[clang] [CIR] Upstream a batch of passing tests from CIR-Incubator (PR #157333)

[Andy Kaylor via cfe-commits]

================
@@ -0,0 +1,264 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir -mmlir --mlir-print-ir-before=cir-lowering-prepare -o %t.cir %s 2>&1 | FileCheck --check-prefixes=CIR-BEFORE,CHECK %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir -mmlir --mlir-print-ir-after=cir-lowering-prepare -o %t.cir %s 2>&1 | FileCheck --check-prefixes=CIR-AFTER,CHECK %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm -o %t.ll %s
+// RUN: FileCheck --input-file=%t.ll --check-prefixes=LLVM,CHECK %s
+
+#include <stdbool.h>
+
+volatile double _Complex cd;
+volatile float _Complex cf;
+volatile int _Complex ci;
+volatile short _Complex cs;
+volatile double sd;
+volatile int si;
+volatile bool b;
+
+void scalar_to_complex() {
+  cd = sd;
+  ci = si;
+  cd = si;
+  ci = sd;
+}
+
+// CHECK-LABEL: @scalar_to_complex()
+
+// CIR-BEFORE: %{{.+}} = cir.cast(float_to_complex, %{{.+}} : !cir.double), !cir.complex<!cir.double>
+
+//      CIR-AFTER: %[[#REAL:]] = cir.load volatile{{.*}}  %{{.+}} : !cir.ptr<!cir.double>, !cir.double
+// CIR-AFTER-NEXT: %[[#IMAG:]] = cir.const #cir.fp<0.000000e+00> : !cir.double
+// CIR-AFTER-NEXT: %{{.+}} = cir.complex.create %[[#REAL]], %[[#IMAG]] : !cir.double -> !cir.complex<!cir.double>
+
+// CIR-BEFORE: %{{.+}} = cir.cast(int_to_complex, %{{.+}} : !s32i), !cir.complex<!s32i>
+
+//      CIR-AFTER: %[[#REAL:]] = cir.load volatile{{.*}}  %{{.+}} : !cir.ptr<!s32i>, !s32i
+// CIR-AFTER-NEXT: %[[#IMAG:]] = cir.const #cir.int<0> : !s32i
+// CIR-AFTER-NEXT: %{{.+}} = cir.complex.create %[[#REAL]], %[[#IMAG]] : !s32i -> !cir.complex<!s32i>
+
+//      CIR-BEFORE: %[[#A:]] = cir.cast(int_to_float, %{{.+}} : !s32i), !cir.double
+// CIR-BEFORE-NEXT: %{{.+}} = cir.cast(float_to_complex, %[[#A]] : !cir.double), !cir.complex<!cir.double>
+
+//      CIR-AFTER: %[[#A:]] = cir.load volatile{{.*}}  %{{.+}} : !cir.ptr<!s32i>, !s32i
+// CIR-AFTER-NEXT: %[[#REAL:]] = cir.cast(int_to_float, %[[#A]] : !s32i), !cir.double
+// CIR-AFTER-NEXT: %[[#IMAG:]] = cir.const #cir.fp<0.000000e+00> : !cir.double
+// CIR-AFTER-NEXT: %{{.+}} = cir.complex.create %[[#REAL]], %[[#IMAG]] : !cir.double -> !cir.complex<!cir.double>
+
+//      CIR-BEFORE: %[[#A:]] = cir.cast(float_to_int, %{{.+}} : !cir.double), !s32i
+// CIR-BEFORE-NEXT: %{{.+}} = cir.cast(int_to_complex, %[[#A]] : !s32i), !cir.complex<!s32i>
+
+//      CIR-AFTER: %[[#A:]] = cir.load volatile{{.*}}  %{{.+}} : !cir.ptr<!cir.double>, !cir.double
+// CIR-AFTER-NEXT: %[[#REAL:]] = cir.cast(float_to_int, %[[#A]] : !cir.double), !s32i
+// CIR-AFTER-NEXT: %[[#IMAG:]] = cir.const #cir.int<0> : !s32i
+// CIR-AFTER-NEXT: %{{.+}} = cir.complex.create %[[#REAL]], %[[#IMAG]] : !s32i -> !cir.complex<!s32i>
+
+//      LLVM: %[[#REAL:]] = load volatile double, ptr @sd, align 8
+// LLVM-NEXT: %[[#A:]] = insertvalue { double, double } {{.*}}, double %[[#REAL]], 0
+// LLVM-NEXT: %{{.+}} = insertvalue { double, double } %[[#A]], double 0.000000e+00, 1
+
+//      LLVM: %[[#REAL:]] = load volatile i32, ptr @si, align 4
----------------
andykaylor wrote:

One of the significant differences in how we're implementing tests upstream is that when we have a test that checks lowering to LLVM IR like this we also add checks for direct LLVM IR generation without going through CIR as a comparison.  You can find examples in any test with `OGCG` checks.

Maybe remove any tests that are checking LLVM IR output and update them in this way in a later PR?

https://github.com/llvm/llvm-project/pull/157333