summaryrefslogtreecommitdiff
path: root/src/tuning/kernels/xgemm_direct.hpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/tuning/kernels/xgemm_direct.hpp')
-rw-r--r--src/tuning/kernels/xgemm_direct.hpp9
1 files changed, 9 insertions, 0 deletions
diff --git a/src/tuning/kernels/xgemm_direct.hpp b/src/tuning/kernels/xgemm_direct.hpp
index ecb10bc6..baa063c0 100644
--- a/src/tuning/kernels/xgemm_direct.hpp
+++ b/src/tuning/kernels/xgemm_direct.hpp
@@ -135,6 +135,15 @@ std::vector<Constraint> XgemmDirectSetConstraints(const int V) {
}
return constraints;
}
+template <typename T>
+LocalMemSizeInfo XgemmDirectComputeLocalMemSize(const int) {
+ return {
+ [] (std::vector<size_t> v) -> size_t {
+ return GetBytes(PrecisionValue<T>()) * ((v[0]*(v[0] + v[1]) + v[0]*(v[0] + v[2])));
+ },
+ {"WGD", "PADA", "PADB"}
+ };
+}
// Sets the kernel's arguments
template <typename T>