Commit Graph

2488 Commits

Author SHA1 Message Date
Michal Mrozek
e08668b982 refactor: remove not needed code.
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-03-07 13:02:29 +01:00
Zbigniew Zdanowicz
fe377b6597 feature: add out cmd pointer to store data imm encoder
Related-To: NEO-10385

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-07 12:52:08 +01:00
Mateusz Jablonski
a22cefdaed fix: don't program dummy blit when no copy command was programmed
Related-To: NEO-9996, NEO-10571
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-03-07 09:01:16 +01:00
Zbigniew Zdanowicz
781ef3a113 feature: update command list update capability
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-06 19:49:59 +01:00
Wenbin Lu
a0faad6558 feature: report multi-hop fabric connections
In additon to physical connections we should also report multi-hop
logical connections (MDFI + XeLink) as have positive bandwidth.

Use a modified BFS algorithm to try to find a path between fabric
vertices that are not directly connected together because the KMD always
try to use MDFI link first, then go to XeLink.

Multi-hop connections are bi-directional but might not be symmetric, so
for every pair of vertices A & B that are not directly connected, we
need to try to find both `A -> B` and `B -> A`.

Related-To: GSD-7126

Signed-off-by: Wenbin Lu <wenbin.lu@intel.com>
2024-03-06 06:36:48 +01:00
Dunajski, Bartosz
7bb23818e6 fix: RelaxedOrdering handling in BcsSplit path
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-03-05 18:06:45 +01:00
Mateusz Jablonski
8b4bca89bc refactor: move ray tracing support capability to release helper
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-03-05 17:30:57 +01:00
Zbigniew Zdanowicz
b37c57c92b feature: add wait command list argument
Related-To: NEO-10065

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-05 15:44:20 +01:00
Mateusz Hoppe
5589e03e2c fix: pass imgInfo by reference
- Related-To: NEO-10352

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-03-05 10:54:59 +01:00
Zbigniew Zdanowicz
2d5d5ec768 fix: correct event host signaling in tbx mode
Related-To: NEO-10065

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-05 08:59:47 +01:00
Dunajski, Bartosz
1d59ffba82 fix: handle in-order allocator destruction
Related-To: NEO-10572

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-03-04 11:16:55 +01:00
Fabian Zwolinski
52430762ac fix: cl_cache L0 env vars + refactor code structure
Added support for new Compiler Cache
environment variables in Level Zero.

Moved
`opencl/source/compiler_interface/default_cache_config.cpp`
`level_zero/core/source/compiler_interface/default_cache_config.cpp`
to shared directory
`source/compiler_interface/default_cache_config.cpp`

Switched enabling cache by default from per OS to per API.
Changed default state of cl_cache in Level Zero to disabled.

Related-To: NEO-10045
Signed-off-by: Fabian Zwolinski <fabian.zwolinski@intel.com>
2024-03-01 17:35:08 +01:00
Dunajski, Bartosz
51ae76a25f refactor: improve handling of in-order atomic signaling
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-03-01 15:59:25 +01:00
Zbigniew Zdanowicz
8840b6d02f feature: add timestamp postsync command list argument
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-01 14:21:54 +01:00
Dunajski, Bartosz
ea2ad550a1 refactor: improve handling duplicated in-order host storage
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-03-01 13:54:28 +01:00
Mateusz Jablonski
80aa55a3cb fix: don't program dummy blit prior to MI_FLUSH_DW without postsync
add missing dummy blits before MI_FLUSH_DW with postsync

Related-To: NEO-9996
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-03-01 12:36:13 +01:00
Dunajski, Bartosz
bbe1043f08 feature: initial support for pooling in-order counter allocations
Related-To: NEO-10507

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-03-01 12:08:02 +01:00
Zbigniew Zdanowicz
4baee71974 refactor: add out cmd argument to semaphore wait encoder
Related-To: NEO-10065

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-03-01 11:43:36 +01:00
Compute-Runtime-Validation
5fce046877 Revert "fix: don't program dummy blit prior to MI_FLUSH_DW without postsync"
This reverts commit d796fb559d.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
2024-03-01 01:43:35 +01:00
Zbigniew Zdanowicz
7d2bed17d9 refactor: change driver and context file names to unique
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-29 16:20:10 +01:00
Zbigniew Zdanowicz
ce216eacf4 refactor: add kernel count setter to level zero event
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-29 15:07:27 +01:00
Zbigniew Zdanowicz
61d1fb9ef6 refactor: unify getter of event packets to wait
Related-To: NEO-10065

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-29 14:43:55 +01:00
Mateusz Jablonski
d796fb559d fix: don't program dummy blit prior to MI_FLUSH_DW without postsync
add missing dummy blits before MI_FLUSH_DW with postsync

Related-To: NEO-9996
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2024-02-29 13:30:09 +01:00
Wenbin Lu
18d263b21c fix: handle smaller-than-actual count of sub-allocations correctly
If the input count is zero or is greater than the number of sub-allocations,
the driver shall update the value with the total number of sub-allocations.

If count is less than the number of sub-allocations available, then driver
shall only retrieve properties for that number of sub-allocations.

Related-To: NEO-9812

Signed-off-by: Wenbin Lu <wenbin.lu@intel.com>
2024-02-29 10:40:50 +01:00
Dunajski, Bartosz
9d2a914b03 fix: unregister csr clients during Event reset
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-29 08:44:20 +01:00
Michal Mrozek
64232ec370 fix: choose proper csr for low priority immediate command lists
Resolves: NEO-10168

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-02-28 12:45:02 +01:00
Mateusz Hoppe
98787b6b71 feature: bindless image extension - ImageView fixes
- set imgInfo fields correctly for image view
- add image view testcase to zello_bindless_kernel test

Related-To: NEO-10352

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-02-27 19:11:20 +01:00
Dunajski, Bartosz
8483a922a7 feature: simplify in-order allocation overflow
Related-To: NEO-10507

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-27 18:25:26 +01:00
Dunajski, Bartosz
fd91543d89 fix: incorrect signaling in-order counter during region copy
Related-To: NEO-7966

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-27 14:11:50 +01:00
Zbigniew Zdanowicz
0bf8e8727e refactor: add output buffer argument to store register to memory encoder
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-27 13:31:55 +01:00
He, Wenju
798c068ed1 feature: copy surface state to ssh in bindless image initialization
The surface state needs to be copied to global ssh either during image
initialization for sycl bindless image that could be passed to kernel
through memory.
Privously the copy is done in KernelImp::setArgImage. However, the
function won't be called for sycl bindless image.

Related-To: NEO-10352

Signed-off-by: He, Wenju <wenju.he@intel.com>
2024-02-27 09:35:52 +01:00
John Falkowski
cebedb50c7 feature: get kernel max group size
Implementing support for ze_kernel_max_group_size_ext_properties_t

Resolves: NEO-9617

Signed-off-by: John Falkowski <john.falkowski@intel.com>
2024-02-27 08:40:55 +01:00
Zbigniew Zdanowicz
0e503b60dc feature: store post sync command in out postsync command pointer
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-26 15:07:52 +01:00
Zbigniew Zdanowicz
e16b5378da refactor: add output list of patched commands to launch parameters
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-26 13:40:35 +01:00
Zbigniew Zdanowicz
36ca1bbe2d feature: store post sync command in out postsync command pointer
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-26 12:37:45 +01:00
Dunajski, Bartosz
1b9f3f11a0 fix: disable CB Event implicit conversion if dcFlush is required
Related-To: NEO-8145

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-26 11:26:38 +01:00
Dunajski, Bartosz
3341de7a0d Revert "fix: disable in-order barrier skipping if dcFlush is required".
This reverts commit 8e75e06574.

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-24 21:09:14 +01:00
Zbigniew Zdanowicz
6011eb147f refactor: add skip event residency flag to compacted event
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-23 15:45:54 +01:00
Dunajski, Bartosz
8e75e06574 fix: disable in-order barrier skipping if dcFlush is required
Related-To: NEO-7966

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-23 12:41:04 +01:00
Michal Mrozek
9f1648b57a fix: use non zero value for event completion.
upon device reset memory gets reset to 0, which can cause
false event completion.

Related-To: NEO-7765
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-02-22 18:50:45 +01:00
Dunajski, Bartosz
000f776741 feature: skip barrier programming for in-order regualr cmdlists
Related-To: NEO-7966

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-22 15:57:44 +01:00
Fabian Zwolinski
6b2281e66c fix: make sure builtins async init is complete when finishing driver init
Related-To: NEO-10045
Signed-off-by: Fabian Zwolinski <fabian.zwolinski@intel.com>
2024-02-20 21:15:57 +01:00
Zbigniew Zdanowicz
db9d0c9a58 fix: remove heap resource leak after command list reset
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-20 14:38:19 +01:00
Michal Mrozek
0edb025cf7 performance: apply padding only for last kernel.
not needed for all kernels, only for last one.
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
2024-02-20 14:30:14 +01:00
Zbigniew Zdanowicz
ca8406f97e refactor: add event residency parameter to command list append kernel call
Related-To: NEO-10064

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2024-02-20 12:13:40 +01:00
Mateusz Hoppe
d9b662a735 fix: ignore ZET_ENABLE_PROGRAM_DEBUGGING when system misconfigured
- if debug not enabled in the system, ignore env var for enabling
debugging

Resolves: NEO-10370

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-02-19 15:54:17 +01:00
Dunajski, Bartosz
f404f3ceb1 refactor: initial support for per platform extension string
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-19 14:47:30 +01:00
Dunajski, Bartosz
8dc3364d56 fix: append paths for standalone CB Event
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
2024-02-16 18:14:31 +01:00
Slawomir Milczarek
1a00d96735 fix: Fixed timeout resolution for zeCommandQueueSynchronize
According to Level Zero specification: timeout [in] if non-zero,
then indicates the maximum time (in nanoseconds) to yield
before returning ZE_RESULT_SUCCESS or ZE_RESULT_NOT_READY

Related-To: NEO-8927

Signed-off-by: Slawomir Milczarek <slawomir.milczarek@intel.com>
2024-02-16 15:39:00 +01:00
Mateusz Hoppe
ce079c62d5 feature: bindless images extension - pitched ptr support
- add support for allocating image from pitched pointer allocated from
device memory
- extend blackbox test with bindless image using USM memory

Related-To: NEO-10352

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2024-02-16 15:30:12 +01:00