[X86] Implement TargetLowering::getScalingFactorCost hook.

Scaling factors are not free on X86 because every "complex" addressing mode
breaks the related instruction into 2 allocations instead of 1.

<rdar://problem/16730541>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207301 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index ddf1d2d73c8f7b5988c63f7e484f130881dc94be..7d195fcdae9be4390e6de83d02d3b622afa0238c 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -20851,3 +20851,16 @@ X86TargetLowering::getRegForInlineAsmConstraint(const std::string &Constraint,
 
   return Res;
 }
 
   return Res;
 }

+
+int X86TargetLowering::getScalingFactorCost(const AddrMode &AM,
+                                            Type *Ty) const {
+  // Scaling factors are not free at all.
+  // An indexed folded instruction, i.e., inst (reg1, reg2, scale),
+  // will take 2 allocations instead of 1 for plain addressing mode,
+  // i.e. inst (reg1).
+  if (isLegalAddressingMode(AM, Ty))
+    // Scale represents reg2 * scale, thus account for 1
+    // as soon as we use a second register.
+    return AM.Scale != 0;
+  return -1;
+}

diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 6eb0069d63b39afb25010b90e86a2128a8ddf160..a598d12cc0b0ddef4a78ea1dc35987eb4b3922b3 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -682,6 +682,12 @@ namespace llvm {
     /// the immediate into a register.
     bool isLegalAddImmediate(int64_t Imm) const override;
 
     /// the immediate into a register.
     bool isLegalAddImmediate(int64_t Imm) const override;
 

+    /// \brief Return the cost of the scaling factor used in the addressing
+    /// mode represented by AM for this target, for a load/store
+    /// of the specified type.
+    /// If the AM is supported, the return value must be >= 0.
+    /// If the AM is not supported, it returns a negative value.
+    int getScalingFactorCost(const AddrMode &AM, Type *Ty) const override;

 
     bool isVectorShiftByScalarCheap(Type *Ty) const override;
 
 
     bool isVectorShiftByScalarCheap(Type *Ty) const override;
 

diff --git a/test/CodeGen/X86/avoid_complex_am.ll b/test/CodeGen/X86/avoid_complex_am.ll
new file mode 100644 (file)
index 0000000..0b7a13d
--- /dev/null
+++ b/test/CodeGen/X86/avoid_complex_am.ll
@@ -0,0 +1,37 @@
+; RUN: opt -S -loop-reduce < %s | FileCheck %s
+; Complex addressing mode are costly.
+; Make loop-reduce prefer unscaled accesses.
+; <rdar://problem/16730541>
+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-macosx"
+
+define void @mulDouble(double* nocapture %a, double* nocapture %b, double* nocapture %c) {
+; CHECK: @mulDouble
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+; CHECK: [[IV:%[^ ]+]] = phi i64 [ [[IVNEXT:%[^,]+]], %for.body ], [ 0, %entry ]
+; Only one induction variable should have been generated.
+; CHECK-NOT: phi
+  %indvars.iv = phi i64 [ 1, %entry ], [ %indvars.iv.next, %for.body ]
+  %tmp = add nsw i64 %indvars.iv, -1
+  %arrayidx = getelementptr inbounds double* %b, i64 %tmp
+  %tmp1 = load double* %arrayidx, align 8
+; The induction variable should carry the scaling factor: 1 * 8 = 8.
+; CHECK: [[IVNEXT]] = add nuw nsw i64 [[IV]], 8
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %arrayidx2 = getelementptr inbounds double* %c, i64 %indvars.iv.next
+  %tmp2 = load double* %arrayidx2, align 8
+  %mul = fmul double %tmp1, %tmp2
+  %arrayidx4 = getelementptr inbounds double* %a, i64 %indvars.iv
+  store double %mul, double* %arrayidx4, align 8
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+; Comparison should be 19 * 8 = 152.
+; CHECK: icmp eq i32 {{%[^,]+}}, 152
+  %exitcond = icmp eq i32 %lftr.wideiv, 20
+  br i1 %exitcond, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body
+  ret void
+}

diff --git a/test/CodeGen/X86/masked-iv-safe.ll b/test/CodeGen/X86/masked-iv-safe.ll
index 4a4d178f6e4181c7ca1b630eb0df0839642aa0a9..7f61e10f5f685e2e268ea3417f578e53d684f724 100644 (file)
--- a/test/CodeGen/X86/masked-iv-safe.ll
+++ b/test/CodeGen/X86/masked-iv-safe.ll
@@ -5,7 +5,7 @@
 
 ; CHECK-LABEL: count_up
 ; CHECK-NOT: {{and|movz|sar|shl}}
 
 ; CHECK-LABEL: count_up
 ; CHECK-NOT: {{and|movz|sar|shl}}

-; CHECK: inc
+; CHECK: addq $8,

 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @count_up(double* %d, i64 %n) nounwind {
 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @count_up(double* %d, i64 %n) nounwind {


@@ -71,7 +71,7 @@ return:
 
 ; CHECK-LABEL: count_up_signed
 ; CHECK-NOT: {{and|movz|sar|shl}}
 
 ; CHECK-LABEL: count_up_signed
 ; CHECK-NOT: {{and|movz|sar|shl}}

-; CHECK: inc
+; CHECK: addq $8,

 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @count_up_signed(double* %d, i64 %n) nounwind {
 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @count_up_signed(double* %d, i64 %n) nounwind {


@@ -174,7 +174,7 @@ return:
 
 ; CHECK-LABEL: another_count_down
 ; CHECK-NOT: {{and|movz|sar|shl}}
 
 ; CHECK-LABEL: another_count_down
 ; CHECK-NOT: {{and|movz|sar|shl}}

-; CHECK: decq
+; CHECK: addq $-8,

 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @another_count_down(double* %d, i64 %n) nounwind {
 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @another_count_down(double* %d, i64 %n) nounwind {


@@ -242,7 +242,7 @@ return:
 
 ; CHECK-LABEL: another_count_down_signed
 ; CHECK-NOT: {{and|movz|sar|shl}}
 
 ; CHECK-LABEL: another_count_down_signed
 ; CHECK-NOT: {{and|movz|sar|shl}}

-; CHECK: decq
+; CHECK: addq $-8,

 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @another_count_down_signed(double* %d, i64 %n) nounwind {
 ; CHECK-NOT: {{and|movz|sar|shl}}
 ; CHECK: jne
 define void @another_count_down_signed(double* %d, i64 %n) nounwind {