RLSL Allows Running A Subset Of Rust On Vulkan/SPIR-V Enabled GPUs

[phoronix]

Phoronix: RLSL Allows Running A Subset Of Rust On Vulkan/SPIR-V Enabled GPUs

There was a recent Khronos meet-up in Munich where Maik Klein of Embark Studios talked about their work on bringing a sub-set of the Rust programming language to Vulkan (SPIR-V) enabled GPUs...

http://www.phoronix.com/scan.php?page=news_item&px=RLSL-Rust-On-SPIR-V-GPUs

[jacob]

Generally speaking I feel that one of Rust's greatest weaknesses in 2019 is the limited range of target architectures and platforms it supports, especially compared to C/C++. Any progress on that front is very positive news.

[grigi]

Looked at the slides, and something like this, if it takes off well, would practically eliminate the bad shaders that are buggy and only works well on a single vendor, or sometimes a single generation from a single vendor.

Yay for progress!

[wizard69]

Generally speaking I feel that one of Rust's greatest weaknesses in 2019 is the limited range of target architectures and platforms it supports, especially compared to C/C++. Any progress on that front is very positive news.

I suspect that this has a lot to do with Rust not being widely accepted. The reality is standardization has done a lot for C++ to put it into the realm of “ professional languages”. It is a mind set that is hard to dissuade. I still run into IT professionals that think of Linux as a toy operating system not worth their time.

personally I find the Rust community to be a bit of a turn off too. The two things that the Rust world could do to address the support problem are to formally standardize the language, hopefully more rigorously than C++ and ditch the fanboy promotions.

[jacob]

I suspect that this has a lot to do with Rust not being widely accepted. The reality is standardization has done a lot for C++ to put it into the realm of “ professional languages”. It is a mind set that is hard to dissuade. I still run into IT professionals that think of Linux as a toy operating system not worth their time.

personally I find the Rust community to be a bit of a turn off too. The two things that the Rust world could do to address the support problem are to formally standardize the language, hopefully more rigorously than C++ and ditch the fanboy promotions.

I don't think Rust is struggling with acceptance, to the contrary I would say for such a relatively recent and somewhat unusual language its adoption has been stellar. The community is also doing what it can to help overcome some of its hurdles, such as the working group that has been set up to help Rust integrate better with existing build systems. Fanboys definitely exist, but they are not specific to Rust: Go has them, Python has them and C++ absolutely has them big time. They are also not really a nuisance, nothing stops anyone from simply ignoring them. The definition-by-implementation is probably the most major downside, some efforts are being made (including somewhat vague plans to develop at least one alternative non-toy compiler based on GCC) but it will certainly take at least a few years before there is some tangible progress.

As far as support for various targets goes, the problem IMO is technical. The LLVM back-end can theoretically be set up to generate code for a much wider spectrum of platforms, but the rigorous testing and validation it would require to officially support them would take resources and manpower the Rust community simply doesn't have (yet?). It's also a reality, whether we like it or not, that some architectures matter and some simply don't. Between Tier1-Tier3 support, Rust now works on Windows, Linux and MacOS on AMD64, Aarch64, MIPS and POWER (in various combinations of thereof) and quite frankly, that covers what 99% of the RELEVANT users (e.g. those who actually pay for the OS and/or the compiler, or otherwise contribute to it) are using. The rest may simply not be worth the effort. But the fact remains that for many potential targets especially in the IoT and embedded areas (which are theoretically one of Rust's selling points), only C/C++ compilers and toolchains are usable in practice.

[jabl]

I don't think a formal (ISO, ECMA, or similar) standard for Rust would make sense at this point. It made sense for stuff like Fortran, C, or C++ where you had multiple independent (proprietary) compilers, and there was a need for a consensus how they would work in order to enable portable source code.

For Rust you so far have one canonical and open source implementation that everyone uses. Now, if there ever will be other serious efforts at independent Rust compilers, then a formal spec might make sense. But so far it doesn't seem a particularly useful way to spend manpower.

[mvniekerk86]

I had a look at the repo - not much activity since April. And the gitter channel is also silent since July.

Dead project?

[stompcrash]

What ever happened to Ruby?

[fuzz]

What ever happened to Ruby?

Huh? Do you mean something other than the Ruby language? If not, how is the Ruby language relevant here?

[stompcrash]

Huh? Do you mean something other than the Ruby language? If not, how is the Ruby language relevant here?

How about, "fad languages which start with the letter 'R'?"