[HLSL] Make bool in hlsl i32

[spall]

make a bool's memory representation i32 in hlsl
add new test
fix broken test
Closes #122932

[llvmbot]

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-hlsl

Author: Sarah Spall (spall)

Changes

make a bool's memory representation i32 in hlsl
add new test
fix broken test
Closes #122932

---

Full diff: https://github.com/llvm/llvm-project/pull/122977.diff

3 Files Affected:

• (modified) clang/lib/Basic/TargetInfo.cpp (+1)
• (added) clang/test/CodeGenHLSL/Bool.hlsl (+12)
• (modified) clang/test/SemaHLSL/BuiltIns/asfloat-errors.hlsl (-9)

diff --git a/clang/lib/Basic/TargetInfo.cpp b/clang/lib/Basic/TargetInfo.cpp
index 86befb1cbc74fc..c0bf4e686cf03c 100644
--- a/clang/lib/Basic/TargetInfo.cpp
+++ b/clang/lib/Basic/TargetInfo.cpp
@@ -425,6 +425,7 @@ void TargetInfo::adjust(DiagnosticsEngine &Diags, LangOptions &Opts) {
   // HLSL explicitly defines the sizes and formats of some data types, and we
   // need to conform to those regardless of what architecture you are targeting.
   if (Opts.HLSL) {
+    BoolWidth = BoolAlign = 32;
     LongWidth = LongAlign = 64;
     if (!Opts.NativeHalfType) {
       HalfFormat = &llvm::APFloat::IEEEsingle();
diff --git a/clang/test/CodeGenHLSL/Bool.hlsl b/clang/test/CodeGenHLSL/Bool.hlsl
new file mode 100644
index 00000000000000..fb0f32b11241d4
--- /dev/null
+++ b/clang/test/CodeGenHLSL/Bool.hlsl
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s
+
+// CHECK-LABEL: define noundef i1 {{.*}}fn{{.*}}(i1 noundef %x)
+// CHECK: [[X:%.*]] = alloca i32, align 4
+// CHECK-NEXT: [[Y:%.*]] = zext i1 {{%.*}} to i32
+// CHECK-NEXT: store i32 [[Y]], ptr [[X]], align 4
+// CHECK-NEXT: [[Z:%.*]] = load i32, ptr [[X]], align 4
+// CHECK-NEXT: [[L:%.*]] = trunc i32 [[Z]] to i1
+// CHECK-NEXT: ret i1 [[L]]
+bool fn(bool x) {
+  return x;
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/asfloat-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/asfloat-errors.hlsl
index c2dd9e272e0937..f5f223943b4cdd 100644
--- a/clang/test/SemaHLSL/BuiltIns/asfloat-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/asfloat-errors.hlsl
@@ -24,12 +24,3 @@ float test_float_half(half p1) {
     // expected-note@hlsl/hlsl_detail.h:* {{candidate template ignored: could not match 'vector<half, N>' against 'half'}}
     // expected-note@hlsl/hlsl_detail.h:* {{candidate template ignored: substitution failure [with U = float, T = half]: no type named 'Type'}}
 }
-
-
-float test_float_half(bool p1) {
-    return asfloat(p1);
-    // expected-error@hlsl/hlsl_intrinsics.h:* {{no matching function for call to 'bit_cast'}}
-    // expected-note@-2 {{in instantiation of function template specialization 'hlsl::asfloat<bool>'}}
-    // expected-note@hlsl/hlsl_detail.h:* {{candidate template ignored: could not match 'vector<bool, N>' against 'bool'}}
-    // expected-note@hlsl/hlsl_detail.h:* {{candidate template ignored: substitution failure [with U = float, T = bool]: no type named 'Type'}}
-}

[bob80905]

Presumably, now that p1 is 32 bits, it can be cast to an int for asfloat, so this doesn't emit any errors, right?

[spall]

Yeah no errors were being emitted, causing the old test to fail (due to lack of errors).