confirmed compiler sanity when skip A-LDS

example/01_gemm/gemm_wmma_fp16.cpp

@@ -36,14 +36,14 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmWmma_CShuffle
            CElementOp,    
            GemmDefault,   
            256,         // BlockSize
-           128,         // MPerBlock
-           128,         // NPerBlock
+           256,         // MPerBlock
+           32,          // NPerBlock
            32,          // KPerBlock
            8,           // K1
            16,          // MPerWmma
            16,          // NPerWmma
-           1,           // M Repeat
-           8,           // N-Repeat
+           4,           // M Repeat
+           1,           // N-Repeat
            S<4, 64, 1>,     
            S<1, 0, 2>,     
            S<1, 0, 2>,              
@@ -51,15 +51,15 @@ using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmWmma_CShuffle
            8,              
            8,      
            true,     
-           S<4, 64, 1>,     
+           S<4, 32, 1>,     
            S<1, 0, 2>,     
            S<1, 0, 2>,             
            2,              
            8,              
            8,      
            true,           
-           1,           // C shuffle (M Repeat) Per store
-           4,           // C shuffle (N Repeat) Per store
+           4,           // C shuffle (M Repeat) Per store
+           1,           // C shuffle (N Repeat) Per store
            S<1, 64, 1,  4>,               
            8>;
 // clang-format on

include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp

@@ -47,7 +47,7 @@ __global__ void
 {
 #if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx1100__) || defined(__gfx1101__) || \
     defined(__gfx1102__))
-    __shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
+    __shared__ char p_shared[GridwiseGemm::SharedMemTrait::lds_size];
 
     GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
                                                   p_b_grid,
@@ -131,8 +131,6 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_wmma
     static constexpr auto I6 = Number<6>{};
     static constexpr auto I7 = Number<7>{};
     
-    static constexpr auto B_K0 = BGridDesc_K0_N_K1{}.GetLength(I0);
-    static constexpr auto B_K1 = BGridDesc_K0_N_K1{}.GetLength(I2);
     // FIX ME: To be deprecated
     static constexpr auto K1 = Number<K1Value>{};
 
@@ -299,6 +297,8 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_wmma
     __host__ __device__ static constexpr auto
     MakeBBlockDescriptor_K0_N0_N1_N2_K1(const BBlockDesc_BK0_N_BK1&)
     {
+        constexpr auto B_K0 = KPerBlock/K1;
+        constexpr auto B_K1 = K1;
         return transform_tensor_descriptor(
             BBlockDesc_BK0_N_BK1{},
             make_tuple(make_pass_through_transform(Number<B_K0>{}),