My progress on real RISC-V hardware

The RISC processors that perform best are the ones that embrace CISC characteristics. I still have a G4 Mac Mini running MorphOS and a Teron Mini derivitave running AmigaOS 4.1FE2 and I am content to leave PPC in the grave.

I have a Pinebook Pro and it is certainly weak but putting a smaller, faster operating system can make up for that performance lag in some instances. MorphOS runs nice and fast on my 1.5 GHz G4 system but even its new web browser doesn’t have a JIT compiler for its JavaScript engine. In most instances the cache characteristics of the operating system only make up for a little bit of the performance lags.

The number one predictor of performance on a cache enhanced CPU is code density. It improves the capabilities of the code cache. That’s why CISC, not RISC, must make a comeback. I miss the 680x0 more than PPC.

1 Like

CISC never went away, complex instruction sets in my opinion have been limiting CPU design and innovation for years. If you want to read up on a phenomenal piece of RISC hardware, look up the Playstation 3’s Cell CPU co-developed by Sony and IBM

3 Likes

I think we’ll just have to agree to disagree. RISC instruction sets are slower because of their bloat. I’m no fan of x86 though.

2 Likes

:+1:, no reason to argue, we’re all here because we love Haiku.

4 Likes

fixed width pipelines i suspect are the actual bottle neck. flexible width pipelines with assignable lanes are probably next gen. or mixed width pipelines, get rid of the concept of a core. dispatch unit selection of pipeline allocation from instruction stream, them allows dispatch to utilize all of the lanes like threads, i think CPUs already do something like this with cache and int units but iirc the pipelines are fixed width per core. would require significant os and compiler and language changes afaict

1 Like

Replied in CPU architecture thread.

1 Like

but its far too expensive for me.

will the situation improve with visionfive 2?

3 Likes

impressive specs on that board !!

1 Like

exactly … they start to aggressively throw themselves into the market becoming competitive, have you noticed the gpu? it’s from imagination, former ex gpu suppliers for Apple … I found this video, they had already announced that they are also working on a RISC-V cpu

2 Likes

Hi yes. I backed it. But it has not the power of the laptop of course. But nice still and nice to tinker with. :slight_smile:

Not surprising, since they were doing MIPS before and the MIPS foundation annoucned that they are abandoning the MIPS architecture to support RISC-V instead.

However the PowerVR GPUs are not exactly known for being friendly with open source, I guess this will have to change if we want to see good support in most operating systems (and not just binary blobs for Linux).

4 Likes

we will see, I am optimistic, this time the world is really changing, I am referring especially to the European and US chips act. Besides Imagination, closing the business with Apple, I don’t think it has reasons to create obstacles, now everyone is opening up with opensource drivers even nvidia has given up, they are more convenient for everyone, and if they want to speed up their popularity on these new RISC-V platforms, so much the better that they use opensource divers

If I recall right there was recently a release of a PowerVR graphics driver for older PowerVR graphics units?

Also isnt the M1 APU a PowerVR derivat which gets currently reverse engineered?

I have not deepened the research, but it seems to make sense with the landing on the risc-v architecture…

If I understand correctly this one is planned to have a fully open source vulkan and OGLES3.x driver (without blobs I think). There is some info here: IMG BXE4-32 GPU Open Source Plan - GPU related - RVspace Forum

Edit: Actually I think it will have firmware blobs but open source shader compiler (so we should be able to have a haiku driver in theory!) Open Source GPU Driver - Imagination Developer

1 Like

do you think they chose the same methodology as nvidia? to make closed source firmware and opensource drivers?

i like where riscV is heading, now to get us companies behind it

It seems likely, that is the approach recommended by the Free Software Foundation in their “Respects Your Freedom” recommendations for hardware: if the software is embedded in a device and not upgradeable, it is not software but firmware, and then it doesn’t need to be free.

Personally I think this is a bit stupid in terms of free software and open hardware, but hey, it means the “intelligence” moves inside the graphics card and the drivers get simpler so we can more easily port them to Haiku. So let’s at least enjoy that :slight_smile:

6 Likes

Yea, the approach looks like this: they obfuscate what really happens between hardware and firmware and provide you with connection tubes to do what you want with the drivers … at least there is good that you can build on it a coherent integration with the rest of the operating system … it’s not the best, but it’s definitely better than it was before.

1 Like

to be honest, win the complexity of graphics drivers and hardware, this is a win period. better to let them fix the firmware.