While this does sound somewhat useful to narrow down driver issues or GPU hangs, I don't see how this could be integrated without
In case of RADV, RADV_DEBUG can be used to debug GPU hangs, vmfaults etc. somewhat effectively, not sure if there are similar ways to achieve this on Nvidia.
While this might sound like a great idea, I don't think it's worth the effort.
RADV_DEBUGx1 2018-04
Hi,
don't know if may be useful for this project as most bugs seems to be catched either at runtime by validation layers or at compile time by SPIR-V tools..
anyway in case you don't read the news seems an AMD engineer has released a much similar to Nvidia Aftermath lib a Vulkan "after crash" as open source header and that says works for both Nvidia and AMD GPUs and should also work on Windows and Linux..
EDIT:
http://asawicki.info/news_1677_debugging_vulkan_driver_crash_-_equivalent_of_nvidia_aftermath.html
https://github.com/sawickiap/MISC/blob/master/VulkanAfterCrash.h
also seeing new GDC Vault Nvidia Aftermath talk seems this month is coming AfterMath 2.0 with added Vulkan support (and Linux too)..
it will be via new Vulkan ext. "vk_nv_device_diagnostic_checkpoints" with new functions like vkcmdsetcheckpoint
perhaps @pdaniell-nv can hint us if it can be useful to integrate support of this library for this project..