glDrawArraysInstanced best practice


Is there a document that tells how to avoid performance pitfalls when using instanced rendering on PowerVR?

I remember geometry instancing being slow on several platforms in the past, especially if the base instance had only a low triangle count:


We have a section in this document:



you mean 6.2.3? I know about that. It hasn’t many tips apart from the “use divisor of 2” recommendation. I would rather like to know if any of the things that made instancing slow for other gpus in the past have been eliminated in PowerVR like

  • Being slower than triangle strips for low triangle counts.
  • Not using the post transform cache optimally
  • Expanding instanced calls to full VBOs on cpu side.

© Imagination Technologies Limited. All rights reserved.
Privacy PolicyTerms & ConditionsTrademarksCookies