[llvm] r273105 - [LoadCombine] Combine Loads formed from GEPS with negative indexes

[David Majnemer via llvm-commits]

Author: majnemer
Date: Sun Jun 19 01:14:56 2016
New Revision: 273105

URL: http://llvm.org/viewvc/llvm-project?rev=273105&view=rev
Log:
[LoadCombine] Combine Loads formed from GEPS with negative indexes

Change the underlying offset and comparisons to use int64_t instead of
uint64_t.

Patch by River Riddle!

Differential Revision: http://reviews.llvm.org/D21499

Added:
    llvm/trunk/test/Transforms/LoadCombine/load-combine-negativegep.ll
Modified:
    llvm/trunk/lib/Transforms/Scalar/LoadCombine.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/LoadCombine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoadCombine.cpp?rev=273105&r1=273104&r2=273105&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoadCombine.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoadCombine.cpp Sun Jun 19 01:14:56 2016
@@ -40,7 +40,7 @@ STATISTIC(NumLoadsCombined, "Number of l
 namespace {
 struct PointerOffsetPair {
   Value *Pointer;
-  uint64_t Offset;
+  int64_t Offset;
 };
 
 struct LoadPOPPair {
@@ -102,7 +102,7 @@ PointerOffsetPair LoadCombine::getPointe
       unsigned BitWidth = DL.getPointerTypeSizeInBits(GEP->getType());
       APInt Offset(BitWidth, 0);
       if (GEP->accumulateConstantOffset(DL, Offset))
-        POP.Offset += Offset.getZExtValue();
+        POP.Offset += Offset.getSExtValue();
       else
         // Can't handle GEPs with variable indices.
         return POP;
@@ -138,28 +138,31 @@ bool LoadCombine::aggregateLoads(SmallVe
   LoadInst *BaseLoad = nullptr;
   SmallVector<LoadPOPPair, 8> AggregateLoads;
   bool Combined = false;
-  uint64_t PrevOffset = -1ull;
+  bool ValidPrevOffset = false;
+  int64_t PrevOffset = 0;
   uint64_t PrevSize = 0;
   for (auto &L : Loads) {
-    if (PrevOffset == -1ull) {
+    if (ValidPrevOffset == false) {
       BaseLoad = L.Load;
       PrevOffset = L.POP.Offset;
       PrevSize = L.Load->getModule()->getDataLayout().getTypeStoreSize(
           L.Load->getType());
       AggregateLoads.push_back(L);
+      ValidPrevOffset = true;
       continue;
     }
     if (L.Load->getAlignment() > BaseLoad->getAlignment())
       continue;
-    if (L.POP.Offset > PrevOffset + PrevSize) {
+    int64_t PrevEnd = PrevOffset + PrevSize;
+    if (L.POP.Offset > PrevEnd) {
       // No other load will be combinable
       if (combineLoads(AggregateLoads))
         Combined = true;
       AggregateLoads.clear();
-      PrevOffset = -1;
+      ValidPrevOffset = false;
       continue;
     }
-    if (L.POP.Offset != PrevOffset + PrevSize)
+    if (L.POP.Offset != PrevEnd)
       // This load is offset less than the size of the last load.
       // FIXME: We may want to handle this case.
       continue;

Added: llvm/trunk/test/Transforms/LoadCombine/load-combine-negativegep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoadCombine/load-combine-negativegep.ll?rev=273105&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LoadCombine/load-combine-negativegep.ll (added)
+++ llvm/trunk/test/Transforms/LoadCombine/load-combine-negativegep.ll Sun Jun 19 01:14:56 2016
@@ -0,0 +1,19 @@
+; RUN: opt -basicaa -load-combine -S < %s | FileCheck %s
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define i32 @Load_NegGep(i32* %i){
+  %1 = getelementptr inbounds i32, i32* %i, i64 -1
+  %2 = load i32, i32* %1, align 4
+  %3 = load i32, i32* %i, align 4
+  %4 = add nsw i32 %3, %2
+  ret i32 %4
+; CHECK-LABEL: @Load_NegGep(
+; CHECK:  %[[load:.*]] = load i64
+; CHECK:  %[[combine_extract_lo:.*]] = trunc i64 %[[load]] to i32
+; CHECK:  %[[combine_extract_shift:.*]] = lshr i64 %[[load]], 32
+; CHECK:  %[[combine_extract_hi:.*]] = trunc i64 %[[combine_extract_shift]] to i32
+; CHECK:  %[[add:.*]] = add nsw i32 %[[combine_extract_hi]], %[[combine_extract_lo]]
+}
+
+