[AArch64] allow v8f16 types when FullFP16 is supported

[SjoerdMeijer]

Summary

This adds support for allowing v8f16 vector types, thus avoiding conversions
from/to single precision for these types. This is a follow up patch of
commits r311154 and r312104, which added support for scalars and v4f16
types, respectively.

[samparker]

LGTM. Luckily the +fullfp16 output is less unwiedly....

lib/Target/AArch64/AArch64ISelLowering.cpp

AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::FCOPYSIGN, MVT::f32, Custom);
   if (Subtarget->hasFullFP16())
     setOperationAction(ISD::FCOPYSIGN, MVT::f16, Custom);
   else
     setOperationAction(ISD::FCOPYSIGN, MVT::f16, Promote);
 
   setOperationAction(ISD::FREM,    MVT::f16,   Promote);
   setOperationAction(ISD::FREM,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FREM,    MVT::v8f16, Promote);
   setOperationAction(ISD::FPOW,    MVT::f16,   Promote);
   setOperationAction(ISD::FPOW,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FPOW,    MVT::v8f16, Promote);
   setOperationAction(ISD::FPOWI,   MVT::f16,   Promote);
   setOperationAction(ISD::FCOS,    MVT::f16,   Promote);
   setOperationAction(ISD::FCOS,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FCOS,    MVT::v8f16, Promote);
   setOperationAction(ISD::FSIN,    MVT::f16,   Promote);
   setOperationAction(ISD::FSIN,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FSIN,    MVT::v8f16, Promote);
   setOperationAction(ISD::FSINCOS, MVT::f16,   Promote);
   setOperationAction(ISD::FSINCOS, MVT::v4f16, Promote);
+  setOperationAction(ISD::FSINCOS, MVT::v8f16, Promote);
   setOperationAction(ISD::FEXP,    MVT::f16,   Promote);
   setOperationAction(ISD::FEXP,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FEXP,    MVT::v8f16, Promote);
   setOperationAction(ISD::FEXP2,   MVT::f16,   Promote);
   setOperationAction(ISD::FEXP2,   MVT::v4f16, Promote);
+  setOperationAction(ISD::FEXP2,   MVT::v8f16, Promote);
   setOperationAction(ISD::FLOG,    MVT::f16,   Promote);
   setOperationAction(ISD::FLOG,    MVT::v4f16, Promote);
+  setOperationAction(ISD::FLOG,    MVT::v8f16, Promote);
   setOperationAction(ISD::FLOG2,   MVT::f16,   Promote);
   setOperationAction(ISD::FLOG2,   MVT::v4f16, Promote);
+  setOperationAction(ISD::FLOG2,   MVT::v8f16, Promote);
   setOperationAction(ISD::FLOG10,  MVT::f16,   Promote);
   setOperationAction(ISD::FLOG10,  MVT::v4f16, Promote);
+  setOperationAction(ISD::FLOG10,  MVT::v8f16, Promote);
 
   if (!Subtarget->hasFullFP16()) {
     setOperationAction(ISD::SELECT,      MVT::f16,  Promote);
     setOperationAction(ISD::SELECT_CC,   MVT::f16,  Promote);
     setOperationAction(ISD::SETCC,       MVT::f16,  Promote);
     setOperationAction(ISD::BR_CC,       MVT::f16,  Promote);
     setOperationAction(ISD::FADD,        MVT::f16,  Promote);
     setOperationAction(ISD::FSUB,        MVT::f16,  Promote);
if (!Subtarget->hasFullFP16()) {
     setOperationAction(ISD::SELECT_CC,   MVT::v4f16, Expand);
     setOperationAction(ISD::FTRUNC,      MVT::v4f16, Expand);
     setOperationAction(ISD::FCOPYSIGN,   MVT::v4f16, Expand);
     setOperationAction(ISD::FFLOOR,      MVT::v4f16, Expand);
     setOperationAction(ISD::FCEIL,       MVT::v4f16, Expand);
     setOperationAction(ISD::FRINT,       MVT::v4f16, Expand);
     setOperationAction(ISD::FNEARBYINT,  MVT::v4f16, Expand);
     setOperationAction(ISD::FSQRT,       MVT::v4f16, Expand);
-  }
-
-  
-  // v8f16 is also a storage-only type, so expand it.
-  setOperationAction(ISD::FABS, MVT::v8f16, Expand);
-  setOperationAction(ISD::FADD, MVT::v8f16, Expand);
-  setOperationAction(ISD::FCEIL, MVT::v8f16, Expand);
-  setOperationAction(ISD::FCOPYSIGN, MVT::v8f16, Expand);
-  setOperationAction(ISD::FCOS, MVT::v8f16, Expand);
-  setOperationAction(ISD::FDIV, MVT::v8f16, Expand);
-  setOperationAction(ISD::FFLOOR, MVT::v8f16, Expand);
-  setOperationAction(ISD::FMA, MVT::v8f16, Expand);
-  setOperationAction(ISD::FMUL, MVT::v8f16, Expand);
-  setOperationAction(ISD::FNEARBYINT, MVT::v8f16, Expand);
-  setOperationAction(ISD::FNEG, MVT::v8f16, Expand);
-  setOperationAction(ISD::FPOW, MVT::v8f16, Expand);
-  setOperationAction(ISD::FREM, MVT::v8f16, Expand);
-  setOperationAction(ISD::FROUND, MVT::v8f16, Expand);
-  setOperationAction(ISD::FRINT, MVT::v8f16, Expand);
-  setOperationAction(ISD::FSIN, MVT::v8f16, Expand);
-  setOperationAction(ISD::FSINCOS, MVT::v8f16, Expand);
-  setOperationAction(ISD::FSQRT, MVT::v8f16, Expand);
-  setOperationAction(ISD::FSUB, MVT::v8f16, Expand);
-  setOperationAction(ISD::FTRUNC, MVT::v8f16, Expand);
-  setOperationAction(ISD::SETCC, MVT::v8f16, Expand);
-  setOperationAction(ISD::BR_CC, MVT::v8f16, Expand);
-  setOperationAction(ISD::SELECT, MVT::v8f16, Expand);
-  setOperationAction(ISD::SELECT_CC, MVT::v8f16, Expand);
-  setOperationAction(ISD::FP_EXTEND, MVT::v8f16, Expand);
-  setOperationAction(ISD::FEXP, MVT::v8f16, Expand);
-  setOperationAction(ISD::FEXP2, MVT::v8f16, Expand);
-  setOperationAction(ISD::FLOG, MVT::v8f16, Expand);
-  setOperationAction(ISD::FLOG2, MVT::v8f16, Expand);
-  setOperationAction(ISD::FLOG10, MVT::v8f16, Expand);
+    setOperationAction(ISD::FABS,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FADD,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FCEIL,       MVT::v8f16, Expand);
+    setOperationAction(ISD::FCOPYSIGN,   MVT::v8f16, Expand);
+    setOperationAction(ISD::FDIV,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FFLOOR,      MVT::v8f16, Expand);
+    setOperationAction(ISD::FMA,         MVT::v8f16, Expand);
+    setOperationAction(ISD::FMUL,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FNEARBYINT,  MVT::v8f16, Expand);
+    setOperationAction(ISD::FNEG,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FROUND,      MVT::v8f16, Expand);
+    setOperationAction(ISD::FRINT,       MVT::v8f16, Expand);
+    setOperationAction(ISD::FSQRT,       MVT::v8f16, Expand);
+    setOperationAction(ISD::FSUB,        MVT::v8f16, Expand);
+    setOperationAction(ISD::FTRUNC,      MVT::v8f16, Expand);
+    setOperationAction(ISD::SETCC,       MVT::v8f16, Expand);
+    setOperationAction(ISD::BR_CC,       MVT::v8f16, Expand);
+    setOperationAction(ISD::SELECT,      MVT::v8f16, Expand);
+    setOperationAction(ISD::SELECT_CC,   MVT::v8f16, Expand);
+    setOperationAction(ISD::FP_EXTEND,   MVT::v8f16, Expand);
+  }
 
   // AArch64 has implementations of a lot of rounding-like FP operations.
   for (MVT Ty : {MVT::f32, MVT::f64}) {
     setOperationAction(ISD::FFLOOR, Ty, Legal);
     setOperationAction(ISD::FNEARBYINT, Ty, Legal);
     setOperationAction(ISD::FCEIL, Ty, Legal);
     setOperationAction(ISD::FRINT, Ty, Legal);
     setOperationAction(ISD::FTRUNC, Ty, Legal);

test/CodeGen/AArch64/arm64-vfloatintrinsics.ll

 ; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple -mattr=-fullfp16 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple -mattr=+fullfp16 | FileCheck %s --check-prefix=CHECK-FP16
 
 ;;; Float vectors
 
 %v2f32 = type <2 x float>
 %v4f16 = type <4 x half>
+%v8f16 = type <8 x half>
 
 ; CHECK-LABEL: test_v2f32.sqrt:
 define %v2f32 @test_v2f32.sqrt(%v2f32 %a) {
   ; CHECK: fsqrt.2s
   %1 = call %v2f32 @llvm.sqrt.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.sqrt(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.sqrt:
 ; CHECK:       fsqrt s{{.}}, s{{.}}
 ; CHECK:       fsqrt s{{.}}, s{{.}}
 ; CHECK:       fsqrt s{{.}}, s{{.}}
 ; CHECK:       fsqrt s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.sqrt:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       fsqrt.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.sqrt.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.sqrt(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 fsqrt tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.sqrt:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       fsqrt.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.sqrt.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK: test_v2f32.powi:
 define %v2f32 @test_v2f32.powi(%v2f32 %a, i32 %b) {
   ; CHECK: pow
   %1 = call %v2f32 @llvm.powi.v2f32(%v2f32 %a, i32 %b)
   ret %v2f32 %1
 }
 ; CHECK: test_v2f32.sin:
 define %v2f32 @test_v2f32.sin(%v2f32 %a) {

 define %v4f16 @test_v4f16.fma(%v4f16 %a, %v4f16 %b, %v4f16 %c) {
 ; CHECK-LABEL: test_v4f16.fma:
 ; CHECK:       fmadd s{{.}}, s{{.}}, s{{.}}, s{{.}}
 ; CHECK:       fmadd s{{.}}, s{{.}}, s{{.}}, s{{.}}
 ; CHECK:       fmadd s{{.}}, s{{.}}, s{{.}}, s{{.}}
 ; CHECK:       fmadd s{{.}}, s{{.}}, s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.fma:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       fmla.4h
   %1 = call %v4f16 @llvm.fma.v4f16(%v4f16 %a, %v4f16 %b, %v4f16 %c)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.fma(%v8f16 %a, %v8f16 %b, %v8f16 %c) {
+; Filechecks are unwieldy with 16 fcvt and 8 fma tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.fma:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       fmla.8h
+  %1 = call %v8f16 @llvm.fma.v8f16(%v8f16 %a, %v8f16 %b, %v8f16 %c)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.fabs:
 define %v2f32 @test_v2f32.fabs(%v2f32 %a) {
   ; CHECK: fabs.2s
   %1 = call %v2f32 @llvm.fabs.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.fabs(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.fabs:
 ; CHECK:       fabs s{{.}}, s{{.}}
 ; CHECK:       fabs s{{.}}, s{{.}}
 ; CHECK:       fabs s{{.}}, s{{.}}
 ; CHECK:       fabs s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.fabs:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       fabs.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.fabs.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.fabs(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 fabs tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.fabs:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       fabs.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.fabs.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.floor:
 define %v2f32 @test_v2f32.floor(%v2f32 %a) {
   ; CHECK: frintm.2s
   %1 = call %v2f32 @llvm.floor.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.floor(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.floor:
 ; CHECK:       frintm s{{.}}, s{{.}}
 ; CHECK:       frintm s{{.}}, s{{.}}
 ; CHECK:       frintm s{{.}}, s{{.}}
 ; CHECK:       frintm s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.floor:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       frintm.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.floor.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.floor(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 frintm tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.floor:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       frintm.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.floor.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.ceil:
 define %v2f32 @test_v2f32.ceil(%v2f32 %a) {
   ; CHECK: frintp.2s
   %1 = call %v2f32 @llvm.ceil.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.ceil(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.ceil:
 ; CHECK:       frintp s{{.}}, s{{.}}
 ; CHECK:       frintp s{{.}}, s{{.}}
 ; CHECK:       frintp s{{.}}, s{{.}}
 ; CHECK:       frintp s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.ceil:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       frintp.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.ceil.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.ceil(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 frint tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.ceil:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       frintp.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.ceil.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.trunc:
 define %v2f32 @test_v2f32.trunc(%v2f32 %a) {
   ; CHECK: frintz.2s
   %1 = call %v2f32 @llvm.trunc.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.trunc(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.trunc:
 ; CHECK:       frintz s{{.}}, s{{.}}
 ; CHECK:       frintz s{{.}}, s{{.}}
 ; CHECK:       frintz s{{.}}, s{{.}}
 ; CHECK:       frintz s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.trunc:
 ; CHECK-FP16:       frintz.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.trunc.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.trunc(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 frint tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.trunc:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       frintz.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.trunc.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.rint:
 define %v2f32 @test_v2f32.rint(%v2f32 %a) {
   ; CHECK: frintx.2s
   %1 = call %v2f32 @llvm.rint.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.rint(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.rint:
 ; CHECK:       frintx s{{.}}, s{{.}}
 ; CHECK:       frintx s{{.}}, s{{.}}
 ; CHECK:       frintx s{{.}}, s{{.}}
 ; CHECK:       frintx s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.rint:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       frintx.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.rint.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.rint(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 frint tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.rint:
+; CHECK-FP16:       frintx.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.rint.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 ; CHECK-LABEL: test_v2f32.nearbyint:
 define %v2f32 @test_v2f32.nearbyint(%v2f32 %a) {
   ; CHECK: frinti.2s
   %1 = call %v2f32 @llvm.nearbyint.v2f32(%v2f32 %a)
   ret %v2f32 %1
 }
 define %v4f16 @test_v4f16.nearbyint(%v4f16 %a) {
 ; CHECK-LABEL: test_v4f16.nearbyint:
 ; CHECK:       frinti s{{.}}, s{{.}}
 ; CHECK:       frinti s{{.}}, s{{.}}
 ; CHECK:       frinti s{{.}}, s{{.}}
 ; CHECK:       frinti s{{.}}, s{{.}}
 
 ; CHECK-FP16-LABEL: test_v4f16.nearbyint:
+; CHECK-FP16-NOT:   fcvt
 ; CHECK-FP16:       frinti.4h
 ; CHECK-FP16-NEXT:  ret
   %1 = call %v4f16 @llvm.nearbyint.v4f16(%v4f16 %a)
   ret %v4f16 %1
 }
+define %v8f16 @test_v8f16.nearbyint(%v8f16 %a) {
+; Filechecks are unwieldy with 16 fcvt and 8 frint tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_v8f16.nearbyint:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16:       frinti.8h
+; CHECK-FP16-NEXT:  ret
+  %1 = call %v8f16 @llvm.nearbyint.v8f16(%v8f16 %a)
+  ret %v8f16 %1
+}
 
 declare %v2f32 @llvm.sqrt.v2f32(%v2f32) #0
 declare %v4f16 @llvm.sqrt.v4f16(%v4f16) #0
+declare %v8f16 @llvm.sqrt.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.powi.v2f32(%v2f32, i32) #0
 declare %v2f32 @llvm.sin.v2f32(%v2f32) #0
 declare %v2f32 @llvm.cos.v2f32(%v2f32) #0
 declare %v2f32 @llvm.pow.v2f32(%v2f32, %v2f32) #0
 declare %v2f32 @llvm.exp.v2f32(%v2f32) #0
 declare %v2f32 @llvm.exp2.v2f32(%v2f32) #0
 declare %v2f32 @llvm.log.v2f32(%v2f32) #0
 declare %v2f32 @llvm.log10.v2f32(%v2f32) #0
 declare %v2f32 @llvm.log2.v2f32(%v2f32) #0
 
 declare %v2f32 @llvm.fma.v2f32(%v2f32, %v2f32, %v2f32) #0
 declare %v4f16 @llvm.fma.v4f16(%v4f16, %v4f16, %v4f16) #0
+declare %v8f16 @llvm.fma.v8f16(%v8f16, %v8f16, %v8f16) #0
 
 declare %v2f32 @llvm.fabs.v2f32(%v2f32) #0
 declare %v4f16 @llvm.fabs.v4f16(%v4f16) #0
+declare %v8f16 @llvm.fabs.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.floor.v2f32(%v2f32) #0
 declare %v4f16 @llvm.floor.v4f16(%v4f16) #0
+declare %v8f16 @llvm.floor.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.ceil.v2f32(%v2f32) #0
 declare %v4f16 @llvm.ceil.v4f16(%v4f16) #0
+declare %v8f16 @llvm.ceil.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.trunc.v2f32(%v2f32) #0
 declare %v4f16 @llvm.trunc.v4f16(%v4f16) #0
+declare %v8f16 @llvm.trunc.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.rint.v2f32(%v2f32) #0
 declare %v4f16 @llvm.rint.v4f16(%v4f16) #0
+declare %v8f16 @llvm.rint.v8f16(%v8f16) #0
 
 declare %v2f32 @llvm.nearbyint.v2f32(%v2f32) #0
 declare %v4f16 @llvm.nearbyint.v4f16(%v4f16) #0
+declare %v8f16 @llvm.nearbyint.v8f16(%v8f16) #0
 
 ;;;
 
 %v4f32 = type <4 x float>
 ; CHECK: test_v4f32.sqrt:
 define %v4f32 @test_v4f32.sqrt(%v4f32 %a) {
   ; CHECK: fsqrt.4s
   %1 = call %v4f32 @llvm.sqrt.v4f32(%v4f32 %a)

test/CodeGen/AArch64/fp16-v8-instructions.ll

-; RUN: llc < %s -asm-verbose=false -mtriple=aarch64-none-eabi | FileCheck %s
+; RUN: llc < %s -asm-verbose=false -mtriple=aarch64-none-eabi -mattr=-fullfp16 | FileCheck %s --check-prefix=CHECK-CVT   --check-prefix=CHECK
+; RUN: llc < %s -asm-verbose=false -mtriple=aarch64-none-eabi -mattr=+fullfp16 | FileCheck %s --check-prefix=CHECK-FP16  --check-prefix=CHECK
 
 define <8 x half> @add_h(<8 x half> %a, <8 x half> %b) {
 entry:
-; CHECK-LABEL: add_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fadd
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK: fcvt
+; CHECK-CVT-LABEL: add_h:
+; CHECK-CVT:     fcvt
+; CHECK-CVT:     fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fadd
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT-DAG: fcvt
+; CHECK-CVT:     fcvt
+
+; CHECK-FP16-LABEL: add_h:
+; CHECK-FP16:       fadd  v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:  ret
+
   %0 = fadd <8 x half> %a, %b
   ret <8 x half> %0
 }
 
 
 define <8 x half> @sub_h(<8 x half> %a, <8 x half> %b) {
 entry:
-; CHECK-LABEL: sub_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fsub
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK: fcvt
+; CHECK-CVT-LABEL: sub_h:
+; CHECK-CVT:       fcvt
+; CHECK-CVT:       fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fsub
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT:       fcvt
+
+; CHECK-FP16-LABEL: sub_h:
+; CHECK-FP16:       fsub  v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:  ret
+
   %0 = fsub <8 x half> %a, %b
   ret <8 x half> %0
 }
 
 
 define <8 x half> @mul_h(<8 x half> %a, <8 x half> %b) {
 entry:
-; CHECK-LABEL: mul_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fmul
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK: fcvt
+; CHECK-CVT-LABEL: mul_h:
+; CHECK-CVT:       fcvt
+; CHECK-CVT:       fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fmul
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT:       fcvt
+
+; CHECK-FP16-LABEL: mul_h:
+; CHECK-FP16:       fmul  v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:  ret
+
   %0 = fmul <8 x half> %a, %b
   ret <8 x half> %0
 }
 
 
 define <8 x half> @div_h(<8 x half> %a, <8 x half> %b) {
 entry:
-; CHECK-LABEL: div_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fdiv
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK: fcvt
+; CHECK-CVT-LABEL: div_h:
+; CHECK-CVT:       fcvt
+; CHECK-CVT:       fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fdiv
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT-DAG:   fcvt
+; CHECK-CVT:       fcvt
+
+; CHECK-FP16-LABEL: div_h:
+; CHECK-FP16:       fdiv  v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:  ret
+
   %0 = fdiv <8 x half> %a, %b
   ret <8 x half> %0
 }
 
 
 define <8 x half> @load_h(<8 x half>* %a) {
 entry:
 ; CHECK-LABEL: load_h:

 ; CHECK-DAG: xtn     [[I16:v[0-9]+]].4h, [[LOF32]]
 ; CHECK-DAG: fcvtzu  [[HIF32:v[0-9]+\.4s]], [[HI]]
 ; CHECK-NEXT: xtn2    [[I16]].8h, [[HIF32]]
 ; CHECK-NEXT: ret
   %1 = fptoui<8 x half> %a to <8 x i16>
   ret <8 x i16> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_une(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_une:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp une <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 16 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ueq(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ueq:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ueq <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ugt(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ugt:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ugt <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_uge(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_uge:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp uge <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ult(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ult:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ult <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ule(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ule:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ule <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_uno(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_uno:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp uno <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_one(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_one:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp one <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_oeq(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_oeq:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp oeq <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ogt(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ogt:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ogt <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_oge(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_oge:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp oge <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_olt(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_olt:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp olt <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ole(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ole:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ole <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
-; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests.  Skipped.
 define <8 x i1> @test_fcmp_ord(<8 x half> %a, <8 x half> %b) #0 {
+; FileCheck checks are unwieldy with 16 fcvt and 8 csel tests, so skipped for -fullfp16.
+
+; CHECK-FP16-LABEL: test_fcmp_ord:
+; CHECK-FP16-NOT:   fcvt
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+; CHECK-FP16-DAG:   fcmp  h{{[0-9]}}, h{{[0-9]}}
+
   %1 = fcmp ord <8 x half> %a, %b
   ret <8 x i1> %1
 }
 
 attributes #0 = { nounwind }