The original logic left only one slot for HW processing in the queue.
For some reason there is a race condition on CPU overwrite of the slot
before the current active. The workaround is to avoid the previous to
the current active slot for possible unfinished HW processing.
Change-Id: I565495a8feeaedffc9fc8a505edbee5ff5816975
[ROCm/clr commit: 65ddfcc6a8]
For the fillBuffer shader, if there are two 32bit writes to a MMIO
register, it can get dropped. It has to be a single 64bit write.
Add optimization to fillBuffer to write 64bit and 16bit writes.
Change-Id: I3aa78e027898f8ae01e9c8f09004615673720c2b
[ROCm/clr commit: 21ba34d0fe]
Add a env var ROC_USE_FGS_KERNARG to toggle kernel arg placement
By default its in Fine Grain Kernel arg segment for supported asics.
Change-Id: I3d57ed69a1a4db2b392b0438ead499f3ddca4716
[ROCm/clr commit: e29b9c00ee]
PCMark10 counts the time spent in clCreateKernel as part of execution
time, so as workaround for the PAL path, move code object loading
back to clBuildProgram.
Change-Id: I3b9cf1879ece08ab59f447ec165b0525bc8593a4
[ROCm/clr commit: 1d0364e590]
Pass the device agent specified by the user to the ROCr api instead of passing the device agent attached to the specified stream
Change-Id: I86c98935b9dc404eaa6d47ccdd082a8c3678fb36
[ROCm/clr commit: 169cc857fd]
Fixes Seg fault caused when the attribute hipMemRangeAttributeAccessedBy
is queried using hipMemRangeGetAttribute
Change-Id: I2ceb2267d89bfc31a55d9eae2685610c7ad89b1f
[ROCm/clr commit: 48c1b895c0]
Reuse FillMemory function, that should fix the cache syncs from the host
Change-Id: Ieebec5fc3ed3a322b88d5187c8dca4805ec6f84b
[ROCm/clr commit: 24442be35a]
This patch allows to substitute binary for the opencl program. It supposed to be used as:
1. Run the opencl program with -save-temps.
2. Open the cl temp and find the following text in the program header:
Hash to override:
Source: 0xd66bcfa20e69e605
Source + clang options: 0x656a9dd8aedcbfb6
3. Create config file (ascii text) with a pair(s):
<hash> <path_to_binary_to_substitute>
where hash is the hex value from step 2 (without leading 0x), you can use either hash
depending on what you're going to match:
only the source text of the program or along with it's clang options.
4. Set the env variable AMD_OCL_SUBST_OBJFILE to the path of your config file.
5. Rerun the opencl program.
Change-Id: I977c80fe529ea14458194918c6ddfbe2de6a8857
[ROCm/clr commit: 51cc9c2f8c]
Current logic when creating a buffer view will end up going into the
allocation block. Even though no memory will be allocated, since
owner()->getSvmPtr() is already allocated, we'll still end up
calling updateFreeMemory().
Checking if we're creating a view, will skip the SVM allocation logic
and let us fall into the actual view creation logic. This won't end up
updating the free memory counter.
Change-Id: I1c260a9ef57895130b272ea1246e06e812b25b37
[ROCm/clr commit: f167136918]
The new query MemRangeAttribute::CoherencyMode can return current
coherency mode for the provided memory region. Coherency mode can
be one of the following types: FineGrain, CoarseGrain and
Indeterminate
Change-Id: Ib66feeeb14f57a8b1cc731c65bb3d0276d297ff7
[ROCm/clr commit: 992830bab7]
Redshift sees around a 3x performance uplift this change.
Turning this on for OpenCL might cause unwanted behaviour, due to
apps like RSX running in the background all the time.
Change-Id: I9f32d5f2e05b6697a8aaa9ddf74474b5531bb7e1
[ROCm/clr commit: 2f00782829]
There is a possible race condition when signal reuse can have
access to a destroyed Timestamp object, because the callback
was running asynchronously. Use reference counter and lock
to allow asynchronous timestamp update
Change-Id: I6224f7c62cb0a03a7466fcc512e5e5afb06736fa
[ROCm/clr commit: ec89348291]
Note that this requires base driver CL#2340320+ to have SQ interrupt
functionality enabled by default.
Change-Id: I04b936819ebe1eb7cf5de1db4fafe83af3a1b5f6
[ROCm/clr commit: 4171e9e0a3]
Below logic allocates the host buffer whenever a subbuffer is created
from a SVM allocation. This is only needed for multi-device contexts.
HIP does not support multi-device contexts, hence this logic just ends
up performing unnecessary system allocations.
Change-Id: I8eae635f7c5289c52ef73434218c1658b788a456
[ROCm/clr commit: ffbf368f4c]
Only add Roc path and don't use Pal path.
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Change-Id: I7117e2dc3c3ad4c8d563e9bbdc721f70ddba51fd
[ROCm/clr commit: c6574cb906]
- Device Reset should not purge the allocations that were not by the user
- Addresses QMCPack Test abort due to the removal of all the mem objects during reset
Change-Id: I7b7a123e72bcc985d7e51d17c2382bc618d3e041
[ROCm/clr commit: 924695fb5e]
This reverts commit 9df70fa03ce60d47247eb0e8f278e1f8dbd33d6e.
Reason for revert: need SWDEV-294782 to be resolved before we can enable SQ interrupt support.
Change-Id: I328170b60f1a3aab28c0b1fd3191297a1a51ecb7
[ROCm/clr commit: 6566361144]
Add ref counting to ProfilingSignal class to track the last release.
If a signal was used in the marker, then don't reuse it,
but create a new one for internal usage.
Don't rely on HSA callback for the command status update if there
are no pending dispatches.
Change-Id: I19f14ed9d80acfe79993b343b2187635f8428a20
[ROCm/clr commit: ff15c0893e]
Currently ADL is only being included if we build HSAIL and not LC. The
only configuration that does this is Ellesmere on Linux.
As the appprofile logic doesn't do anything right now, it should be
safe to remove it.
Change-Id: I13ce8c27af08c25b95edcbfd2ee1ce447b0bfd39
[ROCm/clr commit: 6ff4839d76]
Setting AMD_CPU_AFFINITY=1 will keep Async Handler thread within the
bounds set by numactl.
Change-Id: Id01b30df5127d65c29ac072bf74a04986b7128de
[ROCm/clr commit: cd21af757e]
German Andryeyev
Jason Tang
Saleel Kudchadker
Jatin Chaudhary
Sarbojit Sarkar
jujiang
Satyanvesh Dittakavi
vpykhtin
Vladislav Sytchenko
Todd tiantuo Li
anusha GodavarthySurya
Rahul Garg
Alex Xie
Aaron Liu
Sourabh Betigeri
agunashe
pvellien