Commit Graph

3014 Commits

Author SHA1 Message Date
Raiyan Latif
0370c5afa2 fix: Use unique_ptr for managing externalSemaphoreController
Related-To: NEO-11488

Signed-off-by: Raiyan Latif <raiyan.latif@intel.com>
2025-01-16 19:47:47 +01:00
Jaroslaw Warchulski
4dd1c80362 refactor: use enum class instead of enum for DeviceHierarchyMode
Related-To: GSD-9560
Signed-off-by: Jaroslaw Warchulski <jaroslaw.warchulski@intel.com>
2025-01-16 15:49:44 +01:00
Bartosz Dunajski
310c47cad3 fix: add missing CB event cache flush
Related-To: GSD-10565

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2025-01-16 14:14:51 +01:00
Jaroslaw Warchulski
4cdd5671f7 fix: return correct subDevice in DeviceImp::getActiveDevice()
Resolves: NEO-13759
Signed-off-by: Jaroslaw Warchulski <jaroslaw.warchulski@intel.com>
2025-01-16 00:10:58 +01:00
Krzysztof Sprzaczkowski
a17745532c performance: Move preemption allocation init to the first submission
Related-To: NEO-12323
Signed-off-by: Krzysztof Sprzaczkowski <krzysztof.sprzaczkowski@intel.com>
2025-01-15 20:22:50 +01:00
Lukasz Jobczyk
2dd9940f60 Revert "fix: count active modules for enabling per-dispatch private memory"
This reverts commit a483b361f9.

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2025-01-15 15:03:37 +01:00
Zbigniew Zdanowicz
621ceaf9ec feature: add new properties and getters for kernel slm arguments
- add new vector to read slm offset
- add getter to read new vector
- add getter to read slm argument alignment
- add getter to read slm sizes vector

Related-To: NEO-12933

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2025-01-15 14:23:04 +01:00
Jaroslaw Warchulski
6a7bee67e9 refactor: unify device hierarchy mode
Related-To: GSD-9560
Signed-off-by: Jaroslaw Warchulski <jaroslaw.warchulski@intel.com>
2025-01-14 21:08:44 +01:00
Raiyan Latif
84b8622c16 refactor: Move Ext Semaphore Controller to Driver Handle
Related-To: NEO-11488

Signed-off-by: Raiyan Latif <raiyan.latif@intel.com>
2025-01-14 20:58:35 +01:00
Mateusz Hoppe
60495a5b20 performance: optimize memory used for scratch programming
- reserve exact number of slots for scratch surface states in surface
state heaps
- do not use offsets for contexts depending on engine type executing
cmdlists

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2025-01-14 15:47:08 +01:00
Bartosz Dunajski
26ae830014 fix: add missing dcFlush to signal standalone CB event
Related-To: NEO-13752

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2025-01-14 15:22:33 +01:00
Jack Myers
127af1a715 fix: correct result variable scope in initDriver
The `initDriver` function had an issue where a new `result` variable was defined
in a lower scope, causing the initialization result to be discarded. This commit
removes the redundant variable declaration, ensuring that the `result` variable
initialized in the upper scope is correctly set by the
`Driver::get()->initialize` method and properly reflects the driver
initialization status.

Related-To: NEO-13686
Signed-off-by: Jack Myers <jack.myers@intel.com>
2025-01-14 12:43:35 +01:00
Damian Tomczak
9a149b6da5 refactor: useGlobalHeaps naming unification
Related-to: NEO-12737

Signed-off-by: Damian Tomczak <damian.tomczak@intel.com>
2025-01-14 11:01:07 +01:00
Kulkarni, Ashwin Kumar
bf4d4e71c3 feature: adds support for 3 channel Imageview creation
Related-To: NEO-12887

Signed-off-by: Kulkarni, Ashwin Kumar <ashwin.kumar.kulkarni@intel.com>
2025-01-14 03:19:11 +01:00
Compute-Runtime-Validation
e66fce10b5 Revert "performance: Signal inOrder counter with pipe control, part 6"
This reverts commit fa40c1f2fb.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2025-01-14 03:03:08 +01:00
Lukasz Jobczyk
fa40c1f2fb performance: Signal inOrder counter with pipe control, part 6
When on DC flush platform do not signal in order allocation when using
timestamp CB event. Handle wait on that event and synchronization using
in order timestamp node. Flush DC from wait. In order handled by pipe
controls.

Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2025-01-13 17:40:26 +01:00
Jaroslaw Warchulski
a180afa2e6 fix: respect affinity mask with COMBINED hierarchy in OCL
Related-To: GSD-9560
Signed-off-by: Jaroslaw Warchulski <jaroslaw.warchulski@intel.com>
2025-01-13 15:38:39 +01:00
Wenbin Lu
a483b361f9 fix: count active modules for enabling per-dispatch private memory
Related-To: NEO-13086

Signed-off-by: Wenbin Lu <wenbin.lu@intel.com>
2025-01-10 15:03:34 +01:00
Lukasz Jobczyk
3735ccaed7 performance: Signal inOrder counter with pipe control, part 5
When on DC flush platform resolve in order implicit dependency with pipe
control.

Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2025-01-10 12:33:42 +01:00
Compute-Runtime-Validation
04f1e216a5 Revert "performance: Signal inOrder counter with pipe control, part 5"
This reverts commit 7ee8328c1d.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2025-01-10 06:48:41 +01:00
Szymon Morek
f3c9362fc5 fix: check for gpu hang during wait for ring completion
Related-To: NEO-13490

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2025-01-09 18:44:25 +01:00
Bartosz Dunajski
7c209e1ff0 fix: typo
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2025-01-09 18:06:26 +01:00
Lukasz Jobczyk
7ee8328c1d performance: Signal inOrder counter with pipe control, part 5
When on DC flush platform resolve in order implicit dependency with pipe
control.

Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2025-01-09 13:42:05 +01:00
Jack Myers
7f9fadc314 fix: regression caused by tbx fault mngr
Addresses regressions from the reverted merge
of the tbx fault manager for host memory.

Recursive locking of mutex caused deadlock.

To fix, separate tbx fault data from base
cpu fault data, allowing separate mutexes
for each, eliminating recursive locks on
the same mutex.

By separating, we also help ensure that tbx-related
changes don't affect the original cpu fault manager code
paths.

As an added safe guard preventing critical regressions
and avoiding another auto-revert, the tbx fault manager
is hidden behind a new debug flag which is disabled by default.

Related-To: NEO-12268
Signed-off-by: Jack Myers <jack.myers@intel.com>
2025-01-09 07:48:53 +01:00
Raiyan Latif
b8157a2547 fix: Fix ULTs & signal path in External Semaphore
- Temporarily disables Multi-Threaded ULTs, to be
re-enabled in a separate dedicated test binary

Related-To: NEO-11488

Signed-off-by: Raiyan Latif <raiyan.latif@intel.com>
2025-01-09 07:40:16 +01:00
Mateusz Jablonski
bb1a125f0c feature: add support for Panther Lake platform
Related-To: NEO-12803

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2025-01-07 11:39:56 +01:00
Latif, Raiyan
3122ce963e fix: Fix wait path in Experimental External Semaphore
Related-To: NEO-11488

Signed-off-by: Latif, Raiyan <raiyan.latif@intel.com>
2025-01-06 04:05:59 +01:00
Mateusz Jablonski
5eed889c94 refactor: add add_subdirectories in cmake files
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2025-01-03 16:08:21 +01:00
Filip Hazubski
3997d2a6b3 Revert "feature: Append recorded commandlist into immediate (8/N)"
This reverts commit acfeffe274.

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2025-01-03 15:50:52 +01:00
Mateusz Jablonski
c867a5ed5d refactor: remove redundant flag levelZeroSupported
all platforms are supported by L0

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2025-01-03 11:28:11 +01:00
Bartosz Dunajski
af0d7828ce fix: disable implicit CB conversion for wait events
Related-To: GSD-10533

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2025-01-02 18:00:55 +01:00
Zbigniew Zdanowicz
a744aa07bc refactor: add private property logging in test and kernel class
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2025-01-02 15:26:07 +01:00
Mateusz Jablonski
165c294590 refactor: extract methods to setup SCM state per context
per context properties are now set explicitly

Related-To: NEO-12803, NEO-13632
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2025-01-02 15:19:39 +01:00
Kulkarni, Ashwin Kumar
badbecb126 feature: adds support for 3 channel (2/N)
Allows user to create L0 images with 3 channel

Related-To: NEO-12887

Signed-off-by: Kulkarni, Ashwin Kumar <ashwin.kumar.kulkarni@intel.com>
2024-12-31 20:02:13 +01:00
Compute-Runtime-Validation
c27921767b Revert "performance: Signal inOrder counter with pipe control, part 4"
This reverts commit 332002d9ea.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-12-28 02:45:57 +01:00
Lukasz Jobczyk
332002d9ea performance: Signal inOrder counter with pipe control, part 4
When on DC flush platform signal inOrder counter directly with pipe
control. Obtain profiling from CW. Do not clear in order timestamp node
for CB events without host signal scope when only default CW available
as timestamp is already waited before signaling in order counter.

Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-12-27 17:43:12 +01:00
Bartosz Dunajski
dba1204d9e fix: set ioq info to regular event only if duplicated host storage
Related-To: NEO-13596

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-27 15:48:02 +01:00
Maciej Plewka
c703e0bf6e fix: do not mask lsb for higher part of timestamp
Related-To: NEO-12637
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2024-12-27 13:28:45 +01:00
Lukasz Jobczyk
b5f3b0eba9 performance: Signal inOrder counter with pipe control, part 3
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-12-27 11:52:35 +01:00
Mateusz Hoppe
1c8ee97c24 fix: correctly destroy builtin kernels
- fix ContextGroup test - do not keep neoDevice as unique_ptr

Related-To: NEO-12800

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-12-24 13:38:39 +01:00
Lukasz Jobczyk
83ebbb01d3 performance: Add flag to mitigate host visible signal in CB events
Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-12-24 12:35:55 +01:00
Bartosz Dunajski
fc1dd3b178 fix: ensure regular event completion in in-order mode 3
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-24 12:10:11 +01:00
Mateusz Hoppe
3abd99030a refactor: remove not needed code in BuiltinFunctionsLibImpl
- builtins are released with the object, no need for explicit release of
just first builtin in array

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-12-24 09:58:35 +01:00
Aravind Gopalakrishnan
acfeffe274 feature: Append recorded commandlist into immediate (8/N)
Add primary dispatch capability for immediate command list

Related-To: NEO-10356

Signed-off-by: Aravind Gopalakrishnan <aravind.gopalakrishnan@intel.com>
2024-12-20 18:52:33 +01:00
Bartosz Dunajski
db3f268a97 refactor: reorganize l0 api include directories [2/n]
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-20 17:16:53 +01:00
Bartosz Dunajski
54fb82f689 fix: improve event timestamp completion tracking
Related-To: HSD-18041011542

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-20 16:07:27 +01:00
Lukasz Jobczyk
363f2becd4 performance: Signal inOrder counter with pipe control, part 2
When on DC flush platform signal inOrder counter directly with pipe
control. Obtain profiling from CW. Only for immediate in order cmd
lists.

Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-12-20 14:21:34 +01:00
Bartosz Dunajski
47f74b8a7b refactor: reorganize l0 api include directories [1/n]
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-20 12:59:08 +01:00
Kamil Kopryk
c96ad9dcbf fix: patch 64bit row and slice pitch for builtins in L0 heapless
Related-To: NEO-12149
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2024-12-20 10:55:07 +01:00
Compute-Runtime-Validation
61aa75d6a9 Revert "performance: Signal inOrder counter with pipe control, part 2"
This reverts commit 136e5b4f6c.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-12-20 04:38:46 +01:00