LU-3176: optimize getImageDataAtTime #52
Conversation
|
@stepancar did you check that visual output still makes sense? |
src/backends/beamcoder.ts
Outdated
| @@ -476,25 +483,23 @@ export class BeamcoderExtractor extends BaseExtractor implements Extractor { | |||
| return packet as Packet; | |||
| } | |||
|
|
|||
| _resizeFrameData(frame): Uint8ClampedArray { | |||
| _setFrameDataToImageData(frame: beamcoder.Frame, target: Uint8ClampedArray) { | |||
| const components = 4; // 4 components: r, g, b and a | |||
There was a problem hiding this comment.
nit: but should components be passed in? I think it's characteristic to the frame and target and doesn't really seem like something this function should be defining?
There was a problem hiding this comment.
Maybe we can read that from the frame object somehow.
There was a problem hiding this comment.
I added RGBA_PIXEL_SIZE constant
There was a problem hiding this comment.
@stepancar so we could not find the frame component number? from frame?
There was a problem hiding this comment.
I didn't get we need to. This code works only with RGBA formate. in our logic it's hardcoded. If we want to support other formats - we need to extend public interface too.
But I don't think it makes sense. We use RGBA everywhere in our system
src/backends/beamcoder.ts
Outdated
| let rawData = target; | ||
|
|
||
| if (!target) { | ||
| rawData = new Uint8ClampedArray(frame.width * frame.height * 4); |
There was a problem hiding this comment.
We should probably unhardcode the 4 here too. It would make the code more self-explanatory.
There was a problem hiding this comment.
(const components = /* whatever is needed to extract component number from the frame */;)
There was a problem hiding this comment.
@antoineMoPa , I created constant for that
RGBA_PIXEL_SIZE
stepancar
changed the title
|
✅ Linked to Task LU-3176 · Implement outcome from electron exploration |
stepancar
changed the title
It's somthing we wanted to do
we wanted to reuse existing buffer for specific video element instead of re allocating it for every frame
we don't need to copy beamcoder.frame, it's enough for us to create a view of typed array if we can guarantee that beamcoder frame will not be detached,
It improves GC