Getting Focused was Fwd: [lowrisc-dev] Open GPU for the first CPU
lowrisc-dev at aggregator.eu
Tue Feb 17 10:32:27 GMT 2015
Dne 17.2.2015 09:54, Alex Bradbury napsal(a):
> On 17 February 2015 at 07:17, Logan Streondj <streondj at gmail.com> wrote:
>> For an initial release, it would be best to use what is ready.
>> There is an ASIC ready OpenRISC core, and ethernet,
>> only thing that needs to be ported is a UART to ASIC.
>> Then can package that, crowd-fund it, sell it as a web server.
> We went through a selection process when choosing our ISA and
> ultimately RISC-V won out. RISC-V is behind OpenRISC in some aspects
> (system software) but rapidly catching up, and well ahead in others
> (such as regular tapeouts of high performance implementations). We
> want to provide an SoC platform for others to build on and modify, and
> starting out with a design that we intend to obsolete and abandon is
> the opposite of being focused.
>> The team is here for the moment, but if there aren't results and
>> revenue soon, the team will run out of steam.
> We're very aware of the challenges and the timelines involved in
> producing silicon. e.g. we have a 128-core research test chip on a
> 40nm process taping out this summer. I'm not sure why you think we're
> anywhere near running out of steam, on the contrary we're just getting
Cool - what interfaces i.e. buses will it have for connecting external
peripherals? The GPGPU thing might not be viable to do at the moment,
but over time ... the development of it may be done (physically) by
connecting some FPGA to a bus from lowrisc chip.
I've searched some free or opensource GPU projects and some failed
miserably, because of financing. This open model for developing GPU,
i.e. do some design, then try to sell it before it's ready to some
audience doesn't work. The funding can be there (albeit small) but not
from commercial sector, which is what Logan suggested. The other idea -
take some existing design (and port it - because the OpenRisc VGA is in
verilog AFAIK) might be less frustrating and more succesfull in the
Some relevant links
Only the OrSOC seems to be somewhat usable...
One other possibility for the base of development of a GPU would be
optimizing Mesa library to run in manymany threads, that could be
spawned amongst those many cores of a lowrisc chip and render in
software to some framebuffer. Will those lowrisc cores have all RV64G
instruction set? And Hwacha accelerators?
More information about the lowrisc-dev