From 9acbfeba0962370ea7299770c28e2d5f009b9dea Mon Sep 17 00:00:00 2001
From: Edward Chen <18449977+edgchen1@users.noreply.github.com>
Date: Tue, 16 Nov 2021 10:24:46 -0800
Subject: [PATCH] Address some code scan issues. (#9752)

---
 include/onnxruntime/core/common/status.h         | 16 ++++++++--------
 .../training_ops/cuda/optimizer/lamb.cu          |  4 ++--
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/include/onnxruntime/core/common/status.h b/include/onnxruntime/core/common/status.h
index 2f8e75186b..f348e7e653 100644
--- a/include/onnxruntime/core/common/status.h
+++ b/include/onnxruntime/core/common/status.h
@@ -47,7 +47,7 @@ enum StatusCode {
   EP_FAIL = 11
 };
 
-inline const char* StatusCodeToString(StatusCode status) noexcept {
+constexpr const char* StatusCodeToString(StatusCode status) noexcept {
   switch (status) {
     case StatusCode::OK:
       return "SUCCESS";
@@ -79,7 +79,7 @@ inline const char* StatusCodeToString(StatusCode status) noexcept {
 }
 
 #ifdef _WIN32
-inline HRESULT StatusCodeToHRESULT(StatusCode status) noexcept {
+constexpr HRESULT StatusCodeToHRESULT(StatusCode status) noexcept {
     switch (status)
     {
     case StatusCode::OK:
@@ -89,23 +89,23 @@ inline HRESULT StatusCodeToHRESULT(StatusCode status) noexcept {
     case StatusCode::INVALID_ARGUMENT:
         return E_INVALIDARG;
     case StatusCode::NO_SUCHFILE:
-        return __HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND);
+        return HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND);
     case StatusCode::NO_MODEL:
-        return __HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND);
+        return HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND);
     case StatusCode::ENGINE_ERROR:
         return E_FAIL;
     case StatusCode::RUNTIME_EXCEPTION:
         return E_FAIL;
     case StatusCode::INVALID_PROTOBUF:
-        return __HRESULT_FROM_WIN32(ERROR_FILE_CORRUPT);
+        return HRESULT_FROM_WIN32(ERROR_FILE_CORRUPT);
     case StatusCode::MODEL_LOADED:
-        return __HRESULT_FROM_WIN32(ERROR_INTERNAL_ERROR);
+        return HRESULT_FROM_WIN32(ERROR_INTERNAL_ERROR);
     case StatusCode::NOT_IMPLEMENTED:
         return E_NOTIMPL;
     case StatusCode::INVALID_GRAPH:
-        return __HRESULT_FROM_WIN32(ERROR_FILE_CORRUPT);
+        return HRESULT_FROM_WIN32(ERROR_FILE_CORRUPT);
     case StatusCode::EP_FAIL:
-        return __HRESULT_FROM_WIN32(ERROR_INTERNAL_ERROR);
+        return HRESULT_FROM_WIN32(ERROR_INTERNAL_ERROR);
     default:
         return E_FAIL;
     }
diff --git a/orttraining/orttraining/training_ops/cuda/optimizer/lamb.cu b/orttraining/orttraining/training_ops/cuda/optimizer/lamb.cu
index 269b4c7b15..ed289497f5 100644
--- a/orttraining/orttraining/training_ops/cuda/optimizer/lamb.cu
+++ b/orttraining/orttraining/training_ops/cuda/optimizer/lamb.cu
@@ -630,8 +630,8 @@ void LambMultiTensorReductionFunctor<TIn1, TIn2, TOut1, TOut2, TBuf>::operator()
   const int shared_memory_size = thread_count / GPU_WARP_SIZE * 2 * sizeof(TBuf);
 
   // Enforce assumptions used inside this reduction CUDA kernel.
-  ORT_ENFORCE(thread_count % GPU_WARP_SIZE == 0);
-  ORT_ENFORCE((thread_count & (thread_count - 1)) == 0);
+  static_assert(thread_count % GPU_WARP_SIZE == 0, "thread_count must be a multiple of GPU_WARP_SIZE");
+  static_assert((thread_count & (thread_count - 1)) == 0, "thread_count must be a power of two");
 
   const int num_blocks = chunk_group.chunk_count;
   const size_t w_buffer_size = num_blocks * sizeof(TOut1);