CPU read updates L2 with the latest values and requires
invalidation after, because SDMA doesn't use L2 and data can become
out of sync.
Change-Id: I98d1c91ca78a103fa5409e638f97485d62d5b11e
When OCL ROCr backend performs CL_MEM_COPY_HOST_PTR it may attempt
to have access to amd::Memory object it's currently creating,
but it's not ready yet. The logic creates a temporary dummy object
to perform a copy transfer. The new change will make sure runtime
skips allocation of the same device::Memory object second time.
Change-Id: I14c6a00a3941fdcaa6aea299e9f096e4c3f5cadf
Use barrier packets for every profile marker that gets submitted
and use the completion signal to get GPU ts. This gives most accurate
dispatch time. Club cache flushes with profile marker if there is a
pending dispatch that needs cache flush. This optimization saves on
extra barrier and helps wall time
Change-Id: Ib62d6d7aabf4743827b561be6c9c5afa813203da
ROCr is now reporting the actual HW addressing limits for HIP, so OpenCL will have to impose lower limit.
Change-Id: I60c2ce27ed1d1f45f16fb76438965a236ba872c6
OCL can't distinguish different copy types, but ROC profiler
expects SDMA transfer visibility. Add extra code to detect
a transfer with the host memory and substitute OCL command
Change-Id: I5290acd0e10bc082e00c1d4ae1474a075de7f165
We unmap a memory with a different pointer.
ROCr runtime might be confused and silently ignore the unmap request
Change-Id: Ic5a1387a426cf02a985a4ef8ff8ff05e6a870cbf
Enable this optimization when the barrier is disabled, since
reuse requires a signal wait.
Use the size of pending AQL signals as the size of signal pool.
Change-Id: I2754a0f8b67e19d2601c58945e10fdf0e8be1624
Now rocm/rocdevice.cpp also includes comgrctx.hpp, and we don't want to statically link against comgr when buidling shared libs.
Change-Id: Ic330bd860559b3e07b776c951afe6126b0f43f7d
The change reuses HSA signals for dispatches as a wait signal.
Skipping the barrier requires to disable L2 cache for sysmem
allocations and extra tracking for HDP access with the large bar.
ROC_BARRIER_SYNC=0 activates the new logic. Barrier sync is
still used by default.
ROC_ACTIVE_WAIT=1 enables unconditional active wait in ROCr.
The change also consolidated ROCr wait logic under single function.
Change-Id: I6bd1be30aa88258da1b1f9de319ef5a45852afd8
SWDEV-249719 - root cause: queues with custom CU mask are not inserted
into queuePool_ (i.e., queue of reusable HSA queues) of ROC device class
causing a crash when creating hostcall buffers for printf
Change-Id: Ieee7005d9a5a30b3113394ce23ee65927126d0d6
Replace amd::Atomic with std::atomic. Remove make_atomic uses by
converting the variable to std::atomic and making sure the memory
order is relaxed when synchronizes-with is not needed.
Delete utils/atomic.hpp.
Change-Id: I0b36db8d604a8510ac6e36b32885fd16a1b8ccfa
root cause - cooperative queue is not inserted into queuePool_ (HSA queues) of ROC device calss causing a crash when creating hostcall buffers for printf
Change-Id: I3f9aceb4e5fe6a7c7a2a549a4bb0a3511fe02799
Fix a typo with the name define, when compilation wasn't enabled.
Force CPU prefetch if system was forced in runtime
Change-Id: Id4b578f9fa44a45426fdb5d8ecb1da803aa42313
German Andryeyev
Alex Xie
Saleel Kudchadker
Aryan Salmanpour
Jason Tang
Rahul Garg
Jason Tang
Sarbojit Sarkar
Tao Sang
Jatin Chaudhary
kjayapra-amd
Satyanvesh Dittakavi
Laurent Morichetti
Vlad Sytchenko