Forums - Terrible vkCmdDrawIndexedIndirect|Count performance

2 posts / 0 new
or Register
Last post
Terrible vkCmdDrawIndexedIndirect|Count performance
alexander_1
Join Date: 4 Dec 20
Posts: 7
Posted: Mon, 2020-12-07 22:26
Top

The performance of vkCmdDrawIndexedIndirect|Count function is terrible. It's slower than serial vkCmdDrawIndexed and even slower than serial OpenGLES API.

Vulkan app with 16641 DIPs (32k tris) per frame:
Single call of vkCmdDrawIndexed: 16M tris/sec
16641 calls of vkCmdDrawIndexed: 1.45M tris/sec
16641 calls of vkCmdDrawIndexedIndirect: 169K tris/sec
Single call of vkCmdDrawIndexedIndirect with 16641 draws: 216K tris/sec
Single call of vkCmdDrawIndexedIndirectCount with 16641 draws from buffer: 216K tris/sec

OpenGLES app with 16641 DIPs (32k tris) per frame:
Single call of glDrawElements: 16M tris/sec
16641 calls of glDrawElements: 5.8M tris/sec
16641 calls of glDrawElementsIndirect: 4.8M tris/sec

https://drive.google.com/file/d/18uOBLEjQXwMa7JV-1jZUK5iVkB1FWwX1/view
https://drive.google.com/file/d/1EMlW_h4c7MCT6jDvyaCnsvGWY8-JMn40/view

Quote:
You were the chosen one! It was said that you would destroy the Sith, not join them! Bring balance to the Force, not leave it in darkness!


Thank you!

  • Up0
  • Down0
Re: Terrible vkCmdDrawIndexedIndirect|Count performance
AndreyVK_D3D
Profile picture
Join Date: 10 Jun 13
Location: Moscow
Posts: 36
Posted: Tue, 2021-01-05 07:47
Top

Adreno 610(Android 10)

 

OpenGLES app with 16641 DIPs (32k tris) per frame:
Single call of glDrawElements: 7.35M tris/sec
16641 calls of glDrawElements: 3.4M tris/sec
16641 calls of glDrawElementsIndirect: 2.83M tris/sec

Vulkan app with 16641 DIPs (32k tris) per frame:

Single call of vkCmdDrawIndexed: 7M tris/sec
16641 calls of vkCmdDrawIndexed: 1.12M tris/sec
16641 calls of vkCmdDrawIndexedIndirect: 83.56K tris/sec
Single call of vkCmdDrawIndexedIndirect with 16641 draws: 97.2K tris/sec
Single call of vkCmdDrawIndexedIndirectCount with 16641 draws from buffer: 97.2K tris/sec

  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.