Always promote f16

[olista01]

Summary

Bug: https://llvm.org/bugs/show_bug.cgi?id=24549

• Softening support of f16 was never finished, really. Always promote f16 to f32. If f32 is not legal, rely on f32 legalization after promotion.
• Update SoftenFloatRes_FP_EXTEND to look for a promoted operand if the operand's type action is to promote.
• Add the test case from the linked bug.

Ideally, we should test all the cases in
test/CodeGen/ARM/fp16-promote.ll with -vfp2. Doing that needs extra
work:

• some functions in this test don't compile
• all the functions need a new set of CHECKs for when both f32 and f16 are illegal.

[pirama]

Created https://llvm.org/bugs/show_bug.cgi?id=24814 to track testing of fp16-promote.ll with -vfp2.

[pirama]

Ping...

[ab]

lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp

433–434

IIUC, this means we'll do something like:

(f32 (fpext (f16 ...)))
(i32 (fpext (f32 <promoted f16 ...>)))
(i32 (__gnu_h2f_ieee (f32 ...)))
(i32 (__gnu_h2f_ieee (i32 ...)))

even though __gnu_h2f_ieee expects uint16_t. On ARM uint16_t will get promoted to i32, but now we're interpreting a full f32-softened-to-i32 as an i16-promoted-to-i32, no?

Doesn't this mean it's possible to get stuff like:

(i32 (__gnu_h2f_ieee
  (i32 (__gnu_h2f_ieee
    (i32 (load zext from i16 ...))
  ))
))

by legalizing:

(f32 (fpext (f16 (load ...))))

?

[pirama]

lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp

433–434

You are right. I completely missed testing this. Also, Op isn't legalized yet at this point. I'll fix this and upload a new revision.

[olista01]

Hi Pirama,

Are you still working on this? One of our internal tests suites is currently failing for fp16, and this looks like it should fix it (except for the FPEXT issue Ahmed spotted).

Oliver

[pirama]

Hi Oliver, I started on a different activity and completely forgot about this. Hopefully, I'll have some time late next week to look into this. Please feel free to take over this patch if you have the cycles to fix the FPEXT issue.

Thanks,
-Pirama

[olista01]

Updated to correctly handle FPEXT, and get test/CodeGen/ARM/fp16-promote.ll passing. The changes in SelectionDAGBuilder.cpp were required for a few of the test cases, notably those with half-precision arguments (which the clang frontend doesn't currently emit, at least for ARM), and phi nodes.

[ab]

I think this is fine, modulo a couple nits.

Pirama: any comments?

lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

201

&& bitsLT ?

getNode() is pretty involved, and here we'd end up calling it to insert a no-op truncate for every softened f32/f64, right?

test/CodeGen/ARM/fp16-promote.ll

153–156

-LIBCALL/-ALL ?

885–886

Did we start emitting anything else? If not, I'd keep the -NEXT and .fnstart.

[pirama]

LGTM, after ab's changes. Thanks for taking over this patch and also updating the test. Can you also clean up the commit message, now that you updated the test? (Or did you already update it and it is just not visible in Phabricator?)

[olista01]

Thanks, committed revision 252459 with the suggested changes.

test/CodeGen/ARM/fp16-promote.ll

885–886

Yes, we emit a mov here because the soft-float calling convention is different. I'll update the test to make this explicit.

lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp

SDValue DAGTypeLegalizer::SoftenFloatRes_FP_EXTEND(SDNode *N) {
   // entirely possible for both f16 and f32 to be legal, so use the fully
   // hard-float FP_EXTEND rather than FP16_TO_FP.
   if (Op.getValueType() == MVT::f16 && N->getValueType(0) != MVT::f32) {
     Op = DAG.getNode(ISD::FP_EXTEND, SDLoc(N), MVT::f32, Op);
     if (getTypeAction(MVT::f32) == TargetLowering::TypeSoftenFloat)
       SoftenFloatResult(Op.getNode(), 0);
   }
 
+  if (getTypeAction(Op.getValueType()) == TargetLowering::TypePromoteFloat) {
+    Op = GetPromotedFloat(Op);
+    // If the promotion did the FP_EXTEND to the destination type for us,
+    // there's nothing left to do here.
+    if (Op.getValueType() == N->getValueType(0)) {
+      return BitConvertToInteger(Op);
+    }
+  }
+
   RTLIB::Libcall LC = RTLIB::getFPEXT(Op.getValueType(), N->getValueType(0));
   if (getTypeAction(Op.getValueType()) == TargetLowering::TypeSoftenFloat)
     Op = GetSoftenedFloat(Op);
   assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unsupported FP_EXTEND!");
   return TLI.makeLibCall(DAG, LC, NVT, Op, false, SDLoc(N)).first;
 }
 
 // FIXME: Should we just use 'normal' FP_EXTEND / FP_TRUNC instead of special
 // nodes?
 SDValue DAGTypeLegalizer::SoftenFloatRes_FP16_TO_FP(SDNode *N) {
   EVT MidVT = TLI.getTypeToTransformTo(*DAG.getContext(), MVT::f32);
   SDValue Op = N->getOperand(0);
   SDValue Res32 = TLI.makeLibCall(DAG, RTLIB::FPEXT_F16_F32, MidVT, Op,

lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

static SDValue getCopyFromParts(SelectionDAG &DAG, SDLoc DL,
   }
 
   // There is now one part, held in Val.  Correct it to match ValueVT.
   EVT PartEVT = Val.getValueType();
 
   if (PartEVT == ValueVT)
     return Val;
 
+  if (PartEVT.isInteger() && ValueVT.isFloatingPoint()) {
+    // For an FP value in an integer part, we need to truncate to the right
+    // width first.
+    PartEVT = EVT::getIntegerVT(*DAG.getContext(),  ValueVT.getSizeInBits());
+    Val = DAG.getNode(ISD::TRUNCATE, DL, PartEVT, Val);
+  }
+
   if (PartEVT.isInteger() && ValueVT.isInteger()) {
     if (ValueVT.bitsLT(PartEVT)) {
       // For a truncate, see if we have any information to
       // indicate whether the truncated bits will always be
       // zero or sign-extension.
       if (AssertOp != ISD::DELETED_NODE)
         Val = DAG.getNode(AssertOp, DL, PartEVT, Val,
                           DAG.getValueType(ValueVT));
static void getCopyToParts(SelectionDAG &DAG, SDLoc DL,
   }
 
   if (NumParts * PartBits > ValueVT.getSizeInBits()) {
     // If the parts cover more bits than the value has, promote the value.
     if (PartVT.isFloatingPoint() && ValueVT.isFloatingPoint()) {
       assert(NumParts == 1 && "Do not know what to promote to!");
       Val = DAG.getNode(ISD::FP_EXTEND, DL, PartVT, Val);
     } else {
+      if (ValueVT.isFloatingPoint()) {
+        // FP values need to be bitcast, then extended if they are being put
+        // into a larger container.
+        ValueVT = EVT::getIntegerVT(*DAG.getContext(),  ValueVT.getSizeInBits());
+        Val = DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
+      }
       assert((PartVT.isInteger() || PartVT == MVT::x86mmx) &&
              ValueVT.isInteger() &&
              "Unknown mismatch!");
       ValueVT = EVT::getIntegerVT(*DAG.getContext(), NumParts * PartBits);
       Val = DAG.getNode(ExtendKind, DL, ValueVT, Val);
       if (PartVT == MVT::x86mmx)
         Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
     }

lib/CodeGen/TargetLoweringBase.cpp

void TargetLoweringBase::computeRegisterProperties(
   // expand it to i32 and we will be generating soft float library calls.
   if (!isTypeLegal(MVT::f32)) {
     NumRegistersForVT[MVT::f32] = NumRegistersForVT[MVT::i32];
     RegisterTypeForVT[MVT::f32] = RegisterTypeForVT[MVT::i32];
     TransformToType[MVT::f32] = MVT::i32;
     ValueTypeActions.setTypeAction(MVT::f32, TypeSoftenFloat);
   }
 
+  // Decide how to handle f16. If the target does not have native f16 support,
+  // promote it to f32, because there are no f16 library calls (except for
+  // conversions).
   if (!isTypeLegal(MVT::f16)) {
-    // If the target has native f32 support, promote f16 operations to f32.  If
-    // f32 is not supported, generate soft float library calls.
-    if (isTypeLegal(MVT::f32)) {
-      NumRegistersForVT[MVT::f16] = NumRegistersForVT[MVT::f32];
-      RegisterTypeForVT[MVT::f16] = RegisterTypeForVT[MVT::f32];
-      TransformToType[MVT::f16] = MVT::f32;
-      ValueTypeActions.setTypeAction(MVT::f16, TypePromoteFloat);
-    } else {
-      NumRegistersForVT[MVT::f16] = NumRegistersForVT[MVT::i16];
-      RegisterTypeForVT[MVT::f16] = RegisterTypeForVT[MVT::i16];
-      TransformToType[MVT::f16] = MVT::i16;
-      ValueTypeActions.setTypeAction(MVT::f16, TypeSoftenFloat);
-    }
   }
 
   // Loop over all of the vector value types to see which need transformations.
   for (unsigned i = MVT::FIRST_VECTOR_VALUETYPE;
        i <= (unsigned)MVT::LAST_VECTOR_VALUETYPE; ++i) {
     MVT VT = (MVT::SimpleValueType) i;
     if (isTypeLegal(VT))
       continue;

test/CodeGen/ARM/fp16-promote.ll

-; RUN: llc -asm-verbose=false < %s -mattr=+vfp3,+fp16 | FileCheck %s -check-prefix=CHECK-FP16 -check-prefix=CHECK-ALL
-; RUN: llc -asm-verbose=false < %s | FileCheck %s -check-prefix=CHECK-LIBCALL -check-prefix=CHECK-ALL
+; RUN: llc -asm-verbose=false < %s -mattr=+vfp3,+fp16 | FileCheck %s -check-prefix=CHECK-FP16  --check-prefix=CHECK-VFP -check-prefix=CHECK-ALL
+; RUN: llc -asm-verbose=false < %s | FileCheck %s -check-prefix=CHECK-LIBCALL --check-prefix=CHECK-VFP -check-prefix=CHECK-ALL --check-prefix=CHECK-LIBCALL-VFP
+; RUN: llc -asm-verbose=false < %s -mattr=-vfp2 | FileCheck %s --check-prefix=CHECK-LIBCALL -check-prefix=CHECK-NOVFP -check-prefix=CHECK-ALL
 
 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32"
 target triple = "armv7---eabihf"
 
-; CHECK-FP16-LABEL: test_fadd:
+; CHECK-ALL-LABEL: test_fadd:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vadd.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fadd:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vadd.f32
+; CHECK-VFP: vadd.f32
+; CHECK-NOVFP: bl __aeabi_fadd
+; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_fadd(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fadd half %a, %b
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_fsub:
+; CHECK-ALL-LABEL: test_fsub:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vsub.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fsub:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vsub.f32
+; CHECK-VFP: vsub.f32
+; CHECK-NOVFP: bl __aeabi_fsub
+; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_fsub(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fsub half %a, %b
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_fmul:
+; CHECK-ALL-LABEL: test_fmul:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vmul.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fmul
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vmul.f32
+; CHECK-VFP: vmul.f32
+; CHECK-NOVFP: bl __aeabi_fmul
+; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_fmul(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fmul half %a, %b
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_fdiv:
+; CHECK-ALL-LABEL: test_fdiv:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vdiv.f32
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_fdiv
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vdiv.f32
+; CHECK-VFP: vdiv.f32
+; CHECK-NOVFP: bl __aeabi_fdiv
+; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_fdiv(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fdiv half %a, %b
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_frem:
+; CHECK-ALL-LABEL: test_frem:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl fmodf
-; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_frem
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl fmodf
+; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_frem(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = frem half %a, %b
   store half %r, half* %p
   ret void
 }
 
 ; CHECK-ALL-LABEL: test_load_store:
 ; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: ldrh r0, [r0]
-; CHECK-ALL-NEXT: strh r0, [r1]
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL: ldrh {{r[0-9]+}}, [{{r[0-9]+}}]
+; CHECK-ALL: strh {{r[0-9]+}}, [{{r[0-9]+}}]
 define void @test_load_store(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   store half %a, half* %q
   ret void
 }
 
 ; Testing only successfull compilation of function calls.  In ARM ABI, half
 ; args and returns are handled as f32.
define half @test_call(half %a, half %b) #0 {
   %r = call half @test_callee(half %a, half %b)
   ret half %r
 }
 
 ; CHECK-ALL-LABEL: test_call_flipped:
 ; CHECK-ALL-NEXT: .fnstart
 ; CHECK-ALL-NEXT: .save {r11, lr}
 ; CHECK-ALL-NEXT: push {r11, lr}
-; CHECK-ALL-NEXT: vmov.f32 s2, s0
-; CHECK-ALL-NEXT: vmov.f32 s0, s1
-; CHECK-ALL-NEXT: vmov.f32 s1, s2
+; CHECK-VFP-NEXT: vmov.f32 s2, s0
+; CHECK-VFP-NEXT: vmov.f32 s0, s1
+; CHECK-VFP-NEXT: vmov.f32 s1, s2
+; CHECK-NOVFP-NEXT: mov r2, r0
+; CHECK-NOVFP-NEXT: mov r0, r1
+; CHECK-NOVFP-NEXT: mov r1, r2
 ; CHECK-ALL-NEXT: bl test_callee
 ; CHECK-ALL-NEXT: pop {r11, pc}
 define half @test_call_flipped(half %a, half %b) #0 {
   %r = call half @test_callee(half %b, half %a)
   ret half %r
 }
 
 ; CHECK-ALL-LABEL: test_tailcall_flipped:
 ; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: vmov.f32 s2, s0
-; CHECK-ALL-NEXT: vmov.f32 s0, s1
-; CHECK-ALL-NEXT: vmov.f32 s1, s2
+; CHECK-VFP-NEXT: vmov.f32 s2, s0
+; CHECK-VFP-NEXT: vmov.f32 s0, s1
+; CHECK-VFP-NEXT: vmov.f32 s1, s2
+; CHECK-NOVFP-NEXT: mov r2, r0
+; CHECK-NOVFP-NEXT: mov r0, r1
+; CHECK-NOVFP-NEXT: mov r1, r2
 ; CHECK-ALL-NEXT: b test_callee
 define half @test_tailcall_flipped(half %a, half %b) #0 {
   %r = tail call half @test_callee(half %b, half %a)
   ret half %r
 }
 
 ; Optimizer picks %p or %q based on %c and only loads that value
 ; No conversion is needed
 ; CHECK-ALL-LABEL: test_select:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: cmp r2, #0
-; CHECK-ALL-NEXT: movne r1, r0
-; CHECK-ALL-NEXT: ldrh r1, [r1]
-; CHECK-ALL-NEXT: strh r1, [r0]
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL: cmp {{r[0-9]+}}, #0
+; CHECK-NOVFP: movne {{r[0-9]+}}, {{r[0-9]+}}
+; CHECK-VFP: movne {{r[0-9]+}}, {{r[0-9]+}}
+; CHECK-VFP: ldrh {{r[0-9]+}}, [{{r[0-9]+}}]
+; CHECK-NOVFP: ldrh {{r[0-9]+}}, [{{r[0-9]+}}]
+; CHECK-ALL: strh {{r[0-9]+}}, [{{r[0-9]+}}]
 define void @test_select(half* %p, half* %q, i1 zeroext %c) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = select i1 %c, half %a, half %b
   store half %r, half* %p
   ret void
 }
 
 ; Test only two variants of fcmp.  These get translated to f32 vcmpe
 ; instructions anyway.
-; CHECK-FP16-LABEL: test_fcmp_une:
+; CHECK-ALL-LABEL: test_fcmp_une:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcmpe.f32
-; CHECK-FP16: vmrs APSR_nzcv, fpscr
-; CHECK-FP16: movwne
-; CHECK-LIBCALL-LABEL: test_fcmp_une:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcmpe.f32
-; CHECK-LIBCALL: movwne
+; CHECK-VFP: vcmpe.f32
+; CHECK-NOVFP: bl __aeabi_fcmpeq
+; CHECK-FP16: vmrs APSR_nzcv, fpscr
+; CHECK-ALL: movw{{ne|eq}}
 define i1 @test_fcmp_une(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fcmp une half %a, %b
   ret i1 %r
 }
 
-; CHECK-FP16-LABEL: test_fcmp_ueq:
+; CHECK-ALL-LABEL: test_fcmp_ueq:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcmpe.f32
-; CHECK-FP16: vmrs APSR_nzcv, fpscr
-; CHECK-FP16: movweq
-; CHECK-FP16: movwvs
-; CHECK-LIBCALL-LABEL: test_fcmp_ueq:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcmpe.f32
-; CHECK-LIBCALL: movweq
+; CHECK-VFP: vcmpe.f32
+; CHECK-NOVFP: bl __aeabi_fcmpeq
+; CHECK-FP16: vmrs APSR_nzcv, fpscr
+; CHECK-LIBCALL: movw{{ne|eq}}
 define i1 @test_fcmp_ueq(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = fcmp ueq half %a, %b
   ret i1 %r
 }
 
-; CHECK-FP16-LABEL: test_br_cc:
+; CHECK-ALL-LABEL: test_br_cc:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcmpe.f32
-; CHECK-FP16: vmrs APSR_nzcv, fpscr
-; CHECK-FP16: strmi
-; CHECK-FP16: strpl
-; CHECK-LIBCALL-LABEL: test_br_cc:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcmpe.f32
-; CHECK-LIBCALL: strmi
-; CHECK-LIBCALL: strpl
+; CHECK-VFP: vcmpe.f32
+; CHECK-NOVFP: bl __aeabi_fcmplt
+; CHECK-FP16: vmrs APSR_nzcv, fpscr
+; CHECK-VFP: strmi
+; CHECK-VFP: strpl
+; CHECK-NOVFP: strne
+; CHECK-NOVFP: streq
 define void @test_br_cc(half* %p, half* %q, i32* %p1, i32* %p2) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %c = fcmp uge half %a, %b
   br i1 %c, label %then, label %else
 then:
   store i32 0, i32* %p1
   ret void
 else:
   store i32 0, i32* %p2
   ret void
 }
 
 declare i1 @test_dummy(half* %p) #0
-; CHECK-FP16-LABEL: test_phi:
+; CHECK-ALL-LABEL: test_phi:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: [[LOOP:.LBB[1-9_]+]]:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: bl      test_dummy
 ; CHECK-FP16: bne     [[LOOP]]
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_phi:
-; CHECK-LIBCALL: bl __aeabi_h2f
+; CHECK-LIBCALL-VFP: bl __aeabi_h2f
 ; CHECK-LIBCALL: [[LOOP:.LBB[1-9_]+]]:
-; CHECK-LIBCALL: bl __aeabi_h2f
+; CHECK-LIBCALL-VFP: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl test_dummy
 ; CHECK-LIBCALL: bne     [[LOOP]]
-; CHECK-LIBCALL: bl __aeabi_f2h
+; CHECK-LIBCALL-VFP: bl __aeabi_f2h
 define void @test_phi(half* %p) #0 {
 entry:
   %a = load half, half* %p
   br label %loop
 loop:
   %r = phi half [%a, %entry], [%b, %loop]
   %b = load half, half* %p
   %c = call i1 @test_dummy(half* %p)
   br i1 %c, label %loop, label %return
 return:
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_fptosi_i32:
+; CHECK-ALL-LABEL: test_fptosi_i32:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvt.s32.f32
-; CHECK-LIBCALL-LABEL: test_fptosi_i32:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcvt.s32.f32
+; CHECK-VFP: vcvt.s32.f32
+; CHECK-NOVFP: bl __aeabi_f2iz
 define i32 @test_fptosi_i32(half* %p) #0 {
   %a = load half, half* %p, align 2
   %r = fptosi half %a to i32
   ret i32 %r
 }
 
-; CHECK-FP16-LABEL: test_fptosi_i64:
+; CHECK-ALL-LABEL: test_fptosi_i64:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl __aeabi_f2lz
-; CHECK-LIBCALL-LABEL: test_fptosi_i64:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_f2lz
+; CHECK-ALL: bl __aeabi_f2lz
 define i64 @test_fptosi_i64(half* %p) #0 {
   %a = load half, half* %p, align 2
   %r = fptosi half %a to i64
   ret i64 %r
 }
 
-; CHECK-FP16-LABEL: test_fptoui_i32:
+; CHECK-ALL-LABEL: test_fptoui_i32:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvt.u32.f32
-; CHECK-LIBCALL-LABEL: test_fptoui_i32:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcvt.u32.f32
+; CHECK-VFP: vcvt.u32.f32
+; CHECK-NOVFP: bl __aeabi_f2uiz
 define i32 @test_fptoui_i32(half* %p) #0 {
   %a = load half, half* %p, align 2
   %r = fptoui half %a to i32
   ret i32 %r
 }
 
-; CHECK-FP16-LABEL: test_fptoui_i64:
+; CHECK-ALL-LABEL: test_fptoui_i64:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: bl __aeabi_f2ulz
-; CHECK-LIBCALL-LABEL: test_fptoui_i64:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: bl __aeabi_f2ulz
+; CHECK-ALL: bl __aeabi_f2ulz
 define i64 @test_fptoui_i64(half* %p) #0 {
   %a = load half, half* %p, align 2
   %r = fptoui half %a to i64
   ret i64 %r
 }
 
-; CHECK-FP16-LABEL: test_sitofp_i32:
-; CHECK-FP16: vcvt.f32.s32
+; CHECK-ALL-LABEL: test_sitofp_i32:
+; CHECK-VFP: vcvt.f32.s32
+; CHECK-NOVFP: bl __aeabi_i2f
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_sitofp_i32:
-; CHECK-LIBCALL: vcvt.f32.s32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_sitofp_i32(i32 %a, half* %p) #0 {
   %r = sitofp i32 %a to half
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_uitofp_i32:
-; CHECK-FP16: vcvt.f32.u32
+; CHECK-ALL-LABEL: test_uitofp_i32:
+; CHECK-VFP: vcvt.f32.u32
+; CHECK-NOVFP: bl __aeabi_ui2f
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_uitofp_i32:
-; CHECK-LIBCALL: vcvt.f32.u32
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_uitofp_i32(i32 %a, half* %p) #0 {
   %r = uitofp i32 %a to half
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_sitofp_i64:
-; CHECK-FP16: bl __aeabi_l2f
+; CHECK-ALL-LABEL: test_sitofp_i64:
+; CHECK-ALL: bl __aeabi_l2f
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_sitofp_i64:
-; CHECK-LIBCALL: bl __aeabi_l2f
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_sitofp_i64(i64 %a, half* %p) #0 {
   %r = sitofp i64 %a to half
   store half %r, half* %p
   ret void
 }
 
-; CHECK-FP16-LABEL: test_uitofp_i64:
-; CHECK-FP16: bl __aeabi_ul2f
+; CHECK-ALL-LABEL: test_uitofp_i64:
+; CHECK-ALL: bl __aeabi_ul2f
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_uitofp_i64:
-; CHECK-LIBCALL: bl __aeabi_ul2f
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_uitofp_i64(i64 %a, half* %p) #0 {
   %r = uitofp i64 %a to half
   store half %r, half* %p
   ret void
 }
 
 ; CHECK-FP16-LABEL: test_fptrunc_float:

 define float @test_fpextend_float(half* %p) {
   %a = load half, half* %p, align 2
   %r = fpext half %a to float
   ret float %r
 }
 
 ; CHECK-FP16-LABEL: test_fpextend_double:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-FP16: vcvt.f64.f32
 ; CHECK-LIBCALL-LABEL: test_fpextend_double:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vcvt.f64.f32
+; CHECK-VFP: vcvt.f64.f32
+; CHECK-NOVFP: bl __aeabi_f2d
 define double @test_fpextend_double(half* %p) {
   %a = load half, half* %p, align 2
   %r = fpext half %a to double
   ret double %r
 }
 
 ; CHECK-ALL-LABEL: test_bitcast_halftoi16:
 ; CHECK-ALL-NEXT: .fnstart

 declare half @llvm.floor.f16(half %a) #0
 declare half @llvm.ceil.f16(half %a) #0
 declare half @llvm.trunc.f16(half %a) #0
 declare half @llvm.rint.f16(half %a) #0
 declare half @llvm.nearbyint.f16(half %a) #0
 declare half @llvm.round.f16(half %a) #0
 declare half @llvm.fmuladd.f16(half %a, half %b, half %c) #0
 
-; CHECK-FP16-LABEL: test_sqrt:
+; CHECK-ALL-LABEL: test_sqrt:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vsqrt.f32
 ; CHECK-FP16: vcvtb.f16.f32
-; CHECK-LIBCALL-LABEL: test_sqrt:
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vsqrt.f32
+; CHECK-VFP-LIBCALL: vsqrt.f32
+; CHECK-NOVFP: bl sqrtf
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_sqrt(half* %p) #0 {
   %a = load half, half* %p, align 2
   %r = call half @llvm.sqrt.f16(half %a)
   store half %r, half* %p
   ret void
 }
 

 ; CHECK-FP16-LABEL: test_copysign:
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vbsl
 ; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL-LABEL: test_copysign:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vbsl
+; CHECK-VFP-LIBCALL: vbsl
+; CHECK-NOVFP: bfc
+; CHECK-NOVFP: and
+; CHECK-NOVFP: orr
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_copysign(half* %p, half* %q) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %r = call half @llvm.copysign.f16(half %a, half %b)
   store half %r, half* %p
   ret void
 }

 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vcvtb.f32.f16
 ; CHECK-FP16: vmla.f32
 ; CHECK-FP16: vcvtb.f16.f32
 ; CHECK-LIBCALL-LABEL: test_fmuladd:
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
 ; CHECK-LIBCALL: bl __aeabi_h2f
-; CHECK-LIBCALL: vmla.f32
+; CHECK-VFP-LIBCALL: vmla.f32
+; CHECK-NOVFP: bl __aeabi_fmul
 ; CHECK-LIBCALL: bl __aeabi_f2h
 define void @test_fmuladd(half* %p, half* %q, half* %r) #0 {
   %a = load half, half* %p, align 2
   %b = load half, half* %q, align 2
   %c = load half, half* %r, align 2
   %v = call half @llvm.fmuladd.f16(half %a, half %b, half %c)
   store half %v, half* %p
   ret void
 }
 
 ; f16 vectors are not legal in the backend.  Vector elements are not assigned
 ; to the register, but are stored in the stack instead.  Hence insertelement
 ; and extractelement have these extra loads and stores.
 
 ; CHECK-ALL-LABEL: test_insertelement:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: .pad #8
-; CHECK-ALL-NEXT: sub sp, sp, #8
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: mov
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: add
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: add sp, sp, #8
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL: sub sp, sp, #8
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: mov
+; CHECK-ALL-DAG: ldrh
+; CHECK-ALL-DAG: add
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
+; CHECK-ALL: add sp, sp, #8
 define void @test_insertelement(half* %p, <4 x half>* %q, i32 %i) #0 {
   %a = load half, half* %p, align 2
   %b = load <4 x half>, <4 x half>* %q, align 8
   %c = insertelement <4 x half> %b, half %a, i32 %i
   store <4 x half> %c, <4 x half>* %q
   ret void
 }
 
 ; CHECK-ALL-LABEL: test_extractelement:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: .pad #8
-; CHECK-ALL-NEXT: sub sp, sp, #8
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: orr
-; CHECK-ALL-NEXT: str
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: orr
-; CHECK-ALL-NEXT: str
-; CHECK-ALL-NEXT: mov
-; CHECK-ALL-NEXT: add
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: add sp, sp, #8
-; CHECK-ALL-NEXT: bx lr
+; CHECK-VFP: sub sp, sp, #8
+; CHECK-VFP: ldrh
+; CHECK-VFP: ldrh
+; CHECK-VFP: orr
+; CHECK-VFP: str
+; CHECK-VFP: ldrh
+; CHECK-VFP: ldrh
+; CHECK-VFP: orr
+; CHECK-VFP: str
+; CHECK-VFP: mov
+; CHECK-VFP: add
+; CHECK-VFP: ldrh
+; CHECK-VFP: strh
+; CHECK-VFP: add sp, sp, #8
+; CHECK-VFP: bx lr
+; CHECK-NOVFP: ldrh
+; CHECK-NOVFP: strh
+; CHECK-NOVFP: ldrh
+; CHECK-NOVFP: strh
+; CHECK-NOVFP: ldrh
+; CHECK-NOVFP: strh
+; CHECK-NOVFP: ldrh
+; CHECK-NOVFP: strh
+; CHECK-NOVFP: ldrh
 define void @test_extractelement(half* %p, <4 x half>* %q, i32 %i) #0 {
   %a = load <4 x half>, <4 x half>* %q, align 8
   %b = extractelement <4 x half> %a, i32 %i
   store half %b, half* %p
   ret void
 }
 
 ; test struct operations
 
 %struct.dummy = type { i32, half }
 
 ; CHECK-ALL-LABEL: test_insertvalue:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: ldr
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: str
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL-DAG: ldr
+; CHECK-ALL-DAG: ldrh
+; CHECK-ALL-DAG: strh
+; CHECK-ALL-DAG: str
 define void @test_insertvalue(%struct.dummy* %p, half* %q) {
   %a = load %struct.dummy, %struct.dummy* %p
   %b = load half, half* %q
   %c = insertvalue %struct.dummy %a, half %b, 1
   store %struct.dummy %c, %struct.dummy* %p
   ret void
 }
 
 ; CHECK-ALL-LABEL: test_extractvalue:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: ldrh
-; CHECK-ALL-NEXT: strh
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL: .fnstart
+; CHECK-ALL: ldrh
+; CHECK-ALL: strh
 define void @test_extractvalue(%struct.dummy* %p, half* %q) {
   %a = load %struct.dummy, %struct.dummy* %p
   %b = extractvalue %struct.dummy %a, 1
   store half %b, half* %q
   ret void
 }
 
-; CHECK-FP16-LABEL: test_struct_return:
+; CHECK-ALL-LABEL: test_struct_return:
 ; CHECK-FP16: vcvtb.f32.f16
-; CHECK-LIBCALL-LABEL: test_struct_return:
-; CHECK-LIBCALL: bl __aeabi_h2f
+; CHECK-VFP-LIBCALL: bl __aeabi_h2f
+; CHECK-NOVFP-DAG: ldr
+; CHECK-NOVFP-DAG: ldrh
 define %struct.dummy @test_struct_return(%struct.dummy* %p) {
   %a = load %struct.dummy, %struct.dummy* %p
   ret %struct.dummy %a
 }
 
 ; CHECK-ALL-LABEL: test_struct_arg:
-; CHECK-ALL-NEXT: .fnstart
-; CHECK-ALL-NEXT: bx lr
+; CHECK-ALL: bx lr
 define half @test_struct_arg(%struct.dummy %p) {
   %a = extractvalue %struct.dummy %p, 1
   ret half %a
 }
 
 attributes #0 = { nounwind }