[libre-riscv-dev] Vulkanizing

[Atif Zafar]

Thank you Luke. Yes, we got massive support for a unified processor architecture that combines a CPU/GPU/VPU into one device. The idea is that in a separate CPU-GPU combination you spend an inordinate amount of effort shuttling and copying data from one memory bank to another and then have to adjudicate and make updates according to what has been processed and what has not. So by eliminating the separate caches and a message passing architecture we significantly simplify the design.

Then the entire power of the GPU becomes accessible to the CPU. SO for example in a Visual Studio session I should be able to write assembly code for the CPU and GPU in the same piece of code and intermix as needed.

This make it extremely powerful to "invent" new graphics algorithms when you are not limited to pushing triangles. We can do image based rendering, or use quadrilateral primitives or ray tracing all in the same code!

Cheers
Atif
________________________________
From: Luke Kenneth Casson Leighton <lkcl at lkcl.net>
Sent: Wednesday, February 19, 2020 1:30 AM
To: Libre-RISCV General Development <libre-riscv-dev at lists.libre-riscv.org>
Subject: Re: [libre-riscv-dev] Vulkanizing



On Wednesday, February 19, 2020, Immanuel, Yehowshua U <yimmanuel3 at gatech.edu<mailto:yimmanuel3 at gatech.edu>> wrote:
OK. I like this.

I think the investors should be fine with this too.

do let them know:

the stunning level of simplicity was what got massive support from leading experts who have worked for intel, AMD and NVIDIA for decades, at the RISCV Bay Area meetup in Oct 2019.

presented by Atif from Pixilica, the encouragement and feedback he got was "this is exactly the kind of shakeup that the 3D industry needs. please make sure to include VPU cspability as well".

by going for a direct software route, if a developer wants to drop vulkan entirely and do some conpletely innovative algorithms (raytracing, parametric) they have a FULL unix capable processor, one with advanced accelerated hardware opcodes builtin to the *CPU* in order to do so.

consequently it is a strong base for innovation and experimentation that happens to get high performance as well.



Also,

https://libre-riscv.org/3d_gpu/tutorial/

That pages is pretty awesome!!

Who wrote it?

err, me, because Cole needed something.

 i wanted to keep it funny.

l.



--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68