Merge pull request #1450 from LLNL/v2022.10.5-RC

Pull in changes to fix separable compilation for CUDA or HIP.

Author: rhornung67

CMakeLists.txt

@@ -16,7 +16,7 @@ include(CMakeDependentOption)
 # Set version number
 set(RAJA_VERSION_MAJOR 2022)
 set(RAJA_VERSION_MINOR 10)
-set(RAJA_VERSION_PATCHLEVEL 4)
+set(RAJA_VERSION_PATCHLEVEL 5)
 
 if (RAJA_LOADED AND (NOT RAJA_LOADED STREQUAL "${RAJA_VERSION_MAJOR}.${RAJA_VERSION_MINOR}.${RAJA_VERSION_PATCHLEVEL}"))
   message(FATAL_ERROR "You are mixing RAJA versions. Loaded is ${RAJA_LOADED}, expected ${RAJA_VERSION_MAJOR}.${RAJA_VERSION_MINOR}.${RAJA_VERSION_PATCHLEVEL}")

RELEASE_NOTES.md

@@ -19,7 +19,17 @@ Notable changes include:
 
   * Bug fixes/improvements:
 
-Version 2022.10.4 -- Release date 2022-14-01
+Version 2022.10.5 -- Release date 2023-02-28
+============================================
+
+This release fixes an issue that was found after the v2022.10.4 release.
+
+  * Fixes CUDA and HIP separable compilation option that was broken before the 
+    v2022.10.0 release. For the curious reader, the issue was that resources
+    were constructed and calling CUDA/HIP API routines before either runtime
+    was initialized.
+
+Version 2022.10.4 -- Release date 2022-12-14
 ============================================
 
 This release fixes an issue that was found after the v2022.10.3 release.

docs/conf.py

@@ -79,7 +79,7 @@
 
 # General information about the project.
 project = u'RAJA'
-copyright = u'2016-2022, Lawrence Livermore National Security, LLNS'
+copyright = u'2016-2023, Lawrence Livermore National Security, LLNS'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
@@ -88,7 +88,7 @@
 # The short X.Y version.
 version = u'2022.10'
 # The full version, including alpha/beta/rc tags.
-release = u'2022.10.4'
+release = u'2022.10.5'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

include/RAJA/policy/cuda/MemUtils_CUDA.hpp

@@ -118,7 +118,7 @@ namespace detail
 struct cudaInfo {
   cuda_dim_t gridDim{0, 0, 0};
   cuda_dim_t blockDim{0, 0, 0};
-  ::RAJA::resources::Cuda res;
+  ::RAJA::resources::Cuda res{::RAJA::resources::Cuda::CudaFromStream(0,0)}; 
   bool setup_reducers = false;
 #if defined(RAJA_ENABLE_OPENMP)
   cudaInfo* thread_states = nullptr;

include/RAJA/policy/hip/MemUtils_HIP.hpp

@@ -121,7 +121,7 @@ namespace detail
 struct hipInfo {
   hip_dim_t gridDim = 0;
   hip_dim_t blockDim = 0;
-  ::RAJA::resources::Hip res;
+  ::RAJA::resources::Hip res{::RAJA::resources::Hip::HipFromStream(0,0)};
   bool setup_reducers = false;
 #if defined(RAJA_ENABLE_OPENMP)
   hipInfo* thread_states = nullptr;