[CIR][CodeGen] Support trailing_zeros for constant string literals (#617

)

The patch resolves [issue
#248](#248). It can be considered
a subsequent patch to [#373](#373),
where the case of empty strings was processed.

The new patch adds processing for non-empty strings that may contain
trailing zeros, such as:
```
char big_string[100000] = "123";
```
That is converted to
```
@big_string = #cir.const_array<"123" : !cir.array<!s8i x 3>, trailing_zeros> : !cir.array<!s8i x 100000>
```

clang/lib/CIR/CodeGen/CIRGenBuilder.h

@@ -153,15 +153,26 @@ class CIRGenBuilderTy : public CIRBaseBuilderTy {
                             unsigned size = 0) {
     unsigned finalSize = size ? size : str.size();
 
+    size_t lastNonZeroPos = str.find_last_not_of('\0');
     // If the string is full of null bytes, emit a #cir.zero rather than
     // a #cir.const_array.
-    if (str.count('\0') == str.size()) {
+    if (lastNonZeroPos == llvm::StringRef::npos) {
       auto arrayTy = mlir::cir::ArrayType::get(getContext(), eltTy, finalSize);
       return getZeroAttr(arrayTy);
     }
-
-    auto arrayTy = mlir::cir::ArrayType::get(getContext(), eltTy, finalSize);
-    return getConstArray(mlir::StringAttr::get(str, arrayTy), arrayTy);
+    // We will use trailing zeros only if there are more than one zero
+    // at the end
+    int trailingZerosNum =
+        finalSize > lastNonZeroPos + 2 ? finalSize - lastNonZeroPos - 1 : 0;
+    auto truncatedArrayTy = mlir::cir::ArrayType::get(
+        getContext(), eltTy, finalSize - trailingZerosNum);
+    auto fullArrayTy =
+        mlir::cir::ArrayType::get(getContext(), eltTy, finalSize);
+    return mlir::cir::ConstArrayAttr::get(
+        getContext(), fullArrayTy,
+        mlir::StringAttr::get(str.drop_back(trailingZerosNum),
+                              truncatedArrayTy),
+        trailingZerosNum);
   }
 
   mlir::cir::ConstArrayAttr getConstArray(mlir::Attribute attrs,

clang/test/CIR/CodeGen/globals.c

@@ -8,6 +8,8 @@
 
 char string[] = "whatnow";
 // CHECK: cir.global external @string = #cir.const_array<"whatnow\00" : !cir.array<!s8i x 8>> : !cir.array<!s8i x 8>
+char big_string[100000] = "123";
+// CHECK: cir.global external @big_string = #cir.const_array<"123" : !cir.array<!s8i x 3>, trailing_zeros> : !cir.array<!s8i x 100000>
 int sint[] = {123, 456, 789};
 // CHECK: cir.global external @sint = #cir.const_array<[#cir.int<123> : !s32i, #cir.int<456> : !s32i, #cir.int<789> : !s32i]> : !cir.array<!s32i x 3>
 int filler_sint[4] = {1, 2}; // Ensure missing elements are zero-initialized.
@@ -41,7 +43,7 @@ struct {
   char y[3];
   char z[3];
 } nestedString = {"1", "", "\0"};
-// CHECK: cir.global external @nestedString = #cir.const_struct<{#cir.const_array<"1\00\00" : !cir.array<!s8i x 3>> : !cir.array<!s8i x 3>, #cir.zero : !cir.array<!s8i x 3>, #cir.zero : !cir.array<!s8i x 3>}>
+// CHECK: cir.global external @nestedString = #cir.const_struct<{#cir.const_array<"1" : !cir.array<!s8i x 1>, trailing_zeros> : !cir.array<!s8i x 3>, #cir.zero : !cir.array<!s8i x 3>, #cir.zero : !cir.array<!s8i x 3>}>
 
 struct {
   char *name;

clang/test/CIR/CodeGen/string-literals.c

@@ -0,0 +1,24 @@
+// RUN: %clang_cc1 -triple aarch64-none-linux-android21 -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-android21 -fclangir -emit-llvm %s -o %t.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t.ll %s
+
+struct {
+  char x[10];
+  char y[10];
+  char z[10];
+} literals = {"1", "", "\00"};
+
+// CIR-LABEL: @literals
+// CIR:  #cir.const_struct<{
+// CIR:     #cir.const_array<"1" : !cir.array<!s8i x 1>, trailing_zeros> : !cir.array<!s8i x 10>,
+// CIR:     #cir.zero : !cir.array<!s8i x 10>,
+// CIR:     #cir.zero : !cir.array<!s8i x 10>
+// CIR:  }> 
+
+// LLVM-LABEL: @literals
+// LLVM:  global %struct.anon.1 {
+// LLVM:    [10 x i8] c"1\00\00\00\00\00\00\00\00\00",
+// LLVM:    [10 x i8] zeroinitializer,
+// LLVM:    [10 x i8] zeroinitializer
+// LLVM:  }