P4 to Git Change 1599472 by gandryey@gera-w8 on 2018/08/29 12:25:34

SWDEV-79445 - OCL generic changes and code clean-up
	- Move FindLocalWorkSize() logic to the abstraction layer
	- Replace the ROCr path with the common FindLocalWorkSize() functionality

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/device.cpp#227 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#314 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/devkernel.cpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/devkernel.hpp#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#330 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.hpp#132 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.cpp#63 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.hpp#22 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rockernel.cpp#42 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.cpp#36 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.hpp#16 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#67 edit


[ROCm/clr commit: 5ee211e801]
This commit is contained in:
foreman
2018-08-29 12:35:08 -04:00
parent a48fcb65b2
commit 5f5ea9af64
12 changed files with 216 additions and 408 deletions
@@ -43,16 +43,7 @@ Settings::Settings() {
nonCoherentMode = getenv("OPENCL_USE_NC_MEMORY_POLICY");
enableNCMode_ = (nonCoherentMode) ? true : false;
// Determine if user wishes to disable support for
// partial dispatch. By default support for partial
// dispatch is enabled. Users can turn it off for
// devices that do not support this feature.
//
// @note Update appropriate field of device::Settings
char* partialDispatch = nullptr;
partialDispatch = getenv("OPENCL_DISABLE_PARTIAL_DISPATCH");
enablePartialDispatch_ = (partialDispatch) ? false : true;
partialDispatch_ = (partialDispatch) ? false : true;
partialDispatch_ = GPU_PARTIAL_DISPATCH;
commandQueues_ = 100; //!< Field value set to maximum number
//!< concurrent Virtual GPUs for ROCm backend