thank you for your insight. unfortunately, that hasnt seemed to solve the issue specifically when using Vulcan, both when running no AA or MSAA, and both when there were no meshes in the scene and meshes in the scene. were you using DX12?
heres the cpp error iv found for reference.
going there leads me to this function, specifically the line VERIFYVULKANRESULT(Result);
.
overall the function seems to return if the gpu has completed its work or something before more work can be given? im not sure. im interested in setting this to false (or true) to see if i can recreate my memory leak, but doing a single change to my unreal source takes hours to recompile.
bool FFenceManager::WaitForFence(FFence* Fence, uint64 TimeInNanoseconds)
{
#if VULKAN_ENABLE_AGGRESSIVE_STATS
SCOPE_CYCLE_COUNTER(STAT_VulkanWaitFence);
#endif
check(UsedFences.Contains(Fence));
check(Fence->State == FFence::EState::NotReady);
VkResult Result = VulkanRHI::vkWaitForFences(Device->GetInstanceHandle(), 1, &Fence->Handle, true, TimeInNanoseconds);
switch (Result)
{
case VK_SUCCESS:
Fence->State = FFence::EState::Signaled;
return true;
case VK_TIMEOUT:
break;
default:
VERIFYVULKANRESULT(Result);
break;
}
return false;
}