[llvm] [GlobalISel][LLT] Introduce FPInfo for LLT (Enable bfloat, ppc128float and others in GlobalISel) (PR #155107)

[Pierre van Houtryve via llvm-commits]

================
@@ -36,6 +36,41 @@ LLT llvm::getLLTForType(Type &Ty, const DataLayout &DL) {
     // concerned.
     auto SizeInBits = DL.getTypeSizeInBits(&Ty);
     assert(SizeInBits != 0 && "invalid zero-sized type");
+
+    // Return simple scalar
+    if (!AllowExtendedLLT)
+      return LLT::scalar(SizeInBits);
+
+    // Choose more precise LLT variant
+    if (Ty.isFloatingPointTy()) {
+      if (Ty.isHalfTy())
+        return LLT::float16();
+
+      if (Ty.isBFloatTy())
+        return LLT::bfloat16();
+
+      if (Ty.isFloatTy())
+        return LLT::float32();
+
+      if (Ty.isDoubleTy())
+        return LLT::float64();
+
+      if (Ty.isX86_FP80Ty())
+        return LLT::x86fp80();
+
+      if (Ty.isFP128Ty())
+        return LLT::float128();
+
+      if (Ty.isPPC_FP128Ty())
+        return LLT::ppcf128();
+
+      llvm_unreachable("Unhandled LLVM IR floating point type");
+    }
+
+    if (Ty.isIntegerTy()) {
+      return LLT::integer(SizeInBits);
+    }
----------------
Pierre-vh wrote:

```suggestion
    if (Ty.isIntegerTy()) 
      return LLT::integer(SizeInBits);
```

https://github.com/llvm/llvm-project/pull/155107