From da556345dcfd269af0f6a50acb661dc33d99e92d Mon Sep 17 00:00:00 2001
From: Haicheng Wu <haicheng@codeaurora.org>
Date: Thu, 19 Jan 2017 22:28:34 +0000
Subject: [PATCH] [InlineCost] Use TTI to check if GEP is free.

Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.

Differential Revision: https://reviews.llvm.org/D28693

llvm-svn: 292526
---
 llvm/lib/Analysis/InlineCost.cpp        | 20 ++++++++++++++++++--
 llvm/test/Transforms/Inline/gep-cost.ll | 29 +++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/Transforms/Inline/gep-cost.ll

diff --git a/llvm/lib/Analysis/InlineCost.cpp b/llvm/lib/Analysis/InlineCost.cpp
index 4109049..b6c3997 100644
--- a/llvm/lib/Analysis/InlineCost.cpp
+++ b/llvm/lib/Analysis/InlineCost.cpp
@@ -134,6 +134,7 @@ class CallAnalyzer : public InstVisitor<CallAnalyzer, bool> {
   void accumulateSROACost(DenseMap<Value *, int>::iterator CostIt,
                           int InstructionCost);
   bool isGEPOffsetConstant(GetElementPtrInst &GEP);
+  bool isGEPFree(GetElementPtrInst &GEP);
   bool accumulateGEPOffset(GEPOperator &GEP, APInt &Offset);
   bool simplifyCallSite(Function *F, CallSite CS);
   ConstantInt *stripAndComputeInBoundsConstantOffsets(Value *&V);
@@ -331,6 +332,21 @@ bool CallAnalyzer::accumulateGEPOffset(GEPOperator &GEP, APInt &Offset) {
   return true;
 }
 
+/// \brief Use TTI to check whether a GEP is free.
+///
+/// Respects any simplified values known during the analysis of this callsite.
+bool CallAnalyzer::isGEPFree(GetElementPtrInst &GEP) {
+  SmallVector<Value *, 4> Indices;
+  for (User::op_iterator I = GEP.idx_begin(), E = GEP.idx_end(); I != E; ++I)
+    if (Constant *SimpleOp = SimplifiedValues.lookup(*I))
+       Indices.push_back(SimpleOp);
+     else
+       Indices.push_back(*I);
+  return TargetTransformInfo::TCC_Free ==
+         TTI.getGEPCost(GEP.getSourceElementType(), GEP.getPointerOperand(),
+                        Indices);
+}
+
 bool CallAnalyzer::visitAlloca(AllocaInst &I) {
   // Check whether inlining will turn a dynamic alloca into a static
   // alloca and handle that case.
@@ -396,7 +412,7 @@ bool CallAnalyzer::visitGetElementPtr(GetElementPtrInst &I) {
         // Non-constant GEPs aren't folded, and disable SROA.
         if (SROACandidate)
           disableSROA(CostIt);
-        return false;
+        return isGEPFree(I);
       }
 
       // Add the result as a new mapping to Base + Offset.
@@ -422,7 +438,7 @@ bool CallAnalyzer::visitGetElementPtr(GetElementPtrInst &I) {
   // Variable GEPs will require math and will disable SROA.
   if (SROACandidate)
     disableSROA(CostIt);
-  return false;
+  return isGEPFree(I);
 }
 
 bool CallAnalyzer::visitBitCast(BitCastInst &I) {
diff --git a/llvm/test/Transforms/Inline/gep-cost.ll b/llvm/test/Transforms/Inline/gep-cost.ll
new file mode 100644
index 0000000..6738071
--- /dev/null
+++ b/llvm/test/Transforms/Inline/gep-cost.ll
@@ -0,0 +1,29 @@
+; RUN: opt -inline < %s -S -debug-only=inline-cost 2>&1 | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64--linux-gnu"
+
+define void @outer([4 x i32]* %ptr, i32 %i) {
+  call void @inner1([4 x i32]* %ptr, i32 %i)
+  call void @inner2([4 x i32]* %ptr, i32 %i)
+  ret void
+}
+; The gep in inner1() is reg+reg, which is a legal addressing mode for AArch64.
+; Thus, both the gep and ret can be simplified.
+; CHECK: Analyzing call of inner1
+; CHECK: NumInstructionsSimplified: 2
+; CHECK: NumInstructions: 2
+define void @inner1([4 x i32]* %ptr, i32 %i) {
+  %G = getelementptr inbounds [4 x i32], [4 x i32]* %ptr, i32 0, i32 %i
+  ret void
+}
+
+; The gep in inner2() is reg+imm+reg, which is not a legal addressing mode for 
+; AArch64.  Thus, only the ret can be simplified and not the gep.
+; CHECK: Analyzing call of inner2
+; CHECK: NumInstructionsSimplified: 1
+; CHECK: NumInstructions: 2
+define void @inner2([4 x i32]* %ptr, i32 %i) {
+  %G = getelementptr inbounds [4 x i32], [4 x i32]* %ptr, i32 1, i32 %i
+  ret void
+}
-- 
2.7.4