[llvm] r312104 - [AArch64] allow v4f16 types when FullFP16 is supported
Sjoerd Meijer via llvm-commits
llvm-commits at lists.llvm.org
Wed Aug 30 01:38:13 PDT 2017
Author: sjoerdmeijer
Date: Wed Aug 30 01:38:13 2017
New Revision: 312104
URL: http://llvm.org/viewvc/llvm-project?rev=312104&view=rev
Log:
[AArch64] allow v4f16 types when FullFP16 is supported
Support for scalars was committed in r311154, this adds support for allowing
v4f16 vector types (thus avoiding conversions from/to single precision for
these types).
Differential Revision: https://reviews.llvm.org/D37145
Modified:
llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp
llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll
llvm/trunk/test/CodeGen/AArch64/fp16-v4-instructions.ll
Modified: llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp?rev=312104&r1=312103&r2=312104&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp Wed Aug 30 01:38:13 2017
@@ -326,17 +326,27 @@ AArch64TargetLowering::AArch64TargetLowe
else
setOperationAction(ISD::FCOPYSIGN, MVT::f16, Promote);
- setOperationAction(ISD::FREM, MVT::f16, Promote);
- setOperationAction(ISD::FPOW, MVT::f16, Promote);
- setOperationAction(ISD::FPOWI, MVT::f16, Promote);
- setOperationAction(ISD::FCOS, MVT::f16, Promote);
- setOperationAction(ISD::FSIN, MVT::f16, Promote);
- setOperationAction(ISD::FSINCOS, MVT::f16, Promote);
- setOperationAction(ISD::FEXP, MVT::f16, Promote);
- setOperationAction(ISD::FEXP2, MVT::f16, Promote);
- setOperationAction(ISD::FLOG, MVT::f16, Promote);
- setOperationAction(ISD::FLOG2, MVT::f16, Promote);
- setOperationAction(ISD::FLOG10, MVT::f16, Promote);
+ setOperationAction(ISD::FREM, MVT::f16, Promote);
+ setOperationAction(ISD::FREM, MVT::v4f16, Promote);
+ setOperationAction(ISD::FPOW, MVT::f16, Promote);
+ setOperationAction(ISD::FPOW, MVT::v4f16, Promote);
+ setOperationAction(ISD::FPOWI, MVT::f16, Promote);
+ setOperationAction(ISD::FCOS, MVT::f16, Promote);
+ setOperationAction(ISD::FCOS, MVT::v4f16, Promote);
+ setOperationAction(ISD::FSIN, MVT::f16, Promote);
+ setOperationAction(ISD::FSIN, MVT::v4f16, Promote);
+ setOperationAction(ISD::FSINCOS, MVT::f16, Promote);
+ setOperationAction(ISD::FSINCOS, MVT::v4f16, Promote);
+ setOperationAction(ISD::FEXP, MVT::f16, Promote);
+ setOperationAction(ISD::FEXP, MVT::v4f16, Promote);
+ setOperationAction(ISD::FEXP2, MVT::f16, Promote);
+ setOperationAction(ISD::FEXP2, MVT::v4f16, Promote);
+ setOperationAction(ISD::FLOG, MVT::f16, Promote);
+ setOperationAction(ISD::FLOG, MVT::v4f16, Promote);
+ setOperationAction(ISD::FLOG2, MVT::f16, Promote);
+ setOperationAction(ISD::FLOG2, MVT::v4f16, Promote);
+ setOperationAction(ISD::FLOG10, MVT::f16, Promote);
+ setOperationAction(ISD::FLOG10, MVT::v4f16, Promote);
if (!Subtarget->hasFullFP16()) {
setOperationAction(ISD::SELECT, MVT::f16, Promote);
@@ -361,53 +371,39 @@ AArch64TargetLowering::AArch64TargetLowe
setOperationAction(ISD::FMAXNUM, MVT::f16, Promote);
setOperationAction(ISD::FMINNAN, MVT::f16, Promote);
setOperationAction(ISD::FMAXNAN, MVT::f16, Promote);
- }
-
- // v4f16 is also a storage-only type, so promote it to v4f32 when that is
- // known to be safe.
- setOperationAction(ISD::FADD, MVT::v4f16, Promote);
- setOperationAction(ISD::FSUB, MVT::v4f16, Promote);
- setOperationAction(ISD::FMUL, MVT::v4f16, Promote);
- setOperationAction(ISD::FDIV, MVT::v4f16, Promote);
- setOperationAction(ISD::FP_EXTEND, MVT::v4f16, Promote);
- setOperationAction(ISD::FP_ROUND, MVT::v4f16, Promote);
- AddPromotedToType(ISD::FADD, MVT::v4f16, MVT::v4f32);
- AddPromotedToType(ISD::FSUB, MVT::v4f16, MVT::v4f32);
- AddPromotedToType(ISD::FMUL, MVT::v4f16, MVT::v4f32);
- AddPromotedToType(ISD::FDIV, MVT::v4f16, MVT::v4f32);
- AddPromotedToType(ISD::FP_EXTEND, MVT::v4f16, MVT::v4f32);
- AddPromotedToType(ISD::FP_ROUND, MVT::v4f16, MVT::v4f32);
-
- // Expand all other v4f16 operations.
- // FIXME: We could generate better code by promoting some operations to
- // a pair of v4f32s
- setOperationAction(ISD::FABS, MVT::v4f16, Expand);
- setOperationAction(ISD::FCEIL, MVT::v4f16, Expand);
- setOperationAction(ISD::FCOPYSIGN, MVT::v4f16, Expand);
- setOperationAction(ISD::FCOS, MVT::v4f16, Expand);
- setOperationAction(ISD::FFLOOR, MVT::v4f16, Expand);
- setOperationAction(ISD::FMA, MVT::v4f16, Expand);
- setOperationAction(ISD::FNEARBYINT, MVT::v4f16, Expand);
- setOperationAction(ISD::FNEG, MVT::v4f16, Expand);
- setOperationAction(ISD::FPOW, MVT::v4f16, Expand);
- setOperationAction(ISD::FREM, MVT::v4f16, Expand);
- setOperationAction(ISD::FROUND, MVT::v4f16, Expand);
- setOperationAction(ISD::FRINT, MVT::v4f16, Expand);
- setOperationAction(ISD::FSIN, MVT::v4f16, Expand);
- setOperationAction(ISD::FSINCOS, MVT::v4f16, Expand);
- setOperationAction(ISD::FSQRT, MVT::v4f16, Expand);
- setOperationAction(ISD::FTRUNC, MVT::v4f16, Expand);
- setOperationAction(ISD::SETCC, MVT::v4f16, Expand);
- setOperationAction(ISD::BR_CC, MVT::v4f16, Expand);
- setOperationAction(ISD::SELECT, MVT::v4f16, Expand);
- setOperationAction(ISD::SELECT_CC, MVT::v4f16, Expand);
- setOperationAction(ISD::FEXP, MVT::v4f16, Expand);
- setOperationAction(ISD::FEXP2, MVT::v4f16, Expand);
- setOperationAction(ISD::FLOG, MVT::v4f16, Expand);
- setOperationAction(ISD::FLOG2, MVT::v4f16, Expand);
- setOperationAction(ISD::FLOG10, MVT::v4f16, Expand);
+ // promote v4f16 to v4f32 when that is known to be safe.
+ setOperationAction(ISD::FADD, MVT::v4f16, Promote);
+ setOperationAction(ISD::FSUB, MVT::v4f16, Promote);
+ setOperationAction(ISD::FMUL, MVT::v4f16, Promote);
+ setOperationAction(ISD::FDIV, MVT::v4f16, Promote);
+ setOperationAction(ISD::FP_EXTEND, MVT::v4f16, Promote);
+ setOperationAction(ISD::FP_ROUND, MVT::v4f16, Promote);
+ AddPromotedToType(ISD::FADD, MVT::v4f16, MVT::v4f32);
+ AddPromotedToType(ISD::FSUB, MVT::v4f16, MVT::v4f32);
+ AddPromotedToType(ISD::FMUL, MVT::v4f16, MVT::v4f32);
+ AddPromotedToType(ISD::FDIV, MVT::v4f16, MVT::v4f32);
+ AddPromotedToType(ISD::FP_EXTEND, MVT::v4f16, MVT::v4f32);
+ AddPromotedToType(ISD::FP_ROUND, MVT::v4f16, MVT::v4f32);
+
+ setOperationAction(ISD::FABS, MVT::v4f16, Expand);
+ setOperationAction(ISD::FNEG, MVT::v4f16, Expand);
+ setOperationAction(ISD::FROUND, MVT::v4f16, Expand);
+ setOperationAction(ISD::FMA, MVT::v4f16, Expand);
+ setOperationAction(ISD::SETCC, MVT::v4f16, Expand);
+ setOperationAction(ISD::BR_CC, MVT::v4f16, Expand);
+ setOperationAction(ISD::SELECT, MVT::v4f16, Expand);
+ 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);
Modified: llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll?rev=312104&r1=312103&r2=312104&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll (original)
+++ llvm/trunk/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll Wed Aug 30 01:38:13 2017
@@ -1,14 +1,30 @@
-; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple | FileCheck %s
+; 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>
-; CHECK: test_v2f32.sqrt:
+%v4f16 = type <4 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: fsqrt.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.sqrt.v4f16(%v4f16 %a)
+ ret %v4f16 %1
+}
; CHECK: test_v2f32.powi:
define %v2f32 @test_v2f32.powi(%v2f32 %a, i32 %b) {
; CHECK: pow
@@ -63,50 +79,142 @@ define %v2f32 @test_v2f32.log2(%v2f32 %a
%1 = call %v2f32 @llvm.log2.v2f32(%v2f32 %a)
ret %v2f32 %1
}
-; CHECK: test_v2f32.fma:
+; CHECK-LABEL: test_v2f32.fma:
define %v2f32 @test_v2f32.fma(%v2f32 %a, %v2f32 %b, %v2f32 %c) {
- ; CHECK: fma
+ ; CHECK: fmla.2s
%1 = call %v2f32 @llvm.fma.v2f32(%v2f32 %a, %v2f32 %b, %v2f32 %c)
ret %v2f32 %1
}
-; CHECK: test_v2f32.fabs:
+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: fmla.4h
+ %1 = call %v4f16 @llvm.fma.v4f16(%v4f16 %a, %v4f16 %b, %v4f16 %c)
+ ret %v4f16 %1
+}
+; CHECK-LABEL: test_v2f32.fabs:
define %v2f32 @test_v2f32.fabs(%v2f32 %a) {
- ; CHECK: fabs
+ ; CHECK: fabs.2s
%1 = call %v2f32 @llvm.fabs.v2f32(%v2f32 %a)
ret %v2f32 %1
}
-; CHECK: test_v2f32.floor:
+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: fabs.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.fabs.v4f16(%v4f16 %a)
+ ret %v4f16 %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
}
-; CHECK: test_v2f32.ceil:
+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: frintm.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.floor.v4f16(%v4f16 %a)
+ ret %v4f16 %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
}
-; CHECK: test_v2f32.trunc:
+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: frintp.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.ceil.v4f16(%v4f16 %a)
+ ret %v4f16 %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
}
-; CHECK: test_v2f32.rint:
+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
+}
+; 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
}
-; CHECK: test_v2f32.nearbyint:
+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: frintx.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.rint.v4f16(%v4f16 %a)
+ ret %v4f16 %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: frinti.4h
+; CHECK-FP16-NEXT: ret
+ %1 = call %v4f16 @llvm.nearbyint.v4f16(%v4f16 %a)
+ ret %v4f16 %1
+}
declare %v2f32 @llvm.sqrt.v2f32(%v2f32) #0
+declare %v4f16 @llvm.sqrt.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.powi.v2f32(%v2f32, i32) #0
declare %v2f32 @llvm.sin.v2f32(%v2f32) #0
declare %v2f32 @llvm.cos.v2f32(%v2f32) #0
@@ -116,13 +224,27 @@ declare %v2f32 @llvm.exp2.v2f32(%v2f32)
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 %v2f32 @llvm.fabs.v2f32(%v2f32) #0
+declare %v4f16 @llvm.fabs.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.floor.v2f32(%v2f32) #0
+declare %v4f16 @llvm.floor.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.ceil.v2f32(%v2f32) #0
+declare %v4f16 @llvm.ceil.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.trunc.v2f32(%v2f32) #0
+declare %v4f16 @llvm.trunc.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.rint.v2f32(%v2f32) #0
+declare %v4f16 @llvm.rint.v4f16(%v4f16) #0
+
declare %v2f32 @llvm.nearbyint.v2f32(%v2f32) #0
+declare %v4f16 @llvm.nearbyint.v4f16(%v4f16) #0
;;;
Modified: llvm/trunk/test/CodeGen/AArch64/fp16-v4-instructions.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/fp16-v4-instructions.ll?rev=312104&r1=312103&r2=312104&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AArch64/fp16-v4-instructions.ll (original)
+++ llvm/trunk/test/CodeGen/AArch64/fp16-v4-instructions.ll Wed Aug 30 01:38:13 2017
@@ -1,12 +1,17 @@
-; 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-COMMON
+; RUN: llc < %s -asm-verbose=false -mtriple=aarch64-none-eabi -mattr=+fullfp16 | FileCheck %s --check-prefix=CHECK-FP16 --check-prefix=CHECK-COMMON
define <4 x half> @add_h(<4 x half> %a, <4 x half> %b) {
entry:
-; CHECK-LABEL: add_h:
-; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
-; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
-; CHECK: fadd [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
-; CHECK: fcvtn v0.4h, [[RES]]
+; CHECK-CVT-LABEL: add_h:
+; CHECK-CVT-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-CVT-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK-CVT-NEXT: fadd [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK-CVT-NEXT: fcvtn v0.4h, [[RES]]
+
+; CHECK-FP16-LABEL: add_h:
+; CHECK-FP16: fadd v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT: ret
%0 = fadd <4 x half> %a, %b
ret <4 x half> %0
}
@@ -14,20 +19,24 @@ entry:
define <4 x half> @build_h4(<4 x half> %a) {
entry:
-; CHECK-LABEL: build_h4:
-; CHECK: mov [[GPR:w[0-9]+]], #15565
-; CHECK: dup v0.4h, [[GPR]]
+; CHECK-COMMON-LABEL: build_h4:
+; CHECK-COMMON: mov [[GPR:w[0-9]+]], #15565
+; CHECK-COMMON-NEXT: dup v0.4h, [[GPR]]
ret <4 x half> <half 0xH3CCD, half 0xH3CCD, half 0xH3CCD, half 0xH3CCD>
}
define <4 x half> @sub_h(<4 x half> %a, <4 x half> %b) {
entry:
-; CHECK-LABEL: sub_h:
-; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
-; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
-; CHECK: fsub [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
-; CHECK: fcvtn v0.4h, [[RES]]
+; CHECK-CVT-LABEL: sub_h:
+; CHECK-CVT-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-CVT-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK-CVT-NEXT: fsub [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK-CVT-NEXT: fcvtn v0.4h, [[RES]]
+
+; CHECK-FP16-LABEL: sub_h:
+; CHECK-FP16: fsub v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT: ret
%0 = fsub <4 x half> %a, %b
ret <4 x half> %0
}
@@ -35,11 +44,15 @@ entry:
define <4 x half> @mul_h(<4 x half> %a, <4 x half> %b) {
entry:
-; CHECK-LABEL: mul_h:
-; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
-; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
-; CHECK: fmul [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
-; CHECK: fcvtn v0.4h, [[RES]]
+; CHECK-CVT-LABEL: mul_h:
+; CHECK-CVT-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-CVT-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK-CVT-NEXT: fmul [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK-CVT-NEXT: fcvtn v0.4h, [[RES]]
+
+; CHECK-FP16-LABEL: mul_h:
+; CHECK-FP16: fmul v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT: ret
%0 = fmul <4 x half> %a, %b
ret <4 x half> %0
}
@@ -47,11 +60,15 @@ entry:
define <4 x half> @div_h(<4 x half> %a, <4 x half> %b) {
entry:
-; CHECK-LABEL: div_h:
-; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
-; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
-; CHECK: fdiv [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
-; CHECK: fcvtn v0.4h, [[RES]]
+; CHECK-CVT-LABEL: div_h:
+; CHECK-CVT-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-CVT-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK-CVT-NEXT: fdiv [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK-CVT-NEXT: fcvtn v0.4h, [[RES]]
+
+; CHECK-FP16-LABEL: div_h:
+; CHECK-FP16: fdiv v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT: ret
%0 = fdiv <4 x half> %a, %b
ret <4 x half> %0
}
@@ -59,8 +76,9 @@ entry:
define <4 x half> @load_h(<4 x half>* %a) {
entry:
-; CHECK-LABEL: load_h:
-; CHECK: ldr d0, [x0]
+; CHECK-COMMON-LABEL: load_h:
+; CHECK-COMMON: ldr d0, [x0]
+; CHECK-COMMON-NEXT: ret
%0 = load <4 x half>, <4 x half>* %a, align 4
ret <4 x half> %0
}
@@ -68,476 +86,643 @@ entry:
define void @store_h(<4 x half>* %a, <4 x half> %b) {
entry:
-; CHECK-LABEL: store_h:
-; CHECK: str d0, [x0]
+; CHECK-COMMON-LABEL: store_h:
+; CHECK-COMMON: str d0, [x0]
+; CHECK-COMMON-NEXT: ret
store <4 x half> %b, <4 x half>* %a, align 4
ret void
}
define <4 x half> @s_to_h(<4 x float> %a) {
-; CHECK-LABEL: s_to_h:
-; CHECK: fcvtn v0.4h, v0.4s
+; CHECK-COMMON-LABEL: s_to_h:
+; CHECK-COMMON: fcvtn v0.4h, v0.4s
+; CHECK-COMMON-NEXT: ret
%1 = fptrunc <4 x float> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @d_to_h(<4 x double> %a) {
-; CHECK-LABEL: d_to_h:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: ins
-; CHECK-DAG: ins
-; CHECK-DAG: ins
-; CHECK-DAG: ins
+; CHECK-COMMON-LABEL: d_to_h:
+; CHECK-COMMON: mov
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: mov
+; CHECK-COMMON: ins
+; CHECK-COMMON: ins
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: ins
+; CHECK-COMMON: ret
%1 = fptrunc <4 x double> %a to <4 x half>
ret <4 x half> %1
}
define <4 x float> @h_to_s(<4 x half> %a) {
-; CHECK-LABEL: h_to_s:
-; CHECK: fcvtl v0.4s, v0.4h
+; CHECK-COMMON-LABEL: h_to_s:
+; CHECK-COMMON: fcvtl v0.4s, v0.4h
+; CHECK-COMMON-NEXT: ret
%1 = fpext <4 x half> %a to <4 x float>
ret <4 x float> %1
}
define <4 x double> @h_to_d(<4 x half> %a) {
-; CHECK-LABEL: h_to_d:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: ins
-; CHECK-DAG: ins
-; CHECK-DAG: ins
-; CHECK-DAG: ins
+; CHECK-COMMON-LABEL: h_to_d:
+; CHECK-COMMON: mov
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: mov
+; CHECK-COMMON: mov
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: fcvt
+; CHECK-COMMON: ins
+; CHECK-COMMON: ins
+; CHECK-COMMON: mov
+; CHECK-COMMON: ret
%1 = fpext <4 x half> %a to <4 x double>
ret <4 x double> %1
}
define <4 x half> @bitcast_i_to_h(float, <4 x i16> %a) {
-; CHECK-LABEL: bitcast_i_to_h:
-; CHECK: mov v0.16b, v1.16b
+; CHECK-COMMON-LABEL: bitcast_i_to_h:
+; CHECK-COMMON: mov v0.16b, v1.16b
+; CHECK-COMMON-NEXT: ret
%2 = bitcast <4 x i16> %a to <4 x half>
ret <4 x half> %2
}
define <4 x i16> @bitcast_h_to_i(float, <4 x half> %a) {
-; CHECK-LABEL: bitcast_h_to_i:
-; CHECK: mov v0.16b, v1.16b
+; CHECK-COMMON-LABEL: bitcast_h_to_i:
+; CHECK-COMMON: mov v0.16b, v1.16b
+; CHECK-COMMON-NEXT: ret
%2 = bitcast <4 x half> %a to <4 x i16>
ret <4 x i16> %2
}
define <4 x half> @sitofp_i8(<4 x i8> %a) #0 {
-; CHECK-LABEL: sitofp_i8:
-; CHECK-NEXT: shl [[OP1:v[0-9]+\.4h]], v0.4h, #8
-; CHECK-NEXT: sshr [[OP2:v[0-9]+\.4h]], [[OP1]], #8
-; CHECK-NEXT: sshll [[OP3:v[0-9]+\.4s]], [[OP2]], #0
-; CHECK-NEXT: scvtf [[OP4:v[0-9]+\.4s]], [[OP3]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP4]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: sitofp_i8:
+; CHECK-COMMON-NEXT: shl [[OP1:v[0-9]+\.4h]], v0.4h, #8
+; CHECK-COMMON-NEXT: sshr [[OP2:v[0-9]+\.4h]], [[OP1]], #8
+; CHECK-COMMON-NEXT: sshll [[OP3:v[0-9]+\.4s]], [[OP2]], #0
+; CHECK-COMMON-NEXT: scvtf [[OP4:v[0-9]+\.4s]], [[OP3]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP4]]
+; CHECK-COMMON-NEXT: ret
%1 = sitofp <4 x i8> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @sitofp_i16(<4 x i16> %a) #0 {
-; CHECK-LABEL: sitofp_i16:
-; CHECK-NEXT: sshll [[OP1:v[0-9]+\.4s]], v0.4h, #0
-; CHECK-NEXT: scvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: sitofp_i16:
+; CHECK-COMMON-NEXT: sshll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-COMMON-NEXT: scvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-COMMON-NEXT: ret
%1 = sitofp <4 x i16> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @sitofp_i32(<4 x i32> %a) #0 {
-; CHECK-LABEL: sitofp_i32:
-; CHECK-NEXT: scvtf [[OP1:v[0-9]+\.4s]], v0.4s
-; CHECK-NEXT: fcvtn v0.4h, [[OP1]]
+; CHECK-COMMON-LABEL: sitofp_i32:
+; CHECK-COMMON-NEXT: scvtf [[OP1:v[0-9]+\.4s]], v0.4s
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP1]]
+; CHECK-COMMON-NEXT: ret
%1 = sitofp <4 x i32> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @sitofp_i64(<4 x i64> %a) #0 {
-; CHECK-LABEL: sitofp_i64:
-; CHECK-DAG: scvtf [[OP1:v[0-9]+\.2d]], v0.2d
-; CHECK-DAG: scvtf [[OP2:v[0-9]+\.2d]], v1.2d
-; CHECK-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
-; CHECK-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP3]].4s
+; CHECK-COMMON-LABEL: sitofp_i64:
+; CHECK-COMMON-DAG: scvtf [[OP1:v[0-9]+\.2d]], v0.2d
+; CHECK-COMMON-DAG: scvtf [[OP2:v[0-9]+\.2d]], v1.2d
+; CHECK-COMMON-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
+; CHECK-COMMON-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP3]].4s
+; CHECK-COMMON-NEXT: ret
%1 = sitofp <4 x i64> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @uitofp_i8(<4 x i8> %a) #0 {
-; CHECK-LABEL: uitofp_i8:
-; CHECK-NEXT: bic v0.4h, #255, lsl #8
-; CHECK-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
-; CHECK-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: uitofp_i8:
+; CHECK-COMMON-NEXT: bic v0.4h, #255, lsl #8
+; CHECK-COMMON-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-COMMON-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-COMMON-NEXT: ret
%1 = uitofp <4 x i8> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @uitofp_i16(<4 x i16> %a) #0 {
-; CHECK-LABEL: uitofp_i16:
-; CHECK-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
-; CHECK-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: uitofp_i16:
+; CHECK-COMMON-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-COMMON-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-COMMON-NEXT: ret
%1 = uitofp <4 x i16> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @uitofp_i32(<4 x i32> %a) #0 {
-; CHECK-LABEL: uitofp_i32:
-; CHECK-NEXT: ucvtf [[OP1:v[0-9]+\.4s]], v0.4s
-; CHECK-NEXT: fcvtn v0.4h, [[OP1]]
+; CHECK-COMMON-LABEL: uitofp_i32:
+; CHECK-COMMON-NEXT: ucvtf [[OP1:v[0-9]+\.4s]], v0.4s
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP1]]
+; CHECK-COMMON-NEXT: ret
%1 = uitofp <4 x i32> %a to <4 x half>
ret <4 x half> %1
}
define <4 x half> @uitofp_i64(<4 x i64> %a) #0 {
-; CHECK-LABEL: uitofp_i64:
-; CHECK-DAG: ucvtf [[OP1:v[0-9]+\.2d]], v0.2d
-; CHECK-DAG: ucvtf [[OP2:v[0-9]+\.2d]], v1.2d
-; CHECK-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
-; CHECK-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
-; CHECK-NEXT: fcvtn v0.4h, [[OP3]].4s
+; CHECK-COMMON-LABEL: uitofp_i64:
+; CHECK-COMMON-DAG: ucvtf [[OP1:v[0-9]+\.2d]], v0.2d
+; CHECK-COMMON-DAG: ucvtf [[OP2:v[0-9]+\.2d]], v1.2d
+; CHECK-COMMON-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
+; CHECK-COMMON-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
+; CHECK-COMMON-NEXT: fcvtn v0.4h, [[OP3]].4s
+; CHECK-COMMON-NEXT: ret
%1 = uitofp <4 x i64> %a to <4 x half>
ret <4 x half> %1
}
define void @test_insert_at_zero(half %a, <4 x half>* %b) #0 {
-; CHECK-LABEL: test_insert_at_zero:
-; CHECK-NEXT: str d0, [x0]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: test_insert_at_zero:
+; CHECK-COMMON-NEXT: str d0, [x0]
+; CHECK-COMMON-NEXT: ret
%1 = insertelement <4 x half> undef, half %a, i64 0
store <4 x half> %1, <4 x half>* %b, align 4
ret void
}
define <4 x i8> @fptosi_i8(<4 x half> %a) #0 {
-; CHECK-LABEL: fptosi_i8:
-; CHECK-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
-; CHECK-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
-; CHECK-NEXT: xtn v0.4h, [[REG2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: fptosi_i8:
+; CHECK-COMMON-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
+; CHECK-COMMON-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
+; CHECK-COMMON-NEXT: xtn v0.4h, [[REG2]]
+; CHECK-COMMON-NEXT: ret
%1 = fptosi<4 x half> %a to <4 x i8>
ret <4 x i8> %1
}
define <4 x i16> @fptosi_i16(<4 x half> %a) #0 {
-; CHECK-LABEL: fptosi_i16:
-; CHECK-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
-; CHECK-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
-; CHECK-NEXT: xtn v0.4h, [[REG2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: fptosi_i16:
+; CHECK-COMMON-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
+; CHECK-COMMON-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
+; CHECK-COMMON-NEXT: xtn v0.4h, [[REG2]]
+; CHECK-COMMON-NEXT: ret
%1 = fptosi<4 x half> %a to <4 x i16>
ret <4 x i16> %1
}
define <4 x i8> @fptoui_i8(<4 x half> %a) #0 {
-; CHECK-LABEL: fptoui_i8:
-; CHECK-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
+; CHECK-COMMON-LABEL: fptoui_i8:
+; CHECK-COMMON-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
; NOTE: fcvtzs selected here because the xtn shaves the sign bit
-; CHECK-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
-; CHECK-NEXT: xtn v0.4h, [[REG2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-NEXT: fcvtzs [[REG2:v[0-9]+\.4s]], [[REG1]]
+; CHECK-COMMON-NEXT: xtn v0.4h, [[REG2]]
+; CHECK-COMMON-NEXT: ret
%1 = fptoui<4 x half> %a to <4 x i8>
ret <4 x i8> %1
}
define <4 x i16> @fptoui_i16(<4 x half> %a) #0 {
-; CHECK-LABEL: fptoui_i16:
-; CHECK-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
-; CHECK-NEXT: fcvtzu [[REG2:v[0-9]+\.4s]], [[REG1]]
-; CHECK-NEXT: xtn v0.4h, [[REG2]]
-; CHECK-NEXT: ret
+; CHECK-COMMON-LABEL: fptoui_i16:
+; CHECK-COMMON-NEXT: fcvtl [[REG1:v[0-9]+\.4s]], v0.4h
+; CHECK-COMMON-NEXT: fcvtzu [[REG2:v[0-9]+\.4s]], [[REG1]]
+; CHECK-COMMON-NEXT: xtn v0.4h, [[REG2]]
+; CHECK-COMMON-NEXT: ret
%1 = fptoui<4 x half> %a to <4 x i16>
ret <4 x i16> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_une:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, ne
-; CHECK-DAG: csetm {{.*}}, ne
-; CHECK-DAG: csetm {{.*}}, ne
-; CHECK-DAG: csetm {{.*}}, ne
define <4 x i1> @test_fcmp_une(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_une:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, ne
+; CHECK-CVT-DAG: csetm {{.*}}, ne
+; CHECK-CVT-DAG: csetm {{.*}}, ne
+; CHECK-CVT-DAG: csetm {{.*}}, ne
+
+; CHECK-FP16-LABEL: test_fcmp_une:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ne
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ne
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ne
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ne
+; CHECK-FP16: ret
%1 = fcmp une <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ueq:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm [[REG1:w[0-9]+]], eq
-; CHECK-DAG: csetm [[REG2:w[0-9]+]], eq
-; CHECK-DAG: csetm [[REG3:w[0-9]+]], eq
-; CHECK-DAG: csetm [[REG4:w[0-9]+]], eq
-; CHECK-DAG: csinv {{.*}}, [[REG1]], wzr, vc
-; CHECK-DAG: csinv {{.*}}, [[REG2]], wzr, vc
-; CHECK-DAG: csinv {{.*}}, [[REG3]], wzr, vc
-; CHECK-DAG: csinv {{.*}}, [[REG4]], wzr, vc
define <4 x i1> @test_fcmp_ueq(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ueq:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm [[REG1:w[0-9]+]], eq
+; CHECK-CVT-DAG: csetm [[REG2:w[0-9]+]], eq
+; CHECK-CVT-DAG: csetm [[REG3:w[0-9]+]], eq
+; CHECK-CVT-DAG: csetm [[REG4:w[0-9]+]], eq
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG1]], wzr, vc
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG2]], wzr, vc
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG3]], wzr, vc
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG4]], wzr, vc
+
+; CHECK-FP16-LABEL: test_fcmp_ueq:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: ret
%1 = fcmp ueq <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ugt:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, hi
-; CHECK-DAG: csetm {{.*}}, hi
-; CHECK-DAG: csetm {{.*}}, hi
-; CHECK-DAG: csetm {{.*}}, hi
define <4 x i1> @test_fcmp_ugt(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ugt:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, hi
+; CHECK-CVT-DAG: csetm {{.*}}, hi
+; CHECK-CVT-DAG: csetm {{.*}}, hi
+; CHECK-CVT-DAG: csetm {{.*}}, hi
+
+; CHECK-FP16-LABEL: test_fcmp_ugt:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, hi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, hi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, hi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, hi
+; CHECK-FP16: ret
%1 = fcmp ugt <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_uge:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, pl
-; CHECK-DAG: csetm {{.*}}, pl
-; CHECK-DAG: csetm {{.*}}, pl
-; CHECK-DAG: csetm {{.*}}, pl
define <4 x i1> @test_fcmp_uge(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_uge:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, pl
+; CHECK-CVT-DAG: csetm {{.*}}, pl
+; CHECK-CVT-DAG: csetm {{.*}}, pl
+; CHECK-CVT-DAG: csetm {{.*}}, pl
+
+; CHECK-FP16-LABEL: test_fcmp_uge:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, pl
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, pl
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, pl
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, pl
+; CHECK-FP16: ret
%1 = fcmp uge <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ult:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, lt
-; CHECK-DAG: csetm {{.*}}, lt
-; CHECK-DAG: csetm {{.*}}, lt
-; CHECK-DAG: csetm {{.*}}, lt
define <4 x i1> @test_fcmp_ult(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ult:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, lt
+; CHECK-CVT-DAG: csetm {{.*}}, lt
+; CHECK-CVT-DAG: csetm {{.*}}, lt
+; CHECK-CVT-DAG: csetm {{.*}}, lt
+
+; CHECK-FP16-LABEL: test_fcmp_ult:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, lt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, lt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, lt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, lt
+; CHECK-FP16: ret
%1 = fcmp ult <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ule:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, le
-; CHECK-DAG: csetm {{.*}}, le
-; CHECK-DAG: csetm {{.*}}, le
-; CHECK-DAG: csetm {{.*}}, le
define <4 x i1> @test_fcmp_ule(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ule:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, le
+; CHECK-CVT-DAG: csetm {{.*}}, le
+; CHECK-CVT-DAG: csetm {{.*}}, le
+; CHECK-CVT-DAG: csetm {{.*}}, le
+
+; CHECK-FP16-LABEL: test_fcmp_ule:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, le
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, le
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, le
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, le
+; CHECK-FP16: ret
%1 = fcmp ule <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_uno:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, vs
-; CHECK-DAG: csetm {{.*}}, vs
-; CHECK-DAG: csetm {{.*}}, vs
-; CHECK-DAG: csetm {{.*}}, vs
define <4 x i1> @test_fcmp_uno(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_uno:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, vs
+; CHECK-CVT-DAG: csetm {{.*}}, vs
+; CHECK-CVT-DAG: csetm {{.*}}, vs
+; CHECK-CVT-DAG: csetm {{.*}}, vs
+
+; CHECK-FP16-LABEL: test_fcmp_uno:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vs
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vs
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vs
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vs
+; CHECK-FP16: ret
%1 = fcmp uno <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_one:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm [[REG1:w[0-9]+]], mi
-; CHECK-DAG: csetm [[REG2:w[0-9]+]], mi
-; CHECK-DAG: csetm [[REG3:w[0-9]+]], mi
-; CHECK-DAG: csetm [[REG4:w[0-9]+]], mi
-; CHECK-DAG: csinv {{.*}}, [[REG1]], wzr, le
-; CHECK-DAG: csinv {{.*}}, [[REG2]], wzr, le
-; CHECK-DAG: csinv {{.*}}, [[REG3]], wzr, le
-; CHECK-DAG: csinv {{.*}}, [[REG4]], wzr, le
-
define <4 x i1> @test_fcmp_one(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_one:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm [[REG1:w[0-9]+]], mi
+; CHECK-CVT-DAG: csetm [[REG2:w[0-9]+]], mi
+; CHECK-CVT-DAG: csetm [[REG3:w[0-9]+]], mi
+; CHECK-CVT-DAG: csetm [[REG4:w[0-9]+]], mi
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG1]], wzr, le
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG2]], wzr, le
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG3]], wzr, le
+; CHECK-CVT-DAG: csinv {{.*}}, [[REG4]], wzr, le
+
+; CHECK-FP16-LABEL: test_fcmp_one:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: ret
%1 = fcmp one <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_oeq:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, eq
-; CHECK-DAG: csetm {{.*}}, eq
-; CHECK-DAG: csetm {{.*}}, eq
-; CHECK-DAG: csetm {{.*}}, eq
define <4 x i1> @test_fcmp_oeq(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_oeq:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, eq
+; CHECK-CVT-DAG: csetm {{.*}}, eq
+; CHECK-CVT-DAG: csetm {{.*}}, eq
+; CHECK-CVT-DAG: csetm {{.*}}, eq
+
+; CHECK-FP16-LABEL: test_fcmp_oeq:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, eq
+; CHECK-FP16: ret
%1 = fcmp oeq <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ogt:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, gt
-; CHECK-DAG: csetm {{.*}}, gt
-; CHECK-DAG: csetm {{.*}}, gt
-; CHECK-DAG: csetm {{.*}}, gt
define <4 x i1> @test_fcmp_ogt(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ogt:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, gt
+; CHECK-CVT-DAG: csetm {{.*}}, gt
+; CHECK-CVT-DAG: csetm {{.*}}, gt
+; CHECK-CVT-DAG: csetm {{.*}}, gt
+
+; CHECK-FP16-LABEL: test_fcmp_ogt:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, gt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, gt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, gt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, gt
+; CHECK-FP16: ret
%1 = fcmp ogt <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_oge:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, ge
-; CHECK-DAG: csetm {{.*}}, ge
-; CHECK-DAG: csetm {{.*}}, ge
-; CHECK-DAG: csetm {{.*}}, ge
define <4 x i1> @test_fcmp_oge(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_oge:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, ge
+; CHECK-CVT-DAG: csetm {{.*}}, ge
+; CHECK-CVT-DAG: csetm {{.*}}, ge
+; CHECK-CVT-DAG: csetm {{.*}}, ge
+
+; CHECK-FP16-LABEL: test_fcmp_oge:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ge
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ge
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ge
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ge
+; CHECK-FP16: ret
%1 = fcmp oge <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_olt:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, mi
-; CHECK-DAG: csetm {{.*}}, mi
-; CHECK-DAG: csetm {{.*}}, mi
-; CHECK-DAG: csetm {{.*}}, mi
define <4 x i1> @test_fcmp_olt(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_olt:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, mi
+; CHECK-CVT-DAG: csetm {{.*}}, mi
+; CHECK-CVT-DAG: csetm {{.*}}, mi
+; CHECK-CVT-DAG: csetm {{.*}}, mi
+
+; CHECK-FP16-LABEL: test_fcmp_olt:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, mi
+; CHECK-FP16: ret
%1 = fcmp olt <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ole:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, ls
-; CHECK-DAG: csetm {{.*}}, ls
-; CHECK-DAG: csetm {{.*}}, ls
-; CHECK-DAG: csetm {{.*}}, ls
define <4 x i1> @test_fcmp_ole(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ole:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, ls
+; CHECK-CVT-DAG: csetm {{.*}}, ls
+; CHECK-CVT-DAG: csetm {{.*}}, ls
+; CHECK-CVT-DAG: csetm {{.*}}, ls
+
+; CHECK-FP16-LABEL: test_fcmp_ole:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ls
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ls
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ls
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, ls
+; CHECK-FP16: ret
%1 = fcmp ole <4 x half> %a, %b
ret <4 x i1> %1
}
-; Function Attrs: nounwind readnone
-; CHECK-LABEL: test_fcmp_ord:
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: fcvt
-; CHECK-DAG: csetm {{.*}}, vc
-; CHECK-DAG: csetm {{.*}}, vc
-; CHECK-DAG: csetm {{.*}}, vc
-; CHECK-DAG: csetm {{.*}}, vc
define <4 x i1> @test_fcmp_ord(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-CVT-LABEL: test_fcmp_ord:
+; 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-DAG: fcvt
+; CHECK-CVT-DAG: csetm {{.*}}, vc
+; CHECK-CVT-DAG: csetm {{.*}}, vc
+; CHECK-CVT-DAG: csetm {{.*}}, vc
+; CHECK-CVT-DAG: csetm {{.*}}, vc
+
+; CHECK-FP16-LABEL: test_fcmp_ord:
+; CHECK-FP16-NOT: fcvt
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vc
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vc
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vc
+; CHECK-FP16: fcmp h{{.}}, h{{.}}
+; CHECK-FP16: csetm {{.*}}, vc
+; CHECK-FP16: ret
%1 = fcmp ord <4 x half> %a, %b
ret <4 x i1> %1
}
More information about the llvm-commits
mailing list