Reply to thread

Message

<blockquote data-quote="Scythe42" data-source="post: 508277" data-attributes="member: 95833">Update:Got it. It's the scheduler. There are multiple things. First we have some checks in the code where under certain circumstances a frame should be skipped and some where it is still rendered when it shouldn't. There are improper workarounds for some issues in the code. Also the sleep timings are preventing a proper playback because they often schedule a frame so that it will be too late for the vsync in the end. When you are relying on the internal vsync function of DirectShow then you must make sure that the frame arrives early enough because the internal clocking is not accurate enough, but not too early. Also there are some other minor issues I need to check in details.Right now I have some kind of solution, but it needs further tweaking and testing at various rates and needs some stability testing as well before it makes sense that others test a more complete patch. There's no late frame problem anymore, which reduced the number of dropped frames in comparison to the quick fix. But there will always be a few because Windows is not a real-time OS. This can be further improved by providing an own clocking mechanism that's more accurate than the internal one. But one thing after another.</blockquote>

[QUOTE="Scythe42, post: 508277, member: 95833"] [B]Update:[/B] Got it. It's the scheduler. There are multiple things. First we have some checks in the code where under certain circumstances a frame should be skipped and some where it is still rendered when it shouldn't. There are improper workarounds for some issues in the code. Also the sleep timings are preventing a proper playback because they often schedule a frame so that it will be too late for the vsync in the end. When you are relying on the internal vsync function of DirectShow then you must make sure that the frame arrives early enough because the internal clocking is not accurate enough, but not too early. Also there are some other minor issues I need to check in details. Right now I have some kind of solution, but it needs further tweaking and testing at various rates and needs some stability testing as well before it makes sense that others test a more complete patch. There's no late frame problem anymore, which reduced the number of dropped frames in comparison to the quick fix. But there will always be a few because Windows is not a real-time OS. This can be further improved by providing an own clocking mechanism that's more accurate than the internal one. But one thing after another. [/QUOTE]

Verification