diff --git a/cmake/GPU.cmake b/cmake/GPU.cmake
index eb261d1521..a9388bb340 100644
--- a/cmake/GPU.cmake
+++ b/cmake/GPU.cmake
@@ -115,6 +115,20 @@ elseif (CMAKE_Fortran_COMPILER_ID STREQUAL "Flang")
     if (CMAKE_BUILD_TYPE STREQUAL "Debug" OR CMAKE_BUILD_TYPE STREQUAL "RelDebug")
         add_compile_options($<$<COMPILE_LANGUAGE:Fortran>:-O1> $<$<COMPILE_LANGUAGE:Fortran>:-g>)
     endif()
+elseif (CMAKE_Fortran_COMPILER_ID STREQUAL "LLVMFlang")
+    # AMD/LLVM flang (amdflang). These are CPU-build flags only; the OpenMP-offload GPU
+    # build sets its own flags (MFCTargets.cmake) and links via the offload wrapper, so
+    # it is left untouched here.
+    if (NOT (MFC_OpenMP OR MFC_OpenACC))
+        if (CMAKE_BUILD_TYPE STREQUAL "Debug" OR CMAKE_BUILD_TYPE STREQUAL "RelDebug")
+            # Ordinary (non-LTO) objects plus static deps built without -fPIC
+            # (e.g. libfftw3) hit ld.lld's default-PIE "relocation R_X86_64_32
+            # against local symbol". Link a non-PIE executable to accept them.
+            # Release links fine via LTO's link-time PIC codegen, so it is left alone.
+            add_link_options(-no-pie)
+            add_compile_options($<$<COMPILE_LANGUAGE:Fortran>:-O1> $<$<COMPILE_LANGUAGE:Fortran>:-g>)
+        endif()
+    endif()
 elseif (CMAKE_Fortran_COMPILER_ID STREQUAL "Intel")
     add_compile_options($<$<COMPILE_LANGUAGE:Fortran>:-free>)