PR12716: PPC crashes on vector compare

Vector compare using altivec 'vcmpxxx' instructions have as third argument
a vector register instead of CR one, different from integer and float-point
compares. This leads to a failure in code generation, where 'SelectSETCC'
expects a DAG with a CR register and gets vector register instead.

This patch changes the behavior by just returning a DAG with the 
vector compare instruction based on the type. The patch also adds a testcase
for all vector types llvm defines.

It also included a fix on signed 5-bits predicates printing, where
signed values were not handled correctly as signed (char are unsigned by
default for PowerPC). This generates 'vspltisw' (vector splat)
instruction with SIM out of range.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@165419 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp

@@ -136,21 +136,21 @@ void PPCInstPrinter::printPredicateOperand(const MCInst *MI, unsigned OpNo,
 
 void PPCInstPrinter::printS5ImmOperand(const MCInst *MI, unsigned OpNo,
                                        raw_ostream &O) {
-  char Value = MI->getOperand(OpNo).getImm();
+  int Value = MI->getOperand(OpNo).getImm();
   Value = SignExtend32<5>(Value);
   O << (int)Value;
 }
 
 void PPCInstPrinter::printU5ImmOperand(const MCInst *MI, unsigned OpNo,
                                        raw_ostream &O) {
-  unsigned char Value = MI->getOperand(OpNo).getImm();
+  unsigned int Value = MI->getOperand(OpNo).getImm();
   assert(Value <= 31 && "Invalid u5imm argument!");
   O << (unsigned int)Value;
 }
 
 void PPCInstPrinter::printU6ImmOperand(const MCInst *MI, unsigned OpNo,
                                        raw_ostream &O) {
-  unsigned char Value = MI->getOperand(OpNo).getImm();
+  unsigned int Value = MI->getOperand(OpNo).getImm();
   assert(Value <= 63 && "Invalid u6imm argument!");
   O << (unsigned int)Value;
 }

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

@@ -701,10 +701,29 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) {
     }
   }
 
+  SDValue LHS = N->getOperand(0);
+  SDValue RHS = N->getOperand(1);
+
+  // Altivec Vector compare instructions do not set any CR register by default
+  if (LHS.getValueType().isVector()) {
+    unsigned int Opc;
+    if (LHS.getValueType() == MVT::v16i8)
+      Opc = PPC::VCMPEQUB;
+    else if (LHS.getValueType() == MVT::v4i32)
+      Opc = PPC::VCMPEQUW;
+    else if (LHS.getValueType() == MVT::v8i16)
+      Opc = PPC::VCMPEQUH;
+    else if (LHS.getValueType() == MVT::v4f32)
+      Opc = PPC::VCMPEQFP;
+    else
+      llvm_unreachable("Invalid vector compare type: should be expanded by legalize");
+    return CurDAG->SelectNodeTo(N, Opc, LHS.getValueType(), LHS, RHS);
+  }
+
   bool Inv;
   int OtherCondIdx;
   unsigned Idx = getCRIdxForSetCC(CC, Inv, OtherCondIdx);
-  SDValue CCReg = SelectCC(N->getOperand(0), N->getOperand(1), CC, dl);
+  SDValue CCReg = SelectCC(LHS, RHS, CC, dl);
   SDValue IntCR;
 
   // Force the ccreg into CR7.
@@ -717,7 +736,7 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) {
   if (PPCSubTarget.hasMFOCRF() && OtherCondIdx == -1)
     IntCR = SDValue(CurDAG->getMachineNode(PPC::MFOCRF, dl, MVT::i32, CR7Reg,
                                            CCReg), 0);
- else
+  else
     IntCR = SDValue(CurDAG->getMachineNode(PPC::MFCRpseud, dl, MVT::i32,
                                            CR7Reg, CCReg), 0);
 

lib/Target/PowerPC/PPCISelLowering.cpp

@@ -542,7 +542,9 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const {
 }
 
 EVT PPCTargetLowering::getSetCCResultType(EVT VT) const {
-  return MVT::i32;
+  if (!VT.isVector())
+    return MVT::i32;
+  return VT.changeVectorElementTypeToInteger();
 }
 
 //===----------------------------------------------------------------------===//

test/CodeGen/PowerPC/vec_cmp.ll

@@ -0,0 +1,191 @@
+; RUN: llc -mattr=+altivec < %s | FileCheck %s
+
+; Check vector comparisons using altivec.
+
+
+target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v128:128:128-n32:64"
+target triple = "powerpc64-unknown-linux-gnu"
+
+define <2 x i8> @v2si8_cmp(<2 x i8> %x, <2 x i8> %y) nounwind readnone {
+  %cmp = icmp eq <2 x i8> %x, %y
+  %sext = sext <2 x i1> %cmp to <2 x i8>
+  ret <2 x i8> %sext
+}
+; CHECK: v2si8_cmp:
+; CHECK: vcmpequb {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <4 x i8> @v4si8_cmp(<4 x i8> %x, <4 x i8> %y) nounwind readnone {
+  %cmp = icmp eq <4 x i8> %x, %y
+  %sext = sext <4 x i1> %cmp to <4 x i8>
+  ret <4 x i8> %sext
+}
+; CHECK: v4si8_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <8 x i8> @v8si8_cmp(<8 x i8> %x, <8 x i8> %y) nounwind readnone {
+  %cmp = icmp eq <8 x i8> %x, %y
+  %sext = sext <8 x i1> %cmp to <8 x i8>
+  ret <8 x i8> %sext
+}
+; CHECK: v8si8_cmp:
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <16 x i8> @v16si8_cmp(<16 x i8> %x, <16 x i8> %y) nounwind readnone {
+  %cmp = icmp eq <16 x i8> %x, %y
+  %sext = sext <16 x i1> %cmp to <16 x i8>
+  ret <16 x i8> %sext
+}
+; CHECK: v16si8_cmp:
+; CHECK: vcmpequb {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <32 x i8> @v32si8_cmp(<32 x i8> %x, <32 x i8> %y) nounwind readnone {
+  %cmp = icmp eq <32 x i8> %x, %y
+  %sext = sext <32 x i1> %cmp to <32 x i8>
+  ret <32 x i8> %sext
+}
+; CHECK: v32si8_cmp:
+; CHECK: vcmpequb {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequb {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <2 x i16> @v2si16_cmp(<2 x i16> %x, <2 x i16> %y) nounwind readnone {
+  %cmp = icmp eq <2 x i16> %x, %y
+  %sext = sext <2 x i1> %cmp to <2 x i16>
+  ret <2 x i16> %sext
+}
+; CHECK: v2si16_cmp:
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <4 x i16> @v4si16_cmp(<4 x i16> %x, <4 x i16> %y) nounwind readnone {
+  %cmp = icmp eq <4 x i16> %x, %y
+  %sext = sext <4 x i1> %cmp to <4 x i16>
+  ret <4 x i16> %sext
+}
+; CHECK: v4si16_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <8 x i16> @v8si16_cmp(<8 x i16> %x, <8 x i16> %y) nounwind readnone {
+  %cmp = icmp eq <8 x i16> %x, %y
+  %sext = sext <8 x i1> %cmp to <8 x i16>
+  ret <8 x i16> %sext
+}
+; CHECK: v8si16_cmp:
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <16 x i16> @v16si16_cmp(<16 x i16> %x, <16 x i16> %y) nounwind readnone {
+  %cmp = icmp eq <16 x i16> %x, %y
+  %sext = sext <16 x i1> %cmp to <16 x i16>
+  ret <16 x i16> %sext
+}
+; CHECK: v16si16_cmp:
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <32 x i16> @v32si16_cmp(<32 x i16> %x, <32 x i16> %y) nounwind readnone {
+  %cmp = icmp eq <32 x i16> %x, %y
+  %sext = sext <32 x i1> %cmp to <32 x i16>
+  ret <32 x i16> %sext
+}
+; CHECK: v32si16_cmp:
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequh {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <2 x i32> @v2si32_cmp(<2 x i32> %x, <2 x i32> %y) nounwind readnone {
+  %cmp = icmp eq <2 x i32> %x, %y
+  %sext = sext <2 x i1> %cmp to <2 x i32>
+  ret <2 x i32> %sext
+}
+; CHECK: v2si32_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <4 x i32> @v4si32_cmp(<4 x i32> %x, <4 x i32> %y) nounwind readnone {
+  %cmp = icmp eq <4 x i32> %x, %y
+  %sext = sext <4 x i1> %cmp to <4 x i32>
+  ret <4 x i32> %sext
+}
+; CHECK: v4si32_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <8 x i32> @v8si32_cmp(<8 x i32> %x, <8 x i32> %y) nounwind readnone {
+  %cmp = icmp eq <8 x i32> %x, %y
+  %sext = sext <8 x i1> %cmp to <8 x i32>
+  ret <8 x i32> %sext
+}
+; CHECK: v8si32_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <16 x i32> @v16si32_cmp(<16 x i32> %x, <16 x i32> %y) nounwind readnone {
+  %cmp = icmp eq <16 x i32> %x, %y
+  %sext = sext <16 x i1> %cmp to <16 x i32>
+  ret <16 x i32> %sext
+}
+; CHECK: v16si32_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <32 x i32> @v32si32_cmp(<32 x i32> %x, <32 x i32> %y) nounwind readnone {
+  %cmp = icmp eq <32 x i32> %x, %y
+  %sext = sext <32 x i1> %cmp to <32 x i32>
+  ret <32 x i32> %sext
+}
+; CHECK: v32si32_cmp:
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpequw {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <2 x float> @v2f32_cmp(<2 x float> %x, <2 x float> %y) nounwind readnone {
+entry:
+  %cmp = fcmp oeq <2 x float> %x, %y
+  %sext = sext <2 x i1> %cmp to <2 x i32>
+  %0 = bitcast <2 x i32> %sext to <2 x float>
+  ret <2 x float> %0
+}
+; CHECK: v2f32_cmp:
+; CHECK: vcmpeqfp {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <4 x float> @v4f32_cmp(<4 x float> %x, <4 x float> %y) nounwind readnone {
+entry:
+  %cmp = fcmp oeq <4 x float> %x, %y
+  %sext = sext <4 x i1> %cmp to <4 x i32>
+  %0 = bitcast <4 x i32> %sext to <4 x float>
+  ret <4 x float> %0
+}
+; CHECK: v4f32_cmp:
+; CHECK: vcmpeqfp {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+
+
+define <8 x float> @v8f32_cmp(<8 x float> %x, <8 x float> %y) nounwind readnone {
+entry:
+  %cmp = fcmp oeq <8 x float> %x, %y
+  %sext = sext <8 x i1> %cmp to <8 x i32>
+  %0 = bitcast <8 x i32> %sext to <8 x float>
+  ret <8 x float> %0
+}
+; CHECK: v8f32_cmp:
+; CHECK: vcmpeqfp {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}
+; CHECK: vcmpeqfp {{[0-9]+}}, {{[0-9]+}}, {{[0-9]+}}