[PATCH] D54749: Saturating float to int casts: Basics [1/n]

[Tim Northover via Phabricator via llvm-commits]

t.p.northover added inline comments.


================
Comment at: docs/LangRef.rst:13105
+
+      declare i32 @llvm.fptoui.sat.f32.i32(float %f)
+      declare i19 @llvm.fptoui.sat.f64.i19(double %f)
----------------
What's going on here? All the other intrinsics I know specify the return type first in the prototype. I didn't even think it was something you could override.


================
Comment at: lib/CodeGen/SelectionDAG/LegalizeDAG.cpp:2524
+  while (true) {
+    NewOutTy = (MVT::SimpleValueType)(NewOutTy.getSimpleVT().SimpleTy+1);
+    if (TLI.isOperationLegalOrCustom(Opcode, NewOutTy)) {
----------------
Maybe put the normal assertion in here?

        assert(NewOutTy.isInteger() && "Ran out of possibilities!");



================
Comment at: lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp:5686
+                             getValue(I.getArgOperand(0)),
+                             DAG.getValueType(Type)));
+    return nullptr;
----------------
I was expecting this tag to be the scalar type in the vector situation. Does something get neater this way round?


Repository:
  rL LLVM

https://reviews.llvm.org/D54749