P4 to Git Change 1456420 by wchau@wchau_OCL_boltzmann on 2017/09/07 16:56:26

SWDEV-130808 - Back out changelist 1456363 as it cause AMD SDK 2.9.1 Apps test failure when running OpenCL Sanity tests with Brahma build.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_device.cpp#65 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl.hpp#5 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl_ext.h#16 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl.hpp#7 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl_ext.h#32 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl.hpp#4 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl_ext.h#9 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl.hpp#3 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl_ext.h#3 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#289 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#571 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#354 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.hpp#100 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#57 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#31 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#12 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#59 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.cpp#22 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.hpp#10 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#276 edit


[ROCm/clr commit: 946984cba0]

projects/clr/rocclr/runtime/device/gpu/gpudevice.cpp

@@ -487,7 +487,6 @@ void NullDevice::fillDeviceInfo(const CALdeviceattribs& calAttr, const gslMemInf
   info_.maxWorkItemSizes_[0] = info_.maxWorkGroupSize_;
   info_.maxWorkItemSizes_[1] = info_.maxWorkGroupSize_;
   info_.maxWorkItemSizes_[2] = info_.maxWorkGroupSize_;
-  info_.preferredWorkGroupSize_ = settings().preferredWorkGroupSize_;
 
   if (settings().hwLDSSize_ != 0) {
     info_.localMemType_ = CL_LOCAL;

projects/clr/rocclr/runtime/device/gpu/gpusettings.cpp

@@ -57,8 +57,8 @@ Settings::Settings() {
   doublePrecision_ = ::CL_KHR_FP64;
 
   // Fill workgroup info size
-  maxWorkGroupSize_ = 1024;
-  preferredWorkGroupSize_ = 256;
+  // @todo: revisit the 256 limitation on workgroup size
+  maxWorkGroupSize_ = 256;
 
   hostMemDirectAccess_ = HostMemDisable;
 
@@ -431,10 +431,6 @@ void Settings::override() {
     maxWorkGroupSize_ = GPU_MAX_WORKGROUP_SIZE;
   }
 
-  if (GPU_PREFERRED_WORKGROUP_SIZE != 0) {
-    preferredWorkGroupSize_ = GPU_PREFERRED_WORKGROUP_SIZE;
-  }
-
   // Override blit engine type
   if (GPU_BLIT_ENGINE_TYPE != BlitEngineDefault) {
     blitEngine_ = GPU_BLIT_ENGINE_TYPE;

projects/clr/rocclr/runtime/device/gpu/gpusettings.hpp

@@ -78,7 +78,6 @@ class Settings : public device::Settings {
   uint maxRenameSize_;                //!< Maximum size for all renames
   uint hwLDSSize_;                    //!< HW local data store size
   uint maxWorkGroupSize_;             //!< Requested workgroup size for this device
-  uint preferredWorkGroupSize_;       //!< Requested preferred workgroup size for this device
   uint hostMemDirectAccess_;          //!< Enables direct access to the host memory
   amd::LibrarySelector libSelector_;  //!< Select linking libraries for compiler
   uint workloadSplitSize_;            //!< Workload split size