[PowerPC] Always use aggressive interleaving on the A2

On the A2, with an eye toward QPX unaligned-load merging, we should always use
aggressive interleaving. It is generally superior to only using concatenation
unrolling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@246819 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
index 937d9c6..124556e 100644 (file)
--- a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
+++ b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
@@ -197,6 +197,13 @@ void PPCTTIImpl::getUnrollingPreferences(Loop *L,
 }
 
 bool PPCTTIImpl::enableAggressiveInterleaving(bool LoopHasReductions) {
+  // On the A2, always unroll aggressively. For QPX unaligned loads, we depend
+  // on combining the loads generated for consecutive accesses, and failure to
+  // do so is particularly expensive. This makes it much more likely (compared
+  // to only using concatenation unrolling).
+  if (ST->getDarwinDirective() == PPC::DIR_A2)
+    return true;
+
   return LoopHasReductions;
 }
 

diff --git a/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll b/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll
new file mode 100644 (file)
index 0000000..3491e08
--- /dev/null
+++ b/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll
@@ -0,0 +1,40 @@
+; RUN: opt -S -basicaa -loop-vectorize < %s | FileCheck %s
+target datalayout = "E-m:e-i64:64-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+; Function Attrs: nounwind
+define void @foo(double* noalias nocapture %a, double* noalias nocapture readonly %b, double* noalias nocapture readonly %c) #0 {
+entry:
+  br label %for.body
+
+; CHECK-LABEL: @foo
+; CHECK: fmul <4 x double> %{{[^,]+}}, <double 2.000000e+00, double 2.000000e+00, double 2.000000e+00, double 2.000000e+00>
+; CHECK-NEXT: fmul <4 x double> %{{[^,]+}}, <double 2.000000e+00, double 2.000000e+00, double 2.000000e+00, double 2.000000e+00>
+
+for.cond.cleanup:                                 ; preds = %for.body
+  ret void
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds double, double* %b, i64 %indvars.iv
+  %0 = load double, double* %arrayidx, align 8
+  %mul = fmul double %0, 2.000000e+00
+  %mul3 = fmul double %0, %mul
+  %arrayidx5 = getelementptr inbounds double, double* %c, i64 %indvars.iv
+  %1 = load double, double* %arrayidx5, align 8
+  %mul6 = fmul double %1, 3.000000e+00
+  %mul9 = fmul double %1, %mul6
+  %add = fadd double %mul3, %mul9
+  %mul12 = fmul double %0, 4.000000e+00
+  %mul15 = fmul double %mul12, %1
+  %add16 = fadd double %mul15, %add
+  %add17 = fadd double %add16, 1.000000e+00
+  %arrayidx19 = getelementptr inbounds double, double* %a, i64 %indvars.iv
+  store double %add17, double* %arrayidx19, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 1600
+  br i1 %exitcond, label %for.cond.cleanup, label %for.body
+}
+
+attributes #0 = { nounwind "target-cpu"="a2q" }
+