Closed 8tab closed 3 years ago
Apologies for the delay in responding to this. I needed to get confirmation from our driver team on some of the details. I've updated the spec to include this in ze_device_properties_t
. You should see this in the next version. Thank you for your contribution.
Is it already published? If yes, where should I look for it? I couldn't find it here https://spec.oneapi.com/level-zero/latest/core/api.html?highlight=ze_device_properties_t#_CPPv422ze_device_properties_t nor here https://spec.oneapi.com/level-zero/latest/core/api.html?highlight=ze_device_memory_properties_t#_CPPv429ze_device_memory_properties_t.
it has not been published yet. we are working on getting the latest spec published asap.
The current 1.1.2 spec has a uint64_t maxMemAllocSize field in the ze_device_properties_t. https://spec.oneapi.com/level-zero/latest/core/api.html#_CPPv4N22ze_device_properties_t15maxMemAllocSizeE
Can we close this issue?
Yes, thank you
Is it possible to expose maxMemAllocSize via ze_device_memory_properties_t? Currently, level_zero implementation for Intel GPU uses this property to check if size is valid or not (e.g. https://github.com/intel/compute-runtime/blob/master/level_zero/core/source/memory/memory.cpp#L123) and it seems that the user doesn't have a way to know what its value is (besides looking at compute-runtime source code).