vendor/llvm/llvm-trunk-r300422

4 files changed, 143 insertions, 0 deletions

diff --git a/test/Transforms/LoopVectorize/SystemZ/branch-for-predicated-block.ll b/test/Transforms/LoopVectorize/SystemZ/branch-for-predicated-block.ll
new file mode 100644
index 000000000000..d2e594520332
--- /dev/null
+++ b/test/Transforms/LoopVectorize/SystemZ/branch-for-predicated-block.ll
@@ -0,0 +1,38 @@
+; REQUIRES: asserts
+; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -loop-vectorize \
+; RUN:   -force-vector-width=2 -debug-only=loop-vectorize \
+; RUN:   -disable-output < %s 2>&1 | FileCheck %s
+
+; Check costs for branches inside a vectorized loop around predicated
+; blocks. Each such branch will be guarded with an extractelement from the
+; vector compare plus a test under mask instruction. This cost is modelled on
+; the extractelement of i1.
+
+define void @fun(i32* %arr, i64 %trip.count) {
+entry:
+  br label %for.body
+
+for.body:
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
+  %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %indvars.iv
+  %l = load i32, i32* %arrayidx, align 4
+  %cmp55 = icmp sgt i32 %l, 0
+  br i1 %cmp55, label %if.then, label %for.inc
+
+if.then:
+  %sub = sub nsw i32 0, %l
+  store i32 %sub, i32* %arrayidx, align 4
+  br label %for.inc
+
+for.inc:
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, %trip.count
+  br i1 %exitcond, label %for.end.loopexit, label %for.body
+
+for.end.loopexit:
+  ret void
+
+; CHECK: LV: Found an estimated cost of 5 for VF 2 For instruction:   br i1 %cmp55, label %if.then, label %for.inc
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   br label %for.inc
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   br i1 %exitcond, label %for.end.loopexit, label %for.body
+}
diff --git a/test/Transforms/LoopVectorize/SystemZ/lit.local.cfg b/test/Transforms/LoopVectorize/SystemZ/lit.local.cfg
new file mode 100644
index 000000000000..2f3cf7d3f043
--- /dev/null
+++ b/test/Transforms/LoopVectorize/SystemZ/lit.local.cfg
@@ -0,0 +1,2 @@
+if not 'SystemZ' in config.root.targets:
+    config.unsupported = True
diff --git a/test/Transforms/LoopVectorize/SystemZ/load-store-scalarization-cost.ll b/test/Transforms/LoopVectorize/SystemZ/load-store-scalarization-cost.ll
new file mode 100644
index 000000000000..e7096c29b994
--- /dev/null
+++ b/test/Transforms/LoopVectorize/SystemZ/load-store-scalarization-cost.ll
@@ -0,0 +1,33 @@
+; REQUIRES: asserts
+; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -loop-vectorize \
+; RUN:   -force-vector-width=4 -debug-only=loop-vectorize \
+; RUN:   -disable-output -enable-interleaved-mem-accesses=false < %s 2>&1 | \
+; RUN:   FileCheck %s
+;
+; Check that a scalarized load/store does not get a cost for insterts/
+; extracts, since z13 supports element load/store.
+
+define void @fun(i32* %data, i64 %n) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds i32, i32* %data, i64 %i
+  %tmp1 = load i32, i32* %tmp0, align 4
+  %tmp2 = add i32 %tmp1, 1
+  store i32 %tmp2, i32* %tmp0, align 4
+  %i.next = add nuw nsw i64 %i, 2
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+
+; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction:   %tmp1 = load i32, i32* %tmp0, align 4
+; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction:   store i32 %tmp2, i32* %tmp0, align 4
+
+; CHECK: LV: Scalarizing:  %tmp1 = load i32, i32* %tmp0, align 4
+; CHECK: LV: Scalarizing:  store i32 %tmp2, i32* %tmp0, align 4
+}
+
diff --git a/test/Transforms/LoopVectorize/SystemZ/mem-interleaving-costs.ll b/test/Transforms/LoopVectorize/SystemZ/mem-interleaving-costs.ll
new file mode 100644
index 000000000000..5c15ee4f2d9f
--- /dev/null
+++ b/test/Transforms/LoopVectorize/SystemZ/mem-interleaving-costs.ll
@@ -0,0 +1,70 @@
+; REQUIRES: asserts
+; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -loop-vectorize \
+; RUN:   -force-vector-width=4 -debug-only=loop-vectorize \
+; RUN:   -disable-output < %s 2>&1 | FileCheck %s
+;
+; Check that the loop vectorizer performs memory interleaving with accurate
+; cost estimations.
+
+
+; Simple case where just the load is interleaved, because the store group
+; would have gaps.
+define void @fun0(i32* %data, i64 %n) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds i32, i32* %data, i64 %i
+  %tmp1 = load i32, i32* %tmp0, align 4
+  %tmp2 = add i32 %tmp1, 1
+  store i32 %tmp2, i32* %tmp0, align 4
+  %i.next = add nuw nsw i64 %i, 2
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+
+; CHECK: LV: Creating an interleave group with:  %tmp1 = load i32, i32* %tmp0, align 4
+; CHECK: LV: Found an estimated cost of 3 for VF 4 For instruction:   %tmp1 = load i32, i32* %tmp0, align 4
+;        (vl; vl; vperm)
+}
+
+; Interleaving of both load and stores.
+define void @fun1(i32* %data, i64 %n) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds i32, i32* %data, i64 %i
+  %tmp1 = load i32, i32* %tmp0, align 4
+  %i_1  = add i64 %i, 1
+  %tmp2 = getelementptr inbounds i32, i32* %data, i64 %i_1
+  %tmp3 = load i32, i32* %tmp2, align 4
+  store i32 %tmp1, i32* %tmp2, align 4
+  store i32 %tmp3, i32* %tmp0, align 4
+  %i.next = add nuw nsw i64 %i, 2
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+
+; CHECK: LV: Creating an interleave group with:  store i32 %tmp3, i32* %tmp0, align 4
+; CHECK: LV: Inserted:  store i32 %tmp1, i32* %tmp2, align 4
+; CHECK:     into the interleave group with  store i32 %tmp3, i32* %tmp0, align 4
+; CHECK: LV: Creating an interleave group with:  %tmp3 = load i32, i32* %tmp2, align 4
+; CHECK: LV: Inserted:  %tmp1 = load i32, i32* %tmp0, align 4
+; CHECK:     into the interleave group with  %tmp3 = load i32, i32* %tmp2, align 4
+
+; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction:   %tmp1 = load i32, i32* %tmp0, align 4
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %tmp3 = load i32, i32* %tmp2, align 4
+;            (vl; vl; vperm, vpkg)
+
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   store i32 %tmp1, i32* %tmp2, align 4
+; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction:   store i32 %tmp3, i32* %tmp0, align 4
+;            (vmrlf; vmrhf; vst; vst)
+}
+