TheRock/compiler/amd-llvm/llvm/test/CodeGen/LoongArch/lasx/fp-rounding.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s
; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s

;; ceilf
define void @ceil_v8f32(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: ceil_v8f32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrp.s $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <8 x float>, ptr %a0
  %r = call <8 x float> @llvm.ceil.v8f32(<8 x float> %v0)
  store <8 x float> %r, ptr %res
  ret void
}

;; ceil
define void @ceil_v4f64(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: ceil_v4f64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrp.d $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <4 x double>, ptr %a0
  %r = call <4 x double> @llvm.ceil.v4f64(<4 x double> %v0)
  store <4 x double> %r, ptr %res
  ret void
}

;; floorf
define void @floor_v8f32(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: floor_v8f32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrm.s $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <8 x float>, ptr %a0
  %r = call <8 x float> @llvm.floor.v8f32(<8 x float> %v0)
  store <8 x float> %r, ptr %res
  ret void
}

;; floor
define void @floor_v4f64(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: floor_v4f64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrm.d $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <4 x double>, ptr %a0
  %r = call <4 x double> @llvm.floor.v4f64(<4 x double> %v0)
  store <4 x double> %r, ptr %res
  ret void
}

;; truncf
define void @trunc_v8f32(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: trunc_v8f32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrz.s $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <8 x float>, ptr %a0
  %r = call <8 x float> @llvm.trunc.v8f32(<8 x float> %v0)
  store <8 x float> %r, ptr %res
  ret void
}

;; trunc
define void @trunc_v4f64(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: trunc_v4f64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrz.d $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <4 x double>, ptr %a0
  %r = call <4 x double> @llvm.trunc.v4f64(<4 x double> %v0)
  store <4 x double> %r, ptr %res
  ret void
}

;; roundevenf
define void @roundeven_v8f32(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: roundeven_v8f32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrne.s $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <8 x float>, ptr %a0
  %r = call <8 x float> @llvm.roundeven.v8f32(<8 x float> %v0)
  store <8 x float> %r, ptr %res
  ret void
}

;; roundeven
define void @roundeven_v4f64(ptr %res, ptr %a0) nounwind {
; CHECK-LABEL: roundeven_v4f64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    xvld $xr0, $a1, 0
; CHECK-NEXT:    xvfrintrne.d $xr0, $xr0
; CHECK-NEXT:    xvst $xr0, $a0, 0
; CHECK-NEXT:    ret
entry:
  %v0 = load <4 x double>, ptr %a0
  %r = call <4 x double> @llvm.roundeven.v4f64(<4 x double> %v0)
  store <4 x double> %r, ptr %res
  ret void
}

declare <8 x float> @llvm.ceil.v8f32(<8 x float>)
declare <4 x double> @llvm.ceil.v4f64(<4 x double>)
declare <8 x float> @llvm.floor.v8f32(<8 x float>)
declare <4 x double> @llvm.floor.v4f64(<4 x double>)
declare <8 x float> @llvm.trunc.v8f32(<8 x float>)
declare <4 x double> @llvm.trunc.v4f64(<4 x double>)
declare <8 x float> @llvm.roundeven.v8f32(<8 x float>)
declare <4 x double> @llvm.roundeven.v4f64(<4 x double>)