[llvm] r177098 - LoopVectorize: Invert case when we use a vector cmp value to query select cost

[Arnold Schwaighofer]

Author: arnolds
Date: Thu Mar 14 13:54:36 2013
New Revision: 177098

URL: http://llvm.org/viewvc/llvm-project?rev=177098&view=rev
Log:
LoopVectorize: Invert case when we use a vector cmp value to query select cost

We generate a select with a vectorized condition argument when the condition is
NOT loop invariant. Not the other way around.

Added:
    llvm/trunk/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll
Modified:
    llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp

Modified: llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp?rev=177098&r1=177097&r2=177098&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp (original)
+++ llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp Thu Mar 14 13:54:36 2013
@@ -3338,7 +3338,7 @@ LoopVectorizationCostModel::getInstructi
     const SCEV *CondSCEV = SE->getSCEV(SI->getCondition());
     bool ScalarCond = (SE->isLoopInvariant(CondSCEV, TheLoop));
     Type *CondTy = SI->getCondition()->getType();
-    if (ScalarCond)
+    if (!ScalarCond)
       CondTy = VectorType::get(CondTy, VF);
 
     return TTI.getCmpSelInstrCost(I->getOpcode(), VectorTy, CondTy);

Added: llvm/trunk/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll?rev=177098&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll (added)
+++ llvm/trunk/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll Thu Mar 14 13:54:36 2013
@@ -0,0 +1,62 @@
+; RUN: opt < %s  -loop-vectorize -mcpu=core2 -debug-only=loop-vectorize 2>&1 -S | FileCheck %s
+
+; Make sure we use the right select kind when querying select costs.
+
+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-macosx10.8.0"
+
+ at a = common global [2048 x i32] zeroinitializer, align 16
+ at b = common global [2048 x i32] zeroinitializer, align 16
+ at c = common global [2048 x i32] zeroinitializer, align 16
+
+; CHECK: Checking a loop in "scalarselect"
+define void @scalarselect(i1 %cond) nounwind uwtable ssp {
+  br label %1
+
+; <label>:1                                       ; preds = %1, %0
+  %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
+  %2 = getelementptr inbounds [2048 x i32]* @b, i64 0, i64 %indvars.iv
+  %3 = load i32* %2, align 4
+  %4 = getelementptr inbounds [2048 x i32]* @c, i64 0, i64 %indvars.iv
+  %5 = load i32* %4, align 4
+  %6 = add nsw i32 %5, %3
+  %7 = getelementptr inbounds [2048 x i32]* @a, i64 0, i64 %indvars.iv
+; A scalar select has a cost of 1 on core2
+; CHECK: cost of 1 for VF 2 {{.*}}  select i1 %cond, i32 %6, i32 0
+  %sel = select i1 %cond, i32 %6, i32 zeroinitializer
+  store i32 %sel, i32* %7, align 4
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 256
+  br i1 %exitcond, label %8, label %1
+
+; <label>:8                                       ; preds = %1
+  ret void
+}
+
+; CHECK: Checking a loop in "vectorselect"
+define void @vectorselect(i1 %cond) nounwind uwtable ssp {
+  br label %1
+
+; <label>:1                                       ; preds = %1, %0
+  %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
+  %2 = getelementptr inbounds [2048 x i32]* @b, i64 0, i64 %indvars.iv
+  %3 = load i32* %2, align 4
+  %4 = getelementptr inbounds [2048 x i32]* @c, i64 0, i64 %indvars.iv
+  %5 = load i32* %4, align 4
+  %6 = add nsw i32 %5, %3
+  %7 = getelementptr inbounds [2048 x i32]* @a, i64 0, i64 %indvars.iv
+  %8 = icmp ult i64 %indvars.iv, 8
+; A vector select has a cost of 4 on core2
+; CHECK: cost of 4 for VF 2 {{.*}}  select i1 %8, i32 %6, i32 0
+  %sel = select i1 %8, i32 %6, i32 zeroinitializer
+  store i32 %sel, i32* %7, align 4
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 256
+  br i1 %exitcond, label %9, label %1
+
+; <label>:9                                       ; preds = %1
+  ret void
+}
+