change workdivs

alpaka-group · Feb 2, 2024 · 6f4aca3 · 6f4aca3
1 parent 085be5f
commit 6f4aca3
Show file tree

Hide file tree

Showing 3 changed files with 57 additions and 11 deletions.
diff --git a/benchmarks/babelstream/src/AlpakaStream.cpp b/benchmarks/babelstream/src/AlpakaStream.cpp
@@ -49,8 +49,15 @@ struct InitKernel
 template<typename T>
 void AlpakaStream<T>::init_arrays(T initA, T initB, T initC)
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
+
     alpaka::exec<Acc>(
         queue,
         workdiv,
@@ -85,8 +92,14 @@ struct CopyKernel
 template<typename T>
 void AlpakaStream<T>::copy()
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
     alpaka::exec<Acc>(queue, workdiv, CopyKernel{}, alpaka::getPtrNative(d_a), alpaka::getPtrNative(d_c));
     alpaka::wait(queue);
 }
@@ -105,8 +118,14 @@ struct MulKernel
 template<typename T>
 void AlpakaStream<T>::mul()
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
     alpaka::exec<Acc>(queue, workdiv, MulKernel{}, alpaka::getPtrNative(d_b), alpaka::getPtrNative(d_c));
     alpaka::wait(queue);
 }
@@ -124,8 +143,14 @@ struct AddKernel
 template<typename T>
 void AlpakaStream<T>::add()
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
     alpaka::exec<Acc>(
         queue,
         workdiv,
@@ -150,8 +175,14 @@ struct TriadKernel
 template<typename T>
 void AlpakaStream<T>::triad()
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
     alpaka::exec<Acc>(
         queue,
         workdiv,
@@ -176,8 +207,16 @@ struct NstreamKernel
 template<typename T>
 void AlpakaStream<T>::nstream()
 {
-    auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{arraySize / blockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, arraySize);
+
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
+
     alpaka::exec<Acc>(
         queue,
         workdiv,
@@ -222,8 +261,15 @@ struct DotKernel
 template<typename T>
 auto AlpakaStream<T>::dot() -> T
 {
-    auto const workdiv = WorkDiv{dotBlockSize, blockSize, 1};
+    // auto const workdiv = WorkDiv{dotBlockSize, blockSize, 1};
     // auto const workdiv = alpaka::getValidWorkDiv(devAcc, dotBlockSize * blockSize);
+    auto workdiv = alpaka::getValidWorkDiv<Acc>(
+        devAcc,
+        arraySize,
+        alpaka::Vec<Dim, Idx>::ones(),
+        false,
+        alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
+
     alpaka::exec<Acc>(
         queue,
         workdiv,

diff --git a/benchmarks/babelstream/src/AlpakaStream.h b/benchmarks/babelstream/src/AlpakaStream.h
@@ -21,7 +21,7 @@ inline constexpr auto IMPLEMENTATION_STRING = "alpaka";
 using Dim = alpaka::DimInt<1>;
 using Idx = int;
 using Vec = alpaka::Vec<Dim, Idx>;
-using Acc = alpaka::ExampleDefaultAcc<Dim, Idx>;
+using Acc = alpaka::AccGpuCudaRt<Dim, Idx>;
 
 template<typename T>
 struct AlpakaStream : Stream<T>

diff --git a/benchmarks/babelstream/src/main.cpp b/benchmarks/babelstream/src/main.cpp
@@ -88,7 +88,7 @@ void parseArguments(int argc, char* argv[]);
 
 int main(int argc, char* argv[])
 {
-#if !defined(ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLED)
+    // #if !defined(ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLED)
     parseArguments(argc, argv);
 
     if(!output_as_csv)
@@ -102,8 +102,8 @@ int main(int argc, char* argv[])
         run<float>();
     else
         run<double>();
-#else
-#endif
+    // #else
+    // #endif
 }
 
 // Run the 5 main kernels