MuLab vs. Big Sur

[syntonica]

I found a willing sacrifice to take the new Big Sur install to see what all the hullabaloo is about. So far, it seems nice.

As a test, I installed MuLab 8 Free on it. It takes a bit of wrangling. If I remember the order on how, unzip it and move the folder to Applications. Run the Quarantine Remover. Quit and then right-click it and select Open. Now, drop in both the MuLab program and the Rex Shared Library. Now, you can run MuLab in all its glory. I don't recall if you have to do the Open thing the first time here or not, but it is possible to download and run a fresh install.

I only tested a couple of the demos and they worked without issue. Unfortunately, this is a wimpy 2015 MacBook Air 11", so one of the demos ate like 60% of the CPU, way more than I expected.

At least it runs.

My next MacBook, which I may or may not get next March-ish, will probably be a new M1-based MacBook Air. It depends on all the reviews and info that comes out in the meantime. I'm hoping Jo can just flip the compiler switch to include an ARM executable in the bundle and nothing more needs to be done.

[dakkra]

It's not "just a switch". XCode 12 has a "universal" binary compiler but some developers here on KVR have already ran into issues with x86 not working. Some of the big issues are SIMD extensions which are very different instruction sets on ARM (NEON) and x86 (SSE, AVX). We'll see how this M1/ARM thing goes... ARM has never been a good general purpose computing process, but maybe Apple can pull it off if they plan their integration right.

Good to know that Mulab works on Big Sur though I've not yet made the move yet but I'll probably do it soon.

[syntonica]

Meh. The ARM instruction set is fine, just a different way of thinking. I've only taken a cursory glance at Neon, however, to notice that Neon and SSE are two different animals. There's a lot of one-to-one correspondence and a lot of things that are in one but not the other. However, there are libraries available with macros that will translate out to AVX/Neon, depending on the target. Anybody that also develops for iOS probably has a huge start in this regard.

In my experience so far, the clang compiler catches the majority of loops that it can auto-vectorize, as long as you code them properly. This is probably the only case where I recommend coding to bait the compiler rather than just coding clean and letting the optimizer do its thing.

Big Sur seems to be the High Sierra to Catalina's Sierra. There's a GUI redesign and some boring new tracking features, but otherwise, everything else seems to be under the hood. I'm still playing around with it.

On a side note, while testing out other non-musical softwares, I found that in the System Preferences, under Security, the option to whitelist a program is still there as long as you go there immediately after trying to run it. So, third-party developers who don't want to pay their obeisance to Apple are not yet locked out. Yet.