Add support to BBVectorize for vectorizing selects.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154700 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/Transforms/Vectorize.h b/include/llvm/Transforms/Vectorize.h
index 7701ceb4d0b309d4a2b4844e6ddfe0ac373208b2..776c005da6b0fd19f267912dfc2540008a7161ea 100644 (file)
--- a/include/llvm/Transforms/Vectorize.h
+++ b/include/llvm/Transforms/Vectorize.h
@@ -43,6 +43,9 @@ struct VectorizeConfig {
   /// @brief Vectorize the fused-multiply-add intrinsic.
   bool VectorizeFMA;
 
+  /// @brief Vectorize select instructions.
+  bool VectorizeSelect;
+
   /// @brief Vectorize loads and stores.
   bool VectorizeMemOps;
 

diff --git a/lib/Transforms/Vectorize/BBVectorize.cpp b/lib/Transforms/Vectorize/BBVectorize.cpp
index 286b54f2f06730683c181b2d952ae9f4e55f3e9e..b2f8c8f3285c00e9492bacf4881969f1ad82320c 100644 (file)
--- a/lib/Transforms/Vectorize/BBVectorize.cpp
+++ b/lib/Transforms/Vectorize/BBVectorize.cpp
@@ -95,6 +95,10 @@ static cl::opt<bool>
 NoFMA("bb-vectorize-no-fma", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize the fused-multiply-add intrinsic"));
 
+static cl::opt<bool>
+NoSelect("bb-vectorize-no-select", cl::init(false), cl::Hidden,
+  cl::desc("Don't try to vectorize select instructions"));
+
 static cl::opt<bool>
 NoMemOps("bb-vectorize-no-mem-ops", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize loads and stores"));
@@ -552,6 +556,9 @@ namespace {
       Type *DestTy = C->getDestTy();
       if (!DestTy->isSingleValueType() || DestTy->isPointerTy())
         return false;
+    } else if (isa<SelectInst>(I)) {
+      if (!Config.VectorizeSelect)
+        return false;
     } else if (!(I->isBinaryOp() || isa<ShuffleVectorInst>(I) ||
         isa<ExtractElementInst>(I) || isa<InsertElementInst>(I))) {
       return false;
@@ -1894,6 +1901,7 @@ VectorizeConfig::VectorizeConfig() {
   VectorizeCasts = !::NoCasts;
   VectorizeMath = !::NoMath;
   VectorizeFMA = !::NoFMA;
+  VectorizeSelect = !::NoSelect;
   VectorizeMemOps = !::NoMemOps;
   AlignedOnly = ::AlignedOnly;
   ReqChainDepth= ::ReqChainDepth;

diff --git a/test/Transforms/BBVectorize/simple-sel.ll b/test/Transforms/BBVectorize/simple-sel.ll
new file mode 100644 (file)
index 0000000..4daa571
--- /dev/null
+++ b/test/Transforms/BBVectorize/simple-sel.ll
@@ -0,0 +1,30 @@
+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"
+; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -instcombine -gvn -S | FileCheck %s
+
+; Basic depth-3 chain with select
+define double @test1(double %A1, double %A2, double %B1, double %B2, i1 %C1, i1 %C2) {
+; CHECK: @test1
+; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0
+; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0
+; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1
+; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1
+       %X1 = fsub double %A1, %B1
+       %X2 = fsub double %A2, %B2
+; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2
+       %Y1 = fmul double %X1, %A1
+       %Y2 = fmul double %X2, %A2
+; CHECK: %Y1 = fmul <2 x double> %X1, %X1.v.i0.2
+        %Z1 = select i1 %C1, double %Y1, double %B1
+        %Z2 = select i1 %C2, double %Y2, double %B2
+; CHECK: %Z1.v.i0.1 = insertelement <2 x i1> undef, i1 %C1, i32 0
+; CHECK: %Z1.v.i0.2 = insertelement <2 x i1> %Z1.v.i0.1, i1 %C2, i32 1
+; CHECK: %Z1 = select <2 x i1> %Z1.v.i0.2, <2 x double> %Y1, <2 x double> %X1.v.i1.2
+       %R  = fmul double %Z1, %Z2
+; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0
+; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1
+; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2
+       ret double %R
+; CHECK: ret double %R
+}
+
+