Legalizer optimize a pair of div / mod to a call to divrem libcall if they are

not legal. However, it should use a div instruction + mul + sub if divide is
legal. The rem legalization code was missing a check and incorrectly uses a
divrem libcall even when div is legal.

rdar://12481395


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@165778 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 36 insertions, 12 deletions

diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index bb4f365b21..2ec129f730 100644
--- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -3153,6 +3153,8 @@ void SelectionDAGLegalize::ExpandNode(SDNode *Node) {
     Tmp3 = Node->getOperand(1);
     if (TLI.isOperationLegalOrCustom(DivRemOpc, VT) ||
         (isDivRemLibcallAvailable(Node, isSigned, TLI) &&
+         // If div is legal, it's better to do the normal expansion
+         !TLI.isOperationLegalOrCustom(DivOpc, Node->getValueType(0)) &&
          useDivRem(Node, isSigned, false))) {
       Tmp1 = DAG.getNode(DivRemOpc, dl, VTs, Tmp2, Tmp3).getValue(1);
     } else if (TLI.isOperationLegalOrCustom(DivOpc, VT)) {
diff --git a/test/CodeGen/ARM/divmod.ll b/test/CodeGen/ARM/divmod.ll
index 7fbf8f4090..577f8aa7d3 100644
--- a/test/CodeGen/ARM/divmod.ll
+++ b/test/CodeGen/ARM/divmod.ll
@@ -1,10 +1,18 @@
-; RUN: llc < %s -mtriple=arm-apple-ios5.0 -mcpu=cortex-a8 | FileCheck %s
+; RUN: llc < %s -mtriple=arm-apple-ios5.0 -mcpu=cortex-a8 | FileCheck %s -check-prefix=A8
+; RUN: llc < %s -mtriple=arm-apple-ios5.0 -mcpu=swift     | FileCheck %s -check-prefix=SWIFT
+
+; rdar://12481395
 
 define void @foo(i32 %x, i32 %y, i32* nocapture %P) nounwind ssp {
 entry:
-; CHECK: foo:
-; CHECK: bl ___divmodsi4
-; CHECK-NOT: bl ___divmodsi4
+; A8: foo:
+; A8: bl ___divmodsi4
+; A8-NOT: bl ___divmodsi4
+
+; SWIFT: foo:
+; SWIFT: sdiv
+; SWIFT: mls
+; SWIFT-NOT: bl __divmodsi4
   %div = sdiv i32 %x, %y
   store i32 %div, i32* %P, align 4
   %rem = srem i32 %x, %y
@@ -15,9 +23,14 @@ entry:
 
 define void @bar(i32 %x, i32 %y, i32* nocapture %P) nounwind ssp {
 entry:
-; CHECK: bar:
-; CHECK: bl ___udivmodsi4
-; CHECK-NOT: bl ___udivmodsi4
+; A8: bar:
+; A8: bl ___udivmodsi4
+; A8-NOT: bl ___udivmodsi4
+
+; SWIFT: bar:
+; SWIFT: udiv
+; SWIFT: mls
+; SWIFT-NOT: bl __udivmodsi4
   %div = udiv i32 %x, %y
   store i32 %div, i32* %P, align 4
   %rem = urem i32 %x, %y
@@ -32,14 +45,18 @@ entry:
 
 define void @do_indent(i32 %cols) nounwind {
 entry:
-; CHECK: do_indent:
+; A8: do_indent:
+; SWIFT: do_indent:
   %0 = load i32* @flags, align 4
   %1 = and i32 %0, 67108864
   %2 = icmp eq i32 %1, 0
   br i1 %2, label %bb1, label %bb
 
 bb:
-; CHECK: bl ___divmodsi4
+; A8: bl ___divmodsi4
+; SWIFT: sdiv
+; SWIFT: mls
+; SWIFT-NOT: bl __divmodsi4
   %3 = load i32* @tabsize, align 4
   %4 = srem i32 %cols, %3
   %5 = sdiv i32 %cols, %3
@@ -60,9 +77,14 @@ declare i8* @__memset_chk(i8*, i32, i32, i32) nounwind
 ; rdar://11714607
 define i32 @howmany(i32 %x, i32 %y) nounwind {
 entry:
-; CHECK: howmany:
-; CHECK: bl ___udivmodsi4
-; CHECK-NOT: ___udivsi3
+; A8: howmany:
+; A8: bl ___udivmodsi4
+; A8-NOT: ___udivsi3
+
+; SWIFT: howmany:
+; SWIFT: udiv
+; SWIFT: mls
+; SWIFT-NOT: bl __udivmodsi4
   %rem = urem i32 %x, %y
   %div = udiv i32 %x, %y
   %not.cmp = icmp ne i32 %rem, 0