What Is A Proper Usage Of Fence Synchronization In Webgl2?

Looking for some patterns/code examples/best practices of appropriate usage of fences in webgl2 (gl.fenceSync) - best if it would be non blocking of JS thread. var fence = gl.f

Solution 1:

I'm just guessing to be honest, I'm not actually sure how useful syncs are in WebGL2 but I'd think you don't want to block then the pattern would be like this

function main() {
  const gl = document.createElement('canvas').getContext('webgl2');
  if (!gl) {
    return alert('need webgl2');
  }
  
  callbackOnSync(gl, () => {
    console.log("done");
  });
  
  function callbackOnSync(gl, callback) {
    const sync = gl.fenceSync(gl.SYNC_GPU_COMMANDS_COMPLETE, 0);
    gl.flush();  // make sure the sync command is read

    setTimeout(checkSync);  

    function checkSync() {
      const timeout = 0;   // 0 = just check the status
      const bitflags = 0;
      const status = gl.clientWaitSync(sync, bitflags, timeout);
      switch (status) {
        case gl.TIMEOUT_EXPIRED:
          // it's not done, check again next time
          return setTimeout(checkSync);
        case gl.WAIT_FAILED:
          throw new Error('should never get here');
        default:
          // it's done!
          gl.deleteSync(sync);

          callback();
      }
    }
  }
}

main();