[PATCH 0/1] MR6848: mf/evr: Increase the number of prerolled samples to 3.
Fixes some stuttering in Secret of Mana. Native seems to only preroll 2 samples, but it still stutters a bit in Secret of Mana without at least 3 buffers. This is probably caused by a different sample decoding pattern, and 3 buffers in flight should be fine. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6848
From: Rémi Bernon <rbernon(a)codeweavers.com> Fixes some stuttering in Secret of Mana. Native seems to only preroll 2 samples, but it still stutters a bit in Secret of Mana without at least 3 buffers. This is probably caused by a different sample decoding pattern, and 3 buffers in flight should be fine. --- dlls/mf/evr.c | 15 ++++++++++++--- dlls/mf/session.c | 2 +- 2 files changed, 13 insertions(+), 4 deletions(-) diff --git a/dlls/mf/evr.c b/dlls/mf/evr.c index db7053a1405..5ac526c9651 100644 --- a/dlls/mf/evr.c +++ b/dlls/mf/evr.c @@ -64,6 +64,7 @@ struct video_stream LONG refcount; unsigned int id; unsigned int flags; + unsigned int preroll_count; struct video_renderer *parent; IMFMediaEventQueue *event_queue; IMFVideoSampleAllocator *allocator; @@ -420,9 +421,16 @@ static HRESULT WINAPI video_stream_sink_ProcessSample(IMFStreamSink *iface, IMFS if (stream->flags & EVR_STREAM_PREROLLING) { - IMFMediaEventQueue_QueueEventParamVar(stream->event_queue, MEStreamSinkPrerolled, &GUID_NULL, S_OK, NULL); - stream->flags &= ~EVR_STREAM_PREROLLING; - stream->flags |= EVR_STREAM_PREROLLED; + if (stream->preroll_count--) + IMFMediaEventQueue_QueueEventParamVar(stream->event_queue, MEStreamSinkRequestSample, + &GUID_NULL, S_OK, NULL); + else + { + IMFMediaEventQueue_QueueEventParamVar(stream->event_queue, MEStreamSinkPrerolled, + &GUID_NULL, S_OK, NULL); + stream->flags &= ~EVR_STREAM_PREROLLING; + stream->flags |= EVR_STREAM_PREROLLED; + } } } @@ -1528,6 +1536,7 @@ static HRESULT WINAPI video_renderer_preroll_NotifyPreroll(IMFMediaSinkPreroll * IMFMediaEventQueue_QueueEventParamVar(stream->event_queue, MEStreamSinkRequestSample, &GUID_NULL, S_OK, NULL); stream->flags |= EVR_STREAM_PREROLLING; + stream->preroll_count = 3; } LeaveCriticalSection(&stream->cs); } diff --git a/dlls/mf/session.c b/dlls/mf/session.c index 3c989249229..d01f2728782 100644 --- a/dlls/mf/session.c +++ b/dlls/mf/session.c @@ -1798,7 +1798,7 @@ static HRESULT session_append_node(struct media_session *session, IMFTopologyNod &IID_IMFVideoSampleAllocator, (void **)&topo_node->u.sink.allocator))) { if (FAILED(hr = IMFVideoSampleAllocator_InitializeSampleAllocator(topo_node->u.sink.allocator, - 2, media_type))) + 4, media_type))) { WARN("Failed to initialize sample allocator for the stream, hr %#lx.\n", hr); } -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/6848
So, is this basically a workaround for some slowness of the pipeline? Do you know why it helps? Does it stutter when starting playback or all the way through? -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6848#note_88766
I don't know exactly but the stutter started happening after 8ffebcf4ab869bf1067b23f3a67eead0213494c6. It does seem right to me to have more than one buffer in flight in any case, as using a single buffer will just stall the pipeline while the application is handling the current frame. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6848#note_88783
This merge request was closed by Rémi Bernon. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6848
participants (2)
-
Nikolay Sivov (＠nsivov) -
Rémi Bernon