P4 to Git Change 1456363 by wchau@wchau_OCL_boltzmann on 2017/09/07 15:47:03

SWDEV-130808 - Add support of two new queries: CL_DEVICE_PREFERRED_WORK_GROUP_SIZE_AMD, CL_DEVICE_MAX_WORK_GROUP_SIZE_AMD.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_device.cpp#64 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl.hpp#4 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl1.2/CL/cl_ext.h#15 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl.hpp#6 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.0/CL/cl_ext.h#31 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl.hpp#3 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.1/CL/cl_ext.h#8 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl.hpp#2 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/khronos/headers/opencl2.2/CL/cl_ext.h#2 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/device.hpp#288 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#570 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.cpp#353 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusettings.hpp#99 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paldevice.cpp#55 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.cpp#30 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palsettings.hpp#11 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#58 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocsettings.hpp#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/utils/flags.hpp#275 edit


[ROCm/clr commit: 0b90f3418f]

projects/clr/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;