test/CodeGen/RISCV/bfloat.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32I-ILP32
; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV64I-LP64
; RUN: llc -mtriple=riscv32 -mattr=+d -target-abi=ilp32 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32ID-ILP32
; RUN: llc -mtriple=riscv64 -mattr=+d -target-abi=lp64 -verify-machineinstrs < %s  | FileCheck %s -check-prefix=RV64ID-LP64
; RUN: llc -mtriple=riscv32 -mattr=+d -target-abi=ilp32d -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32ID-ILP32D
; RUN: llc -mtriple=riscv64 -mattr=+d -target-abi=lp64d -verify-machineinstrs < %s  | FileCheck %s -check-prefix=RV64ID-LP64D

define bfloat @float_to_bfloat(float %a) nounwind {
; RV32I-ILP32-LABEL: float_to_bfloat:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    call __truncsfbf2@plt
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: float_to_bfloat:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    call __truncsfbf2@plt
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: float_to_bfloat:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32-NEXT:    lui a1, 1048560
; RV32ID-ILP32-NEXT:    or a0, a0, a1
; RV32ID-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: float_to_bfloat:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    addi sp, sp, -16
; RV64ID-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64-NEXT:    lui a1, 1048560
; RV64ID-LP64-NEXT:    or a0, a0, a1
; RV64ID-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    addi sp, sp, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: float_to_bfloat:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    addi sp, sp, -16
; RV32ID-ILP32D-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    lui a1, 1048560
; RV32ID-ILP32D-NEXT:    or a0, a0, a1
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    addi sp, sp, 16
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: float_to_bfloat:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    addi sp, sp, -16
; RV64ID-LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 1048560
; RV64ID-LP64D-NEXT:    or a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    addi sp, sp, 16
; RV64ID-LP64D-NEXT:    ret
  %1 = fptrunc float %a to bfloat
  ret bfloat %1
}

define bfloat @double_to_bfloat(double %a) nounwind {
; RV32I-ILP32-LABEL: double_to_bfloat:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    call __truncdfbf2@plt
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: double_to_bfloat:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    call __truncdfbf2@plt
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: double_to_bfloat:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    call __truncdfbf2@plt
; RV32ID-ILP32-NEXT:    lui a1, 1048560
; RV32ID-ILP32-NEXT:    or a0, a0, a1
; RV32ID-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: double_to_bfloat:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    addi sp, sp, -16
; RV64ID-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    call __truncdfbf2@plt
; RV64ID-LP64-NEXT:    lui a1, 1048560
; RV64ID-LP64-NEXT:    or a0, a0, a1
; RV64ID-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    addi sp, sp, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: double_to_bfloat:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    addi sp, sp, -16
; RV32ID-ILP32D-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    call __truncdfbf2@plt
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    lui a1, 1048560
; RV32ID-ILP32D-NEXT:    or a0, a0, a1
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    addi sp, sp, 16
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: double_to_bfloat:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    addi sp, sp, -16
; RV64ID-LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    call __truncdfbf2@plt
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 1048560
; RV64ID-LP64D-NEXT:    or a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    addi sp, sp, 16
; RV64ID-LP64D-NEXT:    ret
  %1 = fptrunc double %a to bfloat
  ret bfloat %1
}

define float @bfloat_to_float(bfloat %a) nounwind {
; RV32I-ILP32-LABEL: bfloat_to_float:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    slli a0, a0, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_to_float:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    slliw a0, a0, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_to_float:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    slli a0, a0, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_to_float:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    slli a0, a0, 48
; RV64ID-LP64-NEXT:    srli a0, a0, 48
; RV64ID-LP64-NEXT:    slli a0, a0, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_to_float:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    slli a0, a0, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_to_float:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    slli a0, a0, 48
; RV64ID-LP64D-NEXT:    srli a0, a0, 48
; RV64ID-LP64D-NEXT:    slli a0, a0, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ret
  %1 = fpext bfloat %a to float
  ret float %1
}

define double @bfloat_to_double(bfloat %a) nounwind {
; RV32I-ILP32-LABEL: bfloat_to_double:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    slli a0, a0, 16
; RV32I-ILP32-NEXT:    call __extendsfdf2@plt
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_to_double:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    slliw a0, a0, 16
; RV64I-LP64-NEXT:    call __extendsfdf2@plt
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_to_double:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    slli a0, a0, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa5, a0
; RV32ID-ILP32-NEXT:    fcvt.d.s fa5, fa5
; RV32ID-ILP32-NEXT:    fsd fa5, 8(sp)
; RV32ID-ILP32-NEXT:    lw a0, 8(sp)
; RV32ID-ILP32-NEXT:    lw a1, 12(sp)
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_to_double:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    slli a0, a0, 48
; RV64ID-LP64-NEXT:    srli a0, a0, 48
; RV64ID-LP64-NEXT:    slli a0, a0, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa5, a0
; RV64ID-LP64-NEXT:    fcvt.d.s fa5, fa5
; RV64ID-LP64-NEXT:    fmv.x.d a0, fa5
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_to_double:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    slli a0, a0, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa5, a0
; RV32ID-ILP32D-NEXT:    fcvt.d.s fa0, fa5
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_to_double:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    slli a0, a0, 48
; RV64ID-LP64D-NEXT:    srli a0, a0, 48
; RV64ID-LP64D-NEXT:    slli a0, a0, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa5, a0
; RV64ID-LP64D-NEXT:    fcvt.d.s fa0, fa5
; RV64ID-LP64D-NEXT:    ret
  %1 = fpext bfloat %a to double
  ret double %1
}

define bfloat @i16_to_bfloat(i16 %a) nounwind {
; RV32I-ILP32-LABEL: i16_to_bfloat:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: i16_to_bfloat:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: i16_to_bfloat:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    lui a1, 1048560
; RV32ID-ILP32-NEXT:    or a0, a0, a1
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: i16_to_bfloat:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    lui a1, 1048560
; RV64ID-LP64-NEXT:    or a0, a0, a1
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: i16_to_bfloat:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    lui a1, 1048560
; RV32ID-ILP32D-NEXT:    or a0, a0, a1
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: i16_to_bfloat:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    lui a1, 1048560
; RV64ID-LP64D-NEXT:    or a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ret
  %1 = bitcast i16 %a to bfloat
  ret bfloat %1
}

define i16 @bfloat_to_i16(bfloat %a) nounwind {
; RV32I-ILP32-LABEL: bfloat_to_i16:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_to_i16:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_to_i16:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_to_i16:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_to_i16:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_to_i16:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    ret
  %1 = bitcast bfloat %a to i16
  ret i16 %1
}

define bfloat @bfloat_add(bfloat %a, bfloat %b) nounwind {
; RV32I-ILP32-LABEL: bfloat_add:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    slli a0, a0, 16
; RV32I-ILP32-NEXT:    slli a1, a1, 16
; RV32I-ILP32-NEXT:    call __addsf3@plt
; RV32I-ILP32-NEXT:    call __truncsfbf2@plt
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_add:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    slliw a0, a0, 16
; RV64I-LP64-NEXT:    slliw a1, a1, 16
; RV64I-LP64-NEXT:    call __addsf3@plt
; RV64I-LP64-NEXT:    call __truncsfbf2@plt
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_add:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    slli a1, a1, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa5, a1
; RV32ID-ILP32-NEXT:    slli a0, a0, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa4, a0
; RV32ID-ILP32-NEXT:    fadd.s fa5, fa4, fa5
; RV32ID-ILP32-NEXT:    fmv.x.w a0, fa5
; RV32ID-ILP32-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32-NEXT:    lui a1, 1048560
; RV32ID-ILP32-NEXT:    or a0, a0, a1
; RV32ID-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_add:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    addi sp, sp, -16
; RV64ID-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    lui a2, 16
; RV64ID-LP64-NEXT:    addiw a2, a2, -1
; RV64ID-LP64-NEXT:    and a0, a0, a2
; RV64ID-LP64-NEXT:    and a1, a1, a2
; RV64ID-LP64-NEXT:    slli a1, a1, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa5, a1
; RV64ID-LP64-NEXT:    slli a0, a0, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa4, a0
; RV64ID-LP64-NEXT:    fadd.s fa5, fa4, fa5
; RV64ID-LP64-NEXT:    fmv.x.w a0, fa5
; RV64ID-LP64-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64-NEXT:    lui a1, 1048560
; RV64ID-LP64-NEXT:    or a0, a0, a1
; RV64ID-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    addi sp, sp, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_add:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    addi sp, sp, -16
; RV32ID-ILP32D-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    fmv.x.w a1, fa1
; RV32ID-ILP32D-NEXT:    slli a1, a1, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa5, a1
; RV32ID-ILP32D-NEXT:    slli a0, a0, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa4, a0
; RV32ID-ILP32D-NEXT:    fadd.s fa0, fa4, fa5
; RV32ID-ILP32D-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    lui a1, 1048560
; RV32ID-ILP32D-NEXT:    or a0, a0, a1
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    addi sp, sp, 16
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_add:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    addi sp, sp, -16
; RV64ID-LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 16
; RV64ID-LP64D-NEXT:    addiw a1, a1, -1
; RV64ID-LP64D-NEXT:    and a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.x.w a2, fa1
; RV64ID-LP64D-NEXT:    and a1, a2, a1
; RV64ID-LP64D-NEXT:    slli a1, a1, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa5, a1
; RV64ID-LP64D-NEXT:    slli a0, a0, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa4, a0
; RV64ID-LP64D-NEXT:    fadd.s fa0, fa4, fa5
; RV64ID-LP64D-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 1048560
; RV64ID-LP64D-NEXT:    or a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    addi sp, sp, 16
; RV64ID-LP64D-NEXT:    ret
  %1 = fadd bfloat %a, %b
  ret bfloat %1
}

define bfloat @bfloat_load(ptr %a) nounwind {
; RV32I-ILP32-LABEL: bfloat_load:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    lh a1, 0(a0)
; RV32I-ILP32-NEXT:    lh a2, 6(a0)
; RV32I-ILP32-NEXT:    slli a0, a1, 16
; RV32I-ILP32-NEXT:    slli a1, a2, 16
; RV32I-ILP32-NEXT:    call __addsf3@plt
; RV32I-ILP32-NEXT:    call __truncsfbf2@plt
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_load:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    lh a1, 0(a0)
; RV64I-LP64-NEXT:    lh a2, 6(a0)
; RV64I-LP64-NEXT:    slliw a0, a1, 16
; RV64I-LP64-NEXT:    slliw a1, a2, 16
; RV64I-LP64-NEXT:    call __addsf3@plt
; RV64I-LP64-NEXT:    call __truncsfbf2@plt
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_load:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    lhu a1, 6(a0)
; RV32ID-ILP32-NEXT:    lhu a0, 0(a0)
; RV32ID-ILP32-NEXT:    slli a1, a1, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa5, a1
; RV32ID-ILP32-NEXT:    slli a0, a0, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa4, a0
; RV32ID-ILP32-NEXT:    fadd.s fa5, fa4, fa5
; RV32ID-ILP32-NEXT:    fmv.x.w a0, fa5
; RV32ID-ILP32-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32-NEXT:    lui a1, 1048560
; RV32ID-ILP32-NEXT:    or a0, a0, a1
; RV32ID-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_load:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    addi sp, sp, -16
; RV64ID-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    lhu a1, 6(a0)
; RV64ID-LP64-NEXT:    lhu a0, 0(a0)
; RV64ID-LP64-NEXT:    slli a1, a1, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa5, a1
; RV64ID-LP64-NEXT:    slli a0, a0, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa4, a0
; RV64ID-LP64-NEXT:    fadd.s fa5, fa4, fa5
; RV64ID-LP64-NEXT:    fmv.x.w a0, fa5
; RV64ID-LP64-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64-NEXT:    lui a1, 1048560
; RV64ID-LP64-NEXT:    or a0, a0, a1
; RV64ID-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    addi sp, sp, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_load:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    addi sp, sp, -16
; RV32ID-ILP32D-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    lhu a1, 6(a0)
; RV32ID-ILP32D-NEXT:    lhu a0, 0(a0)
; RV32ID-ILP32D-NEXT:    slli a1, a1, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa5, a1
; RV32ID-ILP32D-NEXT:    slli a0, a0, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa4, a0
; RV32ID-ILP32D-NEXT:    fadd.s fa0, fa4, fa5
; RV32ID-ILP32D-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    lui a1, 1048560
; RV32ID-ILP32D-NEXT:    or a0, a0, a1
; RV32ID-ILP32D-NEXT:    fmv.w.x fa0, a0
; RV32ID-ILP32D-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    addi sp, sp, 16
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_load:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    addi sp, sp, -16
; RV64ID-LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    lhu a1, 6(a0)
; RV64ID-LP64D-NEXT:    lhu a0, 0(a0)
; RV64ID-LP64D-NEXT:    slli a1, a1, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa5, a1
; RV64ID-LP64D-NEXT:    slli a0, a0, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa4, a0
; RV64ID-LP64D-NEXT:    fadd.s fa0, fa4, fa5
; RV64ID-LP64D-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 1048560
; RV64ID-LP64D-NEXT:    or a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.w.x fa0, a0
; RV64ID-LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    addi sp, sp, 16
; RV64ID-LP64D-NEXT:    ret
  %1 = load bfloat, ptr %a
  %2 = getelementptr bfloat, ptr %a, i32 3
  %3 = load bfloat, ptr %2
  %4 = fadd bfloat %1, %3
  ret bfloat %4
}

define void @bfloat_store(ptr %a, bfloat %b, bfloat %c) nounwind {
; RV32I-ILP32-LABEL: bfloat_store:
; RV32I-ILP32:       # %bb.0:
; RV32I-ILP32-NEXT:    addi sp, sp, -16
; RV32I-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
; RV32I-ILP32-NEXT:    mv s0, a0
; RV32I-ILP32-NEXT:    slli a0, a1, 16
; RV32I-ILP32-NEXT:    slli a1, a2, 16
; RV32I-ILP32-NEXT:    call __addsf3@plt
; RV32I-ILP32-NEXT:    call __truncsfbf2@plt
; RV32I-ILP32-NEXT:    sh a0, 0(s0)
; RV32I-ILP32-NEXT:    sh a0, 16(s0)
; RV32I-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
; RV32I-ILP32-NEXT:    addi sp, sp, 16
; RV32I-ILP32-NEXT:    ret
;
; RV64I-LP64-LABEL: bfloat_store:
; RV64I-LP64:       # %bb.0:
; RV64I-LP64-NEXT:    addi sp, sp, -16
; RV64I-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
; RV64I-LP64-NEXT:    mv s0, a0
; RV64I-LP64-NEXT:    slliw a0, a1, 16
; RV64I-LP64-NEXT:    slliw a1, a2, 16
; RV64I-LP64-NEXT:    call __addsf3@plt
; RV64I-LP64-NEXT:    call __truncsfbf2@plt
; RV64I-LP64-NEXT:    sh a0, 0(s0)
; RV64I-LP64-NEXT:    sh a0, 16(s0)
; RV64I-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
; RV64I-LP64-NEXT:    addi sp, sp, 16
; RV64I-LP64-NEXT:    ret
;
; RV32ID-ILP32-LABEL: bfloat_store:
; RV32ID-ILP32:       # %bb.0:
; RV32ID-ILP32-NEXT:    addi sp, sp, -16
; RV32ID-ILP32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
; RV32ID-ILP32-NEXT:    mv s0, a0
; RV32ID-ILP32-NEXT:    slli a2, a2, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa5, a2
; RV32ID-ILP32-NEXT:    slli a1, a1, 16
; RV32ID-ILP32-NEXT:    fmv.w.x fa4, a1
; RV32ID-ILP32-NEXT:    fadd.s fa5, fa4, fa5
; RV32ID-ILP32-NEXT:    fmv.x.w a0, fa5
; RV32ID-ILP32-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32-NEXT:    sh a0, 0(s0)
; RV32ID-ILP32-NEXT:    sh a0, 16(s0)
; RV32ID-ILP32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
; RV32ID-ILP32-NEXT:    addi sp, sp, 16
; RV32ID-ILP32-NEXT:    ret
;
; RV64ID-LP64-LABEL: bfloat_store:
; RV64ID-LP64:       # %bb.0:
; RV64ID-LP64-NEXT:    addi sp, sp, -16
; RV64ID-LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
; RV64ID-LP64-NEXT:    mv s0, a0
; RV64ID-LP64-NEXT:    lui a0, 16
; RV64ID-LP64-NEXT:    addiw a0, a0, -1
; RV64ID-LP64-NEXT:    and a1, a1, a0
; RV64ID-LP64-NEXT:    and a0, a2, a0
; RV64ID-LP64-NEXT:    slli a0, a0, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa5, a0
; RV64ID-LP64-NEXT:    slli a1, a1, 16
; RV64ID-LP64-NEXT:    fmv.w.x fa4, a1
; RV64ID-LP64-NEXT:    fadd.s fa5, fa4, fa5
; RV64ID-LP64-NEXT:    fmv.x.w a0, fa5
; RV64ID-LP64-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64-NEXT:    sh a0, 0(s0)
; RV64ID-LP64-NEXT:    sh a0, 16(s0)
; RV64ID-LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
; RV64ID-LP64-NEXT:    addi sp, sp, 16
; RV64ID-LP64-NEXT:    ret
;
; RV32ID-ILP32D-LABEL: bfloat_store:
; RV32ID-ILP32D:       # %bb.0:
; RV32ID-ILP32D-NEXT:    addi sp, sp, -16
; RV32ID-ILP32D-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
; RV32ID-ILP32D-NEXT:    mv s0, a0
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    fmv.x.w a1, fa1
; RV32ID-ILP32D-NEXT:    slli a1, a1, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa5, a1
; RV32ID-ILP32D-NEXT:    slli a0, a0, 16
; RV32ID-ILP32D-NEXT:    fmv.w.x fa4, a0
; RV32ID-ILP32D-NEXT:    fadd.s fa0, fa4, fa5
; RV32ID-ILP32D-NEXT:    call __truncsfbf2@plt
; RV32ID-ILP32D-NEXT:    fmv.x.w a0, fa0
; RV32ID-ILP32D-NEXT:    sh a0, 0(s0)
; RV32ID-ILP32D-NEXT:    sh a0, 16(s0)
; RV32ID-ILP32D-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
; RV32ID-ILP32D-NEXT:    addi sp, sp, 16
; RV32ID-ILP32D-NEXT:    ret
;
; RV64ID-LP64D-LABEL: bfloat_store:
; RV64ID-LP64D:       # %bb.0:
; RV64ID-LP64D-NEXT:    addi sp, sp, -16
; RV64ID-LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
; RV64ID-LP64D-NEXT:    mv s0, a0
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    lui a1, 16
; RV64ID-LP64D-NEXT:    addiw a1, a1, -1
; RV64ID-LP64D-NEXT:    and a0, a0, a1
; RV64ID-LP64D-NEXT:    fmv.x.w a2, fa1
; RV64ID-LP64D-NEXT:    and a1, a2, a1
; RV64ID-LP64D-NEXT:    slli a1, a1, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa5, a1
; RV64ID-LP64D-NEXT:    slli a0, a0, 16
; RV64ID-LP64D-NEXT:    fmv.w.x fa4, a0
; RV64ID-LP64D-NEXT:    fadd.s fa0, fa4, fa5
; RV64ID-LP64D-NEXT:    call __truncsfbf2@plt
; RV64ID-LP64D-NEXT:    fmv.x.w a0, fa0
; RV64ID-LP64D-NEXT:    sh a0, 0(s0)
; RV64ID-LP64D-NEXT:    sh a0, 16(s0)
; RV64ID-LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
; RV64ID-LP64D-NEXT:    addi sp, sp, 16
; RV64ID-LP64D-NEXT:    ret
  %1 = fadd bfloat %b, %c
  store bfloat %1, ptr %a
  %2 = getelementptr bfloat, ptr %a, i32 8
  store bfloat %1, ptr %2
  ret void
}