Fix undefined symbol on CUDA 11.8

The common symbol was compiled as part of a Hopper kernel.

build.toml

@@ -148,7 +148,6 @@ src = [
     "cutlass_w8a8/c3x/scaled_mm_sm90_int8_dispatch.cuh",
     "cutlass_w8a8/c3x/scaled_mm_helper.hpp",
     "cutlass_w8a8/scaled_mm_c3x_sm90.cu",
-    "cutlass_extensions/common.cpp",
     "cutlass_extensions/common.hpp",
     "cutlass_extensions/epilogue/scaled_mm_epilogues_c3x.hpp",
     "cutlass_extensions/epilogue/broadcast_load_epilogue_c3x.hpp",
@@ -212,6 +211,8 @@ src = [
     "cutlass_w8a8/scaled_mm_c2x_sm89_fp8_dispatch.cuh",
     "cutlass_w8a8/scaled_mm_c2x_sm89_int8_dispatch.cuh",
     "cutlass_w8a8/scaled_mm_entry.cu",
+    "cutlass_extensions/common.cpp",
+    "cutlass_extensions/common.hpp",
     "cutlass_extensions/epilogue/scaled_mm_epilogues_c2x.hpp",
     "cutlass_extensions/epilogue/broadcast_load_epilogue_c2x.hpp",
 ]