P4 to Git Change 1456735 by wchau@wchau_OCL_boltzmann on 2017/09/08 11:04:08

SWDEV-130808 - Resubmit the CL#1456363 after finding the AMD SDK 2.9.1 Apps test failure is an app issue.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_device.cpp#66 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl.hpp#6 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl_ext.h#17 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl.hpp#8 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl_ext.h#33 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl.hpp#5 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl_ext.h#10 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl.hpp#4 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl_ext.h#4 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#290 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#572 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#355 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.hpp#101 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#58 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#32 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#13 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#60 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.cpp#23 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.hpp#11 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#277 edit

rocclr/runtime/device/gpu/gpusettings.cpp

@@ -57,8 +57,8 @@ Settings::Settings() {
   doublePrecision_ = ::CL_KHR_FP64;
 
   // Fill workgroup info size
-  // @todo: revisit the 256 limitation on workgroup size
-  maxWorkGroupSize_ = 256;
+  maxWorkGroupSize_ = 1024;
+  preferredWorkGroupSize_ = 256;
 
   hostMemDirectAccess_ = HostMemDisable;
 
@@ -431,6 +431,10 @@ 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;