[RFC] Fix Parallel-RDP on Intel Xe

[stuken]

As noted in #383 ParaLLel-RDP crashes on current Intel Xe drivers. On newer drivers vkGetPhysicalDeviceMemoryProperties calls incorrectly populates the VkPhysicalDeviceMemoryProperties struct values meaning that all contained memoryTypes->propertyFlags fields return 1 instead of the or'd flag values.
This means that the find_memory_type method's lookups will never match based on the expected flags, the rdram buffer is not initialized and it crashes out.

This will force set the memoryTypes index to the appropriate value bypassing the lookup method if the driver vendor is Intel and the version is at least the current most one, 30.0.100.9862.

[loganmc10]

This should be submitted to: https://github.com/Themaister/parallel-rdp

That is the upstream/source repo

[stuken]

Turns out my root cause analysis was totally wrong. vkGetPhysicalDeviceMemoryProperties is returning correctly and it's a whole host of other values not lining up to stop things.

1. reqs.memoryTypeBits on the offending driver is 7
2. host_pointer_props.memoryTypeBits is 1

Those values are &'d and used as the mask value in find_memory_types but the resultant 1 means thanks to the (1u << i) & mask check only the first entry in the memoryTypes array is ever looked at.

Secondly, if we were to pass that check none of the CachedCoherentHostPreferCached case prio entries will fully match the flags as it contains a 4th flag, MEMORY_PROPERTY_DEVICE_LOCAL_BIT.

[m4xw]

deprecated