[Hexagon] Printing packet brackets when asm printing and adding a number of tests that test packet brackets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240051 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp

@@ -11,9 +11,10 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "HexagonMCTargetDesc.h"
+#include "Hexagon.h"
 #include "HexagonMCAsmInfo.h"
 #include "HexagonMCELFStreamer.h"
-#include "HexagonMCTargetDesc.h"
 #include "MCTargetDesc/HexagonInstPrinter.h"
 #include "llvm/MC/MCCodeGenInfo.h"
 #include "llvm/MC/MCContext.h"
@@ -58,6 +59,57 @@ createHexagonMCSubtargetInfo(const Triple &TT, StringRef CPU, StringRef FS) {
   return X;
 }
 
+namespace {
+class HexagonTargetAsmStreamer : public HexagonTargetStreamer {
+  MCContext &Ctx;
+  formatted_raw_ostream &OS;
+  bool isVerboseAsm;
+  MCInstPrinter &IP;
+public:
+  HexagonTargetAsmStreamer(MCStreamer &S,
+                           formatted_raw_ostream &OS, bool isVerboseAsm,
+                           MCInstPrinter &IP)
+      : HexagonTargetStreamer(S), Ctx(Ctx), OS(OS), isVerboseAsm(isVerboseAsm),
+        IP(IP) {}
+  MCStreamer &getStreamer() { return Streamer; }
+  void prettyPrintAsm(MCInstPrinter &InstPrinter, raw_ostream &OS,
+                      const MCInst &Inst, const MCSubtargetInfo &STI) override {
+    assert(HexagonMCInstrInfo::isBundle(Inst));
+    assert(HexagonMCInstrInfo::bundleSize(Inst) <= HEXAGON_PACKET_SIZE);
+    std::string Buffer;
+    {
+      raw_string_ostream TempStream(Buffer);
+      InstPrinter.printInst(&Inst, TempStream, "", STI);
+    }
+    StringRef Contents(Buffer);
+    auto PacketBundle = Contents.rsplit('\n');
+    auto HeadTail = PacketBundle.first.split('\n');
+    auto Preamble = "\t{\n\t\t";
+    auto Separator = "";
+    while(!HeadTail.first.empty()) {
+      OS << Separator;
+      StringRef Inst;
+      auto Duplex = HeadTail.first.split('\v');
+      if(!Duplex.second.empty()){
+        OS << Duplex.first << "\n";
+        Inst = Duplex.second;
+      }
+      else {
+        if(!HeadTail.first.startswith("immext"))
+          Inst = Duplex.first;
+      }
+      OS << Preamble;
+      OS << Inst;
+      HeadTail = HeadTail.second.split('\n');
+      Preamble = "";
+      Separator = "\n\t\t";
+    }
+    if(HexagonMCInstrInfo::bundleSize(Inst) != 0)
+      OS << "\n\t}" << PacketBundle.second;
+  }
+};
+}
+
 namespace {
 class HexagonTargetELFStreamer : public HexagonTargetStreamer {
 public:
@@ -126,6 +178,12 @@ static MCInstPrinter *createHexagonMCInstPrinter(const Triple &T,
     return nullptr;
 }
 
+MCTargetStreamer *createMCAsmTargetStreamer(
+      MCStreamer &S, formatted_raw_ostream &OS, MCInstPrinter *InstPrint,
+      bool IsVerboseAsm) {
+  return new HexagonTargetAsmStreamer(S,  OS, IsVerboseAsm, *InstPrint);
+}
+
 static MCStreamer *createMCStreamer(Triple const &T, MCContext &Context,
                                     MCAsmBackend &MAB, raw_pwrite_stream &OS,
                                     MCCodeEmitter *Emitter, bool RelaxAll) {
@@ -169,6 +227,10 @@ extern "C" void LLVMInitializeHexagonTargetMC() {
   // Register the obj streamer
   TargetRegistry::RegisterELFStreamer(TheHexagonTarget, createMCStreamer);
 
+  // Register the asm streamer
+  TargetRegistry::RegisterAsmTargetStreamer(TheHexagonTarget,
+                                            createMCAsmTargetStreamer);
+
   // Register the MC Inst Printer
   TargetRegistry::RegisterMCInstPrinter(TheHexagonTarget,
                                         createHexagonMCInstPrinter);

test/CodeGen/Hexagon/always-ext.ll

@@ -4,9 +4,11 @@
 ; due to a store that has a must-extend operand.
 
 ; CHECK: CuSuiteAdd.exit.us
+; CHECK: {
 ; CHECK-NOT: call abort
 ; CHECK: memw(##0)
 ; CHECK: memw(r{{[0-9+]}}{{ *}}<<{{ *}}#2{{ *}}+{{ *}}##4)
+; CHECK: }
 
 %struct.CuTest.1.28.31.37.40.43.52.55.67.85.111 = type { i8*, void (%struct.CuTest.1.28.31.37.40.43.52.55.67.85.111*)*, i32, i32, i8*, [23 x i32]* }
 %struct.CuSuite.2.29.32.38.41.44.53.56.68.86.112 = type { i32, [1024 x %struct.CuTest.1.28.31.37.40.43.52.55.67.85.111*], i32 }

test/CodeGen/Hexagon/bugAsmHWloop.ll

@@ -0,0 +1,71 @@
+; RUN: llc -march=hexagon < %s | FileCheck %s
+
+; CHECK: {
+; CHECK: loop0(.LBB
+; CHECK-NOT: loop0(##.LBB
+
+target datalayout = "e-p:32:32:32-i64:64:64-i32:32:32-i16:16:16-i1:32:32-f64:64:64-f32:32:32-v64:64:64-v32:32:32-a0:0-n16:32"
+target triple = "hexagon"
+
+define i32 @q6zip_uncompress(i8* %out_buf, i32* %out_buf_size, i8* %in_buf, i32 %in_buf_size, i8* nocapture %dict, i32 %dict_size) nounwind {
+entry:
+  %0 = bitcast i8* %in_buf to i32*
+  %incdec.ptr = getelementptr inbounds i8, i8* %in_buf, i32 4
+  %1 = load i32, i32* %0, align 4, !tbaa !0
+  %2 = ptrtoint i8* %incdec.ptr to i32
+  %and.i = and i32 %2, 31
+  %sub.i = sub i32 %2, %and.i
+  %3 = inttoptr i32 %sub.i to i8*
+  %add.i = add i32 %in_buf_size, 31
+  %sub2.i = add i32 %add.i, %and.i
+  %div.i = lshr i32 %sub2.i, 5
+  %4 = tail call i32 @llvm.hexagon.A2.combine.ll(i32 32, i32 %div.i) nounwind
+  %5 = tail call i64 @llvm.hexagon.A4.combineir(i32 32, i32 %4) nounwind
+  tail call void asm sideeffect "l2fetch($0,$1)", "r,r,~{memory}"(i8* %3, i64 %5) nounwind, !srcloc !3
+  %6 = ptrtoint i8* %out_buf to i32
+  br label %for.body.i
+
+for.body.i:                                       ; preds = %for.body.i, %entry
+  %i.02.i = phi i32 [ 0, %entry ], [ %inc.i, %for.body.i ]
+  %addr.addr.01.i = phi i32 [ %6, %entry ], [ %add.i14, %for.body.i ]
+  tail call void asm sideeffect "dczeroa($0)", "r"(i32 %addr.addr.01.i) nounwind, !srcloc !4
+  %add.i14 = add i32 %addr.addr.01.i, 32
+  %inc.i = add i32 %i.02.i, 1
+  %exitcond.i = icmp eq i32 %inc.i, 128
+  br i1 %exitcond.i, label %while.cond.preheader, label %for.body.i
+
+while.cond.preheader:                             ; preds = %for.body.i
+  %and = and i32 %1, 3
+  switch i32 %and, label %infloop.preheader [
+    i32 0, label %exit_inflate.split
+    i32 2, label %if.then.preheader
+  ]
+
+if.then.preheader:                                ; preds = %while.cond.preheader
+  br label %if.then
+
+infloop.preheader:                                ; preds = %while.cond.preheader
+  br label %infloop
+
+if.then:                                          ; preds = %if.then.preheader, %if.then
+  tail call void @llvm.prefetch(i8* %incdec.ptr, i32 0, i32 3, i32 1)
+  br label %if.then
+
+exit_inflate.split:                               ; preds = %while.cond.preheader
+  ret i32 0
+
+infloop:                                          ; preds = %infloop.preheader, %infloop
+  br label %infloop
+}
+
+declare void @llvm.prefetch(i8* nocapture, i32, i32, i32) nounwind
+
+declare i64 @llvm.hexagon.A4.combineir(i32, i32) nounwind readnone
+
+declare i32 @llvm.hexagon.A2.combine.ll(i32, i32) nounwind readnone
+
+!0 = !{!"long", !1}
+!1 = !{!"omnipotent char", !2}
+!2 = !{!"Simple C/C++ TBAA"}
+!3 = !{i32 18362}
+!4 = !{i32 18893}

test/CodeGen/Hexagon/cext-valid-packet1.ll

@@ -1,5 +1,4 @@
-; RUN: llc -march=hexagon -mcpu=hexagonv4 < %s | FileCheck %s
+; RUN: llc -march=hexagon < %s | FileCheck %s
-; XFAIL:
 
 ; Check that the packetizer generates valid packets with constant
 ; extended instructions.

test/CodeGen/Hexagon/cext-valid-packet2.ll

@@ -1,44 +1,16 @@
-; RUN: llc -march=hexagon -mcpu=hexagonv4 < %s | FileCheck %s
+; RUN: llc -march=hexagon < %s | FileCheck %s
-; XFAIL:
 ; Check that the packetizer generates valid packets with constant
 ; extended add and base+offset store instructions.
 
-; CHECK: {
+; CHECK: r{{[0-9]+}}{{ *}}={{ *}}add(r{{[0-9]+}},{{ *}}##200000)
-; CHECK-NEXT: r{{[0-9]+}}{{ *}}={{ *}}add(r{{[0-9]+}}, ##{{[0-9]+}})
+; CHECK-NEXT: memw(r{{[0-9]+}}{{ *}}+{{ *}}##12000){{ *}}={{ *}}r{{[0-9]+}}.new
-; CHECK-NEXT: memw(r{{[0-9]+}}+{{ *}}##{{[0-9]+}}){{ *}}={{ *}}r{{[0-9]+}}.new
 ; CHECK-NEXT: }
 
-define i32 @test(i32* nocapture %a, i32* nocapture %b, i32 %c) nounwind {
+define void @test(i32* nocapture %a, i32* nocapture %b, i32 %c) nounwind {
 entry:
-  %add = add nsw i32 %c, 200002
   %0 = load i32, i32* %a, align 4
   %add1 = add nsw i32 %0, 200000
   %arrayidx2 = getelementptr inbounds i32, i32* %a, i32 3000
   store i32 %add1, i32* %arrayidx2, align 4
-  %1 = load i32, i32* %b, align 4
+  ret void
-  %add4 = add nsw i32 %1, 200001
-  %arrayidx5 = getelementptr inbounds i32, i32* %a, i32 1
-  store i32 %add4, i32* %arrayidx5, align 4
-  %arrayidx7 = getelementptr inbounds i32, i32* %b, i32 1
-  %2 = load i32, i32* %arrayidx7, align 4
-  %cmp = icmp sgt i32 %add4, %2
-  br i1 %cmp, label %if.then, label %if.else
-
-if.then:                                          ; preds = %entry
-  %arrayidx8 = getelementptr inbounds i32, i32* %a, i32 2
-  %3 = load i32, i32* %arrayidx8, align 4
-  %arrayidx9 = getelementptr inbounds i32, i32* %b, i32 2000
-  %4 = load i32, i32* %arrayidx9, align 4
-  %sub = sub nsw i32 %3, %4
-  %arrayidx10 = getelementptr inbounds i32, i32* %a, i32 4000
-  store i32 %sub, i32* %arrayidx10, align 4
-  br label %if.end
-
-if.else:                                          ; preds = %entry
-  %arrayidx11 = getelementptr inbounds i32, i32* %b, i32 3200
-  store i32 %add, i32* %arrayidx11, align 4
-  br label %if.end
-
-if.end:                                           ; preds = %if.else, %if.then
-  ret i32 %add
 }

test/CodeGen/Hexagon/checktabs.ll

@@ -0,0 +1,8 @@
+; RUN: llc -march=hexagon < %s | FileCheck --strict-whitespace %s
+; Make sure we are emitting tabs as formatting.
+; CHECK:	{
+; CHECK-NEXT:		{{jump|r}}
+define i32 @foobar(i32 %a, i32 %b) {
+  %1 = add i32 %a, %b
+  ret i32 %1
+}

test/CodeGen/Hexagon/eh_return.ll

@@ -0,0 +1,48 @@
+; RUN: llc -O0 -march=hexagon < %s | FileCheck %s
+; Make sure we generate an exception handling return.
+
+; CHECK:         deallocframe
+; CHECK-NEXT:  }
+; CHECK-NEXT:  {
+; CHECK-NEXT:    r29 = add(r29, r28)
+; CHECK-NEXT:  }
+; CHECK-NEXT:  {
+; CHECK-NEXT:    jumpr r31
+; CHECK-NEXT:  }
+
+target datalayout = "e-p:32:32:32-i64:64:64-i32:32:32-i16:16:16-i1:32:32-f64:64:64-f32:32:32-a0:0-n32"
+target triple = "hexagon-unknown-linux-gnu"
+
+%struct.Data = type { i32, i8* }
+
+define i32 @test_eh_return(i32 %a, i32 %b) nounwind {
+entry:
+  %a.addr = alloca i32, align 4
+  %b.addr = alloca i32, align 4
+  %d = alloca %struct.Data, align 4
+  store i32 %a, i32* %a.addr, align 4
+  store i32 %b, i32* %b.addr, align 4
+  %0 = load i32, i32* %a.addr, align 4
+  %1 = load i32, i32* %b.addr, align 4
+  %cmp = icmp sgt i32 %0, %1
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:                                          ; preds = %entry
+  %2 = load i32, i32* %a.addr, align 4
+  %3 = load i32, i32* %b.addr, align 4
+  %add = add nsw i32 %2, %3
+  ret i32 %add
+
+if.else:                                          ; preds = %entry
+  %call = call i32 @setup(%struct.Data* %d)
+  %_d1 = getelementptr inbounds %struct.Data, %struct.Data* %d, i32 0, i32 0
+  %4 = load i32, i32* %_d1, align 4
+  %_d2 = getelementptr inbounds %struct.Data, %struct.Data* %d, i32 0, i32 1
+  %5 = load i8*, i8** %_d2, align 4
+  call void @llvm.eh.return.i32(i32 %4, i8* %5)
+  unreachable
+}
+
+declare i32 @setup(%struct.Data*)
+
+declare void @llvm.eh.return.i32(i32, i8*) nounwind

test/CodeGen/Hexagon/postinc-offset.ll

@@ -0,0 +1,40 @@
+; RUN: llc -enable-aa-sched-mi -march=hexagon -mcpu=hexagonv5 < %s | FileCheck %s
+
+; CHECK: {
+; CHECK: ={{ *}}memd([[REG0:(r[0-9]+)]]{{ *}}++{{ *}}#8)
+; CHECK-NOT: memw([[REG0]]{{ *}}+{{ *}}#0){{ *}}=
+; CHECK: }
+
+define void @main() #0 {
+cond.end.6:
+  store i32 -1, i32* undef, align 8, !tbaa !0
+  br label %polly.stmt.for.body.i
+
+if.then:
+  unreachable
+
+if.end:
+  ret void
+
+polly.stmt.for.body.i24:
+  %0 = extractelement <2 x i32> %add.ip_vec, i32 1
+  br i1 undef, label %if.end, label %if.then
+
+polly.stmt.for.body.i:
+  %add.ip_vec30 = phi <2 x i32> [ %add.ip_vec, %polly.stmt.for.body.i ], [ zeroinitializer, %cond.end.6 ]
+  %scevgep.phi = phi i32* [ %scevgep.inc, %polly.stmt.for.body.i ], [ undef, %cond.end.6 ]
+  %polly.indvar = phi i32 [ %polly.indvar_next, %polly.stmt.for.body.i ], [ 0, %cond.end.6 ]
+  %vector_ptr = bitcast i32* %scevgep.phi to <2 x i32>*
+  %_p_vec_full = load <2 x i32>, <2 x i32>* %vector_ptr, align 8
+  %add.ip_vec = add <2 x i32> %_p_vec_full, %add.ip_vec30
+  %polly.indvar_next = add nsw i32 %polly.indvar, 2
+  %polly.loop_cond = icmp slt i32 %polly.indvar, 4
+  %scevgep.inc = getelementptr i32, i32* %scevgep.phi, i32 2
+  br i1 %polly.loop_cond, label %polly.stmt.for.body.i, label %polly.stmt.for.body.i24
+}
+
+attributes #0 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"="true" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!0 = !{!"int", !1}
+!1 = !{!"omnipotent char", !2}
+!2 = !{!"Simple C/C++ TBAA"}

test/CodeGen/Hexagon/usr-ovf-dep.ll

@@ -0,0 +1,28 @@
+; RUN: llc -O2 < %s | FileCheck %s
+target datalayout = "e-m:e-p:32:32-i1:32-i64:64-a:0-v32:32-n16:32"
+target triple = "hexagon"
+
+; Check that the two ":sat" instructions are in the same packet.
+; CHECK: foo
+; CHECK: {
+; CHECK: :sat
+; CHECK-NEXT: :sat
+
+target datalayout = "e-m:e-p:32:32-i1:32-i64:64-a:0-v32:32-n16:32"
+target triple = "hexagon"
+
+; Function Attrs: nounwind readnone
+define i32 @foo(i32 %Rs, i32 %Rt, i32 %Ru) #0 {
+entry:
+  %0 = tail call i32 @llvm.hexagon.S2.asr.r.r.sat(i32 %Rs, i32 %Ru)
+  %1 = tail call i32 @llvm.hexagon.S2.asr.r.r.sat(i32 %Rt, i32 %Ru)
+  %add = add nsw i32 %1, %0
+  ret i32 %add
+}
+
+; Function Attrs: nounwind readnone
+declare i32 @llvm.hexagon.S2.asr.r.r.sat(i32, i32) #1
+
+attributes #0 = { nounwind readnone "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone }
+