mirror of
https://github.com/intel/compute-runtime.git
synced 2026-01-12 17:33:00 +08:00
Allocate all event pools in local memory
Currently event pools are allocated in system memory if HOST_VISIBLE flag is being set. With this change such event pool will be allocated in device memory if DC flush is not allowed. Related-To: NEO-7302 Signed-off-by: Szymon Morek <szymon.morek@intel.com>
This commit is contained in:
committed by
Compute-Runtime-Automation
parent
50f9564a18
commit
3bb12ad10a
@@ -76,6 +76,8 @@ ze_result_t EventPoolImp::initialize(DriverHandle *driver, Context *context, uin
|
||||
|
||||
auto &hwHelper = devices[0]->getHwHelper();
|
||||
|
||||
useDeviceAlloc |= !NEO::HwInfoConfig::get(devices[0]->getHwInfo().platform.eProductFamily)->isDcFlushAllowed();
|
||||
|
||||
eventAlignment = static_cast<uint32_t>(hwHelper.getTimestampPacketAllocatorAlignment());
|
||||
eventSize = static_cast<uint32_t>(alignUp(EventPacketsCount::eventPackets * hwHelper.getSingleTimestampPacketSize(), eventAlignment));
|
||||
|
||||
|
||||
Reference in New Issue
Block a user