Intel Graphics Compiler Can Now Be Built For RISC-V

[Quackdoc]

GPU compute might just become immensely important for intel, and could be what turns it around assuming it's not done dumb. Their hardware is already quite good for compute. Not excellent, but if you have a workload that works, it does work well. Expanding their compute capabilities should be a priority for them IMO.

If Intel becomes a RISC-V chip designer leader and bets on it for its future chips rather than x86, and do not interfere negatively with the ISA evolution (meaning no lobbying to add crappy extensions nobody wants except them), they may regain some of my completely lost trust.

Sadly some of the the risc-v designers who were with intel left and started their own new company so I don't think intel is planning on this any time soon. A shame for intel I'm sure, but seeing as how horsecreek didn't pan out of them, you can see why. Intel could still fab risc-v using their fabs assuming they don't just sell them off.

[brad0]

Excellent news, although most AMD cards (including all the RDNA1/2/3 ones) already work on RISC-V as of Linux 6.10 introducing in-kernel FPU support (which amdgpu needs).

Although generally I am an AMD user, more options doesn't hurt. More options also helps even more so with non-Linux OS's.

[ayumu]

Excellent news, although most AMD cards (including all the RDNA1/2/3 ones) already work on RISC-V as of Linux 6.10 introducing in-kernel FPU support (which amdgpu needs).

[brad0]

Again, how do you know that? Why would Intel put effort into making their compiler work on RISC-V unless they were going to have a product that combined that core with Intel graphics?

No common sense? Selling Arc GPU to customers?

The Intel graphics compiler builds on ARM. Intel added ARM support to SVT-AV1, does that mean they're making ARM CPUs? Or just common sense they want people to use their encoder.

There are enough customers that are desperate for GPUs with open source drivers and all of the related components and that are portable.

https://www.phoronix.com/news/Intel-Arc-Graphics-On-ARM

[strtj]

Intel isn't doing that though.

Again, how do you know that? Why would Intel put effort into making their compiler work on RISC-V unless they were going to have a product that combined that core with Intel graphics?

[brad0]

That's an awfully terse, blanket statement. How do you know that? It would seem from this article that Intel is indeed moving into RISC-V space; do you have evidence otherwise?

The article literally pertains to Arc GPUs running on other architectures and has nothing to do with Intel designing CPUs with another architecture. You know they sell a product to customers and want them to use it.

[brad0]

If Intel becomes a RISC-V chip designer leader and bets on it for its future chips rather than x86, and do not interfere negatively with the ISA evolution (meaning no lobbying to add crappy extensions nobody wants except them), they may regain some of my completely lost trust.

Intel isn't doing that though.

[strtj]

Intel is doing no such thing.

That's an awfully terse, blanket statement. How do you know that? It would seem from this article that Intel is indeed moving into RISC-V space; do you have evidence otherwise?

[brad0]

I find it fascinating that despite all of the negative press about Intel lately, and their having to cut workforce as well as (probably) manufacturing capability, that they appear to be branching out into a totally new architecture. I would think that restructuring/refocusing a company as large and focused as Intel would not involve getting into entirely new areas of technology. After all, look at what happened to their forays into entirely new architectures/tech: Itanium, I don't think I have to explain that one; Xscale, sold to Marvell; Larrabee, refocused into Knights Mill/Knights Landing/etc, then put on hiatus with a supposed forthcoming product that never appeared; and Arc, so far not a total failure but I can't imagine it's the success that Intel wanted it to be. Oh, and also Optane, supposed to be totally groundbreaking and it some ways it was, but my understanding is that it never turned a profit.

Intel is doing no such thing.

[strtj]

I find it fascinating that despite all of the negative press about Intel lately, and their having to cut workforce as well as (probably) manufacturing capability, that they appear to be branching out into a totally new architecture. I would think that restructuring/refocusing a company as large and focused as Intel would not involve getting into entirely new areas of technology. After all, look at what happened to their forays into entirely new architectures/tech: Itanium, I don't think I have to explain that one; Xscale, sold to Marvell; Larrabee, refocused into Knights Mill/Knights Landing/etc, then put on hiatus with a supposed forthcoming product that never appeared; and Arc, so far not a total failure but I can't imagine it's the success that Intel wanted it to be. Oh, and also Optane, supposed to be totally groundbreaking and it some ways it was, but my understanding is that it never turned a profit.