fixed

CMakeLists.txt

@@ -543,7 +543,7 @@ ENDIF()
 ENDFOREACH()
 
 add_custom_target(instances DEPENDS utility;${CK_DEVICE_INSTANCES}  SOURCES ${INSTANCE_FILES})
-#add_subdirectory(library)
+add_subdirectory(library)
 
 if(NOT GPU_ARCHS)
    rocm_package_setup_component(tests
@@ -565,7 +565,7 @@ rocm_package_setup_component(profiler
     LIBRARY_NAME composablekernel
     PACKAGE_NAME ckprofiler
 )
-#add_subdirectory(profiler)
+add_subdirectory(profiler)
 
 #if(CK_USE_CODEGEN AND (GPU_TARGETS MATCHES "gfx9" OR GPU_ARCHS))
   #add_subdirectory(codegen)

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp

@@ -1756,7 +1756,7 @@ struct GridwiseGemm_xdl_cshuffle_v3
             static_cast<ADataType*>(p_shared_0), a_block_desc_ak0_m_ak1.GetElementSpaceSize());
 
         auto b_block_buf_ping = make_dynamic_buffer<AddressSpaceEnum::Lds>(
-            static_cast<BDataType*>(static_cast<char*>(p_shared_0) +
+            bit_cast<BDataType*>(static_cast<char*>(p_shared_0) +
                                     a_block_space_size_aligned * sizeof(ADataType)),
             b_block_desc_bk0_n_bk1.GetElementSpaceSize());
 

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v3r1.hpp

@@ -381,8 +381,8 @@ struct ThreadwiseTensorSliceTransfer_v3r1
                       (is_same<f8_t, remove_cvref_t<DstData>>::value &&
                        SrcScalarPerVector % 4 == 0 && DstScalarPerVector % 4 == 0)))
         {
-            static_assert(is_same_v<remove_cvref_t<SrcData>, pk_i4_t>,
-                          "transpose is not allowed for pk_i4_t");
+            //static_assert(is_same_v<remove_cvref_t<SrcData>, pk_i4_t>,
+                          //"transpose is not allowed for pk_i4_t");
 #if 1
             // each transpose does
             // DstScalarPerVector # of src vectors in src_thread_scratch_