Related-To: NEO-6075
After this change driver will fail clBuildProgram/zeModuleCreate api calls
whenever stateful access is discovered on PVC.
This is required since in this case allocation greater than 4GB
will not work.
If user still wants to use stateful addressing mode,
-cl-opt-smaller-than-4GB-buffers-only / -ze-opt-smaller-than-4GB-buffers-only
build option should be passed as build option, but then user can not use
bufers greater than 4GB.
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
This commits removes part of condition requiring requiresImplicitArgs
flag set in kernel descriptor in order to set usesStringMap flag.
Related-To: NEO-6604
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
This patch moves OA specific Metric Streamer implementation
to OA specific classes.
Related-To: LOCI-2905
Signed-off-by: Joshua Santosh Ranjan <joshua.santosh.ranjan@intel.com>
- every isa allocation will have ELF linked
- fix for debug elf from patchtoken binary:
pass relocated ELF when exists
- simplify code
Related-To: NEO-5571
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
In OCL product family of target device is not set
which leads to a fail on validating target device in
ZEBin path.
This change adds function that sets all
necessary fields based on provided hardware info.
Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
Forces extended buffer size by adding pageSize specify by number when
debug flag is >=1 in L0 USM calls
Usage:
ForceExtendedUSMBufferSize=2
size += (2 * pageSize)
Signed-off-by: Krzysztof Gibala <krzysztof.gibala@intel.com>
This patch is to refactor MetricQuery and MetricQueryPool
so that Stall sampling metric could be integrated seamlessly
Related-To: LOCI-2904
Signed-off-by: Joshua Santosh Ranjan <joshua.santosh.ranjan@intel.com>
- In WSL the handle for DMA buf is an NT handle.
- To share and import this memory we check if the
handle is an NT handle before attempting to
load as an FD.
- If the handle is and NT handle, then we open the fd
as an NT handle.
Signed-off-by: Spruit, Neil R <neil.r.spruit@intel.com>
when local ids are generated by HW, use same dim order for runtime generation
move common logic to separated file
Related-To: NEO-5081
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
- Allow flushTask usage for XeHp+ only
- Fix black box test to only use Copy queue if found
Related-To: LOCI-1988
Signed-off-by: Aravind Gopalakrishnan <aravind.gopalakrishnan@intel.com>
Rename Metric Context and move OA specific functions
and members from Metric Context to OA specific class(MetricSource).
This refactoring is done so that additional Metric Source
like Stall Sampling could be Seamlessly implemented.
Related-To: LOCI-2753
Signed-off-by: Joshua Santosh Ranjan <joshua.santosh.ranjan@intel.com>
If kernel has no stateless indirect accesses don't set the
kernelHasIndirectAccess flag.
Don't make resident or migrate if kernel has no indirect accesses.
Changed initial values in KernelAttributes.
Related-To: NEO-6597
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
Instead of creating new allocation per fence, use the task count.
Fence synchronize will wait for task count update.
Related-To: NEO-6634
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
When linear stream created for command container has not enough space
for command and BB_END it will program BB_END and allocate new command
buffer allocation. Pointer returned from getSpace in this case will
return storage from new command buffer allocation.
Related-To: NEO-5707
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
This change uses DRM_IOCTL_I915_GET_RESET_STATS to detect
GPU hangs. When such situation is encountered, then
zeCommandQueueSynchronize returns ZE_RESULT_ERROR_DEVICE_LOST.
Related-To: NEO-5313
Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
- Add getMemoryManagerType to check which memory manager has been init
to determine if Linux + WDDM memory manager is in use.
- Add isNTHandle to test and verify if a handle is an NT handle during
L0 Open IPC Handle.
Signed-off-by: Spruit, Neil R <neil.r.spruit@intel.com>
Prevent accessing local memory from the host.
Create an array on stack with desired value and pass it to filling function.
175x faster call execution for scenario with profiling and device events.
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
If kernel has no stateless indirect accesses don't set the
kernelHasIndirectAccess flag.
Don't make resident or migrate if kernel has no indirect accesses.
Changed initial values in KernelAttributes.
Related-To: NEO-6597
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
* This commits enables by default implicit scaling, but only on PVC B step
* Users can disable this feature by debug flag EnableImplicitScaling=0|
Related-To: NEO-6589
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
Make them resident directly instead of populating residency container
Remove finds, not needed, CSR resolves duplicates at makeResident calls
Observed gain is 32x for 10k indirect allocations.
Co-authored-by: Michal Mrozek <michal.mrozek@intel.com>
Co-authored-by: Dominik Dabek <dominik.dabek@intel.com>
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
