Pistorm32 showcase thread

[Mr-Z]

Originally Posted by Kin Hell

Isn't it just & really....

With a CM4 Module @ 2.3Ghz & running almost 2000 times fater than a Stock A1200, do you really need more?

Yes

[Kin Hell]

Originally Posted by Kin Hell

<snip>
With a CM4 Module @ 2.3Ghz & running almost 2000 times fater than a Stock A1200, do you really need more?

Originally Posted by Mr-Z

Yes

....of course of course!

More P O W E R !

[Karlos]

What we need is a way to expose the SIMD capabilities for things like datatpes, codecs and the like. Forget 2000 times faster....

[tomcat666]

Originally Posted by Karlos

What we need is a way to expose the SIMD capabilities for things like datatpes, codecs and the like. Forget 2000 times faster....

I think everything is already exposed... i.e. the actual RTG P96 display driver, the SD card driver, ... are already written completely in 68k code and use pi's hardware resources. Michal can explain it better

[VladR]

Do I understand this correctly that under Emu68k's RTG there is no CPU-vs-DMA interleaving issue (as there is no Amiga chipset being emulated) ?

Meaning the 68040 code runs at full possible speed of the single core of Cortex A-72 ?

Because that would make it an RTG monster...

Are all typical color depths (8,16,24,32) and RTG resolutions working there under P96 (800x600,1024x768,1280x1024,1600x1200) ?

[Dunny]

Originally Posted by VladR

Do I understand this correctly that under Emu68k's RTG there is no CPU-vs-DMA interleaving issue (as there is no Amiga chipset being emulated) ?

Correct, as far as I understand it. There is some issue with Chip RAM read speeds - it's too fast so has to be slowed down a touch in certain situations.

Meaning the 68040 code runs at full possible speed of the single core of Cortex A-72 ?

Correct. It's not a 68040 though, it just identifies as one.

Because that would make it an RTG monster...

Are all typical color depths (8,16,24,32) and RTG resolutions working there under P96 (800x600,1024x768,1280x1024,1600x1200) ?

Yes it is an RTG monster. I run my WB at 1920x1080x32 for example.

[VladR]

Originally Posted by Dunny

Correct, as far as I understand it. There is some issue with Chip RAM read speeds - it's too fast so has to be slowed down a touch in certain situations.

I reckon that only applies to legacy SW that wasn't written with 2500 MIPS in mind, right ?
ChipRam latency/bandwidth has certainly been a consideration (e.g. framebuffer/bitplane writes) on the poor old legacy HW, but goes out the window with Cortex A-72.

Originally Posted by Dunny

Correct. It's not a 68040 though, it just identifies as one.

Well, I've read recently that they just fixed some FPU-related bugs, so it leads me to believe it implements FPU instructions (maybe just a most-used subset, perhaps?) from 68040, right ?

Since a lot of older demos using FPU run fine already (especially after recent paraj's work), then Emu68k must be able to interleave Integer and FPU ops in parallel (just like regular 68040), correct ?

Do you perhaps mean that some parts of 040 instructions remain currently unimplemented (say, MMU, and some rarely-used ops, etc.) ?

[Dunny]

Originally Posted by VladR

Do you perhaps mean that some parts of 040 instructions remain currently unimplemented (say, MMU, and some rarely-used ops, etc.) ?

I'm no expert but I believe that's so.

[tomcat666]

Another milestone reached by Paraj... this time TBL - Eon demo perfectly working on pistorm32 ... 2500 MIPS 2.2Ghz amiga He even went so far to completely fix the cube part which was only ever meant to work on non-AGA amigas.

[ Show youtube player ]

[Angus]

Very cool.

[mahen]

Wow !