HIP should block the stream until callback is done. This change
will introduce extra marker that will block the queue.
Note: MT path doesn't really require extra marker, but the logic
is identical to avoid extra checks for direct dispatch
Change-Id: Ib90fd2d751adf337f5e43ac6098e84767530233b
This change is tied to the ROCclr cmake rework. ROCclr will now be built
as part of HIP itself.
All the dependencies on comgr and rocr have been moved into ROCclr
itself. This allows HIP to pull in all the required compiler/runtime
dependencies by simply linking against ROCclr. Note that ROCclr needs to
be configured from HIP cmake command line. If not, we will default to
LC + ROCr support.
Since HIP depends on ROCclr and ROCclr depends on OpenCL, HIP will
transitevly pull in all the required OpenCL headers. No need to keep a
local copy of them anymore.
Change-Id: Ib39f4929e88da1248be962490c150b10fc7abd3a
This change is tied to the ROCclr cmake rework. ROCclr will now be built
as part of OpenCL itself.
All the dependencies on comgr and rocr have been moved into ROCclr
itself. This allows OpenCL to pull in all the required compiler/runtime
dependencies by simply linking against ROCclr. Note that ROCclr needs to
be configured from OpenCL cmake command line. If not, we will default to
LC + ROCr support.
Change-Id: I5271302d073aef8cfa617828d218c856711bf9ba
- Make sure barrier with dependent signals issues before queue
index reservation
- Don't issue extra barrier if it's already a barrier command
with dependent signals
Change-Id: I179a8b7adac79eed698f4a4d9eca2606d8e913aa
Linux Pro build uses the LLVM_ROOT variable to
specify where llvm is located. We must use a different
name, HIP_LLVM_ROOT for LLVM's prefix location.
Change-Id: I28baf8672217a438a5dd45829d9af04b3553f615
Add memory filling performance test for large bar device memory,
hip managed memory, conherent/non-coherent host memory and fine
grained device memory
Change-Id: Id3f9bcad45d643e493daf9d5f47b3a012a427178
Temporarily disable __HIPCC_RTC__ in hip_vector_types.h
while the upstream clang headers are outdated on mainline.
Once upstream patch has landed in mainline, revert this
change. This is a workaround for hiprtc testing.
Change-Id: Ib2cf6023b71431bbfbe3c699076caa4f90f7170c
Run with multiple threads to catch possible race condition
Disable running the test by default on CI as this takes lot of time
Change-Id: I1c1a66fd5c72f8d2bf7ad120461384488b46abbd
- waitList was released before the awaitcompletion
causing stream sync after the async copies fail
- Enable the hipPeerToPeer_simple test
Change-Id: I1eeaa7e6320ab38c106902078f034dfec4c3f3ea
Add back the .type directive needed for ELF
format to add the section to symbol table.
For COFF format the symbol is already added
to the symbol table.
Change-Id: I4920e8dae0af23b4dcabd324b47abc228fcab10f
In HSAIL path, kernel akc info is obtained after code object
loading, and kernel signature creation requires the akc info
when one of the kernel argument is a reference object.
Change-Id: I9cdb1dbf2c72f4620b0b6e46a88402a2473c3e97
This reverts commit 932c83fe8f.
Reason for revert: need to update other projects (linuxpro & marnier)
Change-Id: Ib02194d052170e487cd56480eb0d9226895aa580
Always allocate managed memory in ROCclr and backend will decide
what kind of memory to use. Currently ROCclr should fall to
GPU accessible system memory if HMM isn't available
Change-Id: I71328687b2dd4a3f563571f2b755079d82dca433
Use HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE flag for the initial
allocation instead of HSA_AMD_SVM_ATTRIB_AGENT_ACCESSIBLE_IN_PLACE.
Change-Id: Ia52fe205563df1ea916dc2dc81e749e11c16f83d
Joseph Greathouse
cjatin
Anusha GodavarthySurya
Tao Sang
Satyanvesh Dittakavi
German Andryeyev
Vladislav Sytchenko
Jason Tang
Chauncey Hui
Aaron En Ye Shi
jujiang
Ravi C Akkenapally
kjayapra-amd
Payam Ghafari