[Hexagon] Enable .cur formation in MISched for Hexagon V60

Schedule a load and its use in the same packet in MISched. Previously,
isResourceAvailable was returning false for dependences in the same
packet, which prevented MISched from packetizing a load and its use in
the same packet for v60.

Patch by Ikhlas Ajbar.

llvm-svn: 275804

2 files changed, 70 insertions, 0 deletions

diff --git a/llvm/lib/Target/Hexagon/HexagonMachineScheduler.cpp b/llvm/lib/Target/Hexagon/HexagonMachineScheduler.cpp
index b70a6b13b5a..9368a433743 100644
--- a/llvm/lib/Target/Hexagon/HexagonMachineScheduler.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonMachineScheduler.cpp
@@ -158,11 +158,19 @@ bool VLIWResourceModel::isResourceAvailable(SUnit *SU) {
     break;
   }
 
+  MachineFunction &MF = *SU->getInstr()->getParent()->getParent();
+  auto &QII = *MF.getSubtarget<HexagonSubtarget>().getInstrInfo();
+
   // Now see if there are no other dependencies to instructions already
   // in the packet.
   for (unsigned i = 0, e = Packet.size(); i != e; ++i) {
     if (Packet[i]->Succs.size() == 0)
       continue;
+
+    // Enable .cur formation.
+    if (QII.mayBeCurLoad(Packet[i]->getInstr()))
+      continue;
+
     for (SUnit::const_succ_iterator I = Packet[i]->Succs.begin(),
          E = Packet[i]->Succs.end(); I != E; ++I) {
       // Since we do not add pseudos to packets, might as well
diff --git a/llvm/test/CodeGen/Hexagon/v60-cur.ll b/llvm/test/CodeGen/Hexagon/v60-cur.ll
new file mode 100644
index 00000000000..fe24309f5b8
--- /dev/null
+++ b/llvm/test/CodeGen/Hexagon/v60-cur.ll
@@ -0,0 +1,62 @@
+; RUN: llc -march=hexagon < %s | FileCheck %s
+
+; Test that we generate a .cur
+
+; CHECK: v{{[0-9]*}}.cur{{ *}}
+; CHECK: v{{[0-9]*}}.cur{{ *}}
+
+define void @conv3x3_i(i8* noalias nocapture readonly %iptr0, i32 %shift, i32 %width) #0 {
+entry:
+  br i1 undef, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph:
+  br label %for.body
+
+for.body:
+  %iptr0.pn = phi i8* [ %iptr0, %for.body.lr.ph ], [ %iptr0.addr.0121, %for.body ]
+  %j.0115 = phi i32 [ 0, %for.body.lr.ph ], [ %add, %for.body ]
+  %sline000.0114 = phi <16 x i32> [ zeroinitializer, %for.body.lr.ph ], [ %1, %for.body ]
+  %sline100.0113 = phi <16 x i32> [ zeroinitializer, %for.body.lr.ph ], [ zeroinitializer, %for.body ]
+  %iptr0.addr.0121 = getelementptr inbounds i8, i8* %iptr0.pn, i32 64
+  %0 = bitcast i8* %iptr0.addr.0121 to <16 x i32>*
+  %1 = load <16 x i32>, <16 x i32>* %0, align 64, !tbaa !1
+  %2 = load <16 x i32>, <16 x i32>* null, align 64, !tbaa !1
+  %3 = tail call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> %1, <16 x i32> %sline000.0114, i32 4)
+  %4 = tail call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> zeroinitializer, <16 x i32> %sline100.0113, i32 4)
+  %5 = tail call <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32> %2, <16 x i32> zeroinitializer, i32 4)
+  %6 = tail call <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32> %3, <16 x i32> %sline000.0114)
+  %7 = tail call <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32> %5, <16 x i32> zeroinitializer)
+  %8 = tail call <32 x i32> @llvm.hexagon.V6.vrmpybusi(<32 x i32> %6, i32 0, i32 0)
+  %9 = tail call <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32> %8, <32 x i32> zeroinitializer, i32 undef, i32 0)
+  %10 = tail call <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32> %9, <32 x i32> undef, i32 undef, i32 0)
+  %11 = tail call <16 x i32> @llvm.hexagon.V6.hi(<32 x i32> %10)
+  %12 = tail call <16 x i32> @llvm.hexagon.V6.vasrwh(<16 x i32> %11, <16 x i32> undef, i32 %shift)
+  %13 = tail call <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32> undef, <16 x i32> %12)
+  store <16 x i32> %13, <16 x i32>* undef, align 64, !tbaa !1
+  %14 = tail call <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32> zeroinitializer, <32 x i32> %7, i32 undef, i32 1)
+  %15 = tail call <16 x i32> @llvm.hexagon.V6.hi(<32 x i32> %14)
+  %16 = tail call <16 x i32> @llvm.hexagon.V6.vasrwh(<16 x i32> %15, <16 x i32> undef, i32 %shift)
+  %17 = tail call <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32> %16, <16 x i32> undef)
+  store <16 x i32> %17, <16 x i32>* undef, align 64, !tbaa !1
+  %add = add nsw i32 %j.0115, 64
+  %cmp = icmp slt i32 %add, %width
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+declare <16 x i32> @llvm.hexagon.V6.valignbi(<16 x i32>, <16 x i32>, i32) #1
+declare <32 x i32> @llvm.hexagon.V6.vcombine(<16 x i32>, <16 x i32>) #1
+declare <32 x i32> @llvm.hexagon.V6.vrmpybusi(<32 x i32>, i32, i32) #1
+declare <32 x i32> @llvm.hexagon.V6.vrmpybusi.acc(<32 x i32>, <32 x i32>, i32, i32) #1
+declare <16 x i32> @llvm.hexagon.V6.vasrwh(<16 x i32>, <16 x i32>, i32) #1
+declare <16 x i32> @llvm.hexagon.V6.hi(<32 x i32>) #1
+declare <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32>, <16 x i32>) #1
+
+attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvx" }
+attributes #1 = { nounwind readnone }
+
+!1 = !{!2, !2, i64 0}
+!2 = !{!"omnipotent char", !3, i64 0}
+!3 = !{!"Simple C/C++ TBAA"}