WIP: Enabling gfx90a build

example/67_gemm_microscaling/gemm_mx_common.hpp

@@ -19,8 +19,6 @@
 #include "ck/library/utility/fill.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 
-using ScaleDataType = ck::e8m0_bexp_t;
-
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
 

example/67_gemm_microscaling/gemm_mx_fp8.cpp

@@ -13,11 +13,15 @@ using XDataType = ck::e8m0_bexp_t;
 #endif
 using AccDataType      = float;
 using CShuffleDataType = float;
-using CDataType        = float;
+#if 1
+using CDataType = ck::half_t;
+#else
+using CDataType = float;
+#endif
 
 using ALayout = Row;
 using BLayout = Col;
-using CLayout = Row;
+using CLayout = Col;
 
 using CElementOp = PassThrough; // elementwise transformation for C matrix