llvm/test/CodeGen/PowerPC/masked-sdiv.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc -mtriple=powerpc64le < %s | FileCheck %s

; Legal
define <4 x i32> @sdiv_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m) {
; CHECK-LABEL: sdiv_v4i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    xxleqv 32, 32, 32
; CHECK-NEXT:    vspltisw 5, 1
; CHECK-NEXT:    xxsldwi 1, 34, 34, 1
; CHECK-NEXT:    vslw 4, 4, 0
; CHECK-NEXT:    xxswapd 4, 34
; CHECK-NEXT:    xxsldwi 6, 34, 34, 3
; CHECK-NEXT:    mffprwz 4, 1
; CHECK-NEXT:    vsraw 4, 4, 0
; CHECK-NEXT:    xxsel 0, 37, 35, 36
; CHECK-NEXT:    xxsldwi 2, 0, 0, 1
; CHECK-NEXT:    xxswapd 3, 0
; CHECK-NEXT:    xxsldwi 5, 0, 0, 3
; CHECK-NEXT:    mffprwz 3, 2
; CHECK-NEXT:    mffprwz 5, 3
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mffprwz 4, 4
; CHECK-NEXT:    divw 4, 4, 5
; CHECK-NEXT:    mfvsrwz 5, 34
; CHECK-NEXT:    rldimi 4, 3, 32, 0
; CHECK-NEXT:    mffprwz 3, 5
; CHECK-NEXT:    mtfprd 1, 4
; CHECK-NEXT:    mffprwz 4, 6
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mffprwz 4, 0
; CHECK-NEXT:    divw 4, 5, 4
; CHECK-NEXT:    rldimi 4, 3, 32, 0
; CHECK-NEXT:    mtfprd 0, 4
; CHECK-NEXT:    xxmrghd 34, 0, 1
; CHECK-NEXT:    blr
  %res = call <4 x i32> @llvm.masked.sdiv(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m)
  ret <4 x i32> %res
}

define <2 x i64> @sdiv_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m) {
; CHECK-LABEL: sdiv_v2i64:
; CHECK:       # %bb.0:
; CHECK-NEXT:    xxleqv 32, 32, 32
; CHECK-NEXT:    vspltisw 5, 1
; CHECK-NEXT:    mfvsrd 4, 34
; CHECK-NEXT:    xxswapd 2, 34
; CHECK-NEXT:    vsld 4, 4, 0
; CHECK-NEXT:    vsrad 4, 4, 0
; CHECK-NEXT:    vupklsw 5, 5
; CHECK-NEXT:    xxsel 0, 37, 35, 36
; CHECK-NEXT:    mffprd 3, 0
; CHECK-NEXT:    divd 3, 4, 3
; CHECK-NEXT:    mffprd 4, 2
; CHECK-NEXT:    xxswapd 1, 0
; CHECK-NEXT:    mtfprd 0, 3
; CHECK-NEXT:    mffprd 3, 1
; CHECK-NEXT:    divd 3, 4, 3
; CHECK-NEXT:    mtfprd 1, 3
; CHECK-NEXT:    xxmrghd 34, 0, 1
; CHECK-NEXT:    blr
  %res = call <2 x i64> @llvm.masked.sdiv(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m)
  ret <2 x i64> %res
}

; Splitting
define <4 x i64> @sdiv_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m) {
; CHECK-LABEL: sdiv_v4i64:
; CHECK:       # %bb.0:
; CHECK-NEXT:    xxmrglw 32, 38, 38
; CHECK-NEXT:    xxleqv 39, 39, 39
; CHECK-NEXT:    xxmrghw 33, 38, 38
; CHECK-NEXT:    mfvsrd 3, 34
; CHECK-NEXT:    vspltisw 6, 1
; CHECK-NEXT:    mfvsrd 4, 35
; CHECK-NEXT:    xxswapd 2, 34
; CHECK-NEXT:    xxswapd 4, 35
; CHECK-NEXT:    vsld 0, 0, 7
; CHECK-NEXT:    mffprd 5, 2
; CHECK-NEXT:    vsrad 0, 0, 7
; CHECK-NEXT:    vupklsw 6, 6
; CHECK-NEXT:    xxsel 0, 38, 36, 32
; CHECK-NEXT:    vsld 4, 1, 7
; CHECK-NEXT:    mffprd 6, 0
; CHECK-NEXT:    vsrad 4, 4, 7
; CHECK-NEXT:    divd 3, 3, 6
; CHECK-NEXT:    xxswapd 3, 0
; CHECK-NEXT:    mtfprd 0, 3
; CHECK-NEXT:    xxsel 1, 38, 37, 36
; CHECK-NEXT:    mffprd 6, 1
; CHECK-NEXT:    divd 4, 4, 6
; CHECK-NEXT:    mffprd 6, 3
; CHECK-NEXT:    divd 5, 5, 6
; CHECK-NEXT:    mtfprd 2, 5
; CHECK-NEXT:    xxswapd 5, 1
; CHECK-NEXT:    mtfprd 1, 4
; CHECK-NEXT:    mffprd 3, 5
; CHECK-NEXT:    mffprd 4, 4
; CHECK-NEXT:    divd 3, 4, 3
; CHECK-NEXT:    xxmrghd 34, 0, 2
; CHECK-NEXT:    mtfprd 0, 3
; CHECK-NEXT:    xxmrghd 35, 1, 0
; CHECK-NEXT:    blr
  %res = call <4 x i64> @llvm.masked.sdiv(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m)
  ret <4 x i64> %res
}

; Widening
define <2 x i32> @sdiv_v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m) {
; CHECK-LABEL: sdiv_v2i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    addis 3, 2, .LCPI3_0@toc@ha
; CHECK-NEXT:    xxlxor 32, 32, 32
; CHECK-NEXT:    xxsldwi 1, 34, 34, 1
; CHECK-NEXT:    addi 3, 3, .LCPI3_0@toc@l
; CHECK-NEXT:    mffprwz 4, 1
; CHECK-NEXT:    xxswapd 4, 34
; CHECK-NEXT:    xxsldwi 6, 34, 34, 3
; CHECK-NEXT:    lxvd2x 0, 0, 3
; CHECK-NEXT:    xxswapd 37, 0
; CHECK-NEXT:    vperm 4, 0, 4, 5
; CHECK-NEXT:    xxleqv 32, 32, 32
; CHECK-NEXT:    vspltisw 5, 1
; CHECK-NEXT:    vslw 4, 4, 0
; CHECK-NEXT:    vsraw 4, 4, 0
; CHECK-NEXT:    xxsel 0, 37, 35, 36
; CHECK-NEXT:    xxsldwi 2, 0, 0, 1
; CHECK-NEXT:    xxswapd 3, 0
; CHECK-NEXT:    xxsldwi 5, 0, 0, 3
; CHECK-NEXT:    mffprwz 3, 2
; CHECK-NEXT:    mffprwz 5, 3
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mffprwz 4, 4
; CHECK-NEXT:    divw 4, 4, 5
; CHECK-NEXT:    mfvsrwz 5, 34
; CHECK-NEXT:    rldimi 4, 3, 32, 0
; CHECK-NEXT:    mffprwz 3, 5
; CHECK-NEXT:    mtfprd 1, 4
; CHECK-NEXT:    mffprwz 4, 6
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mffprwz 4, 0
; CHECK-NEXT:    divw 4, 5, 4
; CHECK-NEXT:    rldimi 4, 3, 32, 0
; CHECK-NEXT:    mtfprd 0, 4
; CHECK-NEXT:    xxmrghd 34, 0, 1
; CHECK-NEXT:    blr
  %res = call <2 x i32> @llvm.masked.sdiv(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m)
  ret <2 x i32> %res
}

; Promotion
define <4 x i16> @sdiv_v4i16(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m) {
; CHECK-LABEL: sdiv_v4i16:
; CHECK:       # %bb.0:
; CHECK-NEXT:    xxswapd 0, 36
; CHECK-NEXT:    xxsldwi 1, 36, 36, 1
; CHECK-NEXT:    mfvsrwz 3, 36
; CHECK-NEXT:    li 7, 0
; CHECK-NEXT:    xxsldwi 2, 36, 36, 3
; CHECK-NEXT:    mffprwz 4, 0
; CHECK-NEXT:    mffprwz 5, 1
; CHECK-NEXT:    mffprwz 6, 2
; CHECK-NEXT:    mtvsrd 36, 3
; CHECK-NEXT:    mtvsrd 37, 4
; CHECK-NEXT:    mtvsrd 32, 5
; CHECK-NEXT:    mfvsrd 5, 34
; CHECK-NEXT:    rldicl 8, 5, 48, 48
; CHECK-NEXT:    rldicl 9, 5, 32, 48
; CHECK-NEXT:    extsh 8, 8
; CHECK-NEXT:    extsh 9, 9
; CHECK-NEXT:    vmrghh 5, 0, 5
; CHECK-NEXT:    mtvsrd 32, 6
; CHECK-NEXT:    vmrghh 4, 0, 4
; CHECK-NEXT:    mtvsrd 32, 7
; CHECK-NEXT:    clrldi 7, 5, 48
; CHECK-NEXT:    rldicl 5, 5, 16, 48
; CHECK-NEXT:    extsh 7, 7
; CHECK-NEXT:    extsh 5, 5
; CHECK-NEXT:    xxmrglw 1, 36, 37
; CHECK-NEXT:    vspltish 4, 15
; CHECK-NEXT:    vsplth 0, 0, 3
; CHECK-NEXT:    xxspltw 0, 32, 3
; CHECK-NEXT:    vspltish 0, 1
; CHECK-NEXT:    xxmrgld 37, 0, 1
; CHECK-NEXT:    xxswapd 1, 34
; CHECK-NEXT:    vslh 5, 5, 4
; CHECK-NEXT:    mffprd 3, 1
; CHECK-NEXT:    vsrah 4, 5, 4
; CHECK-NEXT:    clrldi 10, 3, 48
; CHECK-NEXT:    rldicl 11, 3, 48, 48
; CHECK-NEXT:    extsh 10, 10
; CHECK-NEXT:    extsh 11, 11
; CHECK-NEXT:    xxsel 0, 32, 35, 36
; CHECK-NEXT:    mffprd 6, 0
; CHECK-NEXT:    clrldi 12, 6, 48
; CHECK-NEXT:    extsh 12, 12
; CHECK-NEXT:    divw 7, 7, 12
; CHECK-NEXT:    rldicl 12, 6, 48, 48
; CHECK-NEXT:    extsh 12, 12
; CHECK-NEXT:    divw 8, 8, 12
; CHECK-NEXT:    xxswapd 2, 0
; CHECK-NEXT:    mffprd 4, 2
; CHECK-NEXT:    rldicl 12, 6, 32, 48
; CHECK-NEXT:    rldicl 6, 6, 16, 48
; CHECK-NEXT:    extsh 6, 6
; CHECK-NEXT:    extsh 12, 12
; CHECK-NEXT:    divw 5, 5, 6
; CHECK-NEXT:    clrldi 6, 4, 48
; CHECK-NEXT:    divw 9, 9, 12
; CHECK-NEXT:    rldicl 12, 3, 32, 48
; CHECK-NEXT:    rldicl 3, 3, 16, 48
; CHECK-NEXT:    extsh 6, 6
; CHECK-NEXT:    extsh 12, 12
; CHECK-NEXT:    extsh 3, 3
; CHECK-NEXT:    divw 6, 10, 6
; CHECK-NEXT:    rldicl 10, 4, 48, 48
; CHECK-NEXT:    extsh 10, 10
; CHECK-NEXT:    mtvsrd 34, 7
; CHECK-NEXT:    divw 10, 11, 10
; CHECK-NEXT:    rldicl 11, 4, 32, 48
; CHECK-NEXT:    rldicl 4, 4, 16, 48
; CHECK-NEXT:    extsh 11, 11
; CHECK-NEXT:    extsh 4, 4
; CHECK-NEXT:    mtvsrd 35, 8
; CHECK-NEXT:    divw 11, 12, 11
; CHECK-NEXT:    divw 3, 3, 4
; CHECK-NEXT:    mtvsrd 36, 9
; CHECK-NEXT:    mtvsrd 37, 5
; CHECK-NEXT:    mtvsrd 32, 6
; CHECK-NEXT:    vmrghh 2, 3, 2
; CHECK-NEXT:    vmrghh 3, 5, 4
; CHECK-NEXT:    mtvsrd 36, 10
; CHECK-NEXT:    mtvsrd 37, 11
; CHECK-NEXT:    xxmrglw 0, 35, 34
; CHECK-NEXT:    vmrghh 4, 4, 0
; CHECK-NEXT:    mtvsrd 32, 3
; CHECK-NEXT:    vmrghh 5, 0, 5
; CHECK-NEXT:    xxmrglw 1, 37, 36
; CHECK-NEXT:    xxmrgld 34, 0, 1
; CHECK-NEXT:    blr
  %res = call <4 x i16> @llvm.masked.sdiv(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m)
  ret <4 x i16> %res
}

; Scalarization
define <1 x i64> @sdiv_v1i164(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m) {
; CHECK-LABEL: sdiv_v1i164:
; CHECK:       # %bb.0:
; CHECK-NEXT:    andi. 5, 5, 1
; CHECK-NEXT:    li 5, 1
; CHECK-NEXT:    iselgt 4, 4, 5
; CHECK-NEXT:    divd 3, 3, 4
; CHECK-NEXT:    blr
  %res = call <1 x i64> @llvm.masked.sdiv(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m)
  ret <1 x i64> %res
}

; Expansion
define <2 x i128> @sdiv_v2i128(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m) nounwind {
; CHECK-LABEL: sdiv_v2i128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    mfocrf 12, 32
; CHECK-NEXT:    stw 12, 8(1)
; CHECK-NEXT:    mflr 0
; CHECK-NEXT:    stdu 1, -128(1)
; CHECK-NEXT:    li 3, 48
; CHECK-NEXT:    std 0, 144(1)
; CHECK-NEXT:    xxswapd 0, 38
; CHECK-NEXT:    xxswapd 1, 37
; CHECK-NEXT:    std 30, 112(1) # 8-byte Folded Spill
; CHECK-NEXT:    li 30, 1
; CHECK-NEXT:    std 29, 104(1) # 8-byte Folded Spill
; CHECK-NEXT:    li 29, 0
; CHECK-NEXT:    stxvd2x 61, 1, 3 # 16-byte Folded Spill
; CHECK-NEXT:    li 3, 64
; CHECK-NEXT:    mfvsrd 4, 35
; CHECK-NEXT:    stxvd2x 62, 1, 3 # 16-byte Folded Spill
; CHECK-NEXT:    li 3, 80
; CHECK-NEXT:    vmr 30, 2
; CHECK-NEXT:    stxvd2x 63, 1, 3 # 16-byte Folded Spill
; CHECK-NEXT:    mffprd 3, 0
; CHECK-NEXT:    vmr 31, 4
; CHECK-NEXT:    andi. 3, 3, 1
; CHECK-NEXT:    mfvsrd 3, 38
; CHECK-NEXT:    crmove 8, 1
; CHECK-NEXT:    andi. 3, 3, 1
; CHECK-NEXT:    mffprd 3, 1
; CHECK-NEXT:    iselgt 5, 3, 30
; CHECK-NEXT:    mfvsrd 3, 37
; CHECK-NEXT:    xxswapd 0, 35
; CHECK-NEXT:    iselgt 6, 3, 29
; CHECK-NEXT:    mffprd 3, 0
; CHECK-NEXT:    bl __divti3
; CHECK-NEXT:    nop
; CHECK-NEXT:    xxswapd 0, 63
; CHECK-NEXT:    mtfprd 1, 3
; CHECK-NEXT:    mtfprd 2, 4
; CHECK-NEXT:    mfvsrd 4, 62
; CHECK-NEXT:    mffprd 3, 0
; CHECK-NEXT:    isel 5, 3, 30, 8
; CHECK-NEXT:    mfvsrd 3, 63
; CHECK-NEXT:    isel 6, 3, 29, 8
; CHECK-NEXT:    xxswapd 0, 62
; CHECK-NEXT:    mffprd 3, 0
; CHECK-NEXT:    xxmrghd 61, 2, 1
; CHECK-NEXT:    bl __divti3
; CHECK-NEXT:    nop
; CHECK-NEXT:    mtfprd 0, 3
; CHECK-NEXT:    li 3, 80
; CHECK-NEXT:    mtfprd 1, 4
; CHECK-NEXT:    ld 30, 112(1) # 8-byte Folded Reload
; CHECK-NEXT:    vmr 3, 29
; CHECK-NEXT:    ld 29, 104(1) # 8-byte Folded Reload
; CHECK-NEXT:    lxvd2x 63, 1, 3 # 16-byte Folded Reload
; CHECK-NEXT:    li 3, 64
; CHECK-NEXT:    lxvd2x 62, 1, 3 # 16-byte Folded Reload
; CHECK-NEXT:    li 3, 48
; CHECK-NEXT:    lxvd2x 61, 1, 3 # 16-byte Folded Reload
; CHECK-NEXT:    xxmrghd 34, 1, 0
; CHECK-NEXT:    addi 1, 1, 128
; CHECK-NEXT:    ld 0, 16(1)
; CHECK-NEXT:    lwz 12, 8(1)
; CHECK-NEXT:    mtlr 0
; CHECK-NEXT:    mtocrf 32, 12
; CHECK-NEXT:    blr
  %res = call <2 x i128> @llvm.masked.sdiv(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m)
  ret <2 x i128> %res
}

; Promotion and widening
define <3 x i10> @sdiv_v3i10(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m) {
; CHECK-LABEL: sdiv_v3i10:
; CHECK:       # %bb.0:
; CHECK-NEXT:    mtfprwz 0, 9
; CHECK-NEXT:    mtfprwz 1, 10
; CHECK-NEXT:    addis 9, 2, .LCPI7_0@toc@ha
; CHECK-NEXT:    addi 9, 9, .LCPI7_0@toc@l
; CHECK-NEXT:    mtvsrwz 38, 8
; CHECK-NEXT:    vspltisw 4, 11
; CHECK-NEXT:    vadduwm 4, 4, 4
; CHECK-NEXT:    lxvd2x 2, 0, 9
; CHECK-NEXT:    xxmrghw 35, 1, 0
; CHECK-NEXT:    mtfprwz 0, 6
; CHECK-NEXT:    lbz 6, 96(1)
; CHECK-NEXT:    mtfprwz 1, 7
; CHECK-NEXT:    mtvsrwz 32, 6
; CHECK-NEXT:    addis 6, 2, .LCPI7_1@toc@ha
; CHECK-NEXT:    addi 6, 6, .LCPI7_1@toc@l
; CHECK-NEXT:    xxswapd 34, 2
; CHECK-NEXT:    xxmrghw 37, 1, 0
; CHECK-NEXT:    mtfprwz 1, 4
; CHECK-NEXT:    lxvd2x 0, 0, 6
; CHECK-NEXT:    vperm 5, 6, 5, 2
; CHECK-NEXT:    mtvsrwz 38, 5
; CHECK-NEXT:    vslw 5, 5, 4
; CHECK-NEXT:    vsraw 5, 5, 4
; CHECK-NEXT:    vperm 3, 0, 3, 2
; CHECK-NEXT:    xxswapd 32, 0
; CHECK-NEXT:    mtfprwz 0, 3
; CHECK-NEXT:    xxland 35, 35, 32
; CHECK-NEXT:    xxleqv 32, 32, 32
; CHECK-NEXT:    vslw 3, 3, 0
; CHECK-NEXT:    vsraw 3, 3, 0
; CHECK-NEXT:    xxmrghw 33, 1, 0
; CHECK-NEXT:    vperm 1, 6, 1, 2
; CHECK-NEXT:    vspltisw 6, 1
; CHECK-NEXT:    xxsel 0, 38, 37, 35
; CHECK-NEXT:    vslw 3, 1, 4
; CHECK-NEXT:    vsraw 3, 3, 4
; CHECK-NEXT:    xxswapd 1, 0
; CHECK-NEXT:    xxsldwi 3, 0, 0, 1
; CHECK-NEXT:    mffprwz 3, 1
; CHECK-NEXT:    xxswapd 2, 35
; CHECK-NEXT:    xxsldwi 4, 35, 35, 1
; CHECK-NEXT:    mffprwz 4, 2
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mffprwz 4, 3
; CHECK-NEXT:    mtfprwz 1, 3
; CHECK-NEXT:    mffprwz 3, 4
; CHECK-NEXT:    divw 3, 3, 4
; CHECK-NEXT:    mfvsrwz 4, 35
; CHECK-NEXT:    mtfprwz 2, 3
; CHECK-NEXT:    mffprwz 3, 0
; CHECK-NEXT:    divw 3, 4, 3
; CHECK-NEXT:    mtvsrwz 35, 3
; CHECK-NEXT:    xxmrghw 36, 2, 1
; CHECK-NEXT:    vperm 2, 3, 4, 2
; CHECK-NEXT:    mfvsrwz 5, 34
; CHECK-NEXT:    xxswapd 0, 34
; CHECK-NEXT:    xxsldwi 1, 34, 34, 1
; CHECK-NEXT:    mffprwz 3, 0
; CHECK-NEXT:    mffprwz 4, 1
; CHECK-NEXT:    blr
  %res = call <3 x i10> @llvm.masked.sdiv(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m)
  ret <3 x i10> %res
}