Mesa 22.3 Lands New "Rusticl" OpenCL 3.0 Implementation

[TNZfr]

Quite interesting ... but what about mesa 22.2 ?
No news since mid-august

[ArchLinux]

It was clear after ArchLinuxs second post that he's just a troll. Stop feeding him.

In my opinion, everyone deserves food.

[mdedetrich]

Without getting into it too much, I've tried it off and on, but Metals doesn't really like my corporate environment (primarily due to network issues) and because it's a language I'm actually not particularly fond of, it's not a language I really use for projects on my personal computers where it would actually work properly.

I don't think the Scala plugin for Intellij is actually bad though, but I have a lot of trouble with both the language itself and how the libraries written in it I have to interface with are designed. At least for Scala 2(.12) (I haven't had a chance to try 3 because it's not compatible with the libraries i need) the entire language is a bundle of duct tape, bubble gum, and chicken wire very poorly attached on top of the JVM (and due to my requirements of being stuck on Java 8 incompatible with equivalent Java functionality in various ways, Scala closures vs Java Lambdas for one), with a type inferencer that is too primitive, and there's also this tendency to include everything and the kitchen sink which wouldn't be so bad if it weren't for the fact that whereas in basically every other language I can make naive guesses based on what functions sound like the right one and that the intellisense supports in what documentation it shows... in Scala there tends to be several functions that do the same exact thing with completely different names, that do it subtly differently in ways that matter.

Thanks for the response, I am a full time Scala developer so its good to get feedback from outside of the Scala world for first impressions and what you say does have truth with varying degrees.

[piotrj3]

I really like that people experiment with Rust.

I know there might be some concerns about bindings etc. But at the same time potential benefits are great and according to what i saw in merge request comments it is having 1/3 rd of overhead that Clover does. Even if experiment would fail there are a lot of benefits to it and so far it looks promising.

[ArchLinux]

Thanks for the response, I am a full time Scala developer so its good to get feedback from outside of the Scala world for first impressions and what you say does have truth with varying degrees.

Kotlin is the word of the day. Kotlin takes an hour to learn. Scala takes many months.

[mdedetrich]

Kotlin is the word of the day. Kotlin takes an hour to learn. Scala takes many months.

Sure, but Kotlin is Java++ where as Scala is a language on its own. There are many things that Scala can do which either aren't possible in Kotlin (HKT's, there is a proposal but it didn't work out see https://github.com/Kotlin/KEEP/pull/87) or is difficult/not ergonomic. Scala takes longer to learn but there is more you can do with it, whether that tradeoff is worth it of course depends.

Actually with the trajectory that modern Java is going I wouldn't be surprised if it eats into Kotlin's demographic especially with Loom/pattern matching on the horizon. This is the problem with languages which are better versions of another language, they usually just eventually die because they don't fullfill a niche (coffeescript is a classic example here). The main reason that Kotlin became popular in the first place is because Android was stuck on an ancient JDK (1.6) which didn't even have lambda's and hence Kotlin was the only sane way to develop on that platform.

[ArchLinux]

Sure, but Kotlin is Java++ where as Scala is a language on its own.

Ooh, I like that sentence.

The main reason that Kotlin became popular in the first place is because Android was stuck on an ancient JDK (1.6) which didn't even have lambda's and hence Kotlin was the only sane way to develop on that platform.

I'm not sure about that? I thought Kotlin came only after OpenJDK, which allowed higher JDK with a higher API?

[mdedetrich]

I'm not sure about that? I thought Kotlin came only after OpenJDK, which allowed higher JDK with a higher API?

So if you looked at the usage statistics of Kotlin in its early days, its primary use was with Android. The Java community back then was fairly split, you either had the very conservative enterprise applications which wouldn't even considered Kotlin back then or you had the Java developers using Java 1.8 which aside from syntax was only marginally worse.

It is true that Kotlin came out after OpenJDK but remember that at that time Android developers were forced to use Java 1.6 (or ugly workarounds) because Dalvik (the JVM running on Android) only supported Java 1.6 and this was mainly due to the Oracle vs Google lawsuit that was happening at the time.

[ArchLinux]

It is true that Kotlin came out after OpenJDK but remember that at that time Android developers were forced to use Java 1.6 (or ugly workarounds) because Dalvik (the JVM running on Android) only supported Java 1.6 and this was mainly due to the Oracle vs Google lawsuit that was happening at the time.

Yeah, but IIRC their Apache Harmony 1.6 Java was replaced with OpenJDK during the lawsuit, which allowed for lambdas etc, long as the Android API was raised high enough (to not have to support the old 1.6 Java Androids). And, IIRC, only after that did they bring in Kotlin (Oct 2017 with a quick Google).

[Eirikr1848]

AMD doesn't care about consumer OpenCL support.
AMD is dropping Polaris (GCN4) OpenCL support.
Users have troubles with OpenCL on AMD GCN1 - GCN3.
Users have OpenCL support for Terascale 2 & 3 only via Clover, which is crap.

The good news is: Terakan should work for some of the Terascale graphics cards!