GlobalISel: Implement widenScalar for g_extract scalar results

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@354293 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h b/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
index 679e3a0c3d2d42ce83dd70a5c0412905aaff84c7..5ad757769cce192c0f6786642a8ce7d165b4f0aa 100644 (file)
--- a/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
+++ b/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
@@ -237,6 +237,10 @@ public:
     return *State.MF;
   }
 
+  const DataLayout &getDataLayout() const {
+    return getMF().getFunction().getParent()->getDataLayout();
+  }
+
   /// Getter for DebugLoc
   const DebugLoc &getDL() { return State.DL; }
 
@@ -465,6 +469,11 @@ public:
   /// \return The newly created instruction.
   MachineInstrBuilder buildSExt(const DstOp &Res, const SrcOp &Op);
 
+  /// Build and insert a G_PTRTOINT instruction.
+  MachineInstrBuilder buildPtrToInt(const DstOp &Dst, const SrcOp &Src) {
+    return buildInstr(TargetOpcode::G_PTRTOINT, {Dst}, {Src});
+  }
+
   /// \return The opcode of the extension the target wants to use for boolean
   /// values.
   unsigned getBoolExtOp(bool IsVec, bool IsFP) const;

diff --git a/include/llvm/IR/DataLayout.h b/include/llvm/IR/DataLayout.h
index 274ec2351d8be700965d3ea398608e5ba1777b9b..869b27b1fdaf42184b927826ef0cc27be1ea6569 100644 (file)
--- a/include/llvm/IR/DataLayout.h
+++ b/include/llvm/IR/DataLayout.h
@@ -345,10 +345,13 @@ public:
     return NonIntegralAddressSpaces;
   }
 
-  bool isNonIntegralPointerType(PointerType *PT) const {
+  bool isNonIntegralAddressSpace(unsigned AddrSpace) const {
     ArrayRef<unsigned> NonIntegralSpaces = getNonIntegralAddressSpaces();
-    return find(NonIntegralSpaces, PT->getAddressSpace()) !=
-           NonIntegralSpaces.end();
+    return find(NonIntegralSpaces, AddrSpace) != NonIntegralSpaces.end();
+  }
+
+  bool isNonIntegralPointerType(PointerType *PT) const {
+    return isNonIntegralAddressSpace(PT->getAddressSpace());
   }
 
   bool isNonIntegralPointerType(Type *Ty) const {

diff --git a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index e8b5b8cd3b527bf77ff1dc35df7ea860acbc72db..38f0e221431ccbc926cb88d9539b0015363cd8ea 100644 (file)
--- a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -846,20 +846,62 @@ LegalizerHelper::widenScalarUnmergeValues(MachineInstr &MI, unsigned TypeIdx,
 LegalizerHelper::LegalizeResult
 LegalizerHelper::widenScalarExtract(MachineInstr &MI, unsigned TypeIdx,
                                     LLT WideTy) {
-  if (TypeIdx != 1)
-    return UnableToLegalize;
-
+  unsigned DstReg = MI.getOperand(0).getReg();
   unsigned SrcReg = MI.getOperand(1).getReg();
   LLT SrcTy = MRI.getType(SrcReg);
+
+  LLT DstTy = MRI.getType(DstReg);
+  unsigned Offset = MI.getOperand(2).getImm();
+
+  if (TypeIdx == 0) {
+    if (SrcTy.isVector() || DstTy.isVector())
+      return UnableToLegalize;
+
+    SrcOp Src(SrcReg);
+    if (SrcTy.isPointer()) {
+      // Extracts from pointers can be handled only if they are really just
+      // simple integers.
+      const DataLayout &DL = MIRBuilder.getDataLayout();
+      if (DL.isNonIntegralAddressSpace(SrcTy.getAddressSpace()))
+        return UnableToLegalize;
+
+      LLT SrcAsIntTy = LLT::scalar(SrcTy.getSizeInBits());
+      Src = MIRBuilder.buildPtrToInt(SrcAsIntTy, Src);
+      SrcTy = SrcAsIntTy;
+    }
+
+    if (DstTy.isPointer())
+      return UnableToLegalize;
+
+    if (Offset == 0) {
+      // Avoid a shift in the degenerate case.
+      MIRBuilder.buildTrunc(DstReg,
+                            MIRBuilder.buildAnyExtOrTrunc(WideTy, Src));
+      MI.eraseFromParent();
+      return Legalized;
+    }
+
+    // Do a shift in the source type.
+    LLT ShiftTy = SrcTy;
+    if (WideTy.getSizeInBits() > SrcTy.getSizeInBits()) {
+      Src = MIRBuilder.buildAnyExt(WideTy, Src);
+      ShiftTy = WideTy;
+    } else if (WideTy.getSizeInBits() > SrcTy.getSizeInBits())
+      return UnableToLegalize;
+
+    auto LShr = MIRBuilder.buildLShr(
+      ShiftTy, Src, MIRBuilder.buildConstant(ShiftTy, Offset));
+    MIRBuilder.buildTrunc(DstReg, LShr);
+    MI.eraseFromParent();
+    return Legalized;
+  }
+
   if (!SrcTy.isVector())
     return UnableToLegalize;
 
-  unsigned DstReg = MI.getOperand(0).getReg();
-  LLT DstTy = MRI.getType(DstReg);
   if (DstTy != SrcTy.getElementType())
     return UnableToLegalize;
 
-  unsigned Offset = MI.getOperand(2).getImm();
   if (Offset % SrcTy.getScalarSizeInBits() != 0)
     return UnableToLegalize;
 

diff --git a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 4487bffacdc72acd5e351370589ceb3d80d19947..58fd948ccc587c40feb72bce820957e77078d4cb 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -516,9 +516,10 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST,
       .widenScalarIf(
           [=](const LegalityQuery &Query) {
             const LLT Ty1 = Query.Types[1];
-            return (Ty1.getScalarSizeInBits() < 16);
+            return Ty1.isVector() && Ty1.getScalarSizeInBits() < 16;
           },
-          LegalizeMutations::widenScalarOrEltToNextPow2(1, 16));
+          LegalizeMutations::widenScalarOrEltToNextPow2(1, 16))
+    .clampScalar(0, S16, S256);
 
   // TODO: vectors of pointers
   getActionDefinitionsBuilder(G_BUILD_VECTOR)

diff --git a/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir b/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir
index c049231a3a436bbae9b7b6e6bd3deb50ca5c91b4..58ed38d616d0f9fdb86852d5d8fe48b360aa48a9 100644 (file)
--- a/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir
+++ b/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir
@@ -523,3 +523,296 @@ body: |
     %1:_(<2 x s16>) = G_EXTRACT %0, 32
     $vgpr0 = COPY %1
 ...
+
+---
+name: test_extract_s8_s16_offset0
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s16_offset0
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s16) = G_TRUNC %0
+    %2:_(s8) = G_EXTRACT %1, 0
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: test_extract_s8_s16_offset1
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s16_offset1
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[C]](s32)
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC1]](s16)
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s16) = G_TRUNC %0
+    %2:_(s8) = G_EXTRACT %1, 1
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: test_extract_s8_s16_offset8
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s16_offset8
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[C]](s32)
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC1]](s16)
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s16) = G_TRUNC %0
+    %2:_(s8) = G_EXTRACT %1, 8
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: test_extract_s8_s32_offset0
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s32_offset0
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 0
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_s32_offset1
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s32_offset1
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 1
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_s32_offset8
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s32_offset8
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 8
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_s32_offset16
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s32_offset16
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 16
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_s32_offset24
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_s32_offset24
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 16
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_p3_offset0
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_p3_offset0
+    ; CHECK: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+    ; CHECK: [[PTRTOINT:%[0-9]+]]:_(s32) = G_PTRTOINT [[COPY]](p3)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[PTRTOINT]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(p3) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 0
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_p3_offset8
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s8_p3_offset8
+    ; CHECK: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+    ; CHECK: [[PTRTOINT:%[0-9]+]]:_(s32) = G_PTRTOINT [[COPY]](p3)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[PTRTOINT]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(p3) = COPY $vgpr0
+    %1:_(s8) = G_EXTRACT %0, 8
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s1_s8_offset0
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s1_s8_offset0
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY1]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_TRUNC %0
+    %2:_(s1) = G_EXTRACT %1, 0
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: test_extract_s1_s8_offset2
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: test_extract_s1_s8_offset2
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[C]](s32)
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC1]](s16)
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_TRUNC %0
+    %2:_(s1) = G_EXTRACT %1, 2
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: test_extract_s8_s64_offset2
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: test_extract_s8_s64_offset2
+    ; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+    ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[C]](s64)
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY]], [[TRUNC]](s32)
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+    ; CHECK: $vgpr0 = COPY [[TRUNC1]](s32)
+    %0:_(s64) = COPY $vgpr0_vgpr1
+    %1:_(s8) = G_EXTRACT %0, 2
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s8_s64_offset16
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: test_extract_s8_s64_offset16
+    ; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+    ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[C]](s64)
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY]], [[TRUNC]](s32)
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+    ; CHECK: $vgpr0 = COPY [[TRUNC1]](s32)
+    %0:_(s64) = COPY $vgpr0_vgpr1
+    %1:_(s8) = G_EXTRACT %0, 16
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s16_s64_offset16
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: test_extract_s16_s64_offset16
+    ; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+    ; CHECK: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[COPY]](s64), 16
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s64) = COPY $vgpr0_vgpr1
+    %1:_(s16) = G_EXTRACT %0, 16
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s16_s64_offset32
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: test_extract_s16_s64_offset32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+    ; CHECK: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[COPY]](s64), 32
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s64) = COPY $vgpr0_vgpr1
+    %1:_(s16) = G_EXTRACT %0, 32
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...
+
+---
+name: test_extract_s16_s64_offset48
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: test_extract_s16_s64_offset48
+    ; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
+    ; CHECK: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[COPY]](s64), 48
+    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
+    ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32)
+    %0:_(s64) = COPY $vgpr0_vgpr1
+    %1:_(s16) = G_EXTRACT %0, 48
+    %2:_(s32) = G_ANYEXT %1
+    $vgpr0 = COPY %2
+...