P4 to Git Change 1182228 by gandryey@gera-dev-w7 on 2015/08/19 12:22:39

EPR #419072 - [OpenCL2.0] Enable 16MB large on device queues
	- Add mask_groups argument into the library for the empty slot spreading

Affected files ...

... //depot/stg/opencl/drivers/opencl/library/hsa/hsail/src/devenq/devenq.h#11 edit
... //depot/stg/opencl/drivers/opencl/library/hsa/hsail/src/devenq/enqueue.cl#10 edit
... //depot/stg/opencl/drivers/opencl/library/hsa/hsail/src/devenq/eprep.cl#6 edit
... //depot/stg/opencl/drivers/opencl/library/hsa/hsail/src/devenq/events.cl#4 edit
... //depot/stg/opencl/drivers/opencl/library/hsa/hsail/src/devenq/schedule.cl#11 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpudevice.cpp#520 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpusched.hpp#18 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/gpu/gpuvirtual.cpp#373 edit

rocclr/runtime/device/gpu/gpusched.hpp

@@ -28,7 +28,7 @@ struct AmdVQueueHeader {
     uint32_t command_counter;   //!< [LRW] The global counter for the submitted commands into the queue
     uint32_t wait_size;         //!< [LRO] The wait list size (in clk_event_t)
     uint32_t arg_size;          //!< [LRO] The size of argument buffer (in bytes)
-    uint32_t reserved0;         //!< For the future usage
+    uint32_t mask_groups;       //!< Processed mask groups by one thread
     uint64_t kernel_table;      //!< [LRO] Pointer to an array with all kernel objects (ulong for each entry)
     uint32_t reserved[2];       //!< For the future usage
 };
@@ -70,8 +70,7 @@ struct SchedulerParam {
     uint64_t    parentAQL;      //!< Host parent AmdAqlWrap packet
     uint32_t    dedicatedQueue; //!< Scheduler uses a dedicated queue
     uint32_t    scratchOffset;  //!< Scratch buffer offset
-    uint32_t    mask_groups;    //!< Processed mask groups by one thread
-    uint32_t    reserved;       //!< Reserved
+    uint32_t    reserved[2];    //!< Reserved
 };
 
 } // namespace gpu