Some OpenCL application need to render libva surface
with drm display and use the same libva surface as
OpenCL kernel output. Such applicatoins need to
use /dev/dri/card0 instead of /dev/dri/renderD128
for getting the display fd.
Related-To: NEO-11714
Signed-off-by: Elaine, Wang <elaine.wang@intel.com>
Removed if condition when calling vm_bind in order to
reflect changed in KMD.
Specifically, dii-4347 to dii-4349 include changes on
strict fence ordering on vm_bind, where user fence has
to be inserted regardless of using pagefault or explicit
residency.
Related-To: NEO-11966
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>
adjust slice count to proper value based on previously calculated
max slices and max subslice counts
Related-To: NEO-12073
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
KMD exposes flat subslice info without indicating slice index
This commit calculates slice count based on total subslice count and
max subslice per slice count that we have from device blob
Related-To: NEO-12073
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
for xe kmd there will be new query for EU per DSS for PVC, LNL, BMG platforms
when new query is available, previous one (currently used in NEO) will be empty
To avoid integration issues this commit adds fallback to setup that value based
on max eu per dss that we get from GuC in device blob
Support for new query is in PR https://github.com/intel/compute-runtime/pull/745
Related-To: NEO-12012
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
once drm specific code queries engine info, caps are stored in drm-agnostic form
Related-To: NEO-10445
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
Calling virtual functions from within constructors is incorrect, the
virtualization mechanism does not work until a constructor completes.
So, instead of performing many operations in constructors, make them
small and use static `create()` methods to construct the object properly
and only then perform additional setup.
Mocks are a prominent example how such practice degrades the code.
A function called from a constructor is basically unmockable. Despite
being overriden by derived type (to alter the execution path as per
needs of a given test-case) the parent versions are always called.
This was making test development more and more painful as at some point
there is no hack to compensate for the lack of standard mechanism.
Related-To: NEO-9754
Signed-off-by: Maciej Bielski <maciej.bielski@intel.com>
- if create VM ioctl fails, fallback to query VM from created context
- in fallback path context's VM will not have flags applied
Related-To: NEO-7813
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
Enabled scratch page by default on PVC by setting
isDisableScratchPagesSupported to false for PVC.
Related-To: GSD-7742
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>
Initialize SIP kernel when shared device is being initialized
instead of api-specific device.
Initialize debugger when shared device is being initialized
instead of during platform or driver initialization.
Add missing makeResident calls for SIP kernel in heapless paths.
Related-To: HSD-18038645398, HSD-18038819112
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
