This should be the sort of red flag to take note of. There’s an LLVM fork for every esoteric architecture now and this sort of thinking will lead to never being able to run your own software on your own hardware again. A reversion to the dark ages of computing.
Linux magically solves this problem how? GPL isn't magic. It doesn't compel contributing upstream. And half of modern driver stacks live in userspace anyways.
There are also so many G.P.L. violations and nothing is done about it.
I think a big issue is also that it's hard to show actual damages with this kind of copyright violation. It's obviously copyright violation but what damages are there really? Also, there are so many dubious cases where it's not clear whether it is a violation or not.
Software Freedom Conservancy have been doing GPL compliance actions for a long time, especially if you consider their staff's previous lawsuit that resulted in OpenWRT existing. Also the more recent Vizio lawsuit is kinda interesting, it aims to enable any recipient of GPLed binaries to sue for GPL compliance.
> And half of modern driver stacks live in userspace anyways
??? I haven't touched hardware whose driver lives in userspace since 2017 and it was a DMX512 controller of a shitty brand
Infineon tricore compiler from hightec. Compilers are actually, IMO, one of the things that are the most easy to have GPL because you can use it internally however you want without releasing the source outside. You could build whatever you want and you don't have to ship it on the final HW. A kernel does not afford you such a thing, you MUST ship it with your product.
Thanks for the example! Your opinion here aligns with mine: GCC's GPL status has manifestly not been an issue for vendors in the past. I think the reason for vendors selecting LLVM has much more to do with the fact that LLVM is easier to develop on than GCC.
Seriously: stop it. It's none of your business what the author's license choice is. You don't know what the author is trying to accomplish by their choice of license. It could be a mindless choice, or it could be an informed choice. Perhaps the author wants to build interest and later plans to switch licenses (it's not like others are likely to fork _and_ do an excellent job of evolving and maintaining the fork). Perhaps the author is looking to get hired. Perhaps the author believes that BSD/MIT licensing is more free than the GPL. You really don't need to be shaming the author for not making the choice you made.
DSP is simply a compute architecture that focuses on mutliply and accumulate operations on particular numerical formats, often either fixed point q15/q31 type values or floats f16/f32.
The basic operation that a NN needs accelerating is... go figure multiply and accumulate with the added activation function.
A DSP contains analog to digital and digital to analog converters plus DMA for fast transfers to main memory and fixed function blocks for finite impulse response and infinite pulse response filters.
The fact that they also support vector operations or matrix multiplication is kind of irrelevant and not a defining characteristic of DSPs. If you want to go that far, then everything is a DSP, because all signals are analog.
Maybe also note that Qualcomm has renamed their Hexagon DSP to Hexagon NN. Likely the change was adding activation functions but otherwise its a VLIW architecture with accelerated MAC operations, aka a DSP architecture.
ASICs bake one algorithm into the chip. DSPs are programmable, like GPUs or CPUs. The thing that historically set them apart were MAC/FMA and zero overhead loops. Then there are all the nice to haves, like built in tables of FFT twiddle factors, helpers for 1D convolution, vector instructions, fixed point arithmetic, etc.
What makes a DSP different from a GPU is the algorithms typically do not scale nicely to large matrices and vectors. For example, recursive filters. They are also usually much cheaper and lower power, and the reason they lost popularity was because Arm MCUs got good enough and economy of scale kicked in.
I've written code for DSPs both in college and professionally. It's much like writing code for CPUs or MCUs (it's all C or C++ at the end of the day). But it's very different from writing compute shaders or designing an ASIC.
I got a 265kf and motherboard for 350. Plenty fast and saves money for the real issue which is GPU costs. Thankfully B580 is actually a pretty good deal as well at 250 compared to green or red options. Team blue has some good deals out there really if you aren't tied to a team color.
They are majority Taiwanese employees for now, but I’m sure they’re hiring Americans to grow and backfill. They just wanted to bootstrap the knowledge and wisdom, but over enough time that can be shared and spread among Americans as well
reply