correct add 1 to the current size, gpuRange as gpuRange
only specifies the end address of the pool, not the actual
size, which causes alignment issue of all the offsets of
allocated objects. Also, a page was added in the beginning
of the limited range memory pool to avoid the base address
of it to be 0x0 that is interpreted as invalid address by
heap allocator (This makes the size reduced by pageSize)
Internal 32bit allocator is also initialized in proper way
with corrected base address.
v2: added 'givenMemoryManagerLimimedRangeAllocator' unit
test
v3: adjust size to be freed when DrmMemoryManager instance
is destroyed to 4GB
v4: - defined external 32bit allocator for limited Range
allocation case.
- softpinning object on the correct GPU address
Change-Id: Idaa0206d4133a1476cceb5a48ff8c8528742c76a
Signed-off-by: dongwonk <dongwon.kim@intel.com>
correct mapping of cpu and gpu address in memory allocation
in case of NonSVM. Also, used only aligned address since offset
is already calculated and written to "allocationOffset".
gpuBaseAddress is programmed with 0 instead of base address of
heap because it represents GPU's address space.
v2: add allocationOffset to the aligned address in allocation
data to point to exact starting address of buffer in two
NonSVM allocation unit tests
Change-Id: I32ef512de64a13459b7c132672f837c5cb210ada
Signed-off-by: dongwonk <dongwon.kim@intel.com>
checks the address + size of buffer object to determine
whether the object is located within 32bit address space
boundary.
v2: changed end year to 2019 in ther license term
v3: added unit test for checking of flag when size of
bo is given.
v4: two different unit tests are created to cover two
different case separately
Change-Id: Ie2df6025fc116aca679dcfe88d858ff240278c39
Signed-off-by: dongwonk <dongwon.kim@intel.com>
The internal 32bit allocator sometimes need CPU address to access
or store data when it is in reduced address space scenario.
Change-Id: I6c0b3f9703ae3e124249b41ad7d81f03ad93ad17
Signed-off-by: Kai Chen <kai.chen@intel.com>
- fill AllocationData in one place
- remove allocateGraphicsMemoryForSVM function
- refactor SVM manager tests
Change-Id: I6f4ecd70503da8031cced50ea98a54162fd8e5d3
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
- remove method allocateGraphicsMemory(size_t size)
- pass allocation type in allocation properties
- set allocation type in allocateGraphicsMemoryInPreferredPool
Change-Id: Ia9296d0ab2f711b7d78aff615cb56b3a246b60ec
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
- make AllocationData a protected structure
- use AllocationProperties instead of AllocationFlags
- refactor methods: allocateGraphicsMemory64kb, allocateGraphicsMemoryForSVM
- call AllocateGraphicsMemoryInPreferredPool in AllocateGraphicsMemory
where there is no host ptr
Change-Id: Ie9ca47b1bccacd00f8486e7d1bf6fb3985e5cb12
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
pass struct with properties to allocate graphics memory methods:
for protected methods use AllocationData
for public methods use AllocationProperties
Change-Id: Ie1c3cb6b5e330bc4adac2ca8b0bf02d30ec76065
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
Code implementation of GPU address allocator for devices with reduced
address space.
Change-Id: Ieb0412c5930fdd71f90741055cf89c0338b01133
Signed-off-by: Kai Chen <kai.chen@intel.com>
- Residency is being controlled by Graphics Allocation.
- Duplicates are now eliminated only for shared resources.
Change-Id: Ib51e2739a07728ae0b94abf6cce2e9981b017111
- add EnableLocalMemory debug variable
- separate OSInterface::osEnableLocalMemory for dll and unit tests
Change-Id: I78a1f60364eece28b30ce3e91418e7d72ba3e0d9
Piotr Fusik