Optimize memcmp(x,y,n)==0 for small n and suitably aligned x/y.

http://reviews.llvm.org/D6952
PR20673

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@246313 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Utils/SimplifyLibCalls.cpp b/lib/Transforms/Utils/SimplifyLibCalls.cpp
index 6bcbde756bc15b65bd460947a2685a4a51601545..a830c9784b9601c08e6fa37ce56c122799fbb8c8 100644 (file)
--- a/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -32,6 +32,7 @@
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Transforms/Utils/BuildLibCalls.h"
+#include "llvm/Transforms/Utils/Local.h"
 
 using namespace llvm;
 using namespace PatternMatch;
@@ -862,6 +863,27 @@ Value *LibCallSimplifier::optimizeMemCmp(CallInst *CI, IRBuilder<> &B) {
     return B.CreateSub(LHSV, RHSV, "chardiff");
   }
 
+  // memcmp(S1,S2,N/8)==0 -> (*(intN_t*)S1 != *(intN_t*)S2)==0
+  if (DL.isLegalInteger(Len * 8) && isOnlyUsedInZeroEqualityComparison(CI)) {
+
+    IntegerType *IntType = IntegerType::get(CI->getContext(), Len * 8);
+    unsigned PrefAlignment = DL.getPrefTypeAlignment(IntType);
+
+    if (getKnownAlignment(LHS, DL, CI) >= PrefAlignment &&
+        getKnownAlignment(RHS, DL, CI) >= PrefAlignment) {
+
+      Type *LHSPtrTy =
+          IntType->getPointerTo(LHS->getType()->getPointerAddressSpace());
+      Type *RHSPtrTy =
+          IntType->getPointerTo(RHS->getType()->getPointerAddressSpace());
+
+      Value *LHSV = B.CreateLoad(B.CreateBitCast(LHS, LHSPtrTy, "lhsc"), "lhsv");
+      Value *RHSV = B.CreateLoad(B.CreateBitCast(RHS, RHSPtrTy, "rhsc"), "rhsv");
+
+      return B.CreateZExt(B.CreateICmpNE(LHSV, RHSV), CI->getType(), "memcmp");
+    }
+  }
+
   // Constant folding: memcmp(x, y, l) -> cnst (all arguments are constant)
   StringRef LHSStr, RHSStr;
   if (getConstantStringInfo(LHS, LHSStr) &&

diff --git a/test/Transforms/InstCombine/memcmp-1.ll b/test/Transforms/InstCombine/memcmp-1.ll
index db15bd66b7156ad563f3fba68891d7f025b6389f..f9ff479e3addf95f387a5659591d522d815d2b0b 100644 (file)
--- a/test/Transforms/InstCombine/memcmp-1.ll
+++ b/test/Transforms/InstCombine/memcmp-1.ll
@@ -2,7 +2,7 @@
 ;
 ; RUN: opt < %s -instcombine -S | FileCheck %s
 
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32:64"
 
 @foo = constant [4 x i8] c"foo\00"
 @hel = constant [4 x i8] c"hel\00"
@@ -70,3 +70,54 @@ define i32 @test_simplify6() {
   ret i32 %ret
 ; CHECK: ret i32 -1
 }
+
+; Check memcmp(mem1, mem2, 8)==0 -> *(int64_t*)mem1 == *(int64_t*)mem2
+
+define i1 @test_simplify7(i64 %x, i64 %y) {
+; CHECK-LABEL: @test_simplify7(
+  %x.addr = alloca i64, align 8
+  %y.addr = alloca i64, align 8
+  store i64 %x, i64* %x.addr, align 8
+  store i64 %y, i64* %y.addr, align 8
+  %xptr = bitcast i64* %x.addr to i8*
+  %yptr = bitcast i64* %y.addr to i8*
+  %call = call i32 @memcmp(i8* %xptr, i8* %yptr, i32 8)
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+; CHECK: %cmp = icmp eq i64 %x, %y
+; CHECK: ret i1 %cmp
+}
+
+; Check memcmp(mem1, mem2, 4)==0 -> *(int32_t*)mem1 == *(int32_t*)mem2
+
+define i1 @test_simplify8(i32 %x, i32 %y) {
+; CHECK-LABEL: @test_simplify8(
+  %x.addr = alloca i32, align 4
+  %y.addr = alloca i32, align 4
+  store i32 %x, i32* %x.addr, align 4
+  store i32 %y, i32* %y.addr, align 4
+  %xptr = bitcast i32* %x.addr to i8*
+  %yptr = bitcast i32* %y.addr to i8*
+  %call = call i32 @memcmp(i8* %xptr, i8* %yptr, i32 4)
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+; CHECK: %cmp = icmp eq i32 %x, %y
+; CHECK: ret i1 %cmp
+}
+
+; Check memcmp(mem1, mem2, 2)==0 -> *(int16_t*)mem1 == *(int16_t*)mem2
+
+define i1 @test_simplify9(i16 %x, i16 %y) {
+; CHECK-LABEL: @test_simplify9(
+  %x.addr = alloca i16, align 2
+  %y.addr = alloca i16, align 2
+  store i16 %x, i16* %x.addr, align 2
+  store i16 %y, i16* %y.addr, align 2
+  %xptr = bitcast i16* %x.addr to i8*
+  %yptr = bitcast i16* %y.addr to i8*
+  %call = call i32 @memcmp(i8* %xptr, i8* %yptr, i32 2)
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+; CHECK: %cmp = icmp eq i16 %x, %y
+; CHECK: ret i1 %cmp
+}