Commit Graph

231 Commits

Author SHA1 Message Date
Bartosz Dunajski b1dea19fbd refactor: move tag initialization to allocator [1/n]
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-12-17 17:53:13 +01:00
Mateusz Jablonski 552930a75f fix: don't setup preemption surface when debugger is active
Related-To: NEO-12878
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-10-08 13:58:11 +02:00
Slawomir Milczarek edeb7bdd4b refactor: Allocate copy source for work partition allocation on heap
No perforamce impact expected since it is initialized once only,
but has the advantage of using custom allocator by overriding malloc.

Related-To: NEO-12846

Signed-off-by: Slawomir Milczarek <slawomir.milczarek@intel.com>
2024-10-01 13:32:55 +02:00
Bartosz Dunajski 5b1bd4b088 refactor: dont mix aub and hw wait prints
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-09-26 08:51:23 +02:00
Kamil Kopryk ec5beaf616 refactor: reduce csr class size
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2024-09-18 13:33:55 +02:00
Lukasz Jobczyk a54a3bf624 performance: Optimize heap handling when mitigate dc flush
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-09-06 04:33:41 +02:00
Compute-Runtime-Validation dc84b163b5 Revert "performance: Optimize heap handling when mitigate dc flush"
This reverts commit 9249c5c65c.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-09-03 08:33:20 +02:00
Szymon Morek e6abfafa16 fix: drain paging fence queue before waiting for resources
Related-To: NEO-12197

If ULLS controller waits for CSR lock, and driver must
wait for resources due to OOM, then draing paging fence queue
directly

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2024-09-03 07:45:25 +02:00
Lukasz Jobczyk 9249c5c65c performance: Optimize heap handling when mitigate dc flush
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-09-02 18:12:19 +02:00
Compute-Runtime-Validation 63528e70a7 Revert "performance: Optimize heap handling when mitigate dc flush"
This reverts commit 1a8149e91c.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-08-30 05:59:25 +02:00
Lukasz Jobczyk 1a8149e91c performance: Optimize heap handling when mitigate dc flush
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-08-29 17:03:55 +02:00
Szymon Morek b8f181d50e performance: remove trim candidate list
Related-To: NEO-11755

Removing trim candidate list reduces overhead
caused by residency handling. Allocations required
for eviction are placed in eviction container managed
by CSR.

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2024-08-23 12:21:50 +02:00
Bartosz Dunajski 696b02bfd3 fix: improve TBX downloading after L0 Event sync
Related-To: HSD-18038498579

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-08-23 10:42:17 +02:00
Mateusz Jablonski 7ac41615cd fix: create thread with function pointer
don't create async thread in neo shared tests

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-08-21 18:02:37 +02:00
Szymon Morek d4c1631ac7 performance: don't wait for paging fence on user thread
Related-To: NEO-12197

Currently for new resources user thread must wait before submitting
actual workload. With this commit, instead of waiting on user thread,
request is sent to background ULLS controller thread and additional
semaphore is programmed. ULLS controller will perform actual wait
and signal semaphore when paging fence reaches required value.

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2024-08-07 08:30:51 +02:00
Bartosz Dunajski 24cfd203ab fix: dont download tbx allocations on heapless first device submission
Related-To: HSD-18039476929

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-08-06 14:03:42 +02:00
Bartosz Dunajski ec34656e0e fix: debug flag to defer first device submission
Related-To: HSD-18039343751

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-08-02 09:31:25 +02:00
Michal Mrozek 5dc01e5764 Revert "performance: remove not needed logic"
This reverts commit 0ec4e9333d.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-07-30 07:21:40 +02:00
Michal Mrozek 0ec4e9333d performance: remove not needed logic
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-07-29 19:11:19 +02:00
Michal Mrozek fdf9916cab performance: do not execute same code twice
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-07-29 17:33:32 +02:00
Bartosz Dunajski c3312f21f7 feature: dont initialize in-order TS nodes
Related-To: NEO-11925

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-07-23 13:31:23 +02:00
Szymon Morek 09cbbc2625 fix: don't evict always resident allocations
Related-To: NEO-12008

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2024-07-15 11:39:00 +02:00
Dominik Dabek 296d2bb148 performance: adjust ulls timeout on power saving
Use l0 power saving setting to also adjust ulls timeout.

Related-To: NEO-10800

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2024-06-17 16:51:55 +02:00
Bartosz Dunajski da9c009b88 feature: assign unique interrupt to queue
Related-To: NEO-8179

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2024-06-07 10:06:31 +02:00
Mrozek, Michal 38d7e02c28 refactor: remove not needed code
Signed-off-by: Mrozek, Michal <michal.mrozek@intel.com>
2024-05-29 13:24:02 +02:00
Szymon Morek 2c2ff5e369 fix: update task count when object is always resident
Related-To: NEO-11537

Signed-off-by: Szymon Morek <szymon.morek@intel.com>
2024-05-27 15:18:00 +02:00
Mateusz Hoppe d35d8727e5 fix: use primaryCsr allocations when csr has primaryCsr set
- global fence allocation
- global stateless heap allocation
- preemption allocation
- debug surface allocation

all above are shared from primary csr

Related-To: NEO-7824


Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-05-07 02:41:24 +02:00
Lukasz Jobczyk c1004b77bf performance: Limit tlb flush in state cache flush to wddm
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2024-05-06 12:45:26 +02:00
Mateusz Jablonski cb2b572e94 feature: add support for null aub mode
In this mode AUB csr will be created, however, no aub file will be created

Related-To: NEO-11097
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-04-09 16:59:42 +02:00
Dominik Dabek 2b964254d6 performance: debug key for adjust ULLS on battery
ULLS controller timeout settings will be adjusted based on ac line
status and lowest queue throttle from submissions.

Lowest queue throttle is reset when controller stops ULLS.

Related-To: NEO-10800

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2024-03-22 14:24:00 +01:00
Mrozek, Michal 10313b7b84 refactor: remove not needed code
Signed-off-by: Mrozek, Michal <michal.mrozek@intel.com>
2024-03-07 18:50:16 +01:00
Mateusz Jablonski a697a3f718 refactor: create new members for storing spill and private memory in scratch
rename private scratch space into scratch space slot 1 as it can be generic

Related-To: NEO-9944
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-01-23 12:42:25 +01:00
Compute-Runtime-Validation f9f9035b95 Revert "refactor: create new members for storing spill and private memory in ...
This reverts commit 87eb5f554a.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-01-23 09:13:00 +01:00
Mateusz Jablonski 87eb5f554a refactor: create new members for storing spill and private memory in scratch
rename private scratch space into scratch space slot 1 as it can be generic

Related-To: NEO-9944
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-01-22 19:48:48 +01:00
Mateusz Jablonski a673347378 fix: don't allocate preemption surface for BCS engines
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-01-19 07:59:24 +01:00
Mateusz Jablonski dd1b9d6abc refactor: correct naming of enum class constants 8/n
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-12-19 08:18:18 +01:00
Mateusz Jablonski 739d181026 refactor: correct naming of enum class constants 6/n
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-12-13 14:48:52 +01:00
Mateusz Jablonski b182917d9d refactor: correct naming of allocation types
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-12-11 16:23:37 +01:00
Mateusz Jablonski e57d372608 refactor: correct naming of enum class constants
EngineGroupType
BlitDirection
PostBlitMode
WaitStatus

Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-12-01 11:53:51 +01:00
Mateusz Jablonski c9664e6bad refactor: rename global debug manager to debugManager
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2023-11-30 13:00:59 +01:00
Kamil Kopryk ce7298d512 feature: Add heapless mode programming in ocl
Related-To: NEO-7621
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
2023-11-24 12:53:39 +01:00
Dominik Dabek 6562828095 performance: prealloc internal heap on mtl
Preallocate 1 internal heap allocation per csr on mtl

Related-To: NEO-8152

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2023-11-17 13:36:21 +01:00
Dominik Dabek 7a6fc209dd performance: prealloc cmdbuffer on mtl
Preallocate 2 command buffers allocations per command queue initialized
on MTL.

Related-To: NEO-8152

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2023-11-08 17:21:13 +01:00
Michal Mrozek 80e3632c41 performance: do not free deferred allocations when not needed
Level Zero doesn't use deferred allocations so no point in paying the
price to check for them in cleanAllocationList.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2023-10-27 18:02:43 +02:00
Dominik Dabek 39cf653959 performance(ocl): cmd buffer prealloc per cmdqueue
Add mechanism to preallocate cmd buffer allocations in command stream
receiver reusable allocations list per command queue initialized.

This should limit additional allocations during hot loop.

Needs to be enabled in subsequent commits by setting product helper
method.

Related-To: NEO-8152

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2023-10-27 16:56:29 +02:00
Dunajski, Bartosz 80d0c74605 fix: track registered CSR clients 2
Related-To: NEO-8884

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-09-25 09:35:55 +02:00
Compute-Runtime-Validation ade538ce54 Revert "fix: track registered CSR clients"
This reverts commit 53f635e392.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2023-09-24 10:07:26 +02:00
Dunajski, Bartosz 53f635e392 fix: track registered CSR clients
In L0 its not possible to track objects relations. For example CmdList
may be removed before Event.
In such case, Event needs to safely skip unregister call, without
accessing CmdList/CmdQueue object.

Related-To: NEO-8884

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-09-22 09:42:55 +02:00
Dunajski, Bartosz 7562842a58 refactor: remove LogicalStateHelper
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2023-09-13 10:29:53 +02:00
Maciej Plewka 3b3e17e738 performance: Use vector for private allocs to reuse
Related-To: HSD-18033105655, HSD-18033153203

Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2023-09-04 13:34:38 +02:00