P4 to Git Change 1458035 by wchau@wchau_OCL_boltzmann on 2017/09/12 16:31:24

SWDEV-130808 - set the local sizes to preferredWorkGroupSize_ when clEnqueueNDRange is not given and the kernel does not have required workgroup sizes.

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpukernel.cpp#320 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#411 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palkernel.cpp#36 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#56 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rockernel.cpp#26 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#43 edit

rocclr/runtime/device/pal/palkernel.cpp

@@ -659,7 +659,7 @@ bool HSAILKernel::init(amd::hsa::loader::Symbol* sym, bool finalize) {
     workGroupInfo_.size_ = workGroupInfo_.compileSize_[0] * workGroupInfo_.compileSize_[1] *
         workGroupInfo_.compileSize_[2];
   } else {
-    workGroupInfo_.size_ = dev().info().maxWorkGroupSize_;
+    workGroupInfo_.size_ = dev().info().preferredWorkGroupSize_;
   }
 
   // Pull out printf metadata from the ELF
@@ -793,7 +793,7 @@ void HSAILKernel::findLocalWorkSize(size_t workDim, const amd::NDRange& gblWorkS
             (workDim > 1) && ((dev().settings().partialDispatch_) ||
                               (((gblWorkSize[0] % 16) == 0) && ((gblWorkSize[1] % 16) == 0)))) {
           // Use 8x8 workgroup size if kernel has image writes
-          if (flags_.imageWriteEna_ || (thrPerGrp != dev().info().maxWorkGroupSize_)) {
+          if (flags_.imageWriteEna_ || (thrPerGrp != dev().info().preferredWorkGroupSize_)) {
             lclWorkSize[0] = 8;
             lclWorkSize[1] = 8;
           } else {
@@ -1540,7 +1540,7 @@ bool LightningKernel::init(amd::hsa::loader::Symbol* symbol) {
     workGroupInfo_.size_ = workGroupInfo_.compileSize_[0] * workGroupInfo_.compileSize_[1] *
         workGroupInfo_.compileSize_[2];
   } else {
-    workGroupInfo_.size_ = dev().info().maxWorkGroupSize_;
+    workGroupInfo_.size_ = dev().info().preferredWorkGroupSize_;
   }
 
   initPrintf(programMD->mPrintf);