New Apple Silicon

[Gavin Hird]

On 10/27/2021 at 9:59 AM, Jenna Huntsman said:

LL don't have the manpower (or, lets face it: money) to rewrite the viewer for approximately 5% of SL users to have slightly better performance, while cutting everyone else out of the loop.

The motivation for re-writing for Metal is not the 5% (actually 15% according to LL) Mac users, but to address the very large i-device market.

Remember, the M1 processor is essentially the same as the A14 that sits in the latest iPhones and iPads - devices which are perfectly capable of running a viewer both when it comes to processor, memory and screen. Obviously the UI must be updated for touch or at least controller. 

[Jenna Huntsman]

6 hours ago, Gavin Hird said:

The motivation for re-writing for Metal is not the 5% (actually 15% according to LL) Mac users, but to address the very large i-device market.

Remember, the M1 processor is essentially the same as the A14 that sits in the latest iPhones and iPads - devices which are perfectly capable of running a viewer both when it comes to processor, memory and screen. Obviously the UI must be updated for touch or at least controller. 

LL's handling of mobile is separate from the viewer development team, moreover as far as I know LL don't have any plans to overhaul the UI anytime soon.

LL were developing a chat client for mobile up until recently, but they scrapped it to "go in a new direction" (whatever that may mean, although I wouldn't expect it to be a whole viewer port to mobile; or if it is, it'd likely leverage OpenGL / OpenGL ES which works on both Android and iOS (and, conveniently wouldn't require an engine rewrite)).

[Aishagain]

7 hours ago, Gavin Hird said:

The motivation for re-writing for Metal is not the 5% (actually 15% according to LL) Mac users, but to address the very large i-device market.

Remember, the M1 processor is essentially the same as the A14 that sits in the latest iPhones and iPads - devices which are perfectly capable of running a viewer

So we are to rewrite our access to SL to facilitate 15% of the user-base.  Also a company who regularly instructs its users on what they may or may not do?

That might be a little harsh but it speaks to the essential issue:  just who is to dictate the future direction of virtual  worlds.

No thankyou.

[Lyssa Greymoon]

28 minutes ago, Aishagain said:

So we are to rewrite our access to SL to facilitate 15% of the user-base. 

OpenGL is a dead end, and SL is hamstrung by its reliance on obsolete technology. This isn't just a Mac thing, it's everyone using a reasonably modern computer.

[anitabush]

52 minutes ago, Aishagain said:

So we are to rewrite our access to SL to facilitate 15% of the user-base.  Also a company who regularly instructs its users on what they may or may not do?

That might be a little harsh but it speaks to the essential issue:  just who is to dictate the future direction of virtual  worlds.

No thankyou.

You seem to have avoided the second part of what gavin wrote there (accidentally I’m sure). Even ignoring the possibility of market share for macs increasing if there were a decent viewer, the iOS market is quite large to say the least.

[Jenna Huntsman]

18 minutes ago, Lyssa Greymoon said:

OpenGL is a dead end, and SL is hamstrung by its reliance on obsolete technology. This isn't just a Mac thing, it's everyone using a reasonably modern computer.

OpenGL is still alive - SL performs badly on modern hardware because it just isn't optimised for it. The main issue is that the viewer is for the most part single threaded, sometimes in places where it could be multithreaded.

That's not to say we shouldn't port the viewer to a more modern & efficient API (Vulkan), as this would facilitate the viewer's renderer to be multithreaded (And would be able to run on Apple devices through MoltenVK, which is very close to Vulkan 1.1 conformance and likely would be conformant before a Vulkan viewer is introduced)

13 minutes ago, anitabush said:

You seem to have avoided the second part of what gavin wrote there (accidentally I’m sure). Even ignoring the possibility of market share for macs increasing if there were a decent viewer, the iOS market is quite large to say the least.

Actually, the US market share for Macs has been stagnant since 2015. It's been hovering at around 13% of the market since then.

iOS has seen growth in the US market, however only has around 16% of the worldwide market (compared to Android at 40 - 42%)

Stats from: https://gs.statcounter.com/os-market-share

[Gavin Hird]

4 hours ago, Jenna Huntsman said:

LL's handling of mobile is separate from the viewer development team, moreover as far as I know LL don't have any plans to overhaul the UI anytime soon.

LL were developing a chat client for mobile up until recently, but they scrapped it to "go in a new direction" (whatever that may mean, although I wouldn't expect it to be a whole viewer port to mobile; or if it is, it'd likely leverage OpenGL / OpenGL ES which works on both Android and iOS (and, conveniently wouldn't require an engine rewrite)).

The Xcode tools makes it possible to write the same application that runs both on macOS, iPadOS and iOS with relatively small variations in the UI code. 

Your opinion of this is significantly outdated. 

[Gavin Hird]

3 hours ago, Aishagain said:

So we are to rewrite our access to SL to facilitate 15% of the user-base.  Also a company who regularly instructs its users on what they may or may not do?

That might be a little harsh but it speaks to the essential issue:  just who is to dictate the future direction of virtual  worlds.

No thankyou.

No, you would widen your potential user base by about 1 billion users. If you get 1 million of those being active, you have doubled the current user community. 

[Gavin Hird]

2 hours ago, Lyssa Greymoon said:

OpenGL is a dead end, and SL is hamstrung by its reliance on obsolete technology. This isn't just a Mac thing, it's everyone using a reasonably modern computer.

Exactly. 

So you have to pick your target from what is your profit (or other) motive. Mac, iPad and iOS users are usually both willing and able to pay. At the other end of the scale is the whiny Linux base who don't want to pay for anything but the graphics card and CPU, but otherwise don't want to pay for anything in-world. 

Allegedly 25% of the creators in SL are Mac users. 

[Aishagain]

I'm not going to harp on about this ...much...but as long as I've been reading these fora, and as long as I've been using computer hardware (since about 1985), I have been forcibly reminded time and again of just how proscriptive and monolithically secretive the Apple Corporations is.  Now Microsoft is almost as bad, but the open source community has always, it seemed to me, been at odds with Apple;  their "Walled Garden" approach is legendary.  Only now are they apparently relenting on even allowing their kit to be repaired!

So there are billions of potential users out there?  I will wait and see.  I am not holding my breath.

Edited November 18, 2021 by Aishagain
spelling

[Nick0678]

2 minutes ago, Aishagain said:

So there are billions of potential users out there?  I will wait and see.  I am not holding my breath.

Ahhh.. yes.. well LL 's plans for Macs have been the same since 2003 and i am sure won't change..  (=no fks given)

[Gavin Hird]

6 minutes ago, Nick0678 said:

Ahhh.. yes.. well LL 's plans for Macs have been the same since 2003 and i am sure won't change..  (=no fks given)

By my estimation, the minute Apple pull their last Intel machines from marketing (and at the same time zaps OpenGL). LL will announce no support for the Mac viewer. This should happen by this time next year latest.

So, yeah...

[Jenna Huntsman]

15 minutes ago, Gavin Hird said:

The Xcode tools makes it possible to write the same application that runs both on macOS, iPadOS and iOS with relatively small variations in the UI code. 

Your opinion of this is significantly outdated. 

Developing the viewer to use Xcode would require a ground up rewrite, and, once again, does not work on Windows (where the majority of the existing userbase are), nor does it work on Linux (where a lot of the 3rd party SL developer community is).

By that logic, you'd be better off going with GTK (which works on Windows, MacOS and Linux) or QT (which works on Windows, Mac and Linux alongside iOS and Android)

12 minutes ago, Gavin Hird said:

At the other end of the scale is the whiny Linux base who don't want to pay for anything but the graphics card and CPU, but otherwise don't want to pay for anything in-world. 

I mean, this is just outright conjecture.

[Gavin Hird]

6 minutes ago, Jenna Huntsman said:

Developing the viewer to use Xcode would require a ground up rewrite, and, once again, does not work on Windows (where the majority of the existing userbase are), nor does it work on Linux (where a lot of the 3rd party SL developer community is).

By that logic, you'd be better off going with GTK (which works on Windows, MacOS and Linux) or QT (which works on Windows, Mac and Linux alongside iOS and Android)

I mean, this is just outright conjecture.

 

The development of the Mac viewer is to a large extent done in Xcode, and every version of the viewer is compiled from an Xcode project that is generated by cmake via LL's autobuild py scripts.  You obviously have no clue what you are talking about.

Just look to opensim and see what happens when you leave it all to open source and the free ***** Linux user community unwilling to pay for anything (throw in a bunch of Windows users in that bunch too). 

Edited November 18, 2021 by Gavin Hird

[Owen Blaylock]

8 minutes ago, Gavin Hird said:

By my estimation, the minute Apple pull their last Intel machines from marketing (and at the same time zaps OpenGL). LL will announce no support for the Mac viewer. This should happen by this time next year latest.

So, yeah...

Apple continues to include its unsupported OpenGL implementation in MacOS. Apple stopped updating OpenGL five years ago, but they didn’t rip OpenGL out because they know there are apps (professional programs and bespoke software) which depend on OpenGL and OpenCL. Just don’t expect massive performance improvements from improved CPUs and GPUs because Apple’s OpenGL isn’t being optimized for new hardware.

SL runs ok on M1 Macs. Not great, but ok. The OpenGL APIs made the leap to M1, which suggests to me that this situation will persist in MacOS for some time to come. If they wanted OpenGL gone, the transition to ARM would have been the perfect time to say, “Nope, we can no longer support that.”

That said, I don’t see why LL isn’t investing its energies in Vulkan. It’s already cross-platform like OpenGL, and, on the Mac, it is a way to leverage Metal. Instead, they spent how many years chasing a VR pipe dream?

[Nick0678]

For me they should have clients for everyone (especially Android since those dominate the mobile market), of course all that equals to extra cost = $$$$ and since Linden Lab is a small company...   it is what it is...

[Jenna Huntsman]

7 minutes ago, Gavin Hird said:

 

The development of the Mac viewer is to a large extent done in Xcode, and every version of the viewer is compiled from an Xcode project that is generated by cmake via LL's autobuild py scripts.  You obviously have no clue what you are talking about.

Just look to opensim and see what happens when you leave it all to open source and the free ***** Linux user community unwilling to pay for anything (throw in a bunch of Windows users in that bunch too). 

Okay, I did not know that the viewer on Mac was already based on Xcode. I'm sorry and I stand corrected.

Anyway, please do not insult myself or other members (or communities for that matter).

I'd like to point out however that [insert TPV name here] wouldn't exist if the viewer itself wasn't open source. Heck, SL would probably have died if the TPV community abandoned SL when LL themselves were too busy with Sansar.

Also, let's not forget that the software that most people use to create meshes for SL (Blender) is open-source. Not to mention other software for creating textures (GIMP) is also open-source.

OpenSimulator is a project which aims at making both sides of SL open-source. (LL only officially open-sourced the viewer, not the server). That's a neat idea. I don't use it personally. 

 

Anyway, to go around to the main point of the topic:

Could SL be made to run natively on M1 (and derivatives)?

Probably. Some viewers can be compiled for ARM (instead of x86, CoolVL is the example i'm thinking of specifically), but this has only been done on Linux. I don't know if anyone has tried doing so for Mac, but the CoolVL dev (Henri Beauchamp) isn't going to be the one to do it (as he does not have any Apple hardware, nor has an interest in getting any). But, the viewer is open-source so someone could make the required patches and upstream them.

It'd still be running on OpenGL, but it would eliminate Rosetta 2 (which would qualify it as running natively).

[Nick0678]

25 minutes ago, Gavin Hird said:

Just look to opensim and see what happens when you leave it all to open source and the free ***** Linux user community unwilling to pay for anything

What happens to Opensim? As far as i know although i haven't been there for a couple of years it runs well.. Is it buggy now?

[Gavin Hird]

19 minutes ago, Owen Blaylock said:

Apple continues to include its unsupported OpenGL implementation in MacOS. Apple stopped updating OpenGL five years ago, but they didn’t rip OpenGL out because they know there are apps (professional programs and bespoke software) which depend on OpenGL and OpenCL. Just don’t expect massive performance improvements from improved CPUs and GPUs because Apple’s OpenGL isn’t being optimized for new hardware.

SL runs ok on M1 Macs. Not great, but ok. The OpenGL APIs made the leap to M1, which suggests to me that this situation will persist in MacOS for some time to come. If they wanted OpenGL gone, the transition to ARM would have been the perfect time to say, “Nope, we can no longer support that.”

That said, I don’t see why LL isn’t investing its energies in Vulkan. It’s already cross-platform like OpenGL, and, on the Mac, it is a way to leverage Metal. Instead, they spent how many years chasing a VR pipe dream?

It will not continue after Apple only ships Apple Silicon Macs. The current OpenGL implementation runs in Rosetta2 emulation, and the only thing they have done is to write a HAL to the (new) AS GPUs in the same manner as they previously did on the A-processors for the OpenGL ES when it was supported on the i-devices.  

There is no native OpenGL implementation at all on Apple Silicon, and there never will be. 

[Gavin Hird]

10 minutes ago, Nick0678 said:

What happens to Opensim? As far as i know although i haven't been there for a couple of years it runs well.. Is it buggy now?

Actually the current 0.9.2 dev branch is better than it ever was, but Microsoft is creating issues wanting everyone over to .NET5 or higher, and opensim server is written for .NET 4.6 right now, and needs a rather big rewrite to run on .NET5.

In addition mono (for the same reasons) have not been updated for a year because MS also want those users over to .NET5.  So that is obviously an issue for running opensim on Linux/macOS servers.  (MS took over the mono project).

As an example, it is not possible to compile opensim server on debian 11 unless you coerce it to use old mono builds. 64-bit ARM builds you can just forget. 

[Gavin Hird]

49 minutes ago, Jenna Huntsman said:

Okay, I did not know that the viewer on Mac was already based on Xcode. I'm sorry and I stand corrected.

Anyway, please do not insult myself or other members (or communities for that matter).

I'd like to point out however that [insert TPV name here] wouldn't exist if the viewer itself wasn't open source. Heck, SL would probably have died if the TPV community abandoned SL when LL themselves were too busy with Sansar.

Also, let's not forget that the software that most people use to create meshes for SL (Blender) is open-source. Not to mention other software for creating textures (GIMP) is also open-source.

OpenSimulator is a project which aims at making both sides of SL open-source. (LL only officially open-sourced the viewer, not the server). That's a neat idea. I don't use it personally. 

 

Anyway, to go around to the main point of the topic:

Could SL be made to run natively on M1 (and derivatives)?

Probably. Some viewers can be compiled for ARM (instead of x86, CoolVL is the example i'm thinking of specifically), but this has only been done on Linux. I don't know if anyone has tried doing so for Mac, but the CoolVL dev (Henri Beauchamp) isn't going to be the one to do it (as he does not have any Apple hardware, nor has an interest in getting any). But, the viewer is open-source so someone could make the required patches and upstream them.

It'd still be running on OpenGL, but it would eliminate Rosetta 2 (which would qualify it as running natively).

The TPV viewers would not exist unless LL open sourced the viewer code, so that is correct. You do, however significantly overestimate the influence of the TPVs on the survivability of SL. In many respects they have both hampered and slowed down the development of SL.

Additionally, during the development of Sansar, LL added significantly new functionality and support to the viewer code, which was done largely outside TPV influence at all.

 

When it comes to a native ARM viewer, there are two significant issues, of which one affects all ARM builds regardless of who made the hardware: Many of the libraries used by the current viewer are significantly outdated and don't compile on ARM without up to significant effort. 

In the case of Apple, there is also the issue of OpenGL, which simply don't exist, and Apple has no interest in developing and maintaining. The rewrite to Metal is complicated both because the OpenGL code is interwoven in the current viewer code in such a manner that it is hard to entangle. In addition the entire UI is rendered in OpenGL, and not only the 3D view. 

While most of the current viewer is written in C++, to take advantage of that for a Metal version, you have to enable Objective-C++, which is currently not possible given the current state of the code base. The reason for this is that the Metal libraries can only be used with Objective-C or Swift.

Using Objective-C++ is the shortest route, as a Swift rewrite is exactly that – a rewrite, and from the ground up. 

 

Unless LL has a secret viewer project for the Mac, I see no trace of coding activity that is required to support future Macs (or i-devices) once Apple turns off the switch on OpenGL.

As a Mac user in SL and opensim since 2007 and 2008 respectively, it is painful to watch.

(I also develop a viewer for opensim on the same codebase with opensim modifications + Mac specific code for the macOS version).

Edited November 18, 2021 by Gavin Hird

[Nick0678]

17 minutes ago, Gavin Hird said:

As a Mac users in SL ... etc

Non SL related question and off topic. how well does crossover21 runs on it, have you checked?

[RaffaeloMonastir]

Hello,

guess I need some help as my FPS rates are not as good as some of you are reporting here with the M1 and give the fact I am on the M1 Max I would assume quite good performance.

I am on the 16'' MacBookPro M1 Max with 10Core CPU, 32 Core GPU and 32GB of RAM. SSD is 1TB and macOS Monterey 12.0.1

I will add some screen shots. Basically it is no difference for me if I have HiDPi Display activated or not in Firestorm

1st image: FS MediumHigh Charkyn Forest HiDPi on: FPS at 31.2
https://gyazo.com/8e336fe69eaabdad969903340344d85e

2nd image: FS Ultra  Charkyn Forest HDPI on: FPS down to 6.8
https://gyazo.com/835e65a0fe83987e2a81cefd70d009a3

3rd image: FS Low Charkyn Forest HiDPi on: FPS down to 14.2 (very surprising with lowest graphics settings) https://gyazo.com/cf0c10e5cf3def034227d6996f94470d

4th image: FS MediumHigh Charkyn Forest HiDPi off: FPS at 36.4
https://gyazo.com/e1ffc514d49250e891de466d933bbd21

5th image: FS Ultra  Charkyn Forest HDPI off: FPS down to 8.7
https://gyazo.com/8271dcdf884fde291191552ff7c1f027

6th image: FS Low Charkyn Forest HiDPi off: FPS down to 5.6 (even more surprising with lowest graphics settings and HiDPi off) https://gyazo.com/7809383a711361b27f4b1a1788ca8f82

7th image taken at a club: FS MediumHigh Muddys Music Cafe HiDPi on: FPS 10.1
https://gyazo.com/cb4f1accdcce63420cb11c93f5038090

Any support / ideas how I can make FS faster are highly welcome. Thanks to all who are able to support.

PS Here are my "about Firestorm Info":
 

Firestorm 6.4.21 (64531) Jul 22 2021 11:06:53 (64bit / SSE2) (Firestorm-Releasex64) mit Havok-Unterstützung
Versionshinweise

CPU: Apple M1 Max (2400 MHz)
Speicher: 32768 MB
Parallelität: 10
Betriebssystemversion: Mac OS X 10.16.0 Darwin 21.1.0 Darwin Kernel Version 21.1.0: Wed Oct 13 17:33:01 PDT 2021; root:xnu-8019.41.5~1/RELEASE_ARM64_T6000 x86_64
Grafikkartenhersteller: Apple
Grafikkarte: Apple M1 Max
Grafikkartenspeicher: 21845 MB

OpenGL-Version: 2.1 Metal - 76.1
HiDPI-Anzeigemodus: 1

RestrainedLove API: (deaktiviert)
libcurl-Version: libcurl/7.54.1 OpenSSL/1.0.2l zlib/1.2.8 nghttp2/1.40.0
J2C-Decoderversion: KDU v8.1
Audio-Treiberversion: FMOD Studio 2.01.09
Dullahan: 1.7.0.202011160759
  CEF: 81.3.10+gb223419+chromium-81.0.4044.138
  Chromium: 81.0.4044.138
LibVLC-Version: 2.2.8
Voice-Serverversion: Nicht verbunden
Modus: Firestorm
Oberflächendesign: Firestorm (Grey)
Fenstergröße: 3840x2160 px
Schriftart: Deja Vu (96 dpi)
Schriftart Größenanpassung: 0 pt
UI-Skalierung: 1
Draw Distance: 128 m
Bandbreite: 1500 kbit/s
LOD-Faktor: 2
Darstellungsqualität: Mittel-Hoch
Erweitertes Beleuchtungsmodell: Nein
Texturspeicher: 2048 MB (1)
Disk-Cache: Max size 2048.0 MB (100.0% used)
Kompiliert mit Clang, Version 4.2.1 Compatible Apple LLVM 11.0.0 (clang-1100.0.33.12)
18. November 2021 10:52:02 SLT

 

 

Edited November 18, 2021 by RaffaeloMonastir
Added some specs

[RaffaeloMonastir]

On 10/26/2021 at 6:27 PM, anitabush said:

Hopefully some SL user will get their hands on one soon to test. They should be significantly faster than the first m1 macs, especially the 32 core m1 max. 
 

I wish we could get some news on a native apple silicon viewer, using emulation really hampers performance.

See my tests above...not that much promising 😞

[Lyssa Greymoon]

19 minutes ago, RaffaeloMonastir said:

Any support / ideas how I can make FS faster are highly welcome.

HiDPI off, and for water reflections select none, opaque. Mac displays are great and I don't think the performance hit from HiDPI is worth it.