[DirectX] Set module-level flag LowPrecisionPresent in DXIL Shader Flags Analysis

[Icohedron]

Fixes #114561

[llvmbot]

@llvm/pr-subscribers-backend-directx

Author: Deric Cheung (Icohedron)

Changes

Fixes #114561

---

Full diff: https://github.com/llvm/llvm-project/pull/129109.diff

2 Files Affected:

• (modified) llvm/lib/Target/DirectX/DXILShaderFlags.cpp (+16)
• (added) llvm/test/CodeGen/DirectX/ShaderFlags/low-precision.ll (+38)

diff --git a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
index e1cf516758663..98b8c0a916456 100644
--- a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
+++ b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
@@ -99,6 +99,22 @@ void ModuleShaderFlags::updateFunctionFlags(ComputedShaderFlags &CSF,
     }
   }
 
+  auto IsLowPrecisionType = [](const Type *Ty) {
+    return (Ty->isIntegerTy() && Ty->getScalarSizeInBits() == 16) || Ty->isHalfTy();
+  };
+
+  if (!CSF.LowPrecisionPresent)
+    CSF.LowPrecisionPresent = IsLowPrecisionType(I.getType());
+
+  if (!CSF.LowPrecisionPresent) {
+    for (const Value *Op : I.operands()) {
+      if (IsLowPrecisionType(Op->getType())) {
+        CSF.LowPrecisionPresent = true;
+        break;
+      }
+    }
+  }
+
   if (auto *II = dyn_cast<IntrinsicInst>(&I)) {
     switch (II->getIntrinsicID()) {
     default:
diff --git a/llvm/test/CodeGen/DirectX/ShaderFlags/low-precision.ll b/llvm/test/CodeGen/DirectX/ShaderFlags/low-precision.ll
new file mode 100644
index 0000000000000..fd25a165bfe8d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/ShaderFlags/low-precision.ll
@@ -0,0 +1,38 @@
+; RUN: opt -S --passes="print-dx-shader-flags" 2>&1 %s | FileCheck %s
+; RUN: llc %s --filetype=obj -o - | obj2yaml | FileCheck %s --check-prefix=DXC
+
+target triple = "dxil-pc-shadermodel6.7-library"
+
+;CHECK: ; Combined Shader Flags for Module
+;CHECK-NEXT: ; Shader Flags Value: 0x00000020
+;CHECK-NEXT: ;
+;CHECK-NEXT: ; Note: shader requires additional functionality:
+;CHECK-NEXT: ; Note: extra DXIL module flags:
+;CHECK-NEXT: ;       D3D11_1_SB_GLOBAL_FLAG_ENABLE_MINIMUM_PRECISION
+;CHECK-NEXT: ;
+;CHECK-NEXT: ; Shader Flags for Module Functions
+
+;CHECK-LABEL: ; Function add_i16 : 0x00000020
+define i16 @add_i16(i16 %a, i16 %b) #0 {
+  %sum = add i16 %a, %b
+  ret i16 %sum
+}
+
+;CHECK-LABEL: ; Function add_i32 : 0x00000000
+define i32 @add_i32(i32 %a, i32 %b) #0 {
+  %sum = add i32 %a, %b
+  ret i32 %sum
+}
+
+;CHECK-LABEL: ; Function add_half : 0x00000020
+define half @add_half(half %a, half %b) #0 {
+  %sum = fadd half %a, %b
+  ret half %sum
+}
+
+attributes #0 = { convergent norecurse nounwind "hlsl.export"}
+
+; DXC: - Name:            SFI0
+; DXC-NEXT:     Size:            8
+; DXC-NOT:     Flags:
+; DXC: ...

[spall]

Code LGTM but I don't understand the test enough to comment on if that is sufficient.

[hekota]

LGTM!