AMDGPU: Remove dx10-clamp from subtarget features

Since this can be set with s_setreg*, it should not be a subtarget
property. Set a default based on the calling convention, and Introduce
a new amdgpu-dx10-clamp attribute to override this if desired.

Also introduce a new amdgpu-ieee attribute to match.

The values need to match to allow inlining. I think it is OK for the
caller's dx10-clamp attribute to override the callee, but there
doesn't appear to be the infrastructure to do this currently without
definining the attribute in the generic Attributes.td.

Eventually the calling convention lowering will need to insert a mode
switch somewhere for these.

llvm-svn: 357302

2 files changed, 52 insertions, 0 deletions

diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
index b397554e76d..819c06df158 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -1002,6 +1002,19 @@ bool splitMUBUFOffset(uint32_t Imm, uint32_t &SOffset, uint32_t &ImmOffset,
   return true;
 }
 
+SIModeRegisterDefaults::SIModeRegisterDefaults(const Function &F) {
+  *this = getDefaultForCallingConv(F.getCallingConv());
+
+  StringRef IEEEAttr = F.getFnAttribute("amdgpu-ieee").getValueAsString();
+  if (!IEEEAttr.empty())
+    IEEE = IEEEAttr == "true";
+
+  StringRef DX10ClampAttr
+    = F.getFnAttribute("amdgpu-dx10-clamp").getValueAsString();
+  if (!DX10ClampAttr.empty())
+    DX10Clamp = DX10ClampAttr == "true";
+}
+
 namespace {
 
 struct SourceOfDivergence {
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index 174fec47986..2943722963a 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -495,6 +495,45 @@ bool splitMUBUFOffset(uint32_t Imm, uint32_t &SOffset, uint32_t &ImmOffset,
 /// \returns true if the intrinsic is divergent
 bool isIntrinsicSourceOfDivergence(unsigned IntrID);
 
+
+// Track defaults for fields in the MODE registser.
+struct SIModeRegisterDefaults {
+  /// Floating point opcodes that support exception flag gathering quiet and
+  /// propagate signaling NaN inputs per IEEE 754-2008. Min_dx10 and max_dx10
+  /// become IEEE 754- 2008 compliant due to signaling NaN propagation and
+  /// quieting.
+  bool IEEE : 1;
+
+  /// Used by the vector ALU to force DX10-style treatment of NaNs: when set,
+  /// clamp NaN to zero; otherwise, pass NaN through.
+  bool DX10Clamp : 1;
+
+  // TODO: FP mode fields
+
+  SIModeRegisterDefaults() :
+    IEEE(true),
+    DX10Clamp(true) {}
+
+  SIModeRegisterDefaults(const Function &F);
+
+  static SIModeRegisterDefaults getDefaultForCallingConv(CallingConv::ID CC) {
+    SIModeRegisterDefaults Mode;
+    Mode.DX10Clamp = true;
+    Mode.IEEE = AMDGPU::isCompute(CC);
+    return Mode;
+  }
+
+  bool operator ==(const SIModeRegisterDefaults Other) const {
+    return IEEE == Other.IEEE && DX10Clamp == Other.DX10Clamp;
+  }
+
+  // FIXME: Inlining should be OK for dx10-clamp, since the caller's mode should
+  // be able to override.
+  bool isInlineCompatible(SIModeRegisterDefaults CalleeMode) const {
+    return *this == CalleeMode;
+  }
+};
+
 } // end namespace AMDGPU
 } // end namespace llvm