Closed antoniotorresm closed 5 years ago
Actually, I've tried to specifically set the device with index 1 in the sample (seen by vulkaninfo
as Unknown AMD GPU
, just the same card but using amdvlk
instead of RADV
) and now I'm just getting a segmentation fault.
[1] 105065 segmentation fault (core dumped) ./simple
Thank you for testing vuda and reporting back. It seems to be an error from my side as the DescriptorPoolSize does not seem to be specified correctly (apparently this has no implication on my test setups). Try changing, the line in the file pool.hpp from
m_descriptorPoolSize(vk::DescriptorPoolSize(vk::DescriptorType::eStorageBuffer, binding_size)),
to
m_descriptorPoolSize(vk::DescriptorPoolSize(vk::DescriptorType::eStorageBuffer, binding_size * static_cast<uint32_t>(capacity))),
The intended behavior is that the allocation of the descriptor sets should never be able to do any overallocation.
Unfortunately, I don’t have access to a kabini, so I have no possibility of doing a repro. Please let me know what you find. Also, any other feedback on test results and unexpected behavior are welcome.
That worked! Opened a PR with the fix :)
I've just tried to run the "simple" sample on my laptop and I'm getting this error:
Here's the output for
vulkaninfo
:And here is my hardware info:
My OS is Arch Linux with kernel
linux-5.3.8-arch1-1
. I get this problem using bothRADV
(part of Mesa project) andamdvlk
(AMDVLK Open, maintained by AMD). Just for reference, I can run Sascha Willems' Vulkan examples without any issue.Thanks in advance!