trying to make stuff work

CMakeLists.txt

@@ -102,8 +102,8 @@ cmake_dependent_option( UMPIRE_ENABLE_BENCHMARKS "Build Umpire benchmarks" On
 cmake_dependent_option( UMPIRE_ENABLE_EXAMPLES "Build Umpire examples" On
                         "ENABLE_EXAMPLES" Off )
 
-cmake_dependent_option( UMPIRE_ENABLE_DOCS "Build Umpire docs" Off
-                        "ENABLE_DOCS" Off )
+cmake_dependent_option( UMPIRE_ENABLE_DOCS "Build Umpire docs" On
+                        "ENABLE_DOCS" On )
 cmake_dependent_option( UMPIRE_ENABLE_CLANGQUERY "Build Umpire with Clang query" On
                         "ENABLE_CLANGQUERY" Off )
 cmake_dependent_option( UMPIRE_ENABLE_COVERAGE "Build Umpire with Coverage support (with GCC)" On

examples/sap_test.cpp

@@ -36,7 +36,7 @@ __global__ void touch_data(double* data, int len)
 __global__ void do_sleep()
 {
   //sleep - works still at 1000, so keeping it at 100k
-  sleep(1000);
+  sleep(1000000);
 }
 
 __global__ void check_data(double* data, int len)
@@ -62,7 +62,8 @@ __global__ void touch_data_again(double* data, int len)
 int main(int, char**)
 {
   auto& rm = umpire::ResourceManager::getInstance();
-  auto pool = rm.makeAllocator<umpire::strategy::StreamAwareQuickPool>("sap-pool", rm.getAllocator("DEVICE"));
+  //auto pool = rm.makeAllocator<umpire::strategy::StreamAwareQuickPool>("sap-pool", rm.getAllocator("DEVICE"));
+  auto pool = umpire::strategy::StreamAwareQuickPool("sap-pool", 150, rm.getAllocator("DEVICE"), 32, 32);
   int NUM_BLOCKS = NUM_THREADS / BLOCK_SIZE;
 
   cudaStream_t s1, s2;
@@ -78,7 +79,7 @@ int main(int, char**)
   check_data<<<NUM_BLOCKS, BLOCK_SIZE, 0, s1>>>(a, NUM_THREADS);
 
   //deallocate and reallocate a using different streams
-  pool.deallocate(s1, a);
+  pool.deallocate(s1, a, NUM_THREADS * sizeof(double));
   a = static_cast<double*>(pool.allocate(s2, NUM_THREADS * sizeof(double)));
 
   //with stream s2, use memory in reallocated a in kernel
@@ -101,7 +102,7 @@ int main(int, char**)
   std::cout << "Kernel succeeded! Expected result returned" << std::endl;
 
   //final deallocations
-  pool.deallocate(s2, a);
+  pool.deallocate(s2, a, NUM_THREADS * sizeof(double));
   rm.deallocate(b);
   return 0;
 }

src/umpire/strategy/StreamAwareQuickPool.cpp

@@ -43,6 +43,7 @@ StreamAwareQuickPool::~StreamAwareQuickPool()
 
 void* StreamAwareQuickPool::allocate(void* stream, std::size_t bytes)
 {
+/*
   unsigned int size = m_registered_streams.size();
   UMPIRE_LOG(Debug, "Size of registered streams vector is: " << size);
   
@@ -55,6 +56,7 @@ void* StreamAwareQuickPool::allocate(void* stream, std::size_t bytes)
     
   UMPIRE_LOG(Debug, "I did not find a registered stream so I am adding it to vector.");
   m_registered_streams.push_back(stream);
+*/
   return allocate(bytes);
 }
 
@@ -165,11 +167,12 @@ void StreamAwareQuickPool::deallocate(void* stream, void* ptr, std::size_t size)
       m_registered_dealloc.at(i) = deallocate_has_occurred;
     }
   }
-
+/*
   UMPIRE_ERROR(
         runtime_error,
         umpire::fmt::format("Invalid deallocate: {} stream has not been allocated yet", stream));
   deallocate(ptr, size);
+*/
 }
 
 void StreamAwareQuickPool::deallocate(void* ptr, std::size_t UMPIRE_UNUSED_ARG(size))

src/umpire/strategy/StreamAwareQuickPool.hpp

@@ -66,7 +66,7 @@ class StreamAwareQuickPool : public AllocationStrategy, private mixins::AlignedA
 
   ~StreamAwareQuickPool();
 
-  StreamAwareQuickPool(const StreamAwareQuickPool&) = delete;
+  StreamAwareQuickPool(const StreamAwareQuickPool&);
 
  private:
   void* allocate(std::size_t bytes);
@@ -168,7 +168,7 @@ class StreamAwareQuickPool : public AllocationStrategy, private mixins::AlignedA
   const std::size_t m_first_minimum_pool_allocation_size;
   const std::size_t m_next_minimum_pool_allocation_size;
 
-  std::vector<void*> m_registered_streams{0};
+  std::vector<void*> m_registered_streams{};
   std::vector<camp::resources::Event> m_registered_dealloc{};
 
   std::size_t m_total_blocks{0};