came across a weird circumstance recently (and quite independently) where too high 5v meant issues for Ragnarok ppc cards. We mentioned it here on the thread a little while back. What psu are you using and what are your voltages like?

[jkdsteve]

Originally Posted by trixster

@jkdsteve wrangler and I came across a weird circumstance recently (and quite independently) where too high 5v meant issues for Ragnarok ppc cards. We mentioned it here on the thread a little while back. What psu are you using and what are your voltages like?

I’ll check the voltages and report back here.

EDIT: it is 5.02 at the supply and 4.92 at the PCI backplane.

It feels like something weird is going on with CODE and DATA, I have all Sonnet ENV at default:

If I play AmigaAMP using mpega.library thro AHI Paula,I get massive stuttering and wrong speed. If I force AmigaAMP into PPC MEMORY by changing it’s memtaskpri then 68k side of program slows way down but music plays normally.

I expect audio to be weird with PowerPC but the above seems odd. There’s a weird experience using other programs too, like SYSSPEED DRIVE BENCHMARK slowing way (like Zorro speeds) down if INITPPC is started before it. However if I start SYSSPEED then INITPPC the drive benchmark is normal….30MB/s CSPPC.

Does this point to MMU issue or ? I followed all MMU library instructions.

CSPPC is the only accelerator I have that fits the 3000 case. I’m using stuff installed by mmu library and no ppc.library installed.

[trixster]

iirc only certain versions of WinAMP work with powerpc.library - youd have to search back deep into this thread as all chat about that was way back when

[jkdsteve]

Originally Posted by trixster

iirc only certain versions of WinAMP work with powerpc.library - youd have to search back deep into this thread as all chat about that was way back when

I’m using a version (2.x) that has previously worked on my A4000/mediator setup with the same WOS mpega.library.

[jkdsteve]

Well I’m completely at a loss what else to try now, I stripped the whole thing down to 030, Ragnarok (750 with 512MB?) and Voodoo 3 (I tried two different) and it still crashes on PPCWARP3D, Voxelspace.

I even tried A3640 and Warp Engine in place of the CSPPC (after updating mmu-config) with identical result.

It could be down to nitpicking P96 versions or ENV settings or I just have to accept the PPC ain’t gonna fly.

[jkdsteve]

Prmscan before and after INITPPC:

[Hedeon]

17.13b had a check that 040 was required. I think I forgot to implement that in 17.14? In other words: 040 minimum required.

Secondly, if any 68k code gets put into the memory of the PPC card, it will be damn slow. Same with 68k access to PPC card memory (2MB/s or something like that). I guess SysSpeed places it buffer in the wrong memory type. Most programs do it right, a few have a hardcoded patch in the library.

Thirdly, in my experience over the course now of 8 years, most issues regarding crashing are amiga HW related. Mostly voltage/amperes. You haven't shared any crash messages (or p96 version number and I think the prometheus.card version is wrong), so I cannot comment that much on it, but any 68k GURU is mostly 68K HW related. Sadly, such Amigas are not fit to run this PPC HW.

I mean, Voxelspace is nothing special. Mostly 68K with the calculations in PPC. And regarding Warp3D you need the stuff from Aminet. Maybe you have the Mediator libs.

And the last few builds were done on an AA3000+ with Firebird :-)

Put in a mediator and see if that helps.

[jkdsteve]

Originally Posted by Hedeon

17.13b had a check that 040 was required. I think I forgot to implement that in 17.14? In other words: 040 minimum required.

Secondly, if any 68k code gets put into the memory of the PPC card, it will be damn slow. Same with 68k access to PPC card memory (2MB/s or something like that). I guess SysSpeed places it buffer in the wrong memory type. Most programs do it right, a few have a hardcoded patch in the library.

Thirdly, in my experience over the course now of 8 years, most issues regarding crashing are amiga HW related. Mostly voltage/amperes. You haven't shared any crash messages (or p96 version number and I think the prometheus.card version is wrong), so I cannot comment that much on it, but any 68k GURU is mostly 68K HW related. Sadly, such Amigas are not fit to run this PPC HW.

I mean, Voxelspace is nothing special. Mostly 68K with the calculations in PPC. And regarding Warp3D you need the stuff from Aminet. Maybe you have the Mediator libs.

And the last few builds were done on an AA3000+ with Firebird :-)

Put in a mediator and see if that helps.

I made a typo on the card, they are all files from start of thread. This is PPC card from a working mediator setup so I know it works.

When I get chance, I’ll rebuild the OS and libraries from scratch, entirely possible I’ve a mixed setup now I’ve been swapping. P96 should be 2.1b from Aminet and I definitely just place Warp3D from Aminet also.

It’s weird how AmigaAmp behaves the same as SysSpeed though….I’ve never seen that on CSPPC or Mediator.

[jkdsteve]

Originally Posted by Hedeon

17.13b had a check that 040 was required. I think I forgot to implement that in 17.14? In other words: 040 minimum required.

Thirdly, in my experience over the course now of 8 years, most issues regarding crashing are amiga HW related. Mostly voltage/amperes. You haven't shared any crash messages (or p96 version number and I think the prometheus.card version is wrong), so I cannot comment that much on it, but any 68k GURU is mostly 68K HW related. Sadly, such Amigas are not fit to run this PPC HW.

s.

For ref, here's PRMSCAN after initppc:


Prmscan 1.6 by Grzegorz Kraszewski.
PCI cards listing:
-------------------------------------------------
Board in slot 1, function 0
Vendor: [1m3dfx Interactive Inc[0m
Device: [1mVoodoo3 All Voodoo3 chips, 3000[0m
Revision: 1.
Device class 03, subclass 00.
Address range: 62000000 - 63FFFFFF (32 MB).
Address range: 60000000 - 61FFFFFF (32 MB).
Address range: 7FE01000 - 7FE010FF (256 B).
64 kB of ROM at 64010000 - 6401FFFF.
Board driver: [32m3dfxVoodoo.chip[0m.
-------------------------------------------------
Board in slot 2, function 0
Vendor: [1mMotorola[0m
Device: [1munknown ($480B)[0m
Revision: 2.
Device class 06, subclass 00.
Address range: 64000000 - 64000FFF (4 kB).
Address range: 64001000 - 64001FFF (4 kB).
Address range: 70000000 - 7FFFFFFF (256 MB).
Address range: 64040000 - 6407FFFF (256 kB).
Address range: 68000000 - 6FFFFFFF (128 MB).
Board driver: [32mpowerpc.library[0m.
-------------------------------------------------

No gurus, the crashes are from the PPC side and the weird slow behavior makes no sense to me. Is there any way to capture debug memory Allocs or are they private to apps?

I rebuilt P96 and Warp3D from the ground up using Aminet files - they were the same version I already had but I cleaned the old ones out and reinstalled using the apps.

Versions:

Library versions:

PowerPC. library 17.14
Prometheus.library 4.3
Prometheus.card 7.60
3dfxVoodoo.chip 7.16

Picasso v2.0 Aminet archive:
Picasso96API.library 2.300
rtg.library 40.3945

Warp3D v4,2 (all files dated 07/13/01 = 13 Jul 2001)
Warp3D.library 4.2
Warp3DPPC.library 4.2
W3D_AvengerBE.library 4.2
W3D_AvengerLE.library 4.2
W3D_AvengerBE_PPC.library 4.2
W3D_AvengerLE_PPC.library 4.2
W3D_Picasso96.library 4.2
W3D_Picasso96_PPC.library 4.2

I actually managed to capture one today:

[Hedeon]

Use of zero pointer. Also r2 is suspicious. Turn off any Mu stuff in your startup please. (I see MuGuardianAngel?). Also give me the output of a MuScan please after initppc.

And does 68k w3d work?

[jkdsteve]

Originally Posted by Hedeon

Use of zero pointer. Also r2 is suspicious. Turn off any Mu stuff in your startup please. (I see MuGuardianAngel?). Also give me the output of a MuScan please after initppc.

And does 68k w3d work?

Yes, I was using Mu stuff to try and capture stuff.....and Sashimi of course. Here's a clean version without MUFORCE/GA:

Task name: 'gearsppc_PPC' Task address: 7EDA8500
Exception: Data Storage

SRR0: 70874D00 SRR1: 0000F030 MSR: 00003030 HID0: 0000C0A4
PVR: 70000203 DAR: 00000000 DSISR: 42000000 SDR1: 0EF0000F
DEC: 00079DE3 TBU: 00000000 TBL: 071E51D0 XER: 00000000
CR: 48000828 FPSCR: 82024000 LR: 70874CBC CTR: 70876C18

R0-R3: 03C50018 7EE71434 708BC600 708C06E0 IBAT0: 0000003E 00000002
R4-R7: 70874C90 00000000 00000000 708C0A40 IBAT1: 600003FF C0000002
R8-R11: 9FC3B000 9FC3B024 00000000 00100000 IBAT2: 7020003F 00200002
R12-R15: 7055E8A0 708C06E0 7056B588 70572AE8 IBAT3: 00000000 00000000
R16-R19: 705FFDE0 708BE560 705FFDE0 00000000 DBAT0: 0000003E 00000002
R20-R23: 00000000 00000000 705FFDE0 00000000 DBAT1: 600003FF C0000042
R24-R27: 00000000 00000000 00000000 708C06E0 DBAT2: 7020003F 00200022
R28-R31: 00000001 00000000 708BCAB8 7EE71434 DBAT3: 00000000 00000000


68k Warp3D works, following output is with Phase5 CSPPC board but crashes are same with WarpEngine (060).

MuScan 46.1 (02.07.2016) ? THOR

68060 MMU detected.
MMU page size is 0x1000 bytes.

Memory map:
0x00000000 - 0x00000FFF CopyBack Remapped to 0x08238000
0x00001000 - 0x00003FFF CacheInhibit Bundled to 0x080C6000
0x00004000 - 0x001FFFFF CacheInhibit Imprecise
0x00200000 - 0x00BBFFFF CacheInhibit Bundled to 0x080C6000
0x00BC0000 - 0x00BFFFFF CacheInhibit I/O space
0x00C00000 - 0x00D7FFFF CacheInhibit Bundled to 0x080C6000
0x00D80000 - 0x00DFFFFF CacheInhibit I/O space
0x00E00000 - 0x00E8FFFF CacheInhibit Bundled to 0x080C6000
0x00E90000 - 0x00EAFFFF CacheInhibit I/O space
0x00EB0000 - 0x00EFFFFF CacheInhibit Bundled to 0x080C6000
0x00F00000 - 0x00F00FFF CacheInhibit
0x00F01000 - 0x00F21FFF CacheInhibit I/O space
0x00F22000 - 0x00F4FFFF CacheInhibit
0x00F50000 - 0x00F5FFFF CacheInhibit Bundled to 0x080C6000
0x00F60000 - 0x00F7FFFF CacheInhibit
0x00F80000 - 0x00FFFFFF CopyBack
0x01000000 - 0x06FFFFFF CacheInhibit Bundled to 0x080C6000
0x07000000 - 0x0801EFFF CopyBack
0x0801F000 - 0x08021FFF CacheInhibit
0x08022000 - 0x080C4FFF CopyBack
0x080C5000 - 0x08108FFF CacheInhibit Imprecise
0x08109000 - 0x08118FFF CopyBack
0x08119000 - 0x08120FFF CacheInhibit Imprecise
0x08121000 - 0x08123FFF CopyBack
0x08124000 - 0x0822FFFF CacheInhibit Imprecise
0x08230000 - 0x0FEFFFFF CopyBack
0x0FF00000 - 0x0FF7FFFF CacheInhibit
0x0FF80000 - 0x3FFFFFFF CacheInhibit Bundled to 0x080C6000
0x40000000 - 0x4000FFFF CacheInhibit I/O space
0x40010000 - 0x43FFFFFF CacheInhibit Bundled to 0x080C6000
0x44000000 - 0x44FFFFFF CacheInhibit I/O space
0x45000000 - 0x5FFFFFFF CacheInhibit Bundled to 0x080C6000
0x60000000 - 0x60FFFFFF CacheInhibit Imprecise I/O space
0x61000000 - 0x7FFFFFFF CacheInhibit I/O space
0x80000000 - 0xFFEFFFFF CacheInhibit Bundled to 0x080C6000
0xFFF00000 - 0xFFF7FFFF CacheInhibit
0xFFF80000 - 0xFFFF7FFF CacheInhibit Bundled to 0x080C6000
0xFFFF8000 - 0xFFFFFFFF CopyBack Remapped to 0x08230000