Commit Graph

1135 Commits

Author SHA1 Message Date
Dominik Dabek
8d1ad5a4f3 Refactor: use stack vector for root device indices
Stack vector will not cause dynamic allocations in most circumstances
ie. number of root device indices not more than 16

Related-To: NEO-6837

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2022-04-14 14:05:42 +02:00
Neil R Spruit
d6ac8cdd03 Report Unsupported Image Format for invalid types for L0
- If channel type or channel order is invalid for the layout, type, or
swizzle passed by the the user, then return unsupported image format.

Signed-off-by: Neil R Spruit <neil.r.spruit@intel.com>
2022-04-14 00:33:55 +02:00
Jaime Arteaga
3c3dab8fe0 Expose copy engines in parent device with implicit scaling
When using implicit scaling, expose the copy engines from
sub-device 0 in the root device. This to facilitate
programming models of layers above.

Related-To: NEO-6815

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2022-04-14 00:26:14 +02:00
Mateusz Hoppe
13947f3e99 L0 Debugger minor fixes
- Do not make Builtin kernel allocations resident in
Module::initialize()
- Notify debugger with all segment allocations in module
- refactoring: extract logic to dedicated methods

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2022-04-13 14:21:10 +02:00
Brandon Yates
d70b1a2e2a Filter L0 Debugger support by platfom
Related-To: NEO-6678
Signed-off-by: Brandon Yates <brandon.yates@intel.com>
2022-04-13 13:03:40 +02:00
Patryk Wrobel
19dded25ef Detect GPU hang in remaining calls of command queue and list
This change introduces checking of waits status in
CommandQueue and CommandList classes.

Related-To: NEO-6681
Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
2022-04-11 13:31:44 +02:00
Jaime Arteaga
8b88f920ac Use Internal Copy engine from Sub-device for Shared Migrations
This when we have implicit scaling, so fallback to the sub-device
for BCS migration support.

Related-To: LOCI-3014

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2022-04-11 09:51:55 +02:00
Krystian Chmielewski
2c1bfbb5b2 Encode number barriers
When programming number of barriers use BARRIER_SIZE enumeration.
Resolves: NEO-6785

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
2022-04-08 10:32:23 +02:00
Jaime Arteaga
9f79e432bb Clean up driver experimental headers
Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2022-04-08 06:10:25 +02:00
Troels Nielsen
1ee733e191 zeCommandListAppendMemoryFill fails to fill with 2 byte pattern
zeCommandListAppendMemoryFill fails to fill sizes that are not
multiples of four when using a 2 byte pattern.

Fix the problem and add a blackbox test.

Signed-off-by: Troels Nielsen <bn.troels@gmail.com>
2022-04-08 01:08:09 +02:00
Zbigniew Zdanowicz
6f0f15a0b8 Save separate flush call for synchronous command queue
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-07 15:51:17 +02:00
Zbigniew Zdanowicz
dad49bd2ee Fix multi tile race when using signal event from barrier call
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-07 15:23:17 +02:00
Zbigniew Zdanowicz
7d79c64c35 Change event memory layout for multi tile capable platforms
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-07 10:44:14 +02:00
Zbigniew Zdanowicz
fd45ac133d Unify append calls
Related-To: NEO-6242

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-07 10:18:55 +02:00
Mateusz Hoppe
c5e4bec253 Fix ISA allocation transfer
- use MemoryTransferHelper::transferMemoryToAllocation function
to copy ISA to allocation

Related-To: NEO-6862

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2022-04-07 09:57:49 +02:00
Compute-Runtime-Validation
b88e16252e Revert "Change event memory layout for multi tile capable platforms"
This reverts commit 616dd55789.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2022-04-07 07:55:34 +02:00
Zbigniew Zdanowicz
f4407064a4 Refactor store register mem encoder to include partition parameter
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-06 14:00:56 +02:00
Zbigniew Zdanowicz
616dd55789 Change event memory layout for multi tile capable platforms
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-04-06 11:20:57 +02:00
Jobczyk, Lukasz
335bf2fdb2 Update flushed task count when task count write was dispatched
Signed-off-by: Jobczyk, Lukasz <lukasz.jobczyk@intel.com>
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2022-04-05 11:00:51 +02:00
Filip Hazubski
f09fd527f1 Remove redundant using declarations
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2022-04-04 17:10:21 +02:00
Michal Mrozek
9016772351 No dynamic allocations on hot calls [2/n]
- Prevent FULL vector copy.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2022-04-01 16:22:09 +02:00
Mateusz Hoppe
beff0019d1 SBA tracking for single address space
Related-To: NEO-6539


Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2022-04-01 15:24:11 +02:00
Mateusz Hoppe
4374197c9d Fixes for bindless configuration
- enhance ults code, do not inject memory manager
- fix some issues related to bindless global heap

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2022-04-01 15:08:43 +02:00
Lukasz Jobczyk
6fee7c9de1 Unify the appendWaitOnEvents method
Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2022-03-31 14:48:19 +02:00
Dominik Dabek
308f9ccfff Do not allocate dynamic state heap if not needed
Dynamic state heap is only used for sampler data.

Related-To: NEO-6821

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
2022-03-31 12:13:13 +02:00
Bartosz Dunajski
08e3853982 Debug flag to add extra MI_MEM_FENCE for DirectSubmission
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2022-03-30 16:07:25 +02:00
Zbigniew Zdanowicz
a6d99f416d Use context end offset for partitioned events
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-30 14:20:03 +02:00
Mateusz Hoppe
681c09d314 Extend LinearStream with gpuBase
- added methods to set and query gpu base address

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2022-03-29 15:15:53 +02:00
Zbigniew Zdanowicz
9858438121 Limit multiple partition count to compute command lists
Related-To: NEO-6811

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-29 07:29:08 +02:00
Compute-Runtime-Validation
90c6d7d9c9 Revert "Expose copy engines in parent device with implicit scaling"
This reverts commit 6154fb8ea7.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2022-03-28 22:06:10 +02:00
Jaime Arteaga
6154fb8ea7 Expose copy engines in parent device with implicit scaling
When using implicit scaling, expose the copy engines from
sub-device 0 in the root device. This to facilitate
programming models of layers above.

Related-To: NEO-6815

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2022-03-28 16:48:04 +02:00
Dominik Dabek
fccda460d6 Optimize zeCommandListCreate
Do not allocate heap if command list is copy only.

Related-To: NEO-6821

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>

Co-authored-by: Michal Mrozek <michal.mrozek@intel.com>
2022-03-28 16:31:10 +02:00
Konstanty Misiak
174c27eb31 Fix CFEFusedEUDispatch debug flag
Signed-off-by: Konstanty Misiak <konstanty.misiak@intel.com>
2022-03-28 12:32:05 +02:00
Compute-Runtime-Validation
34a5c6cd28 Revert "Remove programming first SCM from the command list"
This reverts commit 9c06af79ad.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2022-03-28 11:15:35 +02:00
Filip Hazubski
586e8510de Remove unneeded include
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2022-03-26 09:55:25 +01:00
Jim Snow
0a926c7d12 Add Windows image memory export functionality
Adds zeImageGetAllocPropertiesExt function, implementation code shared
shared with zeMemGetAllocProperties moved into common helper function.

Related-To: LOCI-2665

Signed-off-by: Jim Snow <jim.m.snow@intel.com>
2022-03-26 05:11:35 +01:00
Jaime Arteaga
e8a6842b7e Add method to read kernel base address
Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2022-03-25 21:49:22 +01:00
Zbigniew Zdanowicz
e4d8dfcdd9 Implement fence creation flag
Related-To: NEO-6242

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-25 12:04:43 +01:00
Zbigniew Zdanowicz
bb753784cc Optimize post sync operations for fence and task count update
Related-To: NEO-6242

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-25 10:57:10 +01:00
Lukasz Jobczyk
a230f267e1 Poll task count indefinitely on high throttle command queue
Resolves: NEO-6781

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2022-03-25 10:06:16 +01:00
Patryk Wrobel
14954acd12 Detect GPU hang in command list immediate
This change introduces checking of values returned
by blocking calls used in cmdlist_hw_immediate.inl.

Signed-off-by: Patryk Wrobel <patryk.wrobel@intel.com>
Related-To: NEO-6681
2022-03-24 12:42:19 +01:00
Krzysztof Gibala
ebc006ad53 Move SBA related WAs logic from CSR to EncodeWA
Signed-off-by: Krzysztof Gibala <krzysztof.gibala@intel.com>
2022-03-24 12:24:56 +01:00
Zbigniew Zdanowicz
814b0ecb87 Improve wait function for Fence object
Related-To: NEO-6242

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-24 10:48:49 +01:00
Kacper Nowak
1b6a907b13 Code refactor
This commit:
- Optimizes usage of std::unordered map
- Corrects variables naming and output messages
- Minor changes in using std::string methods
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
2022-03-23 23:12:21 +01:00
Milczarek, Slawomir
f03f530327 Extend zeCommandListAppendMemoryPrefetch to migrate to associated device
Related-To: NEO-6740

Signed-off-by: Milczarek, Slawomir <slawomir.milczarek@intel.com>
2022-03-23 14:21:17 +01:00
Baj, Tomasz
1f7fd45251 Move device UUID from L0 to shared
Related-To: NEO-5681

Signed-off-by: Baj, Tomasz <tomasz.baj@intel.com>
2022-03-23 12:52:24 +01:00
Jitendra Sharma
f52f3df274 Add platform specific getter of debug surface size
For different platforms based on number of available threads
and debug surface layout, calculate max debug surface size.

Related-To: NEO-6676
Signed-off-by: Jitendra Sharma <jitendra.sharma@intel.com>
2022-03-22 12:18:40 +01:00
Bartosz Dunajski
05a0664720 Change default lod preclamp mode in sampler_state
Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
2022-03-22 11:24:09 +01:00
Filip Hazubski
9c06af79ad Remove programming first SCM from the command list
For first kernel in command list the required state described with SCM
command is stored and programmed when the command list is executed, only
if current stream state requires reprogramming.

Resolves: NEO-6728

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2022-03-21 11:24:07 +01:00
Zbigniew Zdanowicz
46f4472965 Refactor Fence class interface and implementation
Related-To: NEO-6242

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2022-03-21 10:35:26 +01:00