Some components like tensorflow search for the contents of hip_version.h
For backward compatibility insert actual file contents in wrapper file as non executable code
Improvements: wrapper header template file checked in, instead of generating it on runtime
Redundant codes for unsetting variables removed
Change-Id: I67ee48437f8959d8ed70946412d09be3ed262e58
This code change is to improve error handling.
This code change does not fix issue itself.
Before this code change, hostcallBuffer_ point is initialized in the end of
create() function. If create function fails and returns early,
hostcallBuffer_ point is not initialized. This non-initialized point can
cause access violation when object is destructed.
This code change put the initialization of the pointer in the constructor.
Change-Id: I7fb6e764eb0547196dca03db237e49d3ff0fd06a
Add lock protection for access to the pool list.
Remove destroyed stream from the list of the safe streams
Change-Id: I1863b89bd3f5e188c161227cc790c3adaf72cc58
Adding virtual memory management APIs to rocclr.
The HIP layer will handle virtual allocs on devices.
Change-Id: Ia978f105c2c3fed3959c77580ba228e845105754
This patch adds an option (BUILD_ICD) to toggle the ICD loader build
with rocm-opencl. This isn't useful for official ROCm use, but distros
or users may want build rocm-opencl with their own OCL ICD loader.
The default option is BUILD_ICD=ON.
If this option is set to OFF, an OpenCL OCD loader must be installed
prior to attempting to build rocm-opencl.
E.g. distros already provide ocl icd loader packages:
sudo yum install ocd-icd-devel
sudo apt-get install ocl-icd-opencl-dev
sudo zypper install ocd-icd-devel
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
Change-Id: I6006593eceed543cd829e8949ec2256482815673
Honor hipEventReleaseToDevice and hipEventReleaseToSystem flags.
hipEventRecord would flush caches if no release flags are provided. To
change this behavior set ROC_EVENT_NO_FLUSH=1
Change-Id: I03e41b515b1d0cf963b0c2d5b9901b09e71a0e59
- Add a global cache state for a device to indicate scopes of submitted
AQL packets
- Remove scopes for TS marker if hipEventReleaseToDevice is passed. Set
env ROC_EVENT_NO_FLUSH=1 to use NOP AQL for event records.
It would flush caches by default with system scope release.
- Calling finish() should ensure if caches are flushed, if not queue a
marker
Change-Id: Ibbbdbb1cd7ac61cb35649169212142545be159e0
Use GNUInstallDirs variables to determine the location of BINDIR,
LIBDIR, DOCDIR, and SYSCONFDIR.
Note that CMAKE_INSTALL_LIBDIR is overriden, since the default for RHEL
is lib64, but ROCm packaging wants it to be lib always. Distros or users
can easily override this.
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
Change-Id: Ifb38263dce80d80c5ebb4eacd8e49c76bb013a44
Remove assert for kernel arg size, because COv5 reports a value
bigger than the actual usage in the most of cases
Change-Id: I8e15bc45a9e21b58a5894f9977511ca84408ce61
- Clean up detection by using visual studio macros to detect arch; I
didn't list all possible ARM platforms (can be done later if desired)
- Fixed two incorrect uses of !defined(ATI_ARCH_ARM) to instead use
defined(ATI_ARCH_X86), as they contain X86 specific code
- Fixed one use of __ARM_ARCH_7A__ to use ATI_ARCH_ARM instead
This is an improvement to the fixes in the last patch for SWDEV-323669
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
Change-Id: I8568167293c34ad5331902105877f3ab6e25acb3
This looks unused. This should allow building rocm-opencl without any
dependency on any particular OCL ICD.
Signed-off-by: Jeremy Newton <Jeremy.Newton@amd.com>
Change-Id: Ib9229ede8400fd0e883275659ab99513b03952cb
Alex (Bin) Xie
Chauncey Hui
Saleel Kudchadker
Jason Tang
Ranjith Ramakrishnan
Julia Jiang
Jeremy Newton
Sarbojit Sarkar
Alex Xie
German Andryeyev
Satyanvesh Dittakavi
pghafari
Christophe Paquot
Sourabh Betigeri
sdashmiz
Jaydeep Patel
Hao Yuan
Maxime Chambonnet
Ajay
kjayapra-amd