@Maijestro

Quote:

It clear make no sense play around with slower embedded PowerPC chips, most definitely need to look at something faster, it continue to make sense to make new hardware, then we need look at Power6/7/8 maybe (previous generation). Power9 is probably too expensive.

Yeh, we should not blame you for anything, Trever was warned about putting the PowerPC chip he did on A1222 so, that only his fault if any.

I have thought about what puting ARM chip on AmigaONE-SE/XE cpu module, can be pretty good upgrade, you also bypass the slow onboard memory, you can do it with a faster PowerPC chip as well, it does not make much sense economically, perhaps.

Anyhow until QEMU catches up on 3D and 32bit graphics, there is a argument for real hardware.

Even if ARM/QEMU becomes faster, I don’t think it should stop the pressure of Multicore support on PowerPC. It’s better to get that done, and tested, then after that porting to ARM should be done.
Stopping development of SMP, just to port to ARM, will slowdown development.

For AmigaOS5 the plan was to visualize everything, I guess its all happening just not in the way we expected, only the digital platforms are Apple, Microsoft; Linux and Android. not Amiga.

@Maijestro

So do you emulate misalignment on different instructions for different PowerPC cpu models? As seen people saying they are not getting exception under emulation. So clearly it’s not that exact.

It definitely wont make sese to check, such conditions under emulation, unless your looking for bugs, of course, you can have a option increase compatibility for this corner cases. to find this bugs in software. (But running 20% slower to find software bugs, won’t be something people want under normal conditions.)

Providing none exception on alignment bug, providing exception on no alignment bug.
under reporting or over reporting issues, perhaps wont be useful as well, under development.

In any case if QEMU is faster than a real Pegasus II and AmigaONE-XE/SE/Micro/Sam440/460, then people are going to stop using this old computers, they might update to newer X50x0 / X1222, or simply stick to emulation (in that case there is no issue, no point fixing the bugs). That is if people can afford a expensive Apple M1/M2 laptop, after all they are as expensive as PowerPC hardware, that people are complaining about being overpriced.

@Bugala

You need to boot from CDROM, or USB pen drive, and use media toolbox to reinstall SLB into RDB sector, there tools on PC to install CD ISO on pen drives.

https://www.makeuseof.com/tag/10-tools-make-bootable-usb-iso-file/

"Balena Etcher" is extremely popular.

CD ISO you can download from Hyperion, after you have logged in. if you have registered your copy.

Should be included with the machine, or you can transfer owner ship. Or buy a license from ACube systems.

(Not 100% sure if works, if UBOOT only accepts USB CDROM drives, you can also ask ACube-Systems)

@SinanSam460

Some desktop PowerPC is more tolerant to alignment issues.
AMCC40x0 CPU’s are embedded chips, cut down versions of desktop CPU’s,

Alignment exceptions are present on PA-Semi / X1000 cpu’s as well AltiVec, and reverse endianness instructions.

@white

I have a Apple M2, this way I ask.

@SinanSam460

Quote:

better to change

#ifdef __MORPHOS__

to

#if defined(__MORPHOS__) || defined(__AMIGAOS4__)

that why chages can be pushed back to master,
not that I think it will happen.

As MorphOS uses TinyGL some of changes might not be relevant.

Should be taken as grain of salt.

Perhaps it should be changed from __MORPHOS__ to __TINYGL__ ?

@ddni

I have unplugged all SATA cables, just trying to boot into CFE, not into AmigaOS for now.
I think KMEM is corrupt. As that last thing before [EXCP] that think stands for exception.

it all started when try replacing backup battery, and it did work, try unplugging and plug the memory back in, but forgot to unplug power cable.

@Maijestro

Even worst, no prebuilt binary, no description of what dependencies you need to install.
plus there is different repos with different patches.

Right now QEMU boots into a backscreen, so I guess did something wrong.

[HELO][DRAM]SDRAM: ECC off, Non-ECC DIMM used on channel 0.
[RELO][L1CF][GOLO][GOT ][ZBSS][INIT][MAIN][KMEM][EXCP]

Computer is stuck at EXCP, nothing more happens, any idea what this means?

I all started then tried to reset CFE.
(the Jumper close to backup battery)

I think broke my DDR2 dims, ordered some other dims, they sadly had wrong frequency and Pin layout, sadly their seems to be two different kinds of DD2 buffered and none buffered dims.

tried to boot from backup firmware but it did not work.

@walkero

You can’t roll back anything if you can’t boot into the system.
perhaps a safe boot option should be added to slb2 menu.
actually openfirme boot loader used on CFE is not even have the menu activated by default.

@joerg

There is 3 different versions of WPA, P96, CGX and graphic.Library.

from benchmark, it looked where similar scores, this why think perhaps it’s not doing byte swap, should be more deviation, between the different input / output format.

I think I need to do visual confirmation, or wite a validator.

One more thing, is WPA does not return success / failure, perhaps it possible to add without breaking software.

@jabirulo

"Trap type: Alignment exception"

Means that instruction stfs needs aligned memory.

(R9 is 5B7A754D in HEX, 1534752077 in decimal.
as number ends with 7, and 7 can't be divided by 4 or 8.)

Typical 16 bytes alignment on AlltiVec, doubles are 64bit / 8 bytes.

AllocVecTags() has option for alignment, just remember to free it with FreeVec.

If your replacing malloc() with AllocVecTags you also need to remember to replace free() with FreeVec()

It is odd that happen to end up with nonaligned memory,
I assume at least 2bytes alignment be normal,
perhaps a length is read wrong, if high byte, low byte is swapped, you get a bad length. As result you can get a bad offset.

@Reth

I advise against changing boot priority because it can be tricky to get in kickstart menu. If you ever need to reinstall the OS, boot from CDROM or to recover the system, you get stuck, if you are unable to enter the boot menu.

@smarkusg

And the textures should be 16bit under QEMU with current drivers, as that’s all WB can display,
put perhaps its not working in 32bit, because its using 16bit textures.

Stuffing 2 pixels into one-pixel won’t work so well. Of course, it can be entireness thing as well,
graphics drivers have different BE/LE modes.

I’m not sure WPA converts BE/LE or just scales from 32bit to 16bit or the other way. How is it displayed on workbench.

@MigthyMax

Is there any chance its virtual / private, local data is passed as arguments and end up all sort of places. If swapped out it be bad.

@kas1e

chars are defined as read only, can be trying to modify a string?
is there a example of this behavior? Where does it actually crash?

Can we dump what variable that was defined in the .rodata and what variable is defined writeable area. Did someone wrongfully define something as const?

can it be write protecting it before initializing classes, perhaps.

@Maijestro

Was there some kind of byte swap support in northbridge? Perhaps it does something wrong.

looks like you 4 bytes swaped with next 4 bytes.
(8bit GFX mode only way you get correct colors)

Perhaps its swapping 8bytes instead of swapping typical 4 bytes. Pixel 1,2,3,4,5,6,7,8 resvered as 8,7,6,5,4,3,2,1

whats expected is 1,2,3,4,1,2,3,4 revered as 4,3,2,1,4,3,2,1

If writes directly into bitmap buffer instead of using WritePixelArray, strange things can happen.

Odd things can happen with JIT, for example two 32bit instruction can be combined to from a 64bit instruction, maybe assumption is not always right. A reserved 64bit operation, vs 2 x revered 32bit operations won’t actually produce the same result.

@MartinW

congratulation

@Maijestro


A lot of energy when into finding best version of FFMEG to integrate into mplayer, it was during bunty, so was not 100% free, I also guess fact that mplayer is worked on by many people over years contributes to have well it work, a provided config file provided in program dir, sets lower quality to increase speed, as well.

Other things I found during development, was mplayer read small blocks sizes from disk, and drivers sucked (X1000), as drive spined up, only to spin down, so added some code to make it read larger chunks, but that only kicks in when playing from a DVD, as its lowlevel code.

I know also MPlayer as some other cache features often enabled for network playback, it can be that help, as can reduce loading CPU spikes, but not sure if does that. there are some multi treading in mplayer, it can help even out or reduce, lag.

That’s all I remember, everything else I worked on is irrelevant as your using SDL output, that part of the code I did not work on much.

The version of mplayer I wanted to work on was one lest amount of GUI, it’s not accidental. a lot good work already done, on it before I started on it. My main focus on the general code, was make it failsafe, bug free. (mostly focus was on startup and shutdown code).

@Maijestro

you have two soft reboot's.

cold reboot is CTRL+LALT+RALT.
warm reboot is CTRL+LWIN+RWIN (reboot faster, skips firmware)

Newest RX/HX drivers do not support warm reboot, why I'm using a older driver, 99% of the time.

I expect BIOS code need to be executed to reinitialize the GPU, but that does not happen in Warm Reboot.

you can also reboot from command line.