diff --git a/CMakeLists.txt b/CMakeLists.txt
index abfbc87..32505e2 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -8,7 +8,7 @@ include(FetchContent)
 set(KOKKOS_FETCH_TAG "5.1.1" CACHE STRING "Kokkos git tag to fetch")
 find_package(Kokkos 5.1.1 QUIET)
 if (NOT Kokkos_FOUND)
-    message(STATUS "Kokkos not found locally - fetching tag " "${KOKKOS_FETCH_TAG} via FetchContent")
+    message(STATUS "Kokkos not found locally - fetching tag ${KOKKOS_FETCH_TAG} via FetchContent")
     FetchContent_Declare(kokkos GIT_REPOSITORY https://github.com/kokkos/kokkos.git GIT_TAG ${KOKKOS_FETCH_TAG} GIT_SHALLOW ON)
     FetchContent_MakeAvailable(kokkos)
 else()
diff --git a/kokkos_tree/CMakeLists.txt b/kokkos_tree/CMakeLists.txt
index 738823a..728f26a 100644
--- a/kokkos_tree/CMakeLists.txt
+++ b/kokkos_tree/CMakeLists.txt
@@ -1,17 +1,11 @@
 cmake_minimum_required(VERSION 3.21)
-include(GoogleTest)
 if(NOT PROJECT_NAME)
   project(test_dyn_rank CXX)
   set(CMAKE_CXX_STANDARD 20)
   find_package(Kokkos 5.1.1 REQUIRED)
-
   enable_testing()
 endif()
 
 add_executable(Kokkos_qun_test test_dyn_rank.cpp)
 target_link_libraries(Kokkos_qun_test PRIVATE Kokkos::kokkos GTest::gtest)
-
-add_test(
-  NAME    Kokkos_qun_test
-  COMMAND Kokkos_qun_test
-)
+add_test(NAME Kokkos_qun_test COMMAND Kokkos_qun_test)
diff --git a/kokkos_tree/desul/Compare_Exchange_Cuda.hpp b/kokkos_tree/desul/Compare_Exchange_Cuda.hpp
new file mode 100644
index 0000000..8b13789
--- /dev/null
+++ b/kokkos_tree/desul/Compare_Exchange_Cuda.hpp
@@ -0,0 +1 @@
+
diff --git a/kokkos_tree/desul/desul_test.hpp b/kokkos_tree/desul/desul_test.hpp
index 3f9dabc..8e16c90 100644
--- a/kokkos_tree/desul/desul_test.hpp
+++ b/kokkos_tree/desul/desul_test.hpp
@@ -62,4 +62,16 @@ TEST(DesulAtomicTest, CmpExchFailureOrder_TypeMapping) {
           desul::Impl::cmpexch_failure_memory_order<desul::MemoryOrderSeqCst>,
           desul::MemoryOrderSeqCst>::value));
 }
+TEST(DesulNumericLimitsTest,atomics_include_desul_atomics_Common_hpp_L82 ){
+    constexpr uint32_t desul_max = desul::Impl::numeric_limits_max<uint32_t>::value;
+    constexpr uint32_t std_max = std::numeric_limits<uint32_t>::max();
+    EXPECT_EQ(desul_max, std_max) << "Doesn't match";
+    constexpr uint64_t desul_max64 = desul::Impl::numeric_limits_max<uint64_t>::value;
+    constexpr uint64_t std_max64 = std::numeric_limits<uint64_t>::max();
+    EXPECT_EQ(desul_max64, std_max64) << "Doesn't match";
+    auto val = sizeof(int64_t);
+    EXPECT_EQ(val, 8) << "Doesn't match";
+    ASSERT_TRUE((std::is_same_v<desul::Impl::atomic_compare_exchange_t<int64_t>, int64_t>)) << "Doesn't match";
+}
+
 } // namespace Test
\ No newline at end of file
diff --git a/kokkos_tree/kokkos_core/cuda_Error.hpp b/kokkos_tree/kokkos_core/cuda_Error.hpp
new file mode 100644
index 0000000..e63a43f
--- /dev/null
+++ b/kokkos_tree/kokkos_core/cuda_Error.hpp
@@ -0,0 +1,112 @@
+#include <gtest/gtest.h>
+#include <Kokkos_Core.hpp>
+#include <Cuda/Kokkos_Cuda_Error.hpp>
+#include <stdexcept>
+#include <type_traits> // That's for std::is_void_v
+#include <cuda_runtime_api.h>
+#include <Cuda/Kokkos_Cuda_BlockSize_Deduction.hpp>
+namespace Test{
+TEST(KokkosCudaError, InternalErrorThrowsRuntimeError){
+    EXPECT_THROW({
+        Kokkos::Impl::cuda_internal_error_throw(cudaErrorInvalidValue,"PlasmaKernal",__FILE__,__LINE__);
+    }, std::runtime_error);
+}
+struct InternalCudaFunctor{
+    Kokkos::View<int*,Kokkos::CudaSpace> data;
+    InternalCudaFunctor(Kokkos::View<int*,Kokkos::CudaSpace> d):data(d){}
+    KOKKOS_INLINE_FUNCTION void operator()(const int i) const{
+        data(i) = i * 2;
+    }
+};
+
+TEST(Kokkos_internal, InterCudaFunctorCase01){
+    const int N=100000;
+    using CudaPolicy = Kokkos::RangePolicy<Kokkos::Cuda>;
+    Kokkos::View<int*,Kokkos::CudaSpace> data("data",N);
+    Kokkos::parallel_for("TestCudaInternal",CudaPolicy(0,N),InternalCudaFunctor(data));
+    Kokkos::fence();
+    auto h_data = Kokkos::create_mirror_view_and_copy(Kokkos::HostSpace(),data);
+    for(int i=0;i<N;i++){
+        EXPECT_EQ(h_data(i),i*2)<<"Kernel exec failed at index "<<i;
+    }
+}
+TEST(Kokkos_internal, InterCudaFunctorCase02){
+    const int N= 1028;
+    Kokkos::View<int*,Kokkos::CudaSpace> data("data",N);
+    using CudaPolicy = Kokkos::RangePolicy<Kokkos::Cuda>;
+    using Functor = InternalCudaFunctor;
+    CudaPolicy policy(0,N);
+    Functor functor(data);
+    using ImplParallelFor = Kokkos::Impl::ParallelFor<Functor,CudaPolicy,Kokkos::Cuda>;
+    ImplParallelFor impl_parallel_for(functor,policy);
+    impl_parallel_for.execute();
+    Kokkos::Cuda().fence("Wait for internal launch to finish");
+    auto h_data = Kokkos::create_mirror_view_and_copy(Kokkos::HostSpace(),data);
+    for(int i=0;i<N;i++){
+        EXPECT_EQ(h_data(i),i*2)<<"Kernel exec failed at index "<<i;
+    }
+}
+struct SpecialTag {};
+struct TagDispatchFunctor{
+    Kokkos::View<int*,Kokkos::CudaSpace> data;
+    TagDispatchFunctor(Kokkos::View<int*,Kokkos::CudaSpace> d):data(d){}
+    KOKKOS_INLINE_FUNCTION void operator()(const int i) const{
+        data(i) = i * 2;
+    }
+    KOKKOS_INLINE_FUNCTION void operator()(const SpecialTag&,const int i) const{
+        data(i) = i * 3;
+    }
+};
+TEST(kokkos_internal, InterCudaFunctorCase03){
+    const int N = 12800;
+    using KView = Kokkos::View<int*,Kokkos::CudaSpace>;
+    using ImplParallelFor = Kokkos::Impl::ParallelFor<TagDispatchFunctor,Kokkos::RangePolicy<Kokkos::Cuda>,Kokkos::Cuda>;
+    using CudaPolicy = ImplParallelFor::Policy;
+    KView d("data",N);
+    ImplParallelFor imppf(TagDispatchFunctor(d),CudaPolicy(0,N));
+    imppf.execute();
+    Kokkos::Cuda().fence("Wait for void tag launch");
+    auto h_data = Kokkos::create_mirror_view_and_copy(Kokkos::HostSpace(), d);
+    for(int i = 0; i < N; ++i) {
+        EXPECT_EQ(h_data(i), i * 2) << "Void overload failed!";
+    }
+}
+TEST(kokkos_internal, CudaWarpAllocationGranularity){
+    cudaDeviceProp prop;
+    prop.major = 8;
+    prop.minor = 6;
+    EXPECT_EQ(Kokkos::Impl::cuda_warp_per_sm_allocation_granularity(prop), 4u) << "Expected 4 warps per SM for Ampere";
+    prop.major = 6;
+    prop.minor = 0;
+    EXPECT_EQ(Kokkos::Impl::cuda_warp_per_sm_allocation_granularity(prop), 2u) << "Expected 2 warps per SM for Pascal";
+    prop.major = 6;
+    prop.minor = 1;
+    EXPECT_EQ(Kokkos::Impl::cuda_warp_per_sm_allocation_granularity(prop), 4u) << "Expected 4 warps per SM for Pascal";
+    prop.major = 23;
+    prop.minor = 0;
+    EXPECT_THROW({Kokkos::Impl::cuda_warp_per_sm_allocation_granularity(prop);},std::runtime_error) << "Expected runtime error for unknown architecture";
+}
+TEST(kokkos_internal, CudaBlockSizeDeduction){
+    cudaDeviceProp prop;
+    prop.major = 8;
+    prop.sharedMemPerBlockOptin = 100000;
+    prop.reservedSharedMemPerBlock = 1024;
+    size_t shrd_mem = Kokkos::Impl::get_max_shared_mem_per_block(prop);
+    EXPECT_EQ(shrd_mem, 98976) << "Shared memory per block calculation failed for Ampere";
+}
+TEST(kokkos_internal, CudaMaxWarpPerSmRegisters){
+    cudaDeviceProp prop;
+    cudaFuncAttributes attr;
+    prop.major = 8;
+    prop.minor = 6;
+    prop.warpSize = 32;
+    prop.regsPerBlock = 65536;
+    attr.numRegs = 32;
+    int optimal_wraps = Kokkos::Impl::cuda_max_warps_per_sm_registers(prop,attr);
+    EXPECT_EQ(optimal_warps, 64)<< "Expected 32 warps per SM for Ampere with 32 registers per thread";
+    attr.numRegs = 64;
+    int optimal_wraps = Kokkos::Impl::cuda_max_warps_per_sm_registers(prop,attr);
+    EXPECT_EQ(optimal_warps, 32)<< "Expected 16 warps per SM for Ampere with 64 registers per thread";
+    
+}
+}
\ No newline at end of file
diff --git a/kokkos_tree/test_dyn_rank.cpp b/kokkos_tree/test_dyn_rank.cpp
index bd3afae..28a5503 100644
--- a/kokkos_tree/test_dyn_rank.cpp
+++ b/kokkos_tree/test_dyn_rank.cpp
@@ -2,6 +2,7 @@
 #include "kokkos_core/kokkos_pair_investigation.hpp"
 #include "mdspan/kokkos_raw_mdspan.hpp"
 #include "kokkos_core/test_drank.hpp"
+#include "kokkos_core/cuda_Error.hpp"
 #include <Kokkos_Core.hpp>
 #include <gtest/gtest.h>
 int main(int argc, char **argv) {