rocr: Quiet some ROCr compile warnings

These are mostly AIE related, but there are a couple of others. Change-Id: I549e004772160ca282d4c94dc9d94dd2ccae8b1c [ROCm/ROCR-Runtime commit: 08699069d6]
2024-10-23 14:07:00 -05:00
@@ -144,6 +144,9 @@ void AgentPropTest::QueryAgentProp(hsa_agent_t agent,
    case HSA_DEVICE_TYPE_DSP:
      ss << "DSP) : ";
      break;
+    case HSA_DEVICE_TYPE_AIE:
+      ss << "AIE) : ";
+      break;
  }

  // Print the agent property
@@ -183,6 +183,9 @@ static void PrintAgentNameAndType(hsa_agent_t agent) {
    case HSA_DEVICE_TYPE_DSP:
      std::cout << "DSP)";
      break;
+    case HSA_DEVICE_TYPE_AIE:
+      std::cout << "AIE)";
+      break;
    }
  std::cout << std::endl;
  return;
@@ -397,6 +397,9 @@ static void PrintAgentNameAndType(hsa_agent_t agent) {
    case HSA_DEVICE_TYPE_DSP:
      std::cout << "DSP)";
      break;
+    case HSA_DEVICE_TYPE_AIE:
+      std::cout << "AIE)";
+      break;
    }
  std::cout << std::endl;
  return;
@@ -530,13 +533,13 @@ void MemoryAllocationTest::MemoryAllocateContiguousTest(hsa_agent_t agent,
      accessible_gpus.push_back(gpuIter);
  }

-  void* importedPtr;
+  void* importedPtr = nullptr;
  size_t importedSz;

  ASSERT_SUCCESS(hsa_amd_interop_map_buffer(accessible_gpus.size(), accessible_gpus.data(), dmabuf, 0, &importedSz,
                                                   &importedPtr, 0, NULL));

-  ASSERT_NE((uint64_t)importedPtr, 0);
+  ASSERT_NE(importedPtr, nullptr);
  ASSERT_EQ(importedSz, alloc_size);

  close(dmabuf);
@@ -189,6 +189,10 @@ void MemoryTest::MaxSingleAllocationTest(hsa_agent_t ag,
      case HSA_DEVICE_TYPE_DSP:
        std::cout << "DSP)";
        break;
+      case HSA_DEVICE_TYPE_AIE:
+        std::cout << "AIE)";
+        break;
+
    }
    std::cout << std::endl;
  }
@@ -321,6 +325,9 @@ void MemoryTest::MemAvailableTest(hsa_agent_t ag, hsa_amd_memory_pool_t pool) {
      case HSA_DEVICE_TYPE_DSP:
        std::cout << "DSP)";
        break;
+      case HSA_DEVICE_TYPE_AIE:
+        std::cout << "AIE)";
+        break;
    }
    std::cout << std::endl;
  }
@@ -736,7 +736,9 @@ void VirtMemoryTestBasic::GPUAccessToCPUMemoryTest(hsa_agent_t cpuAgent, hsa_age
  ASSERT_SUCCESS(hsa_amd_vmem_unmap(dev_data, sizeof(*dev_data)));
  ASSERT_SUCCESS(hsa_amd_vmem_handle_release(mem_handle));

-  if (dev_data) ASSERT_SUCCESS(hsa_amd_vmem_address_free(dev_data, sizeof(*dev_data)));
+  if (dev_data) {
+    ASSERT_SUCCESS(hsa_amd_vmem_address_free(dev_data, sizeof(*dev_data)));
+  }

  if (host_data) hsa_memory_free(host_data);
  if (kernArgsVirt) {
@@ -1070,7 +1072,7 @@ void VirtMemoryTestBasic::NonContiguousChunks(hsa_agent_t cpuAgent, hsa_agent_t

  size_t& granule_size = pool_i.alloc_granule;
  size_t alloc_size = granule_size * 512;
-  const uint64_t NUM_BUFFERS = 6;
+  const unsigned NUM_BUFFERS = 6;

  void* addr;
  void* addr_chunks[NUM_BUFFERS];
@@ -170,6 +170,9 @@ static void PrintAgentNameAndType(hsa_agent_t agent) {
    case HSA_DEVICE_TYPE_DSP:
      std::cout << "DSP)";
      break;
+    case HSA_DEVICE_TYPE_AIE:
+      std::cout << "AIE)";
+      break;
    }
  std::cout << std::endl;
  return;
@@ -242,6 +242,9 @@ static void PrintAgentNameAndType(hsa_agent_t agent) {
    case HSA_DEVICE_TYPE_DSP:
      std::cout << "DSP)";
      break;
+    case HSA_DEVICE_TYPE_AIE:
+      std::cout << "AIE)";
+      break;
    }
  std::cout << std::endl;
  return;
@@ -95,13 +95,13 @@ public:

  // AIE agent methods.
  /// @brief Get the number of columns on this AIE agent.
-  int GetNumCols() const { return num_cols_; }
-  void SetNumCols(int num_cols) { num_cols_ = num_cols; }
+  uint32_t GetNumCols() const { return num_cols_; }
+  void SetNumCols(uint32_t num_cols) { num_cols_ = num_cols; }
  /// @brief Get the number of core tile rows on this AIE agent.
-  int GetNumCoreRows() const { return num_core_rows_; }
-  void SetNumCoreRows(int num_core_rows) { num_core_rows_ = num_core_rows; }
+  uint32_t GetNumCoreRows() const { return num_core_rows_; }
+  void SetNumCoreRows(uint32_t num_core_rows) { num_core_rows_ = num_core_rows; }
  /// @brief Get the number of core tiles on this AIE agent.
-  int GetNumCores() const { return num_cols_ * num_core_rows_; }
+  uint32_t GetNumCores() const { return num_cols_ * num_core_rows_; }

 private:
  /// @brief Query the driver to get the region list owned by this agent.
@@ -123,10 +123,10 @@ private:
  const uint32_t max_queues_ = 1;

  /// @brief Number of columns in the AIE array.
-  int num_cols_ = 0;
+  uint32_t num_cols_ = 0;
  /// @brief Number of rows of core tiles in the AIE array. Not all rows in a
  /// column are cores. Some can be memory or shim tiles.
-  int num_core_rows_ = 0;
+  uint32_t num_core_rows_ = 0;
 };

 } // namespace AMD
@@ -116,7 +116,7 @@ public:
                  hsa_fence_scope_t releaseFence = HSA_FENCE_SCOPE_NONE,
                  hsa_signal_t *signal = NULL) override;

-  uint32_t queue_id_ = INVALID_QUEUEID;
+  HSA_QUEUEID queue_id_ = INVALID_QUEUEID;
  /// @brief ID of AIE device on which this queue has been mapped.
  uint32_t node_id_ = std::numeric_limits<uint32_t>::max();
  /// @brief Queue size in bytes.
@@ -346,8 +346,6 @@ class AqlQueue : public core::Queue, private core::LocalSignal, public core::Doo
    static KernelMutex queue_lock_;
    return queue_lock_;
 }
-  // Async scratch single limit - may be modified after init
-  size_t async_scratch_single_limit_;

  static __forceinline int& rtti_id() {
    static int rtti_id_ = 0;
@@ -507,14 +507,14 @@ class GpuAgent : public GpuAgentInt {
  hsa_status_t EnableDmaProfiling(bool enable) override;

  hsa_status_t PcSamplingIterateConfig(hsa_ven_amd_pcs_iterate_configuration_callback_t cb,
-                                       void* cb_data);
-  hsa_status_t PcSamplingCreate(pcs::PcsRuntime::PcSamplingSession& session);
+                                       void* cb_data) override;
+  hsa_status_t PcSamplingCreate(pcs::PcsRuntime::PcSamplingSession& session) override;
  hsa_status_t PcSamplingCreateFromId(HsaPcSamplingTraceId pcsId,
-                                      pcs::PcsRuntime::PcSamplingSession& session);
-  hsa_status_t PcSamplingDestroy(pcs::PcsRuntime::PcSamplingSession& session);
-  hsa_status_t PcSamplingStart(pcs::PcsRuntime::PcSamplingSession& session);
-  hsa_status_t PcSamplingStop(pcs::PcsRuntime::PcSamplingSession& session);
-  hsa_status_t PcSamplingFlush(pcs::PcsRuntime::PcSamplingSession& session);
+                            pcs::PcsRuntime::PcSamplingSession& session) override;
+  hsa_status_t PcSamplingDestroy(pcs::PcsRuntime::PcSamplingSession& session) override;
+  hsa_status_t PcSamplingStart(pcs::PcsRuntime::PcSamplingSession& session) override;
+  hsa_status_t PcSamplingStop(pcs::PcsRuntime::PcSamplingSession& session) override;
+  hsa_status_t PcSamplingFlush(pcs::PcsRuntime::PcSamplingSession& session) override;
  hsa_status_t PcSamplingFlushHostTrapDeviceBuffers(pcs::PcsRuntime::PcSamplingSession& session);

  static void PcSamplingThreadRun(void* agent);
@@ -4,7 +4,8 @@
 #include "inc/hsa_amd_tool.h"
 #include "runtime.h"

-namespace rocr::AMD::tool {
+// namespace rocr::AMD::tool {  // C++17
+namespace rocr { namespace AMD { namespace tool {

 using scratch_alloc_flag = hsa_amd_event_scratch_alloc_flag_t;

@@ -135,7 +136,9 @@ __forceinline void notify_event_scratch_async_reclaim_end(const hsa_queue_t* que
      hsa_amd_tool_event_t{.scratch_async_reclaim_end = &event});
 }

-
-}  // namespace rocr::AMD::tool
+// }  // namespace rocr::AMD::tool
+}  // namespace rocr
+}  // namespace AMD
+}  // namespace tool

 #endif
@@ -559,7 +559,7 @@ BlitSdma<RingIndexTy, HwIndexMonotonic, SizeToCountOffset, useGCR>::SubmitCopyRe
  const bool isGFX12Plus = (agent_->isa()->GetMajorVersion() >= 12);

  // Common and GFX12 packet must match in size to use same code for vector/append.
-  static_assert(sizeof(SDMA_PKT_COPY_LINEAR_RECT) == sizeof(SDMA_PKT_COPY_LINEAR_RECT_GFX12));
+  static_assert(sizeof(SDMA_PKT_COPY_LINEAR_RECT) == sizeof(SDMA_PKT_COPY_LINEAR_RECT_GFX12), "");

  const uint max_pitch = 1 << (isGFX12Plus ? SDMA_PKT_COPY_LINEAR_RECT_GFX12::pitch_bits : SDMA_PKT_COPY_LINEAR_RECT::pitch_bits);