diff --git a/libc/startup/gpu/amdgpu/start.cpp b/libc/startup/gpu/amdgpu/start.cpp
index 89b0be208d2e7f..9d7f04c10b4882 100644
--- a/libc/startup/gpu/amdgpu/start.cpp
+++ b/libc/startup/gpu/amdgpu/start.cpp
@@ -37,8 +37,8 @@ static void call_init_array_callbacks(int argc, char **argv, char **env) {
 
 static void call_fini_array_callbacks() {
   size_t fini_array_size = __fini_array_end - __fini_array_start;
-  for (size_t i = 0; i < fini_array_size; ++i)
-    reinterpret_cast<FiniCallback *>(__fini_array_start[i])();
+  for (size_t i = fini_array_size; i > 0; --i)
+    reinterpret_cast<FiniCallback *>(__fini_array_start[i - 1])();
 }
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/startup/gpu/nvptx/start.cpp b/libc/startup/gpu/nvptx/start.cpp
index 1ff187a577789f..2c09430c5971ef 100644
--- a/libc/startup/gpu/nvptx/start.cpp
+++ b/libc/startup/gpu/nvptx/start.cpp
@@ -35,8 +35,8 @@ static void call_init_array_callbacks(int argc, char **argv, char **env) {
 
 static void call_fini_array_callbacks() {
   size_t fini_array_size = __fini_array_end - __fini_array_start;
-  for (size_t i = 0; i < fini_array_size; ++i)
-    reinterpret_cast<FiniCallback *>(__fini_array_start[i])();
+  for (size_t i = fini_array_size; i > 0; --i)
+    reinterpret_cast<FiniCallback *>(__fini_array_start[i - 1])();
 }
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/utils/gpu/loader/nvptx/Loader.cpp b/libc/utils/gpu/loader/nvptx/Loader.cpp
index e920b65a7e10cc..5388f287063b7f 100644
--- a/libc/utils/gpu/loader/nvptx/Loader.cpp
+++ b/libc/utils/gpu/loader/nvptx/Loader.cpp
@@ -83,7 +83,6 @@ Expected<void *> get_ctor_dtor_array(const void *image, const size_t size,
   // for destructors.
   llvm::sort(ctors, [](auto x, auto y) { return x.second < y.second; });
   llvm::sort(dtors, [](auto x, auto y) { return x.second < y.second; });
-  llvm::reverse(dtors);
 
   // Allocate host pinned memory to make these arrays visible to the GPU.
   CUdeviceptr *dev_memory = reinterpret_cast<CUdeviceptr *>(allocator(