[PATCH] D20736: [AMDGPU] Implement AMDGPU relocations in lld

[Tom Stellard via llvm-commits]

tstellarAMD added inline comments.

================
Comment at: ELF/Target.cpp:1259-1269
@@ -1262,2 +1258,13 @@
                                    uint64_t Val) const {
-  llvm_unreachable("not implemented");
+  switch (Type) {
+  case R_AMDGPU_32_LOW:
+    write32le(Loc, Val);
+    break;
+  case R_AMDGPU_32_HIGH:
+    write32le(Loc, (Val >> 32) & 0xFFFFFFFF);
+    break;
+  case R_AMDGPU_64:
+    write64le(Loc, Val);
+    break;
+  default:
+    fatal("unrecognized reloc " + Twine(Type));
----------------
Does this function actually get called for absolute relocations?  I thought it was only called when the relocated value could be computed by the linker.


http://reviews.llvm.org/D20736