intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-25 13:33:02 +08:00
Code Issues Packages Projects Releases Wiki Activity
11,425 Commits 42 Branches 655 Tags
1fedaee18c38afcd9d16aa2bca1541e2672f8310
Commit Graph

460 Commits

Author SHA1 Message Date
Lukasz Jobczyk
1fedaee18c Wait on timestamps in Event::wait 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-10-10 13:04:21 +02:00
Compute-Runtime-Validation
1c73031a9b Revert "Wait on timestamps in Event::wait" 
This reverts commit f91047f2ee.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
 2022-10-08 06:17:33 +02:00
Lukasz Jobczyk
f91047f2ee Wait on timestamps in Event::wait 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-10-07 12:04:45 +02:00
Mateusz Hoppe
2f24ef6855 Error from clEnqueueNDRangeKernel() for too big group counts 
Resolves: NEO-6976

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
 2022-10-05 12:35:16 +02:00
Mateusz Jablonski
23eff82d0a OCL: optimize creating printf buffer 
Dont create printf buffer when kernel doesnt require it

Related-To: HSD-18023825570
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
 2022-09-30 12:25:57 +02:00
Dominik Dabek
d8b7d56160 Copy host ptr on cpu if possible in clCreateBuffer 
use cpu copy with locked pointer if possible
because this is faster than copy on gpu
limit to buffers of size at most 64kb

Related-To: NEO-7332

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2022-09-27 17:54:06 +02:00
Lukasz Jobczyk
e00a5e0a1e Release main copy engine from device and sub device 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-09-20 17:12:34 +02:00
Milczarek, Slawomir
0192e8038f Check for GPU hang in path with wait for timestamps 
Related-To: NEO-6868

Signed-off-by: Milczarek, Slawomir <slawomir.milczarek@intel.com>
 2022-09-19 15:01:46 +02:00
Lukasz Jobczyk
f65d2aeb87 Split copy along single dimension 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-09-08 12:39:06 +02:00
Mateusz Hoppe
6ad16c9e72 Missing support for OCL debugging with L0 
- make resident debugSurface
- allocate debugSurface with correct allocation type and size
- notify cmdQ create/destroy

Related-To: NEO-7075

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
 2022-09-07 13:30:05 +02:00
Lukasz Jobczyk
a9b3b8137f Release locks before wait on blocking BCS split calls 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-09-05 11:20:23 +02:00
Lukasz Jobczyk
0d6bef0753 Add BCS split to api specific config 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-09-05 11:13:10 +02:00
Compute-Runtime-Validation
20f49481f2 Revert "Enable BCS split WA in OCL" 
This reverts commit d672920121.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
 2022-09-03 20:25:10 +02:00
Lukasz Jobczyk
d672920121 Enable BCS split WA in OCL 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-09-01 16:26:00 +02:00
Zbigniew Zdanowicz
c3f7e40a8d Rename special pipeline select mode to systolic 
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-08-31 22:16:26 +02:00
Lukasz Jobczyk
399758ef17 Change default engines for BCS split 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-08-31 21:34:36 +02:00
Lukasz Jobczyk
a3dedcc7ee Add minimal transfer size for BCS split 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-08-31 21:28:40 +02:00
Dunajski, Bartosz
595cfebaef Refactor PIPE_CONTROL programming 
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
 2022-08-23 13:55:25 +02:00
Lukasz Jobczyk
82e29fd048 Add bcs split control mask 
Introduce debug variable to control which engines
the tranfser will be split into

Related-To: NEO-7173

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-08-19 13:21:10 +02:00
Krystian Chmielewski
18adbed233 feat(zebin): add thread scheduling mode support 
Resolves: NEO-7197

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
 2022-08-18 16:10:18 +02:00
Lukasz Jobczyk
b10b3ed9dd Add initial enqueue bcs split infrastructure 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-08-18 15:02:08 +02:00
Katarzyna Cencelewska
fea9c9aca7 Add test to detect potential race 
also add lock inside initialGpgpu

Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-08-12 09:42:58 +02:00
Rafal Maziejuk
ed0c36117e Apply heuristics when setting TG dispatch size on XE_HPC_CORE 
The default TG dispatch size can be changed
to a better value based on number of threads in TG or
currently available amount of threads on GPU.
Decision on what TG dispatch size should be are based on
implemented heuristics.

Signed-off-by: Rafal Maziejuk <rafal.maziejuk@intel.com>
Related-To: NEO-6989
 2022-08-08 16:43:10 +02:00
Compute-Runtime-Validation
b38c750cc4 Revert "Add finish before command queue is released" 
This reverts commit 50fae92ea2.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
 2022-08-05 19:25:23 +02:00
Cencelewska, Katarzyna
61510e9a92 Revert optimization of gpgpu csr's mutex lock in the enqueue blit 
optimization available under flag
ForceCsrLockInBcsEnqueueOnlyForGpgpuSubmission

Related-To: NEO-7011
Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
 2022-08-04 16:42:50 +02:00
Dunajski, Bartosz
a3903c385e Remove HW types from synchronization interface 
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
 2022-07-25 13:59:26 +02:00
Kamil Kopryk
d4d54f5093 Cleanup includes 
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
 2022-07-25 09:58:38 +02:00
Dominik Dabek
e88b24eb37 Remove flush on svmmap 
No longer needed, problem was with reusing mem obj allocation.

Related-To: NEO-6948

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2022-07-20 17:26:15 +02:00
Mateusz Hoppe
50fae92ea2 Add finish before command queue is released 
Related-To: NEO-5279

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
 2022-07-20 12:16:27 +02:00
Zbigniew Zdanowicz
3f8c19eec9 Limit system memory flag in builtin kernels to destination argument 
Related-To: NEO-6959

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-07-19 10:58:51 +02:00
Bartosz Dunajski
52b00a11b0 Remove LSH from CommandQueue 
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2022-07-19 08:47:02 +02:00
Zbigniew Zdanowicz
82a6f9e7b2 Use compute walker system fence for system memory or events in use 
Related-To: NEO-6959

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-07-13 11:55:32 +02:00
Zbigniew Zdanowicz
e07f9f0698 Add kernel algorithm to check any argument is using system memory 
Related-To: NEO-6959

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-07-12 13:08:49 +02:00
Zbigniew Zdanowicz
461a2eb8c7 Refactor interface to hardware interface 
Related-To: NEO-6959

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-07-05 11:54:33 +02:00
Bartosz Dunajski
4fb4a1d77b Add LogicalStateHelper getter for CommandQueue. 
Refactor Kernel handling

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2022-07-04 13:21:15 +02:00
Lukasz Jobczyk
7a1cac0674 Fix EnableCmdQRoundRobindEngineAssign flag 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-06-27 11:38:52 +02:00
Dominik Dabek
39dfaf64bc Fix: Flush on svmMap on multi device 2/n 
Previous fix was causing the runtime to get buffer size
without gfx allocation, causing a seg fault.

This commit moves the fix logic to enqueue handler,
only changing the enqueueProperties.

Related-To: NEO-6948

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2022-06-21 11:38:28 +02:00
Maciej Plewka
6ab6e1abff Fix mutex order for event task and move args to gpu 
This commit fixes problem with untransfered shared usm memory to gpu
when there is submit to gpu trigerred by user event. Also there is a fix
for dead lock problem caused by mixed orders of locking mutexes in csr
and in direct submission controller.

Related-To: NEO-6762

Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
 2022-06-21 11:28:25 +02:00
Lukasz Jobczyk
f98c6b1a8b Disable round robin engine assign on PVC 
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
 2022-06-20 15:24:18 +02:00
Dominik Dabek
ae61f0368a Fix: Flush on svmMap on multi device 
On multi device contexts we allocate svm in system memory.

This caused svmMap calls to assume no copy is needed and
no cache flush was sent to gpu and data modified by earlier
gpu commands was not being visible on host.

This change will add pipecontrol with dcFlush on svmMap
calls when task count from wait is enabled.

Related-To: NEO-6948

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2022-06-20 12:24:43 +02:00
Bartosz Dunajski
939d109362 Add LogicalStateHelper class 
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
 2022-06-14 16:57:16 +02:00
Maciej Plewka
213dc2fe24 Make CPU copy for read buffer when host ptr is write combined on DG2 
With this commit on DG2 32bit driver will check if passed host ptr for
clEnqueueReadBuffer is write combined memory. If check will be true copy
will be make on CPU.

Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
 2022-06-13 21:23:21 +02:00
Dominik Dabek
0a3c960d61 Enable update task count from wait on DG2 
Related-To: NEO-6948

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
 2022-06-06 11:41:27 +02:00
Zbigniew Zdanowicz
8431234845 Change interface to method programing additional fields of command 
Related-To: NEO-6959

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
 2022-05-26 21:32:59 +02:00
Katarzyna Cencelewska
8ff34fea29 Fix typo 
-engineGroupTyp to engineGroupType
-remove gap between commits to use same desc.ordinal in createCommandQueue

Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-05-24 12:15:12 +02:00
Sebastian Luzynski
b8cf0c757a Notify gtpin onCommandBufferComplete 
Notify gtpin onContextDestroy before SVM Allocations are deleted.

Resolves: NEO-6985

Signed-off-by: Sebastian Luzynski <sebastian.jozef.luzynski@intel.com>
 2022-05-20 16:42:13 +02:00
Michal Mrozek
cbd73d2d9c Do not switch to batched mode when update task count form wait is enabled. 
There is not need to enable batching in this case as we do not have
synchronization points between enqueues.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
 2022-05-19 13:16:20 +02:00
Artur Harasimiuk
3f04769f07 style: configure readability-identifier-naming.FunctionCase 
Signed-off-by: Artur Harasimiuk <artur.harasimiuk@intel.com>
 2022-05-17 20:55:56 +02:00
Katarzyna Cencelewska
96e1eb7467 Move variables baseDieRev and baseDieA0Masked from xe_hpc to pvc 
Pvc specific variables should be located in pvc struct

Related-To: NEO-6738
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
 2022-05-17 12:19:16 +02:00
Patryk Wrobel
9b2ad0c5df Detect GPU hangs in flushBcsTask() 
This change introduces detection of GPU hangs in flushBcsTask()
function. The new code has been covered with ULTs.

Related-To: NEO-6681
Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
 2022-05-12 16:17:16 +02:00