Performance counter sampling rate

[DerRM]

Hello everyone,

thank you for all the samples and great resources to learn an efficiently use Vulkan. I have a question about the hwcpipe performance counters. I'm not really sure I really understand the values that are printed out.

Most of the counters e.g. StatIndex::gpu_cycles or StatIndex::gpu_ext_read_bytes claim to be a value to be in per second intervals. But as I looked through the code I couldn't find any indicator that those are indeed sampled per second. They seem to be sampled per frame.

Am I missing something in the logic here?

Edit: Another indicator that those values are sampled by frame is the comparison of the values on my device (Mali-G78) and the sample README from the Vulkan docs which is a Mali-G76, so I would assume the values to be much closer and also more in favor of my device:

My device

Vulkan Docs

Best regards,
DerRM