[LoopVectorizer] Count dependencies of consecutive pointers as uniforms

For the purpose of calculating the cost of the loop at various vectorization
factors, we need to count dependencies of consecutive pointers as uniforms
(which means that the VF = 1 cost is used for all overall VF values).

For example, the TSVC benchmark function s173 has:
  ...
  %3 = add nsw i64 %indvars.iv, 16000
  %arrayidx8 = getelementptr inbounds %struct.GlobalData* @global_data, i64 0, i32 0, i64 %3
  ...
and we must realize that the add will be a scalar in order to correctly deduce
it to be profitable to vectorize this on PowerPC with VSX enabled. In fact, all
dependencies of a consecutive pointer must be a scalar (uniform), and so we
simply need to add all consecutive pointers to the worklist that currently
detects collects uniforms.

Fixes PR19296.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205387 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index dc1ca5c..9a98c44 100644 (file)
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -3656,6 +3656,16 @@ void LoopVectorizationLegality::collectLoopUniforms() {
   // Start with the conditional branch and walk up the block.
   Worklist.push_back(Latch->getTerminator()->getOperand(0));
 
+  // Also add all consecutive pointer values; these values will be uniform
+  // after vectorization (and subsequent cleanup) and, until revectorization is
+  // supported, all dependencies must also be uniform.
+  for (Loop::block_iterator B = TheLoop->block_begin(),
+       BE = TheLoop->block_end(); B != BE; ++B)
+    for (BasicBlock::iterator I = (*B)->begin(), IE = (*B)->end();
+         I != IE; ++I)
+      if (I->getType()->isPointerTy() && isConsecutivePtr(I))
+        Worklist.insert(Worklist.end(), I->op_begin(), I->op_end());
+
   while (Worklist.size()) {
     Instruction *I = dyn_cast<Instruction>(Worklist.back());
     Worklist.pop_back();

diff --git a/test/Transforms/LoopVectorize/PowerPC/lit.local.cfg b/test/Transforms/LoopVectorize/PowerPC/lit.local.cfg
new file mode 100644 (file)
index 0000000..2e46300
--- /dev/null
+++ b/test/Transforms/LoopVectorize/PowerPC/lit.local.cfg
@@ -0,0 +1,4 @@
+targets = set(config.root.targets_to_build.split())
+if not 'PowerPC' in targets:
+    config.unsupported = True
+

diff --git a/test/Transforms/LoopVectorize/PowerPC/vsx-tsvc-s173.ll b/test/Transforms/LoopVectorize/PowerPC/vsx-tsvc-s173.ll
new file mode 100644 (file)
index 0000000..6cd9c4d
--- /dev/null
+++ b/test/Transforms/LoopVectorize/PowerPC/vsx-tsvc-s173.ll
@@ -0,0 +1,51 @@
+; RUN: opt < %s -mcpu=pwr7 -mattr=+vsx -loop-vectorize -instcombine -S | FileCheck %s
+target datalayout = "E-m:e-i64:64-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+%struct.GlobalData = type { [32000 x float], [3 x i32], [4 x i8], [32000 x float], [5 x i32], [12 x i8], [32000 x float], [7 x i32], [4 x i8], [32000 x float], [11 x i32], [4 x i8], [32000 x float], [13 x i32], [12 x i8], [256 x [256 x float]], [17 x i32], [12 x i8], [256 x [256 x float]], [19 x i32], [4 x i8], [256 x [256 x float]], [23 x i32], [4 x i8], [256 x [256 x float]] }
+
+@global_data = external global %struct.GlobalData, align 16
+@ntimes = external hidden unnamed_addr global i32, align 4
+
+define signext i32 @s173() #0 {
+entry:
+  %0 = load i32* @ntimes, align 4
+  %cmp21 = icmp sgt i32 %0, 0
+  br i1 %cmp21, label %for.cond1.preheader, label %for.end12
+
+for.cond1.preheader:                              ; preds = %for.end, %entry
+  %nl.022 = phi i32 [ %inc11, %for.end ], [ 0, %entry ]
+  br label %for.body3
+
+for.body3:                                        ; preds = %for.body3, %for.cond1.preheader
+  %indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.body3 ]
+  %arrayidx = getelementptr inbounds %struct.GlobalData* @global_data, i64 0, i32 0, i64 %indvars.iv
+  %1 = load float* %arrayidx, align 4
+  %arrayidx5 = getelementptr inbounds %struct.GlobalData* @global_data, i64 0, i32 3, i64 %indvars.iv
+  %2 = load float* %arrayidx5, align 4
+  %add = fadd float %1, %2
+  %3 = add nsw i64 %indvars.iv, 16000
+  %arrayidx8 = getelementptr inbounds %struct.GlobalData* @global_data, i64 0, i32 0, i64 %3
+  store float %add, float* %arrayidx8, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 16000
+  br i1 %exitcond, label %for.end, label %for.body3
+
+for.end:                                          ; preds = %for.body3
+  %inc11 = add nsw i32 %nl.022, 1
+  %4 = load i32* @ntimes, align 4
+  %mul = mul nsw i32 %4, 10
+  %cmp = icmp slt i32 %inc11, %mul
+  br i1 %cmp, label %for.cond1.preheader, label %for.end12
+
+for.end12:                                        ; preds = %for.end, %entry
+  ret i32 0
+
+; CHECK-LABEL: @s173
+; CHECK: load <4 x float>*
+; CHECK: add i64 %index, 16000
+; CHECK: ret i32 0
+}
+
+attributes #0 = { nounwind }
+