diff --git a/clang/tools/amdgpu-arch/AMDGPUArch.cpp b/clang/tools/amdgpu-arch/AMDGPUArch.cpp
index caf8562c4e9dc2..308bdeca243aa1 100644
--- a/clang/tools/amdgpu-arch/AMDGPUArch.cpp
+++ b/clang/tools/amdgpu-arch/AMDGPUArch.cpp
@@ -17,33 +17,16 @@
 #include <string>
 #include <vector>
 
-#if defined(__has_include)
-#if __has_include("hsa/hsa.h")
-#define HSA_HEADER_FOUND 1
-#include "hsa/hsa.h"
-#elif __has_include("hsa.h")
-#define HSA_HEADER_FOUND 1
-#include "hsa.h"
-#else
-#define HSA_HEADER_FOUND 0
-#endif
-#else
-#define HSA_HEADER_FOUND 0
-#endif
-
-#if !HSA_HEADER_FOUND
-// Forward declaration of the status enumeration used by HSA.
+#if DYNAMIC_HSA
 typedef enum {
   HSA_STATUS_SUCCESS = 0x0,
 } hsa_status_t;
 
-// Forward declaration of the device types used by HSA.
 typedef enum {
   HSA_DEVICE_TYPE_CPU = 0,
   HSA_DEVICE_TYPE_GPU = 1,
 } hsa_device_type_t;
 
-// Forward declaration of the agent information types we use.
 typedef enum {
   HSA_AGENT_INFO_NAME = 0,
   HSA_AGENT_INFO_DEVICE = 17,
@@ -56,8 +39,7 @@ typedef struct hsa_agent_s {
 hsa_status_t (*hsa_init)();
 hsa_status_t (*hsa_shut_down)();
 hsa_status_t (*hsa_agent_get_info)(hsa_agent_t, hsa_agent_info_t, void *);
-hsa_status_t (*hsa_iterate_agents)(hsa_status_t (*callback)(hsa_agent_t,
-                                                            void *),
+hsa_status_t (*hsa_iterate_agents)(hsa_status_t (*)(hsa_agent_t, void *),
                                    void *);
 
 constexpr const char *DynamicHSAPath = "libhsa-runtime64.so";
@@ -86,6 +68,16 @@ llvm::Error loadHSA() {
   return llvm::Error::success();
 }
 #else
+
+#if defined(__has_include)
+#if __has_include("hsa/hsa.h")
+#include "hsa/hsa.h"
+#elif __has_include("hsa.h")
+#include "hsa.h"
+#endif
+#include "hsa/hsa.h"
+#endif
+
 llvm::Error loadHSA() { return llvm::Error::success(); }
 #endif
 
diff --git a/clang/tools/amdgpu-arch/CMakeLists.txt b/clang/tools/amdgpu-arch/CMakeLists.txt
index a5361e65758b1b..d687d1691b7a6c 100644
--- a/clang/tools/amdgpu-arch/CMakeLists.txt
+++ b/clang/tools/amdgpu-arch/CMakeLists.txt
@@ -15,4 +15,6 @@ find_package(hsa-runtime64 QUIET 1.2.0 HINTS ${CMAKE_INSTALL_PREFIX} PATHS /opt/
 if (${hsa-runtime64_FOUND})
   set_target_properties(amdgpu-arch PROPERTIES INSTALL_RPATH_USE_LINK_PATH ON)
   clang_target_link_libraries(amdgpu-arch PRIVATE hsa-runtime64::hsa-runtime64)
+else()
+  target_compile_definitions(amdgpu-arch PRIVATE "DYNAMIC_HSA")
 endif()
diff --git a/clang/tools/nvptx-arch/CMakeLists.txt b/clang/tools/nvptx-arch/CMakeLists.txt
index 1b6362c8f19511..9b31cdd772dcf3 100644
--- a/clang/tools/nvptx-arch/CMakeLists.txt
+++ b/clang/tools/nvptx-arch/CMakeLists.txt
@@ -21,4 +21,6 @@ endif()
 if (CUDA_FOUND AND cuda-library)
   target_include_directories(nvptx-arch PRIVATE ${CUDA_INCLUDE_DIRS})
   target_link_libraries(nvptx-arch PRIVATE ${cuda-library})
+else()
+  target_compile_definitions(nvptx-arch PRIVATE "DYNAMIC_CUDA")
 endif()
diff --git a/clang/tools/nvptx-arch/NVPTXArch.cpp b/clang/tools/nvptx-arch/NVPTXArch.cpp
index acde975a34dc4c..6509e42fee4e6a 100644
--- a/clang/tools/nvptx-arch/NVPTXArch.cpp
+++ b/clang/tools/nvptx-arch/NVPTXArch.cpp
@@ -17,18 +17,7 @@
 #include <cstdio>
 #include <memory>
 
-#if defined(__has_include)
-#if __has_include("cuda.h")
-#include "cuda.h"
-#define CUDA_HEADER_FOUND 1
-#else
-#define CUDA_HEADER_FOUND 0
-#endif
-#else
-#define CUDA_HEADER_FOUND 0
-#endif
-
-#if !CUDA_HEADER_FOUND
+#if DYNAMIC_CUDA
 typedef enum cudaError_enum {
   CUDA_SUCCESS = 0,
   CUDA_ERROR_NO_DEVICE = 100,
@@ -74,28 +63,31 @@ llvm::Error loadCUDA() {
   return llvm::Error::success();
 }
 #else
+
+#include "cuda.h"
 llvm::Error loadCUDA() { return llvm::Error::success(); }
+
 #endif
 
 static int handleError(CUresult Err) {
   const char *ErrStr = nullptr;
   CUresult Result = cuGetErrorString(Err, &ErrStr);
   if (Result != CUDA_SUCCESS)
-    return EXIT_FAILURE;
+    return 1;
   fprintf(stderr, "CUDA error: %s\n", ErrStr);
-  return EXIT_FAILURE;
+  return 1;
 }
 
 int main(int argc, char *argv[]) {
   // Attempt to load the NVPTX driver runtime.
   if (llvm::Error Err = loadCUDA()) {
     logAllUnhandledErrors(std::move(Err), llvm::errs());
-    return EXIT_FAILURE;
+    return 1;
   }
 
   if (CUresult Err = cuInit(0)) {
     if (Err == CUDA_ERROR_NO_DEVICE)
-      return EXIT_SUCCESS;
+      return 0;
     else
       return handleError(Err);
   }
@@ -104,7 +96,7 @@ int main(int argc, char *argv[]) {
   if (CUresult Err = cuDeviceGetCount(&Count))
     return handleError(Err);
   if (Count == 0)
-    return EXIT_SUCCESS;
+    return 0;
   for (int DeviceId = 0; DeviceId < Count; ++DeviceId) {
     CUdevice Device;
     if (CUresult Err = cuDeviceGet(&Device, DeviceId))
@@ -120,5 +112,5 @@ int main(int argc, char *argv[]) {
 
     printf("sm_%d%d\n", Major, Minor);
   }
-  return EXIT_SUCCESS;
+  return 0;
 }