Electron中低延迟向渲染进程传输大YUV数据的方案咨询

我正在开发一个视频相关项目，出于安全考虑已禁用nodeIntegration并启用contextIsolation。业务逻辑在utility process中处理，解码后的YUV数据需发送至renderer process，通过VideoFrame和canvas完成渲染。但处理6K分辨率视频时，YUV数据量极大，传输至渲染进程的耗时过长，无法达到30fps的目标帧率。设备配置为Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz。

性能测试代码 #

setInterval(()=> {
  const beforeAlloc = performance.now()
  const data = new Uint8Array(4000_0000)  // large array
  // const data = new Uint8Array(4000)
  const afterAlloc = performance.now()

  const beforeSend = performance.now()
  port.postMessage(data) // the port is MessageChannelMain.port
  const afterSend = performance.now()

  const now = performance.now()
  // alloc's time was almost negligible, while postMessage took more than 90% of the time
  console.log(`alloc: ${parseInt(afterAlloc - beforeAlloc)}, send: ${parseInt(afterSend - beforeSend)}, time since last: ${parseInt(now - lastTime)}`)
  lastTime = now
}, 1000 / 30)

测试结果 #

传输大数据时的输出 #

alloc: 0, send: 77, time since last: 77
alloc: 3, send: 76, time since last: 80
alloc: 3, send: 82, time since last: 85
alloc: 0, send: 84, time since last: 85
alloc: 3, send: 78, time since last: 82
alloc: 3, send: 95, time since last: 99
...

传输小数据时的输出 #

alloc: 0, send: 0, time since last: 46
alloc: 0, send: 0, time since last: 46
alloc: 0, send: 0, time since last: 46
alloc: 0, send: 0, time since last: 48
alloc: 0, send: 0, time since last: 46
alloc: 0, send: 0, time since last: 46
...

当前困境与疑问 #

为避免阻塞main process，我选择在utility process中处理逻辑并通过MessageChannelMain通信，但效率仍不理想。目前能想到的高效方案是启用nodeIntegration并关闭contextIsolation，让renderer process直接访问YUV数据，但非必要不想这么做。

想请教：

• 是否有其他高效的跨进程大数据传输方法？
• 我的实现思路是否存在问题？
• 需要实时展示处理后的帧，还有其他解决方案吗？

内容的提问来源于stack exchange，提问作者li wakin