update to replace module inputs, move pass down

src/auto_contiguous.cpp

@@ -76,7 +76,10 @@ void auto_contiguous::apply(module& m) const
 
         if(new_args != args)
         {
-            m.replace_instruction(ins, ins->get_operator(), new_args);
+            if(not ins->module_inputs().empty())
+                m.replace_instruction(ins, ins->get_operator(), new_args, ins->module_inputs());
+            else
+                m.replace_instruction(ins, ins->get_operator(), new_args);
         }
         // shape s = ins->get_shape();
         // if(not s.dynamic() and not s.standard() and s.elements() != 0)

src/targets/gpu/target.cpp

@@ -132,7 +132,6 @@ std::vector<pass> target::get_passes(migraphx::context& gctx, const compile_opti
         dead_code_elimination{},
         prefuse_ops{},
         dead_code_elimination{},
-        auto_contiguous{},
         optimize_module{},
         fuse_pointwise{},
         dead_code_elimination{},
@@ -144,6 +143,7 @@ std::vector<pass> target::get_passes(migraphx::context& gctx, const compile_opti
         dead_code_elimination{},
         enable_pass(mlir_enabled(), fuse_mlir{&ctx}),
         dead_code_elimination{},
+        auto_contiguous{},
         lowering{&ctx, options.offload_copy},
         eliminate_contiguous{"gpu::contiguous"},
         dead_code_elimination{},