Merge pull request #1 from CNugteren/master

fork up to date

2 files changed, 6 insertions, 6 deletions

diff --git a/src/routines/level3/xgemm.cpp b/src/routines/level3/xgemm.cpp
index 30e5999c..f4611aba 100644
--- a/src/routines/level3/xgemm.cpp
+++ b/src/routines/level3/xgemm.cpp
@@ -280,11 +280,11 @@ void Xgemm<T>::GemmDirect(const size_t m, const size_t n, const size_t k,
   kernel.SetArgument(16, static_cast<int>(b_conjugate));
 
   // Computes the global and local thread sizes
-  const auto m_ceiled = Ceil(m, db_["WGD"]);
-  const auto n_ceiled = Ceil(n, db_["WGD"]);
+  //const auto m_ceiled = Ceil(m, db_["WGD"]);
+  //const auto n_ceiled = Ceil(n, db_["WGD"]);
   const auto global = std::vector<size_t>{
-    (m_ceiled * db_["MDIMCD"]) / db_["WGD"],
-    (n_ceiled * db_["NDIMCD"]) / db_["WGD"]
+    CeilDiv(m * db_["MDIMCD"], db_["WGD"]),
+    CeilDiv(n * db_["NDIMCD"], db_["WGD"])
   };
   const auto local = std::vector<size_t>{db_["MDIMCD"], db_["NDIMCD"]};
 
diff --git a/src/routines/levelx/xgemmbatched.cpp b/src/routines/levelx/xgemmbatched.cpp
index 0fea1922..ee8448d2 100644
--- a/src/routines/levelx/xgemmbatched.cpp
+++ b/src/routines/levelx/xgemmbatched.cpp
@@ -94,8 +94,8 @@ void XgemmBatched<T>::DoGemmBatched(const Layout layout, const Transpose a_trans
 
   // Tests the matrices for validity
   for (auto batch = size_t{0}; batch < batch_count; ++batch) {
-    TestMatrixA(a_one, a_two, a_buffer, a_offsets[batch], a_ld);
-    TestMatrixB(b_one, b_two, b_buffer, b_offsets[batch], b_ld);
+    TestMatrixA(a_one, a_two, a_buffer, a_offsets[batch], a_ld, false); // don't test for invalid LD
+    TestMatrixB(b_one, b_two, b_buffer, b_offsets[batch], b_ld, false); // don't test for invalid LD
     TestMatrixC(c_one, c_two, c_buffer, c_offsets[batch], c_ld);
   }