IPC from Renderer to Browser with a shared memory handle

[wkbittner]

Does CEF currently support sending cross IPC messages ( renderer to browser) with a shared memory handle as an argument? If not - can I add that feature?

[wkbittner]

I am currently using CefBinaryValue - is it implemented using Shared Memory handles? Also, is there a way to avoid the Copies to create, and get the data and do something like VideoFrame does in Chromium by allowing to wrap external shared memory?

[magreenblatt]

I am currently using CefBinaryValue - is it implemented using Shared Memory handles? Also, is there a way to avoid the Copies to create, and get the data and do something like VideoFrame does in Chromium by allowing to wrap external shared memory?

Shared memory functionality is not exposed by CEF. It does exist in Chromium in various forms.

[wkbittner]

I am going to add it to CefListValue for IPC messages - I need to anyway as there is a mem cpy in and out of CefBinaryValue and I am send mass amounts of data cross IPC. How is CefSharedMemory? Any requests?

[magreenblatt]

CefListValue uses base::Value internally so you won't easily be able to add a new type. What Chromium object/API are you planning to use when implementing your shared memory support?

[wkbittner]

Hi,
I have built a custom video rendering pipeline for real time streaming video - I am passing already decoded video frames into a custom part of chromium. Would be awesome if there was info on how to build a mojo interface then implement it on cef - that would be ideal I am thinking - if not I just would like to pass a shared memory handle cross renderer->browser process to chromium - The reason it is being done like this because I am built the rendering pipeline to be used in JS by normal webrtc calls to manipulate/add/remove video streams and though the enumerate devices seems to be able to enumerate cross IPC - it was too complicated to build initially and easier to leave in browser process. Right now I use CefBinaryValue to send in a message to browser process but that adds two memory copies which is really expensive to - imagine - 5 video streams at lets say approx 400 Mb/s of cross IPC frames ( back of envelope but of course varies) so the memory copies are huge to us.