If flag "LogUsmReuse" is set, usm reuse will log operations to csv file.
Each line will contain: timestamp, host/device, operation type,
allocation size, true/false whether operation succeeded.
This data can then be used to produce graphs and help in analyzing
usm reuse in a particular workload.
Related-To: NEO-6893
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
When is2MBLocalMemAlignmentEnabled returns true,
increase pool size for builtins from 64k to 2MB.
Additionally, set appropriate alignment for kernel ISA heap allocations.
Additionally, configure isaAllocationPageSize based on productHelper
Related-To: NEO-12287
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
- Updated `isAllocTbxFaultable` to exclude `gpuTimestampDeviceBuffer` from being
faultable.
- Replaced `SpinLock` with `RecursiveSpinLock` in `CpuPageFaultManager` and
`TbxPageFaultManager` to allow recursive locking.
- Added unit tests to verify the correct handling of `gpuTimestampDeviceBuffer`
in `TbxCommandStreamTests`.
Related-To: NEO-13748
Signed-off-by: Jack Myers <jack.myers@intel.com>
Related-To: NEO-13574
User can read image with row pitch larger than region width.
In such cases, don't override memory which is out of region
but within row pitch.
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
- Updated `isAllocTbxFaultable` to exclude `gpuTimestampDeviceBuffer` from being
faultable.
- Replaced `SpinLock` with `RecursiveSpinLock` in `CpuPageFaultManager` and
`TbxPageFaultManager` to allow recursive locking.
- Added unit tests to verify the correct handling of `gpuTimestampDeviceBuffer`
in `TbxCommandStreamTests`.
Related-To: NEO-13748
Signed-off-by: Jack Myers <jack.myers@intel.com>
Refactor buffer pool allocator to support configurable
SmallBuffersParams based on product helper capabilities.
This patch enables setting custom pool
parameters instead of using fixed static values.
For devices with 2MB local memory alignment enabled
(is2MBLocalMemAlignmentEnabled),
use larger pool configuration:
- Pool size: 16MB (up from 2MB)
- Threshold: 2MB (up from 1MB)
- Alignment: 64KB (unchanged)
- Starting offset: 64KB (unchanged)
This improves memory utilization for devices supporting larger memory
alignments
while maintaining original parameters for other devices.
Key changes:
- Moved params from static template to instance member
- Added SmallBuffersParams struct with default/large configs
- Added constructor and setter methods for params configuration
Related-To: NEO-12287
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
Related-To: NEO-13529
Also, add size threshold on iGPU on Linux,
and disable staging if imported host ptr could
be reused
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
Adding helpers for find first greater equal
value in lookup array. Useful when finding
minimum SLM encoding that satisfies requirements.
Related-To: NEO-12747
Signed-off-by: Chodor, Jaroslaw <jaroslaw.chodor@intel.com>
restored the possibility of different writing modes in the FileLogger
mocked writeToFile in TestFileLogger
Related-To: NEO-7006
Signed-off-by: Marcel Skierkowski <marcel.skierkowski@intel.com>
Track amount of created buffer pools per device. Do not allocate extra
pools if limit is reached. New contexts will have pooling disabled if
limit is reached on device.
Related-To: NEO-13461
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
reducing the number of tests that have interactions with filesystem.
writeDataToFile() saves filename and content in std::map.
fileExistsHasSize() checks if file was previously written to virtualFileSystem
loadDataFromVirtualFile() fetches data from std::map based on filename
Related-To: NEO-7006
Signed-off-by: Marcel Skierkowski <marcel.skierkowski@intel.com>
getBaseAddress was incorrectly returning pLeftBound which changes after
memory allocation.
Added baseAddress field to store and return initial address value.
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
getBaseAddress was incorrectly returning pLeftBound which changes after
memory allocation.
Added baseAddress field to store and return initial address value.
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
This PR handles the situation in which a component
has reserved a front window space for itself in the external heap,
so that the Compute Runtime cannot access this area.
In such a situation, we perform the following steps:
1. reserve 4GB chunk in heapStandard
2. split our chunk into 2 parts: heapFrontWindow, heapRegular
3. from this point on, map all linearStream allocations in reserved 4GB
chunk
Patch applies to Windows and WSL.
Patch only applies when the bindless global allocator is enabled.
Related-To: HSD-16025889919
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
- freed pointer cannot overlap with freed chunks, therefore there is no way
to merge it with freed chunks if they overlap
Signed-off-by: Mrozek, Michal <michal.mrozek@intel.com>
