GlobalISel: Add basic legalization for G_BITREVERSE

[paquette]

LGTM

[arsenm]

r370979

lib/CodeGen/GlobalISel/LegalizerHelper.cpp

MIRBuilder.buildInstr(TargetOpcode::G_LSHR)
       .addDef(ShrReg)
       .addUse(DstExt)
       .addUse(ShiftAmtReg);
 
     MIRBuilder.buildTrunc(DstReg, ShrReg);
     Observer.changedInstr(MI);
     return Legalized;
   }
+  case TargetOpcode::G_BITREVERSE: {
+    Observer.changingInstr(MI);
+
+    Register DstReg = MI.getOperand(0).getReg();
+    LLT Ty = MRI.getType(DstReg);
+    unsigned DiffBits = WideTy.getScalarSizeInBits() - Ty.getScalarSizeInBits();
+
+    Register DstExt = MRI.createGenericVirtualRegister(WideTy);
+    widenScalarSrc(MI, WideTy, 1, TargetOpcode::G_ANYEXT);
+    MI.getOperand(0).setReg(DstExt);
+    MIRBuilder.setInsertPt(MIRBuilder.getMBB(), ++MIRBuilder.getInsertPt());
+
+    auto ShiftAmt = MIRBuilder.buildConstant(WideTy, DiffBits);
+    auto Shift = MIRBuilder.buildLShr(WideTy, DstExt, ShiftAmt);
+    MIRBuilder.buildTrunc(DstReg, Shift);
+    Observer.changedInstr(MI);
+    return Legalized;
+  }
   case TargetOpcode::G_ADD:
   case TargetOpcode::G_AND:
   case TargetOpcode::G_MUL:
   case TargetOpcode::G_OR:
   case TargetOpcode::G_XOR:
   case TargetOpcode::G_SUB:
     // Perform operation at larger width (any extension is fines here, high bits
     // don't affect the result) and then truncate the result back to the
LegalizerHelper::fewerElementsVector(MachineInstr &MI, unsigned TypeIdx,
   case G_FFLOOR:
   case G_FRINT:
   case G_INTRINSIC_ROUND:
   case G_INTRINSIC_TRUNC:
   case G_FCOS:
   case G_FSIN:
   case G_FSQRT:
   case G_BSWAP:
+  case G_BITREVERSE:
   case G_SDIV:
   case G_SMIN:
   case G_SMAX:
   case G_UMIN:
   case G_UMAX:
   case G_FMINNUM:
   case G_FMAXNUM:
   case G_FMINNUM_IEEE:

lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

getActionDefinitionsBuilder({G_CTLZ, G_CTLZ_ZERO_UNDEF,
     .legalFor({{S32, S32}, {S32, S64}})
     .clampScalar(0, S32, S32)
     .clampScalar(1, S32, S64)
     .scalarize(0)
     .widenScalarToNextPow2(0, 32)
     .widenScalarToNextPow2(1, 32);
 
   // TODO: Expand for > s32
-  getActionDefinitionsBuilder(G_BSWAP)
+  getActionDefinitionsBuilder({G_BSWAP, G_BITREVERSE})
     .legalFor({S32})
     .clampScalar(0, S32, S32)
     .scalarize(0);
 
   if (ST.has16BitInsts()) {
     if (ST.hasVOP3PInsts()) {
       getActionDefinitionsBuilder({G_SMIN, G_SMAX, G_UMIN, G_UMAX})
         .legalFor({S32, S16, V2S16})

test/CodeGen/AMDGPU/GlobalISel/legalize-bitreverse.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: bitreverse_s8
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: bitreverse_s8
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY1]]
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITREVERSE]], [[C]](s32)
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY2]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s8) = G_TRUNC %0
+    %2:_(s8) = G_BITREVERSE %1
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: bitreverse_s16
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: bitreverse_s16
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY1]]
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITREVERSE]], [[C]](s32)
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY2]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s16) = G_TRUNC %0
+    %2:_(s16) = G_BITREVERSE %1
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: bitreverse_s24
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: bitreverse_s24
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY1]]
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITREVERSE]], [[C]](s32)
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: $vgpr0 = COPY [[COPY2]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s24) = G_TRUNC %0
+    %2:_(s24) = G_BITREVERSE %1
+    %3:_(s32) = G_ANYEXT %2
+    $vgpr0 = COPY %3
+...
+
+---
+name: bitreverse_s32
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: bitreverse_s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY]]
+    ; CHECK: $vgpr0 = COPY [[BITREVERSE]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = G_BITREVERSE %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: bitreverse_v2s16
+
+body: |
+  bb.0:
+    liveins: $vgpr0
+    ; CHECK-LABEL: name: bitreverse_v2s16
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
+    ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY1]]
+    ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITREVERSE]], [[C]](s32)
+    ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
+    ; CHECK: [[BITREVERSE1:%[0-9]+]]:_(s32) = G_BITREVERSE [[COPY2]]
+    ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITREVERSE1]], [[C]](s32)
+    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+    ; CHECK: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+    %0:_(<2 x s16>) = COPY $vgpr0
+    %1:_(<2 x s16>) = G_BITREVERSE %0
+    $vgpr0 = COPY %1
+...
+
+---
+name: bitreverse_v2s32
+
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1
+    ; CHECK-LABEL: name: bitreverse_v2s32
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[BITREVERSE:%[0-9]+]]:_(s32) = G_BITREVERSE [[UV]]
+    ; CHECK: [[BITREVERSE1:%[0-9]+]]:_(s32) = G_BITREVERSE [[UV1]]
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[BITREVERSE]](s32), [[BITREVERSE1]](s32)
+    ; CHECK: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
+    %1:_(<2 x s32>) = G_BITREVERSE %0
+    $vgpr0_vgpr1 = COPY %1
+...