intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-27 07:44:16 +08:00
Code Issues Packages Projects Releases Wiki Activity
9,815 Commits 42 Branches 655 Tags
a74ae8f6affea215e19c823664ede2581ea90da7
Commit Graph

376 Commits

Author SHA1 Message Date
Lukasz Jobczyk
a74ae8f6af Implement timestamp wait for events 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-17 17:57:30 +01:00
Lukasz Jobczyk
c8ba97e492 Restore gpgpu csr's mutex lock in the enqueue blit 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-10 13:36:46 +01:00
Lukasz Jobczyk
f91ae9d59c Add multithread enqueue blit OOQ test 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-07 11:17:27 +01:00
Compute-Runtime-Validation
e526cc470b Revert "Add multithread enqueue blit OOQ test" 
This reverts commit 0919cad885.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
 2022-03-03 16:06:15 +01:00
Lukasz Jobczyk
999c6424a4 While enqueue blit do not flush gpgpu if already flushed 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-03 13:01:57 +01:00
Patryk Wrobel
f193efec2f Remove additional memory allocations for surfaces container 
In constructor of CommandComputeKernel we had been doing multiple allocations
of memory on heap due to lack of call to std::vector copy-constructor or reserve
member function.

Furthermore, in production code there is only one place, where we create objects
of this type and we redundantly copy the local variable, which could be moved.

This change:
- ensures that constructor of CommandComputeKernel performs single allocation
in the worst case; in the best case, it does not allocate memory due to usage
of std::move on input parameter
- steals the memory of the local variable in place of usage of the constructor
to remove redundant copying and memory allocations
- uses reserve() method to reduce the number of allocations during creation
of this local variable

Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
 2022-03-03 12:07:36 +01:00
Lukasz Jobczyk
0919cad885 Add multithread enqueue blit OOQ test 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-02 14:18:58 +01:00
Michal Mrozek
bfacd14b61 Remove not needed code. 
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-03-02 13:10:15 +01:00
Lukasz Jobczyk
ea574d9b39 Optimize enqueue blit mutex 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-01 14:43:29 +01:00
Lukasz Jobczyk
3c30e1b02b Add AssignBCSAtEnqueue debug flag 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-03-01 12:43:36 +01:00
Patryk Wrobel
0ecc7c5e3b Detect GPU hangs in clFinish 
This change introduces detection of GPU hangs in
clFinish function as well as unit tests to cover
the new code.

Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
 2022-02-28 19:07:36 +01:00
Konstanty Misiak
cf1bc3a2ba Disable EU fusion based on kernel properties from compiler 
Related-To: NEO-6633

Signed-off-by: Konstanty Misiak <konstanty.misiak@intel.com>
 2022-02-28 18:50:38 +01:00
Patryk Wrobel
7f729b7f89 Detect GPU hang in clWaitForEvents 
This change:
- moves NEO::WaitStatus to a separate file
- enables detection of GPU hang in clWaitForEvents
- adjusts most of blocking calls in CommandStreamReceiver to return WaitStatus
- adds ULTs to cover the new code

Related-To: NEO-6681
Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
 2022-02-23 13:33:09 +01:00
Bartosz Dunajski
4b0d986876 Move AllocationType enum out of GraphicsAllocation class 
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2022-02-04 17:49:09 +01:00
Lukasz Jobczyk
9ff1307b4b Fix optimize timestamp packet dependiencies 
-program barrier after global fence allocation is programmed
-do not double barrier timestamp in blit enqueue
-flush GPGPU while submitting to BCS when barrier requested

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-02-03 16:27:09 +01:00
Bartosz Dunajski
52c6973e89 Rename blitBuffer method 
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2022-02-03 15:02:55 +01:00
Michal Mrozek
8f85d4b8f8 Add debug variable to override message simd. 
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-02-03 11:07:42 +01:00
Konstanty Misiak
65255ea38f Allow pointer copy when copying from host to svm 
Related-To: NEO-5737

Signed-off-by: Konstanty Misiak <konstanty.misiak@intel.com>
 2022-02-02 13:08:35 +01:00
Mateusz Hoppe
3e66f21df6 Remove incorrect Debug break 
- kernels using debug surface does not need to use
only bindful buffers

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
 2022-01-28 13:51:02 +01:00
Maciej Plewka
f8c104feaa Use fw declaration of IndirectHeap in CommandContainer 
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
 2022-01-26 13:30:26 +01:00
Rafal Maziejuk
f064f7dd67 Check if origin + region doesn't exceed image dimensions 
Signed-off-by: Rafal Maziejuk <rafal.maziejuk@intel.com>
Related-To: NEO-6137
 2022-01-25 15:13:15 +01:00
Katarzyna Cencelewska
3d9e1ea3a5 Remove device enqueue part 11 
- remove templates from queue functions

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-25 12:40:56 +01:00
Michal Mrozek
27c43b27f3 Remove not needed method. 
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-01-20 15:02:19 +01:00
Katarzyna Cencelewska
a06fbd2077 Remove device enqueue part 10 
- remove DeviceQueue

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-19 17:41:06 +01:00
Katarzyna Cencelewska
97765cd39f Remove device enqueue part 7 
- mainly remove BlockKernelManager and ReflectionSurfaceHelper

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-19 10:41:57 +01:00
Michal Mrozek
a8b8a61c87 Fix marker calls in non profiling path. 
When waiting , if we guard blit copy operation,
we need to wait for bcs as well.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-01-18 15:46:50 +01:00
Michal Mrozek
dd63f96836 Fix dependency resolution when marker with profiling is programmed. 
We need to make sure it properly follows dependencies.
Before the change, code was not programming semaphores.

Related-to: NEO-6444

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-01-18 11:48:38 +01:00
Katarzyna Cencelewska
59683ec491 Remove device enqueue part 6 
- isParentKernel, peekParentKernel, parentKernel
- structs: AUBParentKernelFixture, MockParentKernel,
ParentKernelCommandQueueFixture

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-17 13:56:29 +01:00
Katarzyna Cencelewska
d2818aaea2 Remove device enqueue part 5 
-remove scheduler and builtin_kernels_simulation

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-13 14:15:26 +01:00
Katarzyna Cencelewska
a9ebb49fb5 Remove device enqueue part 1 
Remove:
-tests with matcher for device enqueue
-classes: MockDeviceQueueHw, DeviceQueueHw, SchedulerSimulation,
DeviceQueueHwTest, KernelArgDevQueueTest
-functions: forceDispatchScheduler, processDeviceEnqueue, dispatchScheduler

Related-To: NEO-6559
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-01-11 15:29:49 +01:00
Dominik Dabek
6c40c712d8 Check workDim when enqueueing kernel 
Return CL_INVALID_WORK_DIMENSION when workDim is more than
maxWorkItemDimensions.

Related-To: NEO-6159

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2021-12-31 15:42:09 +01:00
Filip Hazubski
5be4d89b73 Rename function 
Rename MemorySynchronizationCommands::isDcFlushAllowed
to MemorySynchronizationCommands::getDcFlushEnable

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-30 17:03:22 +01:00
Lukasz Jobczyk
418a8854b2 Skip timestamp wait when enqueue non kernel operation 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-30 10:02:07 +01:00
Lukasz Jobczyk
95585a81f7 Optimize timestamp packet dependencies 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-30 09:48:36 +01:00
Lukasz Jobczyk
b59b0b6b36 Download timestamps before checking completion 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-28 08:14:27 +01:00
Lukasz Jobczyk
14e338e669 Revert "Optimize timestamp packet dependencies" 
This reverts commit c365b422963917e7b882f9db985969c036f0fa3f.

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-23 14:29:29 +01:00
Filip Hazubski
f4c151cce5 Refactor PipeControlArgs struct 
Remove struct PipeControlArgsBase

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-22 17:13:16 +01:00
Filip Hazubski
9a450d1b74 Pass hwInfo to appendMiFlushDw 
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-22 15:22:47 +01:00
Filip Hazubski
0fd685541d Add isDcFlushAllowed function to HwInfoConfig 
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-21 18:29:43 +01:00
Filip Hazubski
6d439f88bb Explicitly set dcFlushEnable value 
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-21 12:21:11 +01:00
Mateusz Jablonski
66bf806018 Remove magic number from set/getBatchBufferStartAddressGraphicsaddress methods 
rename methods to set/getBatchBufferStartAddress

Related-To: NEO-6466
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
 2021-12-16 19:03:01 +01:00
Lukasz Jobczyk
1f0c58d0bf Refactor timestamp wait mechanism 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-16 13:22:19 +01:00
Bartosz Dunajski
fe8cb190bc Remove deprecated code 
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2021-12-14 11:36:10 +01:00
Zbigniew Zdanowicz
01348451db Add multi tile barrier to marker command on multi tile device 
Related-To: NEO-6262

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2021-12-13 18:14:39 +01:00
Filip Hazubski
64aec8dc4e Rename functions 
Rename getEngines to getAllEngines
Rename engines to allEngines
Rename getEngineGroups to getRegularEngineGroups
Rename engineGroups to regularEngineGroups

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-13 14:27:02 +01:00
Michal Mrozek
ebb16c8b74 Add new debug key to control dependency resolution. 
Currently only supported scenario is one in order queue.
Instead of resolving dependencies via semaphores, do this with pipe controls.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2021-12-09 17:26:38 +01:00
Filip Hazubski
f137e927c5 Add EngineUsageHint debug variable 
Related-To: NEO-6219

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
 2021-12-09 16:06:23 +01:00
Rafal Maziejuk
d5f3ac37bf Add KernelExecutionType argument to encodeAdditionalWalkerFields method 
Signed-off-by: Rafal Maziejuk <rafal.maziejuk@intel.com>
 2021-12-08 12:00:42 +01:00
Lukasz Jobczyk
0c49fa8203 Add missing cache flush for blocked path 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2021-12-07 12:36:34 +01:00
Maciej Dziuban
e3bb526067 Optimize timestamp packet dependencies 
- Clear dependencies even if last engine changed
- Do no program semaphore waiting for blit when blit is submitted with gpgpu
- Track barrier timestamps to correctly synchronize blits in OOQ

Related-To: NEO-6444
Signed-off-by: Maciej Dziuban <maciej.dziuban@intel.com>
 2021-12-07 11:47:50 +01:00