diff --git a/benchmarks/async_alloc/async_alloc.cpp b/benchmarks/async_alloc/async_alloc.cpp
index 075702891d9..bb3f287fce5 100644
--- a/benchmarks/async_alloc/async_alloc.cpp
+++ b/benchmarks/async_alloc/async_alloc.cpp
@@ -14,7 +14,7 @@ std::vector<std::pair<size_t, double>> inner_loop_times;
 //
 std::pair<double, double> test(bool up) {
   int iters      = 50;
-  size_t minimum = 8 / sizeof(float);  // 64K
+  size_t minimum = 8 / sizeof(float);
 
   size_t gb = 1024 * 1024 * 1024 / sizeof(float);  // number of floats per GiB
   size_t maximum = gb;  // on 32 bit, we make 1GiB the max
@@ -30,8 +30,10 @@ std::pair<double, double> test(bool up) {
   }
 
   Kokkos::Timer first_alloc_timer;
-  {  // Prime the pump - first long alloc -- Time it.
-    Kokkos::View<float *, MemorySpace> dummy("unlabeled", 64);
+  {  // Prime the pump - first long alloc -- Time it
+     // 64 bytes is an arbitrary number here. .
+    Kokkos::View<float *, MemorySpace> dummy(
+        Kokkos::view_alloc(Kokkos::WithoutInitializing, "unlabeled"), 64);
   }
   double first_alloc_time = first_alloc_timer.seconds();
 
@@ -43,7 +45,8 @@ std::pair<double, double> test(bool up) {
   for (size_t num : sizes) {
     inner_loop_timer.reset();
     for (int i = 0; i < iters; i++) {
-      Kokkos::View<float *, MemorySpace> a("unlabeled", num);
+      Kokkos::View<float *, MemorySpace> a(
+          Kokkos::view_alloc(Kokkos::WithoutInitializing, "unlabeled"), num);
     }
     double inner_loop_time = inner_loop_timer.seconds();
 
@@ -67,7 +70,7 @@ int main(int argc, char *argv[]) {
 
   // Check the env var for reporting
   char *env_string = getenv("KOKKOS_CUDA_MEMPOOL_SIZE");
-  std::cout << "Async Malloc Benchmark: KOKKOS_CUDA_MEMPOOL_SIZE is ";
+  std::cout << "# Async Malloc Benchmark: KOKKOS_CUDA_MEMPOOL_SIZE is ";
 
   if (env_string == nullptr)
     std::cout << "not set,";
@@ -75,28 +78,26 @@ int main(int argc, char *argv[]) {
     std::cout << " " << env_string << ",";
 
   if (up)
-    std::cout << " memory cycling upwards \n";
+    std::cout << "# memory cycling upwards \n";
   else
-    std::cout << " memory_cycling downwards \n";
+    std::cout << "# memory_cycling downwards \n";
   std::cout << std::flush;
 
   Kokkos::initialize(argc, argv);
 
-  inner_loop_times.reserve(34);
-
   // Love structured bindings?
   const auto [first_alloc_time, alloc_loop_time] = test(up);
 
-  std::cout << "First Alloc: " << 64 << " bytes, " << first_alloc_time
+  if (!up) std::reverse(inner_loop_times.begin(), inner_loop_times.end());
+
+  std::cout << "# First Alloc: " << 64 << " bytes, " << first_alloc_time
             << " sec\n";
-  std::cout << "Test Alloc Loop Total: " << alloc_loop_time << " sec\n";
-  std::cout << "Alloc Loop Timings:\n";
-  std::cout << "===================\n";
+  std::cout << "# Test Alloc Loop Total: " << alloc_loop_time << " sec\n";
+  std::cout << "# Alloc Loop Timings:\n";
+  std::cout << "# ===================\n";
 
   std::cout << "# size (B) \t time (sec) \n";
   std::cout << "# -----------------------\n";
-  std::sort(inner_loop_times.begin(), inner_loop_times.end(),
-            [=](const auto &a, const auto &b) { return a.first < b.first; });
   for (auto pair : inner_loop_times) {
     std::cout << pair.first << ", " << pair.second << "\n";
   }
diff --git a/core/src/Cuda/Kokkos_CudaSpace.cpp b/core/src/Cuda/Kokkos_CudaSpace.cpp
index 3f46866143f..603d06af00b 100644
--- a/core/src/Cuda/Kokkos_CudaSpace.cpp
+++ b/core/src/Cuda/Kokkos_CudaSpace.cpp
@@ -34,7 +34,6 @@
 #include <unordered_set>
 #include <limits>
 
-// #include <Cuda/Kokkos_Cuda_BlockSize_Deduction.hpp>
 #include <impl/Kokkos_Error.hpp>
 
 #include <impl/Kokkos_Tools.hpp>
@@ -233,15 +232,15 @@ bool initializeMempool(const int device_id, const cudaStream_t stream,
 
   // Handle exception in case the string is unconvertible
   try {
-    requested_size = static_cast<size_t>(std::stod(mempool_size_string));
+    requested_size = std::stod(mempool_size_string);
   } catch (...) {
     std::cerr << "Unable to convert " << mempool_size_string
               << " to a number\n";
     return false;
   }
 
-  // Check for non-positive size memory requests
-  if (requested_size <= 0) {
+  // Check for negative size memory requests (zero is allowed)
+  if (requested_size < 0) {
     std::cerr << "Negative amount of memory requested in allocation\n";
     return false;
   }
@@ -250,17 +249,18 @@ bool initializeMempool(const int device_id, const cudaStream_t stream,
   requested_size *= factor;
 
   // Check we are not asking for memory that is greater than what size_t can
-  // hold. Since requested can be the larger I convert the maximum of size_t
-  // to a double and compare those
+  // hold. Since the requested size can be the larger I convert the maximum of
+  // size_t to a double and compare those
   double max_size_t = static_cast<double>(std::numeric_limits<size_t>::max());
   if (requested_size > max_size_t) {
     std::cerr << "Requested amount of memory " << requested_size << " exceeds "
-              << "maximum alloatable size " << max_size_t << "\n";
+              << "maximum allocatable size " << max_size_t << "\n";
     return false;
   }
 
   // At this point requested_size should be appropriate
-  // neither too big nor negative.
+  // neither too big nor negative - even with the ceiling it should
+  // not be bigger than max_size_t. Safe to cast
   size_t n_bytes = static_cast<size_t>(std::ceil(requested_size));
 
   // We set up the default memory pool