Add thread count to CPU blocks accelerators (#2338)

Add thread count of a CPU as m_multiProcessorCount for CPU blocks accelerators.
alpaka-group · Aug 4, 2024 · 2d652dd · 2d652dd
1 parent dbc5ebe
commit 2d652dd
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 2 deletions.
diff --git a/include/alpaka/acc/AccCpuOmp2Blocks.hpp b/include/alpaka/acc/AccCpuOmp2Blocks.hpp
@@ -32,6 +32,7 @@
 
 // Implementation details.
 #include "alpaka/acc/Tag.hpp"
+#include "alpaka/core/ClipCast.hpp"
 #include "alpaka/core/Concepts.hpp"
 #include "alpaka/dev/DevCpu.hpp"
 
@@ -135,7 +136,7 @@ namespace alpaka
             ALPAKA_FN_HOST static auto getAccDevProps(DevCpu const& dev) -> alpaka::AccDevProps<TDim, TIdx>
             {
                 return {// m_multiProcessorCount
-                        static_cast<TIdx>(1),
+                        alpaka::core::clipCast<TIdx>(omp_get_max_threads()),
                         // m_gridBlockExtentMax
                         Vec<TDim, TIdx>::all(std::numeric_limits<TIdx>::max()),
                         // m_gridBlockCountMax

diff --git a/include/alpaka/acc/AccCpuTbbBlocks.hpp b/include/alpaka/acc/AccCpuTbbBlocks.hpp
@@ -32,6 +32,7 @@
 
 // Implementation details.
 #include "alpaka/acc/Tag.hpp"
+#include "alpaka/core/ClipCast.hpp"
 #include "alpaka/core/Concepts.hpp"
 #include "alpaka/dev/DevCpu.hpp"
 
@@ -40,6 +41,8 @@
 
 #ifdef ALPAKA_ACC_CPU_B_TBB_T_SEQ_ENABLED
 
+#    include <tbb/tbb.h>
+
 namespace alpaka
 {
     template<typename TDim, typename TIdx, typename TKernelFnObj, typename... TArgs>
@@ -127,7 +130,7 @@ namespace alpaka
             ALPAKA_FN_HOST static auto getAccDevProps(DevCpu const& dev) -> AccDevProps<TDim, TIdx>
             {
                 return {// m_multiProcessorCount
-                        static_cast<TIdx>(1),
+                        alpaka::core::clipCast<TIdx>(tbb::this_task_arena::max_concurrency()),
                         // m_gridBlockExtentMax
                         Vec<TDim, TIdx>::all(std::numeric_limits<TIdx>::max()),
                         // m_gridBlockCountMax