AArch64: correct CodeGen of MOVZ/MOVK combinations.

According to the AArch64 ELF specification (4.6.8), it's the
assembler's responsibility to make sure the shift amount is correct in
relocated MOVZ/MOVK instructions.

This wasn't being obeyed by either the MCJIT CodeGen or RuntimeDyldELF
(which happened to work out well for JIT tests). This commit should
make us compliant in this area.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185360 91177308-0d34-0410-b5e6-96231b3b80d8

test/CodeGen/AArch64/movw-shift-encoding.ll

@@ -0,0 +1,14 @@
+; RUN: llc -mtriple=aarch64-linux-gnu < %s -show-mc-encoding -code-model=large | FileCheck %s
+
+@var = global i32 0
+
+; CodeGen should ensure that the correct shift bits are set, because the linker
+; isn't going to!
+
+define i32* @get_var() {
+  ret i32* @var
+; CHECK: movz    x0, #:abs_g3:var        // encoding: [A,A,0xe0'A',0xd2'A']
+; CHECK: movk    x0, #:abs_g2_nc:var     // encoding: [A,A,0xc0'A',0xf2'A']
+; CHECK: movk    x0, #:abs_g1_nc:var     // encoding: [A,A,0xa0'A',0xf2'A']
+; CHECK: movk    x0, #:abs_g0_nc:var     // encoding: [A,A,0x80'A',0xf2'A']
+}