[llvm] [AMDGPU][GlobalISel] Add regbankselect rules for G_FSHR (PR #159818)

[Petar Avramovic via llvm-commits]

================
@@ -0,0 +1,30 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 < %s | FileCheck %s
+
+define amdgpu_ps void @uniform_fshr_i32(i32 inreg %lhs, i32 inreg %rhs, i32 inreg %amt, ptr %resptr) {
+; CHECK-LABEL: uniform_fshr_i32:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; CHECK-NEXT:    v_alignbit_b32 v2, s0, s1, v2
+; CHECK-NEXT:    flat_store_b32 v[0:1], v2
+; CHECK-NEXT:    s_endpgm
+  %vres = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
+  store i32 %vres, ptr %resptr
+  ret void
----------------
petar-avramovic wrote:

Just one more nit, have salu use for uniform test
```suggestion
  %vres = call i32 @llvm.fshr.i32(i32 %lhs, i32 %rhs, i32 %amt)
  %add = add i32 %vres, %vres
  store i32 %add, ptr %resptr
  ret void
```

https://github.com/llvm/llvm-project/pull/159818