make a couple of changes to the standard pass pipeline:

1. Only run the early (in the module pass pipe) instcombine/simplifycfg
   if the "unit at a time" passes they are cleaning up after runs.

2. Move the "clean up after the unroller" pass to the very end of the
   function-level pass pipeline.  Loop unroll uses instsimplify now,
   so it doesn't create a ton of trash.  Moving instcombine later allows
   it to clean up after opportunities are exposed by GVN, DSE, etc.

3. Introduce some phase ordering tests for things that are specifically
   intended to be simplified by the full optimizer as a whole.

This resolves PR2338, and is progress towards PR6627, which will be 
generating code that looks similar to test2.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130241 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/Support/StandardPasses.h

@@ -72,6 +72,7 @@ namespace llvm {
                                                 Pass *InliningPass) {
     createStandardAliasAnalysisPasses(PM);
 
+    // If all optimizations are disabled, just run the always-inline pass.
     if (OptimizationLevel == 0) {
       if (InliningPass)
         PM->add(InliningPass);
@@ -83,9 +84,10 @@ namespace llvm {
       
       PM->add(createIPSCCPPass());              // IP SCCP
       PM->add(createDeadArgEliminationPass());  // Dead argument elimination
-    }
+      
-    PM->add(createInstructionCombiningPass());  // Clean up after IPCP & DAE
+      PM->add(createInstructionCombiningPass());// Clean up after IPCP & DAE
       PM->add(createCFGSimplificationPass());   // Clean up after IPCP & DAE
+    }
     
     // Start of CallGraph SCC passes.
     if (UnitAtATime && HaveExceptions)
@@ -120,7 +122,6 @@ namespace llvm {
     PM->add(createLoopDeletionPass());          // Delete dead loops
     if (UnrollLoops)
       PM->add(createLoopUnrollPass());          // Unroll small loops
-    PM->add(createInstructionCombiningPass());  // Clean up after the unroller
     if (OptimizationLevel > 1)
       PM->add(createGVNPass());                 // Remove redundancies
     PM->add(createMemCpyOptPass());             // Remove memcpy / form memset
@@ -134,6 +135,7 @@ namespace llvm {
     PM->add(createDeadStoreEliminationPass());  // Delete dead stores
     PM->add(createAggressiveDCEPass());         // Delete dead instructions
     PM->add(createCFGSimplificationPass());     // Merge & remove BBs
+    PM->add(createInstructionCombiningPass());  // Clean up after everything.
 
     if (UnitAtATime) {
       PM->add(createStripDeadPrototypesPass()); // Get rid of dead prototypes

test/Transforms/PhaseOrdering/basic.ll

@@ -0,0 +1,73 @@
+; RUN: opt -O3 -S %s | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "x86_64-apple-macosx10.6.7"
+
+declare i8* @malloc(i64)
+declare void @free(i8*)
+
+
+; PR2338
+define void @test1() nounwind ssp {
+  %retval = alloca i32, align 4
+  %i = alloca i8*, align 8
+  %call = call i8* @malloc(i64 1)
+  store i8* %call, i8** %i, align 8
+  %tmp = load i8** %i, align 8
+  store i8 1, i8* %tmp
+  %tmp1 = load i8** %i, align 8
+  call void @free(i8* %tmp1)
+  ret void
+
+; CHECK: @test1
+; CHECK-NEXT: ret void
+}
+
+
+; PR6627 - This whole nasty sequence should be flattened down to a single
+; 32-bit comparison.
+define void @test2(i8* %arrayidx) nounwind ssp {
+entry:
+  %xx = bitcast i8* %arrayidx to i32*
+  %x1 = load i32* %xx, align 4
+  %tmp = trunc i32 %x1 to i8
+  %conv = zext i8 %tmp to i32
+  %cmp = icmp eq i32 %conv, 127
+  br i1 %cmp, label %land.lhs.true, label %if.end
+
+land.lhs.true:                                    ; preds = %entry
+  %arrayidx4 = getelementptr inbounds i8* %arrayidx, i64 1
+  %tmp5 = load i8* %arrayidx4, align 1
+  %conv6 = zext i8 %tmp5 to i32
+  %cmp7 = icmp eq i32 %conv6, 69
+  br i1 %cmp7, label %land.lhs.true9, label %if.end
+
+land.lhs.true9:                                   ; preds = %land.lhs.true
+  %arrayidx12 = getelementptr inbounds i8* %arrayidx, i64 2
+  %tmp13 = load i8* %arrayidx12, align 1
+  %conv14 = zext i8 %tmp13 to i32
+  %cmp15 = icmp eq i32 %conv14, 76
+  br i1 %cmp15, label %land.lhs.true17, label %if.end
+
+land.lhs.true17:                                  ; preds = %land.lhs.true9
+  %arrayidx20 = getelementptr inbounds i8* %arrayidx, i64 3
+  %tmp21 = load i8* %arrayidx20, align 1
+  %conv22 = zext i8 %tmp21 to i32
+  %cmp23 = icmp eq i32 %conv22, 70
+  br i1 %cmp23, label %if.then, label %if.end
+
+if.then:                                          ; preds = %land.lhs.true17
+  %call25 = call i32 (...)* @doo()
+  br label %if.end
+
+if.end:                                           ; preds = %if.then, %land.lhs.true17, %land.lhs.true9, %land.lhs.true, %entry
+  ret void
+
+; CHECK: @test2
+; CHECK: %x1 = load i32* %xx, align 4
+; CHECK-NEXT: icmp eq i32 %x1, 1179403647
+; CHECK-NEXT: br i1 {{.*}}, label %if.then, label %if.end 
+}
+
+declare i32 @doo(...)
+

test/Transforms/ScalarRepl/dg.exp

@@ -1,3 +1,3 @@
 load_lib llvm.exp
 
-RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]]
+RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll}]]