[GlobalISel][AArch64] Legalize vector G_FPOW

[paquette]

Summary

This instruction is legalized in the same way as G_FSIN, G_FCOS, G_FLOG10, etc.

Update legalize-pow.mir and arm64-vfloatintrinsics.ll to reflect the change.

llvm/trunk/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
   case TargetOpcode::G_FCOS:
   case TargetOpcode::G_FSIN:
   case TargetOpcode::G_FLOG10:
   case TargetOpcode::G_FLOG:
   case TargetOpcode::G_FLOG2:
   case TargetOpcode::G_FSQRT:
   case TargetOpcode::G_FEXP:
   case TargetOpcode::G_FEXP2:
+  case TargetOpcode::G_FPOW:
     assert(TypeIdx == 0);
     Observer.changingInstr(MI);
 
     for (unsigned I = 1, E = MI.getNumOperands(); I != E; ++I)
       widenScalarSrc(MI, WideTy, I, TargetOpcode::G_FPEXT);
 
     widenScalarDst(MI, WideTy, 0, TargetOpcode::G_FPTRUNC);
     Observer.changedInstr(MI);

llvm/trunk/lib/Target/AArch64/AArch64LegalizerInfo.cpp

AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST) {
   getActionDefinitionsBuilder({G_SMULH, G_UMULH}).legalFor({s32, s64});
 
   getActionDefinitionsBuilder({G_UADDE, G_USUBE, G_SADDO, G_SSUBO, G_UADDO})
       .legalFor({{s32, s1}, {s64, s1}});
 
   getActionDefinitionsBuilder({G_FADD, G_FSUB, G_FMA, G_FMUL, G_FDIV, G_FNEG})
     .legalFor({s32, s64, v2s64, v4s32, v2s32});
 
-  getActionDefinitionsBuilder({G_FREM, G_FPOW}).libcallFor({s32, s64});
+  getActionDefinitionsBuilder(G_FREM).libcallFor({s32, s64});
 
   getActionDefinitionsBuilder({G_FCEIL, G_FABS, G_FSQRT, G_FFLOOR})
       // If we don't have full FP16 support, then scalarize the elements of
       // vectors containing fp16 types.
       .fewerElementsIf(
           [=, &ST](const LegalityQuery &Query) {
             const auto &Ty = Query.Types[0];
             return Ty.isVector() && Ty.getElementType() == s16 &&
                    !ST.hasFullFP16();
           },
           [=](const LegalityQuery &Query) { return std::make_pair(0, s16); })
       // If we don't have full FP16 support, then widen s16 to s32 if we
       // encounter it.
       .widenScalarIf(
           [=, &ST](const LegalityQuery &Query) {
             return Query.Types[0] == s16 && !ST.hasFullFP16();
           },
           [=](const LegalityQuery &Query) { return std::make_pair(0, s32); })
       .legalFor({s16, s32, s64, v2s32, v4s32, v2s64, v2s16, v4s16, v8s16});
 
   getActionDefinitionsBuilder(
-      {G_FCOS, G_FSIN, G_FLOG10, G_FLOG, G_FLOG2, G_FEXP, G_FEXP2})
+      {G_FCOS, G_FSIN, G_FLOG10, G_FLOG, G_FLOG2, G_FEXP, G_FEXP2, G_FPOW})
       // We need a call for these, so we always need to scalarize.
       .scalarize(0)
       // Regardless of FP16 support, widen 16-bit elements to 32-bits.
       .minScalar(0, s32)
       .libcallFor({s32, s64, v2s32, v4s32, v2s64});
 
   getActionDefinitionsBuilder(G_INSERT)
       .unsupportedIf([=](const LegalityQuery &Query) {

llvm/trunk/test/CodeGen/AArch64/GlobalISel/legalize-pow.mir

bb.0.entry:
     %2:_(s32) = COPY $s2
     %3:_(s32) = COPY $s3
     %4:_(s64) = G_FPOW %0, %1
     $x0 = COPY %4(s64)
     %5:_(s32) = G_FPOW %2, %3
     $w0 = COPY %5(s32)
 
 ...
+---
+name:            test_v4f16.pow
+alignment:       2
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4f16.pow
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $d1
+    ; CHECK: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK: [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; CHECK: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
+    ; CHECK: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT1]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY2]](s32)
+    ; CHECK: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
+    ; CHECK: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT2]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT3]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY3]](s32)
+    ; CHECK: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
+    ; CHECK: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT4]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT5]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY4]](s32)
+    ; CHECK: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
+    ; CHECK: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT6]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT7]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY5]](s32)
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16)
+    ; CHECK: $d0 = COPY [[BUILD_VECTOR]](<4 x s16>)
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:_(<4 x s16>) = COPY $d0
+    %1:_(<4 x s16>) = COPY $d1
+    %2:_(<4 x s16>) = G_FPOW %0, %1
+    $d0 = COPY %2(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8f16.pow
+alignment:       2
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8f16.pow
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $q1
+    ; CHECK: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
+    ; CHECK: [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16), [[UV12:%[0-9]+]]:_(s16), [[UV13:%[0-9]+]]:_(s16), [[UV14:%[0-9]+]]:_(s16), [[UV15:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<8 x s16>)
+    ; CHECK: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
+    ; CHECK: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV8]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT1]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY2]](s32)
+    ; CHECK: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
+    ; CHECK: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV9]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT2]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT3]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY3]](s32)
+    ; CHECK: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
+    ; CHECK: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV10]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT4]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT5]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY4]](s32)
+    ; CHECK: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
+    ; CHECK: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV11]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT6]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT7]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY5]](s32)
+    ; CHECK: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
+    ; CHECK: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[UV12]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT8]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT9]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY6]](s32)
+    ; CHECK: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
+    ; CHECK: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[UV13]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT10]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT11]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY7]](s32)
+    ; CHECK: [[FPEXT12:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
+    ; CHECK: [[FPEXT13:%[0-9]+]]:_(s32) = G_FPEXT [[UV14]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT12]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT13]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY8:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY8]](s32)
+    ; CHECK: [[FPEXT14:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
+    ; CHECK: [[FPEXT15:%[0-9]+]]:_(s32) = G_FPEXT [[UV15]](s16)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[FPEXT14]](s32)
+    ; CHECK: $s1 = COPY [[FPEXT15]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY9:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[COPY9]](s32)
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16), [[FPTRUNC4]](s16), [[FPTRUNC5]](s16), [[FPTRUNC6]](s16), [[FPTRUNC7]](s16)
+    ; CHECK: $q0 = COPY [[BUILD_VECTOR]](<8 x s16>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<8 x s16>) = COPY $q0
+    %1:_(<8 x s16>) = COPY $q1
+    %2:_(<8 x s16>) = G_FPOW %0, %1
+    $q0 = COPY %2(<8 x s16>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v2f32.pow
+alignment:       2
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2f32.pow
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV]](s32)
+    ; CHECK: $s1 = COPY [[UV2]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV1]](s32)
+    ; CHECK: $s1 = COPY [[UV3]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY3]](s32)
+    ; CHECK: $d0 = COPY [[BUILD_VECTOR]](<2 x s32>)
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:_(<2 x s32>) = COPY $d0
+    %1:_(<2 x s32>) = COPY $d1
+    %2:_(<2 x s32>) = G_FPOW %0, %1
+    $d0 = COPY %2(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4f32.pow
+alignment:       2
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4f32.pow
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
+    ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<4 x s32>)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV]](s32)
+    ; CHECK: $s1 = COPY [[UV4]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV1]](s32)
+    ; CHECK: $s1 = COPY [[UV5]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV2]](s32)
+    ; CHECK: $s1 = COPY [[UV6]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $s0 = COPY [[UV3]](s32)
+    ; CHECK: $s1 = COPY [[UV7]](s32)
+    ; CHECK: BL &powf, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit $s1, implicit-def $s0
+    ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32)
+    ; CHECK: $q0 = COPY [[BUILD_VECTOR]](<4 x s32>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<4 x s32>) = COPY $q0
+    %1:_(<4 x s32>) = COPY $q1
+    %2:_(<4 x s32>) = G_FPOW %0, %1
+    $q0 = COPY %2(<4 x s32>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v2f64.pow
+alignment:       2
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2f64.pow
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1
+    ; CHECK: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>)
+    ; CHECK: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $d0 = COPY [[UV]](s64)
+    ; CHECK: $d1 = COPY [[UV2]](s64)
+    ; CHECK: BL &pow, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $d0, implicit $d1, implicit-def $d0
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $d0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $d0 = COPY [[UV1]](s64)
+    ; CHECK: $d1 = COPY [[UV3]](s64)
+    ; CHECK: BL &pow, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $d0, implicit $d1, implicit-def $d0
+    ; CHECK: [[COPY3:%[0-9]+]]:_(s64) = COPY $d0
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY2]](s64), [[COPY3]](s64)
+    ; CHECK: $q0 = COPY [[BUILD_VECTOR]](<2 x s64>)
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:_(<2 x s64>) = COPY $q0
+    %1:_(<2 x s64>) = COPY $q1
+    %2:_(<2 x s64>) = G_FPOW %0, %1
+    $q0 = COPY %2(<2 x s64>)
+    RET_ReallyLR implicit $q0

llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll

define %v4f16 @test_v4f16.cos(%v4f16 %a) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v4f16.cos:
   ; CHECK-COUNT-4: bl cosf
   ; GISEL-LABEL:   test_v4f16.cos:
   ; GISEL-COUNT-4: bl cosf
   %1 = call %v4f16 @llvm.cos.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+
+; FALLBACK-NOT: remark{{.*}}test_v4f16.pow
 define %v4f16 @test_v4f16.pow(%v4f16 %a, %v4f16 %b) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v4f16.pow:
+  ; GISEL-LABEL:   test_v4f16.pow:
   ; CHECK-COUNT-4: bl pow
+  ; GISEL-COUNT-4: bl pow
   %1 = call %v4f16 @llvm.pow.v4f16(%v4f16 %a, %v4f16 %b)
   ret %v4f16 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v4f16.exp
 define %v4f16 @test_v4f16.exp(%v4f16 %a) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v4f16.exp:
define %v8f16 @test_v8f16.sqrt(%v8f16 %a) {
   ; GISEL-FP16-NEXT:      ret
   %1 = call %v8f16 @llvm.sqrt.v8f16(%v8f16 %a)
   ret %v8f16 %1
 }
 define %v8f16 @test_v8f16.powi(%v8f16 %a, i32 %b) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v8f16.powi:
   ; CHECK-COUNT-8: bl __powi
+  ; GISEL-LABEL:   test_v8f16.powi:
+  ; GISEL-COUNT-8: bl __powi
   %1 = call %v8f16 @llvm.powi.v8f16(%v8f16 %a, i32 %b)
   ret %v8f16 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v8f16.sin
 define %v8f16 @test_v8f16.sin(%v8f16 %a) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v8f16.sin:
define %v8f16 @test_v8f16.cos(%v8f16 %a) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v8f16.cos:
   ; CHECK-COUNT-8: bl cosf
   ; GISEL-LABEL:   test_v8f16.cos:
   ; GISEL-COUNT-8: bl cosf
   %1 = call %v8f16 @llvm.cos.v8f16(%v8f16 %a)
   ret %v8f16 %1
 }
+
+; FALLBACK-NOT: remark{{.*}}test_v8f16.pow
 define %v8f16 @test_v8f16.pow(%v8f16 %a, %v8f16 %b) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v8f16.pow:
   ; CHECK-COUNT-8: bl pow
+  ; GISEL-LABEL:   test_v8f16.pow:
+  ; GISEL-COUNT-8: bl pow
   %1 = call %v8f16 @llvm.pow.v8f16(%v8f16 %a, %v8f16 %b)
   ret %v8f16 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v8f16.exp
 define %v8f16 @test_v8f16.exp(%v8f16 %a) {
   ; This operation is expanded, whether with or without +fullfp16.
   ; CHECK-LABEL:   test_v8f16.exp:

 ; FALLBACK-NOT: remark{{.*}}test_v2f32.cos
 ; CHECK: test_v2f32.cos:
 define %v2f32 @test_v2f32.cos(%v2f32 %a) {
   ; CHECK: cos
   ; GISEL: cos
   %1 = call %v2f32 @llvm.cos.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
+
+; FALLBACK-NOT: remark{{.*}}test_v2f32.pow
 ; CHECK: test_v2f32.pow:
+; GISEL-LABEL: test_v2f32.pow:
 define %v2f32 @test_v2f32.pow(%v2f32 %a, %v2f32 %b) {
   ; CHECK: pow
+  ; GISEL: pow
   %1 = call %v2f32 @llvm.pow.v2f32(%v2f32 %a, %v2f32 %b)
   ret %v2f32 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v2f32.exp
 ; CHECK: test_v2f32.exp:
 ; GISEL: test_v2f32.exp:
 define %v2f32 @test_v2f32.exp(%v2f32 %a) {

 ; FALLBACK-NOT: remark{{.*}}test_v4f32.cos
 ; CHECK: test_v4f32.cos:
 define %v4f32 @test_v4f32.cos(%v4f32 %a) {
   ; CHECK: cos
   ; GISEL: cos
   %1 = call %v4f32 @llvm.cos.v4f32(%v4f32 %a)
   ret %v4f32 %1
 }
+
+; FALLBACK-NOT: remark{{.*}}test_v4f32.pow
 ; CHECK: test_v4f32.pow:
+; GISEL-LABEL: test_v4f32.pow:
 define %v4f32 @test_v4f32.pow(%v4f32 %a, %v4f32 %b) {
   ; CHECK: pow
+  ; GISEL: pow
   %1 = call %v4f32 @llvm.pow.v4f32(%v4f32 %a, %v4f32 %b)
   ret %v4f32 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v4f32.exp
 ; CHECK: test_v4f32.exp:
 ; GISEL: test_v4f32.exp:
 define %v4f32 @test_v4f32.exp(%v4f32 %a) {

 ; FALLBACK-NOT: remark{{.*}}test_v2f64.cos
 ; CHECK: test_v2f64.cos:
 define %v2f64 @test_v2f64.cos(%v2f64 %a) {
   ; CHECK: cos
   ; GISEL: cos
   %1 = call %v2f64 @llvm.cos.v2f64(%v2f64 %a)
   ret %v2f64 %1
 }
+
+; FALLBACK-NOT: remark{{.*}}test_v2f64.pow
 ; CHECK: test_v2f64.pow:
+; GISEL-LABEL: test_v2f64.pow:
 define %v2f64 @test_v2f64.pow(%v2f64 %a, %v2f64 %b) {
   ; CHECK: pow
+  ; GISEL: pow
   %1 = call %v2f64 @llvm.pow.v2f64(%v2f64 %a, %v2f64 %b)
   ret %v2f64 %1
 }
 
 ; FALLBACK-NOT: remark{{.*}}test_v2f64.exp
 ; CHECK: test_v2f64.exp:
 ; GISEL: test_v2f64.exp:
 define %v2f64 @test_v2f64.exp(%v2f64 %a) {