[clang] [CIR] Fix Codegen for Comparison of Complex bin ops (PR #185316)

[Henrich Lauko via cfe-commits]

================
@@ -1867,3 +1867,83 @@ bool ne_float_and_float_complex(float a, float _Complex b) {
 // OGCG: %[[REAL_CMP:.*]] = fcmp une float %[[TMP_A]], %[[B_REAL]]
 // OGCG: %[[IMAG_CMP:.*]] = fcmp une float 0.000000e+00, %[[B_IMAG]]
 // OGCG: %[[RESULT:.*]] = or i1 %[[REAL_CMP]], %[[IMAG_CMP]]
+
+void compare_two_complex_bin_ops() {
+  double _Complex a;
+  double _Complex b;
+  bool c = a + b != b + a;
+}
+
+// CIR: %[[A_ADDR:.*]] = cir.alloca !cir.complex<!cir.double>, !cir.ptr<!cir.complex<!cir.double>>, ["a"]
+// CIR: %[[B_ADDR:.*]] = cir.alloca !cir.complex<!cir.double>, !cir.ptr<!cir.complex<!cir.double>>, ["b"]
+// CIR: %[[C_ADDR:.*]] = cir.alloca !cir.bool, !cir.ptr<!cir.bool>, ["c", init]
+// CIR: %[[TMP_A:.*]] = cir.load {{.*}} %[[A_ADDR]] : !cir.ptr<!cir.complex<!cir.double>>, !cir.complex<!cir.double>
+// CIR: %[[TMP_B:.*]] = cir.load {{.*}} %[[B_ADDR]] : !cir.ptr<!cir.complex<!cir.double>>, !cir.complex<!cir.double>
+// CIR: %[[COMPLEX_AB:.*]] = cir.complex.add %[[TMP_A]], %[[TMP_B]] : !cir.complex<!cir.double>
+// CIR: %[[TMP_B:.*]] = cir.load {{.*}} %[[B_ADDR]] : !cir.ptr<!cir.complex<!cir.double>>, !cir.complex<!cir.double>
+// CIR: %[[TMP_A:.*]] = cir.load {{.*}} %[[A_ADDR]] : !cir.ptr<!cir.complex<!cir.double>>, !cir.complex<!cir.double>
+// CIR: %[[COMPLEX_BA:.*]] = cir.complex.add %[[TMP_B]], %[[TMP_A]] : !cir.complex<!cir.double>
+// CIR: %[[RESULT:.*]] = cir.cmp(ne, %[[COMPLEX_AB]], %[[COMPLEX_BA]]) : !cir.complex<!cir.double>, !cir.bool
----------------
xlauko wrote:

Please rabase after I land https://github.com/llvm/llvm-project/pull/185114 it will break format here

https://github.com/llvm/llvm-project/pull/185316