PMB: Run the whole-program-devirt pass during LTO at --lto-O0.

[pcc]

Summary

The whole-program-devirt pass needs to run at -O0 because only it
knows about the llvm.type.checked.load intrinsic: it needs to both
lower the intrinsic itself and handle it in the summary.

[tejohnson]

Looks like the new pass manager needs a similar fix. Test?

[pcc]

In D33571#765174, @tejohnson wrote:

Looks like the new pass manager needs a similar fix.

None of the CFI/whole-program-devirt stuff seems like it will work with the new PM because it isn't threading the pointer to the summary index through to the passes :(

Test?

I guess we could add an "end to end" test for this with llvm-lto2.

[pcc]

add test case

[mehdi_amini]

LGTM.

llvm/trunk/lib/Transforms/IPO/PassManagerBuilder.cpp

void PassManagerBuilder::populateLTOPassManager(legacy::PassManagerBase &PM) {
   if (LibraryInfo)
     PM.add(new TargetLibraryInfoWrapperPass(*LibraryInfo));
 
   if (VerifyInput)
     PM.add(createVerifierPass());
 
   if (OptLevel != 0)
     addLTOOptimizationPasses(PM);
+  else {
+    // The whole-program-devirt pass needs to run at -O0 because only it knows
+    // about the llvm.type.checked.load intrinsic: it needs to both lower the
+    // intrinsic itself and handle it in the summary.
+    PM.add(createWholeProgramDevirtPass(ExportSummary, nullptr));
+  }
 
   // Create a function that performs CFI checks for cross-DSO calls with targets
   // in the current module.
   PM.add(createCrossDSOCFIPass());
 
   // Lower type metadata and the type.test intrinsic. This pass supports Clang's
   // control flow integrity mechanisms (-fsanitize=cfi*) and needs to run at
   // link time if CFI is enabled. The pass does nothing if CFI is disabled.

llvm/trunk/test/LTO/Resolution/X86/type-checked-load.ll

+; RUN: llvm-as -o %t %s
+; RUN: llvm-lto2 run %t -O0 -r %t,foo,px -o %t2
+
+; This just tests that we don't crash when compiling this test case.
+; It means that the wholeprogramdevirt pass must have run and lowered
+; the llvm.type.checked.load call.
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define {i8*, i1} @foo(i8* %ptr) {
+  %pair = call {i8*, i1} @llvm.type.checked.load(i8* %ptr, i32 16, metadata !"foo")
+  ret {i8*, i1} %pair
+}
+
+declare {i8*, i1} @llvm.type.checked.load(i8* %ptr, i32 %offset, metadata %type)