[PATCH] D28546: LowerTypeTests: Implement importing of type identifiers.

[Peter Collingbourne via Phabricator via llvm-commits]

pcc marked an inline comment as done.
pcc added inline comments.


================
Comment at: llvm/lib/Transforms/IPO/LowerTypeTests.cpp:735
+LowerTypeTestsModule::TypeIdLowering
+LowerTypeTestsModule::importTypeId(StringRef TypeId) {
+  TypeTestResolution &TTRes = Summary->getTypeIdSummary(TypeId).TTRes;
----------------
eugenis wrote:
> This function does not import SizeM1BitWidth.
> I realize it is unused, but this could be a surprise for someone later, comments in TypeIdLowering say that it should be set for ByteArray, Inline, AllOnes.
> Do we even need this field? It looks like it could be reconstructed from SizeM1.
> 
> 
No, we don't need it. Removed in r292647.


================
Comment at: llvm/lib/Transforms/IPO/LowerTypeTests.cpp:746
+    auto *GV = dyn_cast<GlobalVariable>(C);
+    if (!GV || GV->getVisibility() == GlobalValue::HiddenVisibility)
+      return C;
----------------
eugenis wrote:
> I don't get this condition. Why would this variable exist and be non-hidden?
If `getOrInsertGlobal` created the global, it would have default visibility. Added a comment.


================
Comment at: llvm/lib/Transforms/IPO/LowerTypeTests.cpp:758
+    if (AbsWidth == PtrWidth)
+      SetAbsRange(~0ull, ~0ull);
+    else if (AbsWidth)
----------------
eugenis wrote:
> should this say 0ull, ~0ull ?
"~0ull, ~0ull" means the full set, whereas "0ull, ~0ull" means the full set except for the value ~0ull. From http://llvm.org/docs/LangRef.html#range-metadata : "The pair a,b represents the range [a,b).".

See also http://llvm-cs.pcc.me.uk/lib/IR/ConstantRange.cpp#237

 Added a comment.


https://reviews.llvm.org/D28546