diff options
| author | Hal Finkel <hfinkel@anl.gov> | 2015-09-03 23:23:00 +0000 |
|---|---|---|
| committer | Hal Finkel <hfinkel@anl.gov> | 2015-09-03 23:23:00 +0000 |
| commit | 75afa2b6b61e26ccc560732d33dcad3c1fc9f871 (patch) | |
| tree | 9e41a4f4eace0abdf236fb757e7345c2fa6252a9 /llvm/test/Transforms/LoopVectorize/PowerPC | |
| parent | b47128aaf37bc027d66a0bc331beaad4a84da4dd (diff) | |
| download | bcm5719-llvm-75afa2b6b61e26ccc560732d33dcad3c1fc9f871.tar.gz bcm5719-llvm-75afa2b6b61e26ccc560732d33dcad3c1fc9f871.zip | |
[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.
llvm-svn: 246819
Diffstat (limited to 'llvm/test/Transforms/LoopVectorize/PowerPC')
| -rw-r--r-- | llvm/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll | 40 |
1 files changed, 40 insertions, 0 deletions
diff --git a/llvm/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll b/llvm/test/Transforms/LoopVectorize/PowerPC/agg-interleave-a2.ll new file mode 100644 index 00000000000..3491e08bbaa --- /dev/null +++ b/llvm/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" } + |
