NewGVN: Handle coercion of constant stores, loads, memory insts.

[• dberlin]

Summary

Depends on D30928.

This adds support for coercion of stores and memory instructions that do not require insertion to process.
Another few tests down.
I added the relevant tests from rle.ll

[• dberlin]

lib/Transforms/Scalar/NewGVN.cpp

814

i'll fix the comment formatting.

818

waiting on a bootstrap to check this.

829

and this :)

[• dberlin]

Update to fix bugs discovered during a few 3-stages

lib/Transforms/Scalar/NewGVN.cpp

 #include "llvm/Support/Debug.h"
 #include "llvm/Support/DebugCounter.h"
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/Transforms/Scalar/GVNExpression.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
 #include "llvm/Transforms/Utils/MemorySSA.h"
 #include "llvm/Transforms/Utils/PredicateInfo.h"
+#include "llvm/Transforms/Utils/VNCoercion.h"
 #include <unordered_map>
 #include <utility>
 #include <vector>
 using namespace llvm;
 using namespace PatternMatch;
 using namespace llvm::GVNExpression;
+using namespace llvm::VNCoercion;
 #define DEBUG_TYPE "newgvn"
 
 STATISTIC(NumGVNInstrDeleted, "Number of instructions deleted");
 STATISTIC(NumGVNBlocksDeleted, "Number of blocks deleted");
 STATISTIC(NumGVNOpsSimplified, "Number of Expressions simplified");
 STATISTIC(NumGVNPhisAllSame, "Number of PHIs whos arguments are all the same");
 STATISTIC(NumGVNMaxIterations,
           "Maximum Number of iterations it took to converge GVN");
private:
   // Value number an Instruction or MemoryPhi.
   void valueNumberMemoryPhi(MemoryPhi *);
   void valueNumberInstruction(Instruction *);
 
   // Symbolic evaluation.
   const Expression *checkSimplificationResults(Expression *, Instruction *,
                                                Value *);
   const Expression *performSymbolicEvaluation(Value *);
+  const Expression *performSymbolicLoadCoercion(Type *, Value *, LoadInst *,
+                                                Instruction *, MemoryAccess *);
   const Expression *performSymbolicLoadEvaluation(Instruction *);
   const Expression *performSymbolicStoreEvaluation(Instruction *);
   const Expression *performSymbolicCallEvaluation(Instruction *);
   const Expression *performSymbolicPHIEvaluation(Instruction *);
   const Expression *performSymbolicAggrValueEvaluation(Instruction *);
   const Expression *performSymbolicCmpEvaluation(Instruction *);
   const Expression *performSymbolicPredicateInfoEvaluation(Instruction *);
 
if (LoadInst *LI = dyn_cast<LoadInst>(SI->getValueOperand())) {
           (lookupMemoryAccessEquiv(
                MSSA->getMemoryAccess(LI)->getDefiningAccess()) == StoreRHS))
         return createVariableExpression(LI);
     }
   }
   return createStoreExpression(SI, StoreAccess);
 }
 
+// See if we can extract the value of a loaded pointer from a load, a store, or
+// a memory instruction.
+const Expression *
+NewGVN::performSymbolicLoadCoercion(Type *LoadType, Value *LoadPtr,
+                                    LoadInst *LI, Instruction *DepInst,
+                                    MemoryAccess *DefiningAccess) {
+  assert((!LI || LI->isSimple()) && "Not a simple load");
+
+  if (auto *DepSI = dyn_cast<StoreInst>(DepInst)) {
+    // Can't forward from non-atomic to atomic without violating memory model.
+    // Don't need to coerce if they are the same type
+    if (LI->isAtomic() > DepSI->isAtomic() ||
+        LoadType == DepSI->getValueOperand()->getType())
+      return nullptr;
+    //    auto *LoadAddressLeader = lookupOperandLeader(LoadPtr);
+    int Offset = analyzeLoadFromClobberingStore(LoadType, LoadPtr, DepSI, DL);
+    if (Offset >= 0) {
+      if (auto *C =
+              dyn_cast<Constant>(lookupOperandLeader(DepSI->getValueOperand())))
+        return createConstantExpression(
+            getConstantStoreValueForLoad(C, Offset, LoadType, DL));
+    }
+
+  } else if (LoadInst *DepLI = dyn_cast<LoadInst>(DepInst)) {
+    // Can't forward from non-atomic to atomic without violating memory model.
+    // FIXME: Check why DepLI == LI would occur
+    if (DepLI == LI || LI->isAtomic() > DepLI->isAtomic())
+      return nullptr;
+    int Offset = analyzeLoadFromClobberingLoad(LoadType, LoadPtr, DepLI, DL);
+    if (Offset >= 0) {
+      if (auto *C = dyn_cast<Constant>(lookupOperandLeader(DepLI)))
+        if (auto *PossibleConstant =
+                getConstantLoadValueForLoad(C, Offset, LoadType, DL))
+          return createConstantExpression(PossibleConstant);
+    }
+
+  } else if (MemIntrinsic *DepMI = dyn_cast<MemIntrinsic>(DepInst)) {
+    int Offset = analyzeLoadFromClobberingMemInst(LoadType, LoadPtr, DepMI, DL);
+    if (Offset >= 0) {
+      if (auto *PossibleConstant =
+              getConstantMemInstValueForLoad(DepMI, Offset, LoadType, DL))
+        return createConstantExpression(PossibleConstant);
+    }
+  }
+  // If this load really doesn't depend on anything, then we must be loading an
+  // undef value.  This can happen when loading for a fresh allocation with no
+  // intervening stores, for example.
+  else if (isa<AllocaInst>(DepInst) || isMallocLikeFn(DepInst, TLI))
+    return createConstantExpression(UndefValue::get(LoadType));
+
+  // If this load occurs either right after a lifetime begin,
+  // then the loaded value is undefined.
+  else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(DepInst)) {
+    if (II->getIntrinsicID() == Intrinsic::lifetime_start)
+      return createConstantExpression(UndefValue::get(LoadType));
+  }
+  // If this load follows a calloc (which zero initializes memory),
+  // then the loaded value is zero
+  else if (isCallocLikeFn(DepInst, TLI)) {
+    return createConstantExpression(Constant::getNullValue(LoadType));
+  }
+
+  return nullptr;
+}
+
 const Expression *NewGVN::performSymbolicLoadEvaluation(Instruction *I) {
   auto *LI = cast<LoadInst>(I);
 
   // We can eliminate in favor of non-simple loads, but we won't be able to
   // eliminate the loads themselves.
   if (!LI->isSimple())
     return nullptr;
 
   Value *LoadAddressLeader = lookupOperandLeader(LI->getPointerOperand());
   // Load of undef is undef.
   if (isa<UndefValue>(LoadAddressLeader))
     return createConstantExpression(UndefValue::get(LI->getType()));
 
   MemoryAccess *DefiningAccess = MSSAWalker->getClobberingMemoryAccess(I);
 
   if (!MSSA->isLiveOnEntryDef(DefiningAccess)) {
     if (auto *MD = dyn_cast<MemoryDef>(DefiningAccess)) {
       Instruction *DefiningInst = MD->getMemoryInst();
       // If the defining instruction is not reachable, replace with undef.
       if (!ReachableBlocks.count(DefiningInst->getParent()))
         return createConstantExpression(UndefValue::get(LI->getType()));
+      // This will handle stores and memory insts.
+      // We only do if it the defining access has a different type.
+      const Expression *CoercionResult =
+          performSymbolicLoadCoercion(LI->getType(), LI->getPointerOperand(),
+                                      LI, DefiningInst, DefiningAccess);
+      if (CoercionResult)
+        return CoercionResult;
     }
   }
 
   const Expression *E =
       createLoadExpression(LI->getType(), LI->getPointerOperand(), LI,
                            lookupMemoryAccessEquiv(DefiningAccess));
   return E;
 }

test/Transforms/NewGVN/calloc-load-removal.ll

-; XFAIL: *
 ; RUN: opt -S -basicaa -newgvn < %s | FileCheck %s
 ; RUN: opt -S -basicaa -newgvn -disable-simplify-libcalls < %s | FileCheck %s -check-prefix=CHECK_NO_LIBCALLS
 ; Check that loads from calloc are recognized as being zero.
 
 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-S128"
 
 ; Function Attrs: nounwind uwtable
 define i32 @test1() {

test/Transforms/NewGVN/fold-const-expr.ll

-; XFAIL: *
 ; GVN failed to do constant expression folding and expanded
 ; them unfolded in many places, producing exponentially large const
 ; expressions. As a result, the compilation never fisished.
 ; This test checks that we are folding constant expression
 ; PR 28418
 ; RUN: opt -newgvn -S < %s | FileCheck %s
 ;; NewGVN fails this due to not having load coercion
 %2 = type { i32, i32, i32, i32, i32 }

test/Transforms/NewGVN/malloc-load-removal.ll

-; XFAIL: *
 ; RUN: opt -S -basicaa -newgvn < %s | FileCheck %s
 ; RUN: opt -S -basicaa -newgvn -disable-simplify-libcalls < %s | FileCheck %s -check-prefix=CHECK_NO_LIBCALLS
 ; PR13694
 
 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-S128"
 target triple = "x86_64-apple-macosx10.8.0"
 
 declare i8* @malloc(i64) nounwind

test/Transforms/NewGVN/pr17732.ll

-; XFAIL: *
 ; RUN: opt -newgvn -S -o - < %s | FileCheck %s
-; NewGVN fails this due to missing load coercion
 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-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
 %struct.with_array = type { [2 x i8], i32, i8 }
 %struct.with_vector = type { <2 x i8>, i32, i8 }
 
 @main.obj_with_array = private unnamed_addr constant { [2 x i8], i32, i8, [3 x i8] } { [2 x i8] zeroinitializer, i32 0, i8 1, [3 x i8] undef }, align 4
 @array_with_zeroinit = common global %struct.with_array zeroinitializer, align 4

test/Transforms/NewGVN/rle.ll

+; RUN: opt < %s -data-layout="e-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-n8:16:32" -basicaa -newgvn -S -die | FileCheck %s
+; RUN: opt < %s -data-layout="E-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-n32"      -basicaa -newgvn -S -die | FileCheck %s
+; memset -> i16 forwarding.
+define signext i16 @memset_to_i16_local(i16* %A) nounwind ssp {
+entry:
+  %conv = bitcast i16* %A to i8*
+  tail call void @llvm.memset.p0i8.i64(i8* %conv, i8 1, i64 200, i32 1, i1 false)
+  %arrayidx = getelementptr inbounds i16, i16* %A, i64 42
+  %tmp2 = load i16, i16* %arrayidx
+  ret i16 %tmp2
+; CHECK-LABEL: @memset_to_i16_local(
+; CHECK-NOT: load
+; CHECK: ret i16 257
+}
+
+@GCst = constant {i32, float, i32 } { i32 42, float 14., i32 97 }
+@GCst_as1 = addrspace(1) constant {i32, float, i32 } { i32 42, float 14., i32 97 }
+
+; memset -> float forwarding.
+define float @memcpy_to_float_local(float* %A) nounwind ssp {
+entry:
+  %conv = bitcast float* %A to i8*                ; <i8*> [#uses=1]
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %conv, i8* bitcast ({i32, float, i32 }* @GCst to i8*), i64 12, i32 1, i1 false)
+  %arrayidx = getelementptr inbounds float, float* %A, i64 1 ; <float*> [#uses=1]
+  %tmp2 = load float, float* %arrayidx                   ; <float> [#uses=1]
+  ret float %tmp2
+; CHECK-LABEL: @memcpy_to_float_local(
+; CHECK-NOT: load
+; CHECK: ret float 1.400000e+01
+}
+; memcpy from address space 1
+define float @memcpy_to_float_local_as1(float* %A) nounwind ssp {
+entry:
+  %conv = bitcast float* %A to i8*                ; <i8*> [#uses=1]
+  tail call void @llvm.memcpy.p0i8.p1i8.i64(i8* %conv, i8 addrspace(1)* bitcast ({i32, float, i32 } addrspace(1)* @GCst_as1 to i8 addrspace(1)*), i64 12, i32 1, i1 false)
+  %arrayidx = getelementptr inbounds float, float* %A, i64 1 ; <float*> [#uses=1]
+  %tmp2 = load float, float* %arrayidx                   ; <float> [#uses=1]
+  ret float %tmp2
+; CHECK-LABEL: @memcpy_to_float_local_as1(
+; CHECK-NOT: load
+; CHECK: ret float 1.400000e+01
+}
+
+; PR6642
+define i32 @memset_to_load() nounwind readnone {
+entry:
+  %x = alloca [256 x i32], align 4                ; <[256 x i32]*> [#uses=2]
+  %tmp = bitcast [256 x i32]* %x to i8*           ; <i8*> [#uses=1]
+  call void @llvm.memset.p0i8.i64(i8* %tmp, i8 0, i64 1024, i32 4, i1 false)
+  %arraydecay = getelementptr inbounds [256 x i32], [256 x i32]* %x, i32 0, i32 0 ; <i32*>
+  %tmp1 = load i32, i32* %arraydecay                   ; <i32> [#uses=1]
+  ret i32 %tmp1
+; CHECK-LABEL: @memset_to_load(
+; CHECK: ret i32 0
+}
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
+declare void @llvm.memcpy.p0i8.p1i8.i64(i8* nocapture, i8 addrspace(1)* nocapture, i64, i32, i1) nounwind