Commit Graph

395 Commits

Author SHA1 Message Date
Lukasz Jobczyk
18bfc3be8b Revert "Extend ZE_AFFINITY_MASK to OCL"
This reverts commit 452b53fb4d02ebf4205018b9fd982422e2d85f30.

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2021-01-21 10:31:40 +01:00
Daria Hinz
5a05d7643c Correct event reset
Signed-off-by: Daria Hinz <daria.hinz@intel.com>
2021-01-20 19:22:07 +01:00
Lukasz Jobczyk
d9b6280511 Extend ZE_AFFINITY_MASK to OCL
Resolves: NEO-5438

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
2021-01-19 14:10:44 +01:00
Maciej Plewka
24a88d22aa Fix scratch in bindless mode
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2021-01-18 15:46:18 +01:00
Jaime Arteaga
05b5ad37ea Initialize kernel private surface when kernel is created
This instead of when the associated module is created, to avoid
allocating memory for kernels that are never created nor used.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-13 17:22:40 +01:00
Jaime Arteaga
5c9d43e618 Revert "Remove unnecessary flush in event profiling"
This reverts commit a0db607083.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-10 22:59:41 +01:00
Jaime Arteaga
08655a315c Revert "Initialize kernel private surface when kernel is created"
This reverts commit be2a87fe98.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-10 22:56:32 +01:00
Jaime Arteaga
26b036ab97 Improve zeCommandListAppendMemoryFill Performance (2)
Add missing kernel for remainder kernel when pattern size is 1.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-09 08:41:17 +01:00
Jaime Arteaga
a0db607083 Remove unnecessary flush in event profiling
Remove flush before appending appendWriteKernelTimestamp, since
the associated PC is only to ensure kernel execution has completed
in the device, so flush to the host is not required.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-09 02:40:45 +01:00
Jaime Arteaga
be2a87fe98 Initialize kernel private surface when kernel is created
This instead of when the associated module is created, to avoid
allocating memory for kernels that are never created nor used.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-08 19:22:17 +01:00
Mateusz Hoppe
6f69bd3560 Initialize SIP kernel in Device::create()
Related-To: NEO-4878

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2021-01-08 15:46:23 +01:00
Slawomir Milczarek
55f3c8f134 Add resolve capability for compressed USM device allocations
Related-To: NEO-5107

Signed-off-by: Slawomir Milczarek <slawomir.milczarek@intel.com>
2021-01-08 12:16:26 +01:00
Jaime Arteaga
479d01c118 Improve zeCommandListAppendMemoryFill Performance
Improve L0 fill operations by copying the pattern using
two kernels: one that copies four bytes at a time, and one
that takes care of the remainder. Additionally, a new
allocation is created to fill up at least a cacheline.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-08 07:24:55 +01:00
Daria Hinz
4591101541 Implementation of profiling for copy/fill API's
Signed-off-by: Daria Hinz <daria.hinz@intel.com>
2021-01-07 09:43:00 +01:00
vinod krishna
7d29a0342a maxHardwareContexts now reports 64K
Related-To: NEO-1582

Signed-off-by: vinod krishna <vinod.krishna@intel.com>
2021-01-06 07:04:55 +01:00
Maciej Plewka
3ca77a6cbe Program sba for global bindless heaps
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2021-01-04 14:23:47 +01:00
Jaime Arteaga
d7ea713c5f Revert "Initialize kernel immutable data when kernel is created"
This reverts commit a6ac10088c.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2021-01-04 11:11:39 +01:00
Young Jin Yoon
e09ac446c4 Mask bit 0 of timestamp for event profiling
Related-to: LOCI-1161
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>
2020-12-31 23:51:12 +01:00
Kamil Diedrich
e0da0e1868 Fix for block load and store
Signed-off-by: Kamil Diedrich <kamil.diedrich@intel.com>
2020-12-31 12:28:26 +01:00
Mateusz Jablonski
1b7d7afc07 Refactor USM properties
store reference to std of root device indices and device bitfields
store NEO::Device in USM properties

Related-To: NEO-3691
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
2020-12-30 11:36:56 +01:00
Young Jin Yoon
06dad67c5a Add a return statement for getCsrForLowPriority
Related-to: LOCI-1124
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>
2020-12-29 21:34:05 +01:00
Jaime Arteaga
a6ac10088c Initialize kernel immutable data when kernel is created
This instead of when the associated module is created, to avoid
allocating memory for kernels that are never created nor used.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-12-29 19:29:10 +01:00
Vinod Tipparaju
bbb0e9d095 Fix to Append signal/ wait events for appendMemoryFill & appendMemoryCopyRegion
Signed-off-by: Vinod Tipparaju <vinod.tipparaju@intel.com>
2020-12-29 07:28:50 +01:00
Young Jin Yoon
571007cfe9 Support low priority command queues
Related-to: LOCI-1124
Signed-off-by: Young Jin Yoon <young.jin.yoon@intel.com>
2020-12-24 05:00:34 +01:00
Vinod Tipparaju
5440c83539 Fix to acquire lock during migrating shared allocation via makeResident() call
This fix is required to guarentee thread safety

Signed-off-by: Vinod Tipparaju <vinod.tipparaju@intel.com>
2020-12-24 04:42:08 +01:00
Jitendra Sharma
7c70a14bc4 Implement ZE_ENABLE_PCI_ID_DEVICE_ORDER environment variable
Signed-off-by: Jitendra Sharma <jitendra.sharma@intel.com>
2020-12-23 12:08:43 +01:00
Slawomir Milczarek
6986d5de0b Add helper functions for memory compression to CSR
Signed-off-by: Slawomir Milczarek <slawomir.milczarek@intel.com>
2020-12-21 17:43:03 +01:00
Daria Hinz
1ef9a1c35f Add extended functionality for timestamps at Event
Related-To: NEO-4584
Signed-off-by: Daria Hinz <daria.hinz@intel.com>
2020-12-17 10:53:16 +01:00
Krystian Chmielewski
4948c39d39 Remove executionEnvironment from KernelInfo
Related-To: NEO-3739
2020-12-16 14:56:57 +01:00
Jim Snow
37cd49330c Implement ZE_CACHE_CONFIG_FLAG_LARGE_DATA for zeKernelSetCacheConfig
Signed-off-by: Jim Snow <jim.m.snow@intel.com>
2020-12-16 07:00:13 +01:00
Pawel Wilma
9bd0c69913 Add ClearColor allocation support for blitter
Related-To: NEO-5175

Signed-off-by: Pawel Wilma <pawel.wilma@intel.com>
2020-12-14 18:22:05 +01:00
Zbigniew Zdanowicz
b32b5784c2 Fix calculations for offseted addresses
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2020-12-11 20:25:54 +01:00
Maciej Plewka
830bcd6987 Scratch using bindless heaps
Signed-off-by: Maciej Plewka <maciej.plewka@intel.com>
2020-12-11 09:38:44 +01:00
Jaime Arteaga
d46bb59508 Migrate shared-allocations made resident with makeMemoryResident
Shared-allocations are currently migrated to GPU by the page-fault
manager when calling executeCommandLists. Allocations to migrate are
taken from the lists container. However, if a shared-allocation
has been made resident with zeContextMakeMemoryResident(), it is not
added to the list container, and hence it is not migrated to device.

So, add a container of resident allocations to the driver and migrate
them along with the other allocations.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-12-11 03:27:42 +01:00
Mateusz Hoppe
d3f3730989 Refactor L0 core tests
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2020-12-09 17:51:15 +01:00
Jaime Arteaga
beb3c5ed05 Add support for global work offset extension in L0
Add experimental extension to set global work offest in L0.
Current L0 specification does not have interface to export
experimental function symbols, so for now, applications need
to find the symbol like with dlsym on Linux.

A blackbox test showing functionality is also added.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-12-09 07:33:40 +01:00
Kacper Nowak
4804eca86d Refactoring Level Zero ULTs in order to use standalone spirv builtins
Signed-off-by: Kacper Nowak <kacper.nowak@intel.com>
2020-12-08 09:19:05 +01:00
Zbigniew Zdanowicz
b4a901ac1c Fix aligned address for USM pointers
Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2020-12-07 15:25:04 +01:00
Zbigniew Zdanowicz
19bea962d0 Use imported host pointers in the driver when available
Related-To: NEO-5126

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2020-12-03 13:52:14 +01:00
Jaime Arteaga
be90b9ff93 Add support for ZE_DEVICE_MEM_ALLOC_FLAG_BIAS_UNCACHED
Add support for device and shared allocations that use the
ZE_DEVICE_MEM_ALLOC_FLAG_BIAS_UNCACHED flag, whether the
kernel using the memory is stateless or statefull.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-12-02 10:43:45 +01:00
Jobczyk, Lukasz
343fd602fa Add method to submit kernel on single subdevice
Signed-off-by: Jobczyk, Lukasz <lukasz.jobczyk@intel.com>
2020-12-01 22:15:05 +01:00
macabral
720ba46548 Register kernel Elf for debugging purpose
Signed-off-by: macabral <matias.a.cabral@intel.com>
2020-12-01 17:16:14 +01:00
Zbigniew Zdanowicz
4602220e62 Add manager of imported host pointers
Related-To: NEO-5126

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
2020-11-30 13:28:25 +01:00
Filip Hazubski
6c4b1f951c Do not force cl_khr_3d_image_writes extension when compiling kernels
Update usage of SUPPORTED_IMAGES flag and do not use images when disabled.
Use SUPPORTED_2_0 only on fully OCL 2.1 conformant platforms.

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
2020-11-27 17:20:46 +01:00
Mateusz Hoppe
8cac8d4b51 Fix possible device nullptr dereference
- abort will be called if device is nullptr

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2020-11-26 22:51:44 +01:00
Jaime Arteaga
ba89a47aea Add L0 extension for extended module descriptor
Add ZE_experimental_module_program extension to L0 driver to support
multiple input modules being passed to zeModuleCreate().

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-11-26 22:44:22 +01:00
Jaime Arteaga
9d632b9c92 Use C++ numeric limits instead of C macro
Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-11-25 12:01:07 +01:00
Mateusz Hoppe
8aacad1854 Create DebuggerL0 only when debugging is supported
Related-To: NEO-5239

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
2020-11-23 18:28:45 +01:00
Kamil Diedrich
626b71e379 Disable Y16 layout
Signed-off-by: Kamil Diedrich <kamil.diedrich@intel.com>
2020-11-23 16:34:55 +01:00
Jaime Arteaga
4432547ff5 Add support for importing memory handles through zeMemAllocDevice
Add support for passing ze_external_memory_import_fd_t extended
structure to zeMemAllocDevice() to allocate a device memory
out of an fd handle.

Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
2020-11-23 04:18:03 +01:00