6 ms·
Like iOS 12 very promising overall from the sound of it, and a genuine step forward from previous years in the fundamentals of speed and stability just as they
by xoa 8y ago
Like iOS 12 very promising overall from the sound of it, and a genuine step forward from previous years in the fundamentals of speed and stability just as they promised to work towards. One continuing (though not new at all to the release) disappointment I have with Apple though:
>The end of OpenGL and OpenCL on the Mac
A solid summary of the situation, but to me where Apple really deserves blame there is not having MoltenGL and MoltenVK (and perhaps some theoretical MoltenCL too) as 1st party projects. I completely understand them wanting their own directly writable and controlled low level graphics language and having that be the interface layer. They are pretty vertically integrated and are now even doing their own GPUs, but even before that on the Mac they have long built the OS very strongly around GPU capabilities (for better and for worse, basic VM experience is very mediocre if you don't have a hardware GPU to passthrough). If anything I'm mildly surprised in retrospect Metal wasn't done earlier, OpenGL wasn't working for them and it's that big a deal. Microsoft invested in their own graphics layer for Windows ages earlier.
But controlling their own low level layer doesn't mean that higher level non-proprietary interfaces shouldn't still be supported on top. Part of the point should be to enable abstraction there. Apple is a big enough player and platform that this shouldn't be a zero-sum game, and their perfectly understandable strategic concerns addressed by Metal don't preclude OpenGL continuing to be officially supported as a layer on top of Metal. Apple could have even done it as another open source project of which they were the primary sponsor even if they wanted to manage it that way.
This may be just another symptom of their unusual startup-type culture which lets them hyperfocus on specific products but reduces their corporate ability to multitask, but it's still a shame. It could have been just about supporting newer and better stuff going forward, not taking something away. Even if MolenVK/MoltenGL ultimately fill some of that void I don't think that should have been relegated to 3rd party only.
- monocasa 8y agoI don't blame them for this. They want to get out of OpenGL because how much of a maintenance nightmare it has become due to non technical concerns. To ship top end OpenGL (and DirectX) drivers these days, it's more or less required that you write tons of code that recognizes the application that's calling your driver and fixes bugs in their code. Nvidia starting doing this, and now all end users see is that programs work on Nvidia and not AMD (or Apple as they write their own drivers), and blame everyone other than Nvidia. The switch to these new Mantle derived APIs is as much if not more about resetting those expectations of responsibilities than the technical benefits that these APIs provide (which AZDO 80/20 ruled most of the way to anyway). Supporting a MoltenGL as a first class item defeats the whole purpose of moving away from OpenGL to begin with from their perspective. OpenCL never got the critical mass in their eyes, and someone else is already taking on MoltenVK.
- ancarda 8y agoWait a sec... My gaming PC (2x AMD R9 290X, 8 GB video memory) is absolutely awful with a lot of (esp. Ubisoft) games like Assassins Creed or Watch_Dogs. I just assumed the games weren't optimized for AMD hardware. That is based partly off the Nvidia logos that are shown during startup and in the graphical settings menu. There's also graphical issues at high resolutions. I've been planning on buying an Nvidia GPU so I can actually play games at a decent frame rate. Are you telling me part (or all?) of the problem is bad drivers? I got the drivers from Windows Update rather than from AMD so they would just update themselves and be the correct ones.
- crooked-v 8y agoNot bad drivers, bad game code.
- cpeterso 8y agoOr AMD drivers that are not adding "do what I mean" hacks to work around bad game code? Sounds like AMD and Nvidia took different approaches to Postel's Law ("Be conservative in what you send, be liberal in what you accept").
- monocasa 8y agoIMO, there's a practical upper limit to how liberal you can be. There are some games that straight up don't call endscene and present and crap like that in DirectX land. Or have bugs in their shaders, expecting the driver ISVs to just completely rewrite them by hand anyway. Nvidia can get away with this because they have something like 10x the engineering staff as AMD's on the GPU sides of their companies. Apple never had a hope of competing in this space at the same level.
- cpeterso 8y agoIf Nvidia is adding workarounds in their drivers for crappy game code, how did those games work before Nvidia added the workaround for their crappy code? :)
- Someone 8y ago”but to me where Apple really deserves blame there is not having MoltenGL and MoltenVK (and perhaps some theoretical MoltenCL too) as 1st party projects” Yeah, they should have learned from OS/2 and that other graphics subsystem it supported for compatibility (sorry, forgot what it was called) Foo on top of bar will always have incompatibilities and performance problems compared to foo on bare metal, but third parties will develop for it anyways if it gives them an easy path to a larger market, and the risk is that bar will become an implementation detail, rather than a competitive advantage. Electron apps are a good recent example. I bet all major OS manufacturers would like to get rid of them, if it meant those applications ran on their OS and not on others.
- pjmlp 8y agoGame consoles and Windows have done just fine with their own APIs, and OpenGL specific talks usually don't win time slots at Game Developers Conference sessions, outside Khronos parallel talks. AAA studios are quite comfortable writing multi-plaform game engines since the 80's.