The patch applies to Level Zero.
Only allocations < 2MB will be fetched from the pool.
Allocations are shared and reused within a given device.
Additionally, I added a new debug flag to control the allocator:
EnableTimestampPoolAllocator
Related-To: NEO-12287
Signed-off-by: Fabian Zwoliński <fabian.zwolinski@intel.com>
Host usm and device usm for igfx checks system memory usage.
Device usm for dgfx checks local memory usage.
If used memory is above limit threshold:
- no new allocations will be saved for reuse
- cleaner will use shorter hold time of 2 seconds
- cleaner will free all eligible allocations, regardless of async
deleter thread having work
Motivation: in case of gfx memory being full, making resident new
allocations will require evictions which leads to massive slowdown on
enqueue calls.
This change aims to minimize cases where extra memory usage from usm
reuse mechanism leads to above situation.
Related-To: NEO-6893, NEO-14160
Signed-off-by: Dominik Dabek <dominik.dabek@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>
Calculate available memory for usm device reuse based as (total device
memory - used memory) * fraction for reuse.
Use sys mem allocs for devices without local memory.
Related-To: NEO-12902
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
Related-To: NEO-13340
When regular copy CSR has enabled direct submission,
stop it before migration on internal CSR.
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
Related-To: NEO-13340
When regular copy CSR has enabled direct submission,
stop it before migration on internal CSR.
Signed-off-by: Szymon Morek <szymon.morek@intel.com>
Add tracking of memory used for usm reuse mechanism when multiple cl
contexts are used.
Tracking for device added to NEO::Device, for host added to
NEO::MemoryManager.
This fixes usm reuse using x% of memory per each context instead of
globally.
Related-To: NEO-13308
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
Dedicated pools for different allocations size ranges.
Additional reused allocations will create their own pools.
Do not reuse allocations >256MB.
Related-To: NEO-6893, NEO-12299, NEO-12349
Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
- add support for contect group with HP copy engine
- choose HP copy engine when available
Related-To: NEO-11983
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
- add support for contect group with HP copy engine
- choose HP copy engine when available
Related-To: NEO-11983
Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
Call the new function after debugger and sip kernel are initialized.
Related-To: HSD-18038645398, HSD-18038819112
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
Initialize SIP kernel when shared device is being initialized
instead of api-specific device.
Initialize debugger when shared device is being initialized
instead of during platform or driver initialization.
Add missing makeResident calls for SIP kernel in heapless paths.
Related-To: HSD-18038645398, HSD-18038819112
Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
