diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7350d72b1..1ccd2caf5 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,8 @@
 - PR #343 Add in option to statically link against cudart
 
 ## Improvements
+
+- PR #354 Add CMake option for per-thread default stream
 - PR #350 Add .clang-format file & format all files
 
 ## Bug Fixes
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 31462ccfb..cc5e1e214 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -86,6 +86,16 @@ include(FeatureSummary)
 include(CheckIncludeFiles)
 include(CheckLibraryExists)
 
+###################################################################################################
+# - per-thread default stream option --------------------------------------------------------------
+# This needs to be defined first so tests and benchmarks can inherit it.
+
+option(PER_THREAD_DEFAULT_STREAM "Build with per-thread default stream" OFF)
+if(PER_THREAD_DEFAULT_STREAM)
+    message(STATUS "Using per-thread default stream")
+    add_compile_definitions(CUDA_API_PER_THREAD_DEFAULT_STREAM)
+endif(PER_THREAD_DEFAULT_STREAM)
+
 ###################################################################################################
 # - add gtest -------------------------------------------------------------------------------------