Allow the application to force storageInfo.localOnly and get the
out-of-memory returned if not possible.
This is a windows-only feature supported on discrete platforms.
Related-To: NEO-13428
Signed-off-by: Maciej Bielski <maciej.bielski@intel.com>
Related-To: NEO-14360
Current gmm usage type of these resources is causing
them to be cached, which is incorrect.
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
Join thread in DllMain (which we are not aware of)
can result in hang occurring in DeferredDeleter, if the library is
freed before FreeLibraryAndExitThread call from within the worker thread,
the thread gets stuck, thus the main thread is stuck on worker->join().
Related-To: NEO-14121
Signed-off-by: Oskar Hubert Weber <oskar.hubert.weber@intel.com>
Join thread in DllMain (which we are not aware of)
can result in hang occurring in DeferredDeleter, if the library is
freed before FreeLibraryAndExitThread call from within the worker thread,
the thread gets stuck, thus the main thread is stuck on worker->join().
Related-To: NEO-14121
Signed-off-by: Oskar Hubert Weber <oskar.hubert.weber@intel.com>
Related-To: NEO-14056
No need to explicitly wait on Windows KMD during make resident as it has
a while loop that does it nevertheless. The KMD wait affects the API
overhead of zeCommandQueueExecuteCommandLists some platforms (MTL, and ARL).
Signed-off-by: Chandio, Bibrak Qamar <bibrak.qamar.chandio@intel.com>
Lock on csr is needed before lock on residency controller to prevent
incorrect lock order. Csr may be locked in waitOnCpu called from trimToBudget,
which may lead to deadlocks
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
Related-To: NEO-13870
Currently all monitor fences are triggering
interrupt due to Notify Enable field.
With this change, such field is programmed
right before KMD wait.
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
Related-To: NEO-13529
Also, add size threshold on iGPU on Linux,
and disable staging if imported host ptr could
be reused
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
on Windows setup ReadOnly flag, it makes there is no ProbeForWrite call within
CreateAllocation2 function call
it doesn't force read-only access from GPU perspective, that access it
defined in MapGpuVa function
Related-To: NEO-12986, HSD-15016952776
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
setting ReadOnly flag makes there is no ProbeForWrite call within
CreateAllocation2 function call
it doesn't make read-only access from GPU perspective, that access it
defined in MapGpuVa function
Related-To: NEO-12986, HSD-15016952776
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
Switching tagAllocation into UC causes regressions in non event
scenarios. It is no longer used as dependant type for semaphores.
Restoring previous GMM_USAGE settings for tag.
Marking events and in order nodes using only timestamp types as they
have proper GMM_USAGE settings already and can be both in smem and lmem.
Resolves: NEO-13847
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
Lukasz Jobczyk