two patches for swdev 517884 (llvm#1449)

Author: jharryma

offload/plugins-nextgen/amdgpu/src/rtl.cpp

@@ -2655,7 +2655,7 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
         OMPX_QueueSize("LIBOMPTARGET_AMDGPU_HSA_QUEUE_SIZE", 512),
         OMPX_DefaultTeamsPerCU("LIBOMPTARGET_AMDGPU_TEAMS_PER_CU", 6),
         OMPX_GenericSpmdTeamsPerCU(
-            "LIBOMPTARGET_AMDGPU_GENERIC_SPMD_TEAMS_PER_CU", 0),
+            "LIBOMPTARGET_AMDGPU_GENERIC_SPMD_TEAMS_PER_CU", 6),
         OMPX_BigJumpLoopTeamsPerCU(
             "LIBOMPTARGET_AMDGPU_BIG_JUMP_LOOP_TEAMS_PER_CU", 0),
         OMPX_LowTripCount("LIBOMPTARGET_AMDGPU_LOW_TRIPCOUNT", 2000),
@@ -2676,7 +2676,7 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
         OMPX_StreamBusyWait("LIBOMPTARGET_AMDGPU_STREAM_BUSYWAIT", 2000000),
         OMPX_UseMultipleSdmaEngines(
             // setting default to true here appears to solve random sdma problem
-            "LIBOMPTARGET_AMDGPU_USE_MULTIPLE_SDMA_ENGINES", false),
+            "LIBOMPTARGET_AMDGPU_USE_MULTIPLE_SDMA_ENGINES", true),
         OMPX_ApuMaps("OMPX_APU_MAPS", false),
         OMPX_EnableGFX90ACoarseGrainUsmMaps(
             "OMPX_ENABLE_GFX90A_COARSE_GRAIN_USM_MAPS", false),