UseListOrder: Visit global values

When predicting use-list order, we visit functions in reverse order
followed by `GlobalValue`s and write out use-lists at the first
opportunity.  In the reader, this will translate to *after* the last use
has been added.

For this to work, we actually need to descend into `GlobalValue`s.
Added a targeted test in `use-list-order.ll` and `RUN` lines to the
newly passing tests in `test/Bitcode`.

There are two remaining failures in `test/Bitcode`:

  - blockaddress.ll: I haven't thought through how to model the way
    block addresses change the order of use-lists (or how to work around
    it).

  - metadata-2.ll: There's an old-style `@llvm.used` global array here
    that I suspect the .ll parser isn't upgrading properly.  When it
    round-trips through bitcode, the .bc reader *does* upgrade it, so
    the extra variable (`i8* null`) has an extra use, and the shuffle
    vector doesn't match.

    I think the fix is to upgrade old-style global arrays (or reject
    them?) in the .ll parser.

This is part of PR5680.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@214321 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Bitcode/Writer/ValueEnumerator.cpp

@@ -211,9 +211,9 @@ static void predictValueUseListOrder(const Value *V, const Function *F,
 
   // Recursive descent into constants.
   if (const Constant *C = dyn_cast<Constant>(V))
-    if (C->getNumOperands() && !isa<GlobalValue>(C))
+    if (C->getNumOperands()) // Visit GlobalValues.
       for (const Value *Op : C->operands())
-        if (isa<Constant>(Op) && !isa<GlobalValue>(Op))
+        if (isa<Constant>(Op)) // Visit GlobalValues.
           predictValueUseListOrder(Op, F, OM, Stack);
 }
 
@@ -241,8 +241,7 @@ static UseListOrderStack predictUseListOrder(const Module *M) {
     for (const BasicBlock &BB : F)
       for (const Instruction &I : BB)
         for (const Value *Op : I.operands())
-          if ((isa<Constant>(*Op) && !isa<GlobalValue>(*Op)) ||
-              isa<InlineAsm>(*Op))
+          if (isa<Constant>(*Op) || isa<InlineAsm>(*Op)) // Visit GlobalValues.
             predictValueUseListOrder(Op, &F, OM, Stack);
     for (const BasicBlock &BB : F)
       for (const Instruction &I : BB)

test/Bitcode/miscInstructions.3.2.ll

@@ -1,4 +1,5 @@
 ; RUN: llvm-dis < %s.bc| FileCheck %s
+; RUN: verify-uselistorder < %s.bc -preserve-bc-use-list-order -num-shuffles=5
 
 ; miscInstructions.3.2.ll.bc was generated by passing this file to llvm-as-3.2.
 ; The test checks that LLVM does not misread miscellaneous instructions of

test/Bitcode/ssse3_palignr.ll

@@ -1,4 +1,5 @@
 ; RUN: opt < %s -S | FileCheck %s
+; RUN: verify-uselistorder < %s -preserve-bc-use-list-order -num-shuffles=5
 ; CHECK-NOT: {@llvm\\.palign}
 
 define <4 x i32> @align1(<4 x i32> %a, <4 x i32> %b) nounwind readnone ssp {

test/Bitcode/use-list-order.ll

@@ -17,6 +17,11 @@
 @var2 = global i3* @target
 @var3 = global i3* @target
 
+; Check use-list order for a global when used both by a global and in a
+; function.
+@globalAndFunction = global i4 4
+@globalAndFunctionGlobalUser = global i4* @globalAndFunction
+
 define i64 @f(i64 %f) {
 entry:
   %sum = add i64 %f, 0
@@ -94,3 +99,9 @@ first:
   %gotosecond = icmp slt i32 %gh, -9
   br i1 %gotosecond, label %second, label %exit
 }
+
+define i4 @globalAndFunctionFunctionUser() {
+entry:
+  %local = load i4* @globalAndFunction
+  ret i4 %local
+}

test/Bitcode/variableArgumentIntrinsic.3.2.ll

@@ -1,4 +1,5 @@
 ; RUN: llvm-dis < %s.bc| FileCheck %s
+; RUN: verify-uselistorder < %s.bc -preserve-bc-use-list-order -num-shuffles=5
 
 ; vaArgIntrinsic.3.2.ll.bc was generated by passing this file to llvm-as-3.2.
 ; The test checks that LLVM does not misread variable argument intrinsic instructions