From bbc7752ec413b130acf534b3d5f55ed857b5497a Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Mon, 15 Jul 2019 08:48:47 +0000
Subject: [PATCH] [LoopVectorize] Pass unfiltered list of arguments to
 getIntrinsicInstCost.

We do not compute the scalarization overhead in getVectorIntrinsicCost
and TTI::getIntrinsicInstrCost requires the full arguments list.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@366049 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Vectorize/LoopVectorize.cpp    |  7 ++---
 .../vector-intrinsic-call-cost.ll             | 30 +++++++++++++++++++
 2 files changed, 32 insertions(+), 5 deletions(-)
 create mode 100644 test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index 22cf9c7db94..46265e3f3e1 100644
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -3149,11 +3149,8 @@ unsigned LoopVectorizationCostModel::getVectorIntrinsicCost(CallInst *CI,
   if (auto *FPMO = dyn_cast<FPMathOperator>(CI))
     FMF = FPMO->getFastMathFlags();
 
-  // Skip operands that do not require extraction/scalarization and do not incur
-  // any overhead.
-  return TTI.getIntrinsicInstrCost(
-      ID, CI->getType(), filterExtractingOperands(CI->arg_operands(), VF), FMF,
-      VF);
+  SmallVector<Value *, 4> Operands(CI->arg_operands());
+  return TTI.getIntrinsicInstrCost(ID, CI->getType(), Operands, FMF, VF);
 }
 
 static Type *smallestIntegerVectorType(Type *T1, Type *T2) {
diff --git a/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll b/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll
new file mode 100644
index 00000000000..fce4d56c2e6
--- /dev/null
+++ b/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll
@@ -0,0 +1,30 @@
+; RUN: opt -S -loop-vectorize -force-vector-width=4 %s | FileCheck %s
+
+; CHECK-LABEL: @test_fshl
+; CHECK-LABEL: vector.body:
+; CHECK-NEXT:    %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
+; CHECK-NEXT:    %broadcast.splatinsert = insertelement <4 x i32> undef, i32 %index, i32 0
+; CHECK-NEXT:    %broadcast.splat = shufflevector <4 x i32> %broadcast.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:    %induction = add <4 x i32> %broadcast.splat, <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    %0 = add i32 %index, 0
+; CHECK-NEXT:    %1 = call <4 x i16> @llvm.fshl.v4i16(<4 x i16> undef, <4 x i16> undef, <4 x i16> <i16 15, i16 15, i16 15, i16 15>)
+; CHECK-NEXT:    %index.next = add i32 %index, 4
+; CHECK-NEXT:    %2 = icmp eq i32 %index.next, %n.vec
+; CHECK-NEXT:     br i1 %2, label %middle.block, label %vector.body, !llvm.loop !0
+;
+define void @test_fshl(i32 %width) {
+entry:
+  br label %for.body9.us.us
+
+for.cond6.for.cond.cleanup8_crit_edge.us.us:      ; preds = %for.body9.us.us
+  ret void
+
+for.body9.us.us:                                  ; preds = %for.body9.us.us, %entry
+  %x.020.us.us = phi i32 [ 0, %entry ], [ %inc.us.us, %for.body9.us.us ]
+  %conv4.i.us.us = tail call i16 @llvm.fshl.i16(i16 undef, i16 undef, i16 15)
+  %inc.us.us = add nuw i32 %x.020.us.us, 1
+  %exitcond50 = icmp eq i32 %inc.us.us, %width
+  br i1 %exitcond50, label %for.cond6.for.cond.cleanup8_crit_edge.us.us, label %for.body9.us.us
+}
+
+declare i16 @llvm.fshl.i16(i16, i16, i16)
-- 
2.50.0