RISC-V Linux Port Pursuing Mainlining In The Kernel

[phoronix]

Phoronix: RISC-V Linux Port Pursuing Mainlining In The Kernel

RISC-V developers believe that while their Linux kernel port isn't yet fully complete, they are hoping to get it mainlined now for this open-source CPU ISA...

http://www.phoronix.com/scan.php?page=news_item&px=RISC-V-Linux-Kernel-Port

[uid313]

I would like to see a single-board computer like the Raspberry Pi but powered by RISC-V.

[wdb974]

I would like to see a single-board computer like the Raspberry Pi but powered by RISC-V.

What would be the advantages of such a board?

[boxie]

What would be the advantages of such a board?

not encumbered by patents, open design. will be interesting to see how well it performs

[GI_Jack]

What would be the advantages of such a board?

If I Recall Correctly, the RISC-V is supposed to be a family of CPUs in three flavors from big to small that use a standardized instruction set, with a whole universe of add on optional instructional sets which are standardized as well. Sort of how x86 wound up, but designed from the start. Extensions are not ad-hoc.

And they are all Patent unencumbered since day one. The aim is a single patent free family of CPUs which anyone can make one. In theory it will drive innovation and standardize chip design, and improve compatibility. Also, its Free as in speech

[blubbaer]

I would like to see a single-board computer like the Raspberry Pi but powered by RISC-V.

But the other components like GPU are not free. A better suggestion would be a SoC with RISC-V cores and an FPGA that is supported by the yosis synthesis tools. This way one could have a single-board computer that consists of only open components.

[pal666]

But the other components like GPU are not free

not every sbc needs gpu
feel free to design free gpu though

[microcode]

not every sbc needs gpu
feel free to design free gpu though

I feel like some work could really be done around the architecture of GPUs, at least when it comes to GPUs on embedded devices. I'm thinking that you could probably achieve most embedded graphics workloads with something much simpler (maybe add some specialized instructions to do graphics stuff on the CPU more efficiently). Maybe it could be like big.LITTLE, where the operating system can schedule a process on any of a set of different types of cores. If there were a set of cores devoted to performing graphics tasks (maybe lower frequency, but huge wide vectors, maybe a large number of cores) which your process would be automatically rescheduled onto when when the CPU traps an unsupported instruction, it would be interesting to see what kind of performance you could get out of that.

[profoundWHALE]

I'm waiting for lowRISC to come out with a board with a GPU, but that is still going to take quite some time.

While it is an ultimate goal to support all the features of a modern commercial SoC, it will require a number of iterations of the design to achieve this. For example, early versions of our SoC will not include a GPU.

OpenRISV also looks very promising

[Falcon1]

I would like to see a single-board computer like the Raspberry Pi but powered by RISC-V.

There is work going on in that direction.

http://lists.phcomp.co.uk/pipermail/arm-netbook/2017-April/013457.html