SWDEV-79445 - OCL generic changes and code clean-up
- Use max number of waves per SIMD in the scratch calculation to allow async kernel execution with the scratch buffer
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#155 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.hpp#43 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#147 edit
[ROCm/clr commit: 34e526d77f]
SWDEV-145570 - Rename OCL_DUMP_CODE_OBJECT to GPU_DUMP_CODE_OBJECT.
Since this is used by both OCL and HIP. Rename to avoid confusion.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/devprogram.cpp#59 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#315 edit
[ROCm/clr commit: 4c8ef11012]
SWDEV-2 - Change OpenCL version number from 2970 to 2971.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2718 edit
[ROCm/clr commit: 3b17b1cbbf]
SWDEV-199667 - [OpenCL][PAL][LC] bruteforce conformance test causes vm fault
- With HWS scratch buffer has to be allocated per each windows scheduling context, because HWS can schedule CB on any pipe and different pipes can't preserve unique wave_id. Recycle PAL queues in order to keep the scratch buffer per scheduling context. The change will also allow to remove Windows/kmd limitation of the scheduling contexts per process. GPU_MAX_HW_QUEUES controls the number of unique PAL queues, default is 4
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#154 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.hpp#42 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#146 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#63 edit
[ROCm/clr commit: 884478d0aa]
SWDEV-2 - Change OpenCL version number from 2969 to 2970.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2717 edit
[ROCm/clr commit: 5e86562499]
SWDEV-2 - Change OpenCL version number from 2968 to 2969.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2716 edit
[ROCm/clr commit: 0036c61a8c]
SWDEV-79445 - OCL generic changes and code clean-up
- Allow async execution with scratch on the same queue. COMPUTE_TMPRING_SIZE.WAVESIZE should be constant across all dispatches.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#151 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.hpp#40 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#145 edit
[ROCm/clr commit: 0a2d4e2dc3]
SWDEV-198556 - [HIP] Use src/dstMemory->getContext instead of host_context.
Also relax the check for P2P copies in case of hipMemcpy(hostMalloced, hipMalloced(dev1), dev0)
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#67 edit
[ROCm/clr commit: 7bbbdabfb0]
SWDEV-79445 - OCL generic changes and code clean-up
- Remove the first queue skipping logic, since KMD no longer reports an extra normal queue
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#144 edit
[ROCm/clr commit: 2fbff434ba]
SWDEV-2 - Change OpenCL version number from 2967 to 2968.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2715 edit
[ROCm/clr commit: 44911178b7]
Compilation error being observed due to popen(), pclose() and setenv() linux calls on windows. Replaced with appropriate calls on windows.
[ROCm/clr commit: 19ad365834]
Compilation error being observed on windows due to aligned_alloc() call. Mapped the call to _aligned_malloc() for windows.
[ROCm/clr commit: 93f5940ef8]
* Removed unwanted #include sys/time.h , gettimeofday() and timeval variables and this also helps avavoid compilation error in windows due to gettimeofday() call equivalent of which is not available in windows
* Changed the Macro name from GPU_PRINT_TIME to MY_LAUNCH_MACRO
[ROCm/clr commit: 36977dc276]
Changed the third arg of the functions __hip_as_write_block and __ockl_as_write_block from ulong to uint64_t so as to fix the compilation error in windows
[ROCm/clr commit: c212c484af]
* add default visibility to most APIs in program_state
* remove unwanted C++ headers
* Add symbol visibility pragmas and compiler flags
* Add visibility attribute to APIs in channel_descriptor and hip_hcc
* remove unused headers
* simplify build flags with hcc
* add pragma visibility hidden to functional_grid_launch
* [CMake] add gfx908 back
[ROCm/clr commit: 3cf2d7d385]
* Add support for hipFunGetAttribute
* Support NVCC path
* Test using sample module_api_global
* Try fixing CI build failure due to hip_prof_gen scan
* Fix for CI build issue
* Resolve conflict
* Rebase and resolve conflicts with master
* Fix build error
* Fix NVCC path build error
[ROCm/clr commit: 8b6317d041]
foreman
Evgeny Mankov
Jenkins
Nick Curtis
Maneesh Gupta
amd-lthakur
ansurya
Rahul Garg
Rene van Oostrum
Siu Chi Chan