path: root/llvm/test/Transforms/InstCombine/fsh.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S | FileCheck %s

declare i16 @llvm.fshl.i16(i16, i16, i16)
declare i16 @llvm.fshr.i16(i16, i16, i16)
declare i32 @llvm.fshl.i32(i32, i32, i32)
declare i33 @llvm.fshr.i33(i33, i33, i33)
declare <2 x i32> @llvm.fshr.v2i32(<2 x i32>, <2 x i32>, <2 x i32>)
declare <2 x i31> @llvm.fshl.v2i31(<2 x i31>, <2 x i31>, <2 x i31>)
declare <3 x i16> @llvm.fshl.v3i16(<3 x i16>, <3 x i16>, <3 x i16>)

declare void @use_v2(<2 x i31>)
declare void @use_v3(<3 x i16>)

; If the shift mask doesn't include any demanded bits, the funnel shift can be eliminated.

define i32 @fshl_mask_simplify1(i32 %x, i32 %y, i32 %sh) {
; CHECK-LABEL: @fshl_mask_simplify1(
; CHECK-NEXT:    ret i32 [[X:%.*]]
;
  %maskedsh = and i32 %sh, 32
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 %maskedsh)
  ret i32 %r
}

define <2 x i32> @fshr_mask_simplify2(<2 x i32> %x, <2 x i32> %y, <2 x i32> %sh) {
; CHECK-LABEL: @fshr_mask_simplify2(
; CHECK-NEXT:    ret <2 x i32> [[Y:%.*]]
;
  %maskedsh = and <2 x i32> %sh, <i32 64, i32 64>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> %maskedsh)
  ret <2 x i32> %r
}

; Negative test.

define i32 @fshl_mask_simplify3(i32 %x, i32 %y, i32 %sh) {
; CHECK-LABEL: @fshl_mask_simplify3(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and i32 [[SH:%.*]], 16
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 [[MASKEDSH]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %maskedsh = and i32 %sh, 16
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 %maskedsh)
  ret i32 %r
}

; Check again with weird bitwidths - the analysis is invalid with non-power-of-2.

define i33 @fshr_mask_simplify1(i33 %x, i33 %y, i33 %sh) {
; CHECK-LABEL: @fshr_mask_simplify1(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and i33 [[SH:%.*]], 64
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshr.i33(i33 [[X:%.*]], i33 [[Y:%.*]], i33 [[MASKEDSH]])
; CHECK-NEXT:    ret i33 [[R]]
;
  %maskedsh = and i33 %sh, 64
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 %maskedsh)
  ret i33 %r
}

; Check again with weird bitwidths - the analysis is invalid with non-power-of-2.

define <2 x i31> @fshl_mask_simplify2(<2 x i31> %x, <2 x i31> %y, <2 x i31> %sh) {
; CHECK-LABEL: @fshl_mask_simplify2(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and <2 x i31> [[SH:%.*]], splat (i31 32)
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> [[Y:%.*]], <2 x i31> [[MASKEDSH]])
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %maskedsh = and <2 x i31> %sh, <i31 32, i31 32>
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> %maskedsh)
  ret <2 x i31> %r
}

; Check again with weird bitwidths - the analysis is invalid with non-power-of-2.

define i33 @fshr_mask_simplify3(i33 %x, i33 %y, i33 %sh) {
; CHECK-LABEL: @fshr_mask_simplify3(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and i33 [[SH:%.*]], 32
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshr.i33(i33 [[X:%.*]], i33 [[Y:%.*]], i33 [[MASKEDSH]])
; CHECK-NEXT:    ret i33 [[R]]
;
  %maskedsh = and i33 %sh, 32
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 %maskedsh)
  ret i33 %r
}

; This mask op is unnecessary.

define i32 @fshl_mask_not_required(i32 %x, i32 %y, i32 %sh) {
; CHECK-LABEL: @fshl_mask_not_required(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 [[SH:%.*]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %maskedsh = and i32 %sh, 31
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 %maskedsh)
  ret i32 %r
}

; This mask op can be reduced.

define i32 @fshl_mask_reduce_constant(i32 %x, i32 %y, i32 %sh) {
; CHECK-LABEL: @fshl_mask_reduce_constant(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and i32 [[SH:%.*]], 1
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 [[MASKEDSH]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %maskedsh = and i32 %sh, 33
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 %maskedsh)
  ret i32 %r
}

; But this mask op is required.

define i32 @fshl_mask_negative(i32 %x, i32 %y, i32 %sh) {
; CHECK-LABEL: @fshl_mask_negative(
; CHECK-NEXT:    [[MASKEDSH:%.*]] = and i32 [[SH:%.*]], 15
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 [[MASKEDSH]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %maskedsh = and i32 %sh, 15
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 %maskedsh)
  ret i32 %r
}

; The transform is not limited to mask ops.

define <2 x i32> @fshr_set_but_not_demanded_vec(<2 x i32> %x, <2 x i32> %y, <2 x i32> %sh) {
; CHECK-LABEL: @fshr_set_but_not_demanded_vec(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[Y:%.*]], <2 x i32> [[SH:%.*]])
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %bogusbits = or <2 x i32> %sh, <i32 32, i32 32>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> %bogusbits)
  ret <2 x i32> %r
}

; Check again with weird bitwidths - the analysis is invalid with non-power-of-2.

define <2 x i31> @fshl_set_but_not_demanded_vec(<2 x i31> %x, <2 x i31> %y, <2 x i31> %sh) {
; CHECK-LABEL: @fshl_set_but_not_demanded_vec(
; CHECK-NEXT:    [[BOGUSBITS:%.*]] = or <2 x i31> [[SH:%.*]], splat (i31 32)
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> [[Y:%.*]], <2 x i31> [[BOGUSBITS]])
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %bogusbits = or <2 x i31> %sh, <i31 32, i31 32>
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> %bogusbits)
  ret <2 x i31> %r
}

; Simplify one undef or zero operand and constant shift amount.

define i32 @fshl_op0_undef(i32 %x) {
; CHECK-LABEL: @fshl_op0_undef(
; CHECK-NEXT:    [[R:%.*]] = lshr i32 [[X:%.*]], 25
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 undef, i32 %x, i32 7)
  ret i32 %r
}

define i32 @fshl_op0_zero(i32 %x) {
; CHECK-LABEL: @fshl_op0_zero(
; CHECK-NEXT:    [[R:%.*]] = lshr i32 [[X:%.*]], 25
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 0, i32 %x, i32 7)
  ret i32 %r
}

define i33 @fshr_op0_undef(i33 %x) {
; CHECK-LABEL: @fshr_op0_undef(
; CHECK-NEXT:    [[R:%.*]] = lshr i33 [[X:%.*]], 7
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 undef, i33 %x, i33 7)
  ret i33 %r
}

define i33 @fshr_op0_zero(i33 %x) {
; CHECK-LABEL: @fshr_op0_zero(
; CHECK-NEXT:    [[R:%.*]] = lshr i33 [[X:%.*]], 7
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 0, i33 %x, i33 7)
  ret i33 %r
}

define i32 @fshl_op1_undef(i32 %x) {
; CHECK-LABEL: @fshl_op1_undef(
; CHECK-NEXT:    [[R:%.*]] = shl i32 [[X:%.*]], 7
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 undef, i32 7)
  ret i32 %r
}

define i32 @fshl_op1_zero(i32 %x) {
; CHECK-LABEL: @fshl_op1_zero(
; CHECK-NEXT:    [[R:%.*]] = shl i32 [[X:%.*]], 7
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 0, i32 7)
  ret i32 %r
}

define i33 @fshr_op1_undef(i33 %x) {
; CHECK-LABEL: @fshr_op1_undef(
; CHECK-NEXT:    [[R:%.*]] = shl i33 [[X:%.*]], 26
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 undef, i33 7)
  ret i33 %r
}

define i33 @fshr_op1_zero(i33 %x) {
; CHECK-LABEL: @fshr_op1_zero(
; CHECK-NEXT:    [[R:%.*]] = shl i33 [[X:%.*]], 26
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 0, i33 7)
  ret i33 %r
}

define <2 x i31> @fshl_op0_zero_splat_vec(<2 x i31> %x) {
; CHECK-LABEL: @fshl_op0_zero_splat_vec(
; CHECK-NEXT:    [[R:%.*]] = lshr <2 x i31> [[X:%.*]], splat (i31 24)
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> zeroinitializer, <2 x i31> %x, <2 x i31> <i31 7, i31 7>)
  ret <2 x i31> %r
}

define <2 x i31> @fshl_op1_undef_splat_vec(<2 x i31> %x) {
; CHECK-LABEL: @fshl_op1_undef_splat_vec(
; CHECK-NEXT:    [[R:%.*]] = shl <2 x i31> [[X:%.*]], splat (i31 7)
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> undef, <2 x i31> <i31 7, i31 7>)
  ret <2 x i31> %r
}

define <2 x i32> @fshr_op0_undef_splat_vec(<2 x i32> %x) {
; CHECK-LABEL: @fshr_op0_undef_splat_vec(
; CHECK-NEXT:    [[R:%.*]] = lshr <2 x i32> [[X:%.*]], splat (i32 7)
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> undef, <2 x i32> %x, <2 x i32> <i32 7, i32 7>)
  ret <2 x i32> %r
}

define <2 x i32> @fshr_op1_zero_splat_vec(<2 x i32> %x) {
; CHECK-LABEL: @fshr_op1_zero_splat_vec(
; CHECK-NEXT:    [[R:%.*]] = shl <2 x i32> [[X:%.*]], splat (i32 25)
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> zeroinitializer, <2 x i32> <i32 7, i32 7>)
  ret <2 x i32> %r
}

define <2 x i31> @fshl_op0_zero_vec(<2 x i31> %x) {
; CHECK-LABEL: @fshl_op0_zero_vec(
; CHECK-NEXT:    [[R:%.*]] = lshr <2 x i31> [[X:%.*]], <i31 30, i31 29>
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> zeroinitializer, <2 x i31> %x, <2 x i31> <i31 -1, i31 33>)
  ret <2 x i31> %r
}

define <2 x i31> @fshl_op1_undef_vec(<2 x i31> %x) {
; CHECK-LABEL: @fshl_op1_undef_vec(
; CHECK-NEXT:    [[R:%.*]] = shl <2 x i31> [[X:%.*]], <i31 1, i31 2>
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> undef, <2 x i31> <i31 -1, i31 33>)
  ret <2 x i31> %r
}

define <2 x i32> @fshr_op0_undef_vec(<2 x i32> %x) {
; CHECK-LABEL: @fshr_op0_undef_vec(
; CHECK-NEXT:    [[R:%.*]] = lshr <2 x i32> [[X:%.*]], <i32 31, i32 1>
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> undef, <2 x i32> %x, <2 x i32> <i32 -1, i32 33>)
  ret <2 x i32> %r
}

define <2 x i32> @fshr_op1_zero_vec(<2 x i32> %x) {
; CHECK-LABEL: @fshr_op1_zero_vec(
; CHECK-NEXT:    [[R:%.*]] = shl <2 x i32> [[X:%.*]], <i32 1, i32 31>
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> zeroinitializer, <2 x i32> <i32 -1, i32 33>)
  ret <2 x i32> %r
}

; Only demand bits from one of the operands.

define i32 @fshl_only_op0_demanded(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_only_op0_demanded(
; CHECK-NEXT:    [[Z:%.*]] = shl i32 [[X:%.*]], 7
; CHECK-NEXT:    [[R:%.*]] = and i32 [[Z]], 128
; CHECK-NEXT:    ret i32 [[R]]
;
  %z = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 7)
  %r = and i32 %z, 128
  ret i32 %r
}

define i32 @fshl_only_op1_demanded(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_only_op1_demanded(
; CHECK-NEXT:    [[Z:%.*]] = lshr i32 [[Y:%.*]], 25
; CHECK-NEXT:    [[R:%.*]] = and i32 [[Z]], 63
; CHECK-NEXT:    ret i32 [[R]]
;
  %z = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 7)
  %r = and i32 %z, 63
  ret i32 %r
}

define i33 @fshr_only_op1_demanded(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_only_op1_demanded(
; CHECK-NEXT:    [[Z:%.*]] = lshr i33 [[Y:%.*]], 7
; CHECK-NEXT:    [[R:%.*]] = and i33 [[Z]], 12392
; CHECK-NEXT:    ret i33 [[R]]
;
  %z = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 7)
  %r = and i33 %z, 12392
  ret i33 %r
}

define i33 @fshr_only_op0_demanded(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_only_op0_demanded(
; CHECK-NEXT:    [[TMP1:%.*]] = lshr i33 [[X:%.*]], 4
; CHECK-NEXT:    [[R:%.*]] = and i33 [[TMP1]], 7
; CHECK-NEXT:    ret i33 [[R]]
;
  %z = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 7)
  %r = lshr i33 %z, 30
  ret i33 %r
}

define <2 x i31> @fshl_only_op1_demanded_vec_splat(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_only_op1_demanded_vec_splat(
; CHECK-NEXT:    [[Z:%.*]] = lshr <2 x i31> [[Y:%.*]], splat (i31 24)
; CHECK-NEXT:    [[R:%.*]] = and <2 x i31> [[Z]], <i31 63, i31 31>
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %z = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> <i31 7, i31 7>)
  %r = and <2 x i31> %z, <i31 63, i31 31>
  ret <2 x i31> %r
}

define i32 @fshl_constant_shift_amount_modulo_bitwidth(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_constant_shift_amount_modulo_bitwidth(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 1)
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 33)
  ret i32 %r
}

define i33 @fshr_constant_shift_amount_modulo_bitwidth(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_constant_shift_amount_modulo_bitwidth(
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshl.i33(i33 [[X:%.*]], i33 [[Y:%.*]], i33 32)
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 34)
  ret i33 %r
}

define i32 @fshl_undef_shift_amount(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_undef_shift_amount(
; CHECK-NEXT:    ret i32 [[X:%.*]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 undef)
  ret i32 %r
}

define i33 @fshr_undef_shift_amount(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_undef_shift_amount(
; CHECK-NEXT:    ret i33 [[Y:%.*]]
;
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 undef)
  ret i33 %r
}

@external_global = external global i8

define i33 @fshr_constant_shift_amount_modulo_bitwidth_constexpr(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_constant_shift_amount_modulo_bitwidth_constexpr(
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshr.i33(i33 [[X:%.*]], i33 [[Y:%.*]], i33 ptrtoint (ptr @external_global to i33))
; CHECK-NEXT:    ret i33 [[R]]
;
  %shamt = ptrtoint ptr @external_global to i33
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 %shamt)
  ret i33 %r
}

define <2 x i32> @fshr_constant_shift_amount_modulo_bitwidth_vec(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @fshr_constant_shift_amount_modulo_bitwidth_vec(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[Y:%.*]], <2 x i32> <i32 30, i32 1>)
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> <i32 34, i32 -1>)
  ret <2 x i32> %r
}

define <2 x i31> @fshl_constant_shift_amount_modulo_bitwidth_vec(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_constant_shift_amount_modulo_bitwidth_vec(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> [[Y:%.*]], <2 x i31> <i31 3, i31 1>)
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> <i31 34, i31 -1>)
  ret <2 x i31> %r
}

define <2 x i31> @fshl_constant_shift_amount_modulo_bitwidth_vec_const_expr(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_constant_shift_amount_modulo_bitwidth_vec_const_expr(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> [[Y:%.*]], <2 x i31> <i31 34, i31 ptrtoint (ptr @external_global to i31)>)
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %shamt = ptrtoint ptr @external_global to i31
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> <i31 34, i31 ptrtoint (ptr @external_global to i31)>)
  ret <2 x i31> %r
}

define <2 x i31> @fshl_undef_shift_amount_vec(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_undef_shift_amount_vec(
; CHECK-NEXT:    ret <2 x i31> [[X:%.*]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> undef)
  ret <2 x i31> %r
}

define <2 x i32> @fshr_undef_shift_amount_vec(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @fshr_undef_shift_amount_vec(
; CHECK-NEXT:    ret <2 x i32> [[Y:%.*]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> undef)
  ret <2 x i32> %r
}

define i32 @rotl_common_demanded(i32 %a0) {
; CHECK-LABEL: @rotl_common_demanded(
; CHECK-NEXT:    [[X:%.*]] = xor i32 [[A0:%.*]], 2
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X]], i32 [[X]], i32 8)
; CHECK-NEXT:    ret i32 [[R]]
;
  %x = xor i32 %a0, 2
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 8)
  ret i32 %r
}

define i33 @rotr_common_demanded(i33 %a0) {
; CHECK-LABEL: @rotr_common_demanded(
; CHECK-NEXT:    [[X:%.*]] = xor i33 [[A0:%.*]], 2
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshl.i33(i33 [[X]], i33 [[X]], i33 25)
; CHECK-NEXT:    ret i33 [[R]]
;
  %x = xor i33 %a0, 2
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %x, i33 8)
  ret i33 %r
}

; The shift modulo bitwidth is the same for all vector elements.

define <2 x i31> @fshl_only_op1_demanded_vec_nonsplat(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_only_op1_demanded_vec_nonsplat(
; CHECK-NEXT:    [[Z:%.*]] = lshr <2 x i31> [[Y:%.*]], splat (i31 24)
; CHECK-NEXT:    [[R:%.*]] = and <2 x i31> [[Z]], <i31 63, i31 31>
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %z = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %y, <2 x i31> <i31 7, i31 38>)
  %r = and <2 x i31> %z, <i31 63, i31 31>
  ret <2 x i31> %r
}

define i32 @rotl_constant_shift_amount(i32 %x) {
; CHECK-LABEL: @rotl_constant_shift_amount(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 1)
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 33)
  ret i32 %r
}

define <2 x i31> @rotl_constant_shift_amount_vec(<2 x i31> %x) {
; CHECK-LABEL: @rotl_constant_shift_amount_vec(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> [[X]], <2 x i31> splat (i31 1))
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> %x, <2 x i31> <i31 32, i31 -1>)
  ret <2 x i31> %r
}

define i33 @rotr_constant_shift_amount(i33 %x) {
; CHECK-LABEL: @rotr_constant_shift_amount(
; CHECK-NEXT:    [[R:%.*]] = call i33 @llvm.fshl.i33(i33 [[X:%.*]], i33 [[X]], i33 32)
; CHECK-NEXT:    ret i33 [[R]]
;
  %r = call i33 @llvm.fshr.i33(i33 %x, i33 %x, i33 34)
  ret i33 %r
}

define <2 x i32> @rotr_constant_shift_amount_vec(<2 x i32> %x) {
; CHECK-LABEL: @rotr_constant_shift_amount_vec(
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 31, i32 1>)
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %x, <2 x i32> <i32 33, i32 -1>)
  ret <2 x i32> %r
}

; Demand bits from both operands -- cannot simplify.

define i32 @fshl_both_ops_demanded(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_both_ops_demanded(
; CHECK-NEXT:    [[Z:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 7)
; CHECK-NEXT:    [[R:%.*]] = and i32 [[Z]], 192
; CHECK-NEXT:    ret i32 [[R]]
;
  %z = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 7)
  %r = and i32 %z, 192
  ret i32 %r
}

define i33 @fshr_both_ops_demanded(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_both_ops_demanded(
; CHECK-NEXT:    [[Z:%.*]] = call i33 @llvm.fshl.i33(i33 [[X:%.*]], i33 [[Y:%.*]], i33 7)
; CHECK-NEXT:    [[R:%.*]] = and i33 [[Z]], 192
; CHECK-NEXT:    ret i33 [[R]]
;
  %z = call i33 @llvm.fshr.i33(i33 %x, i33 %y, i33 26)
  %r = and i33 %z, 192
  ret i33 %r
}

; Both operands are demanded, but there are known bits.

define i32 @fshl_known_bits(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_known_bits(
; CHECK-NEXT:    ret i32 128
;
  %x2 = or i32 %x, 1   ; lo bit set
  %y2 = lshr i32 %y, 1 ; hi bit clear
  %z = call i32 @llvm.fshl.i32(i32 %x2, i32 %y2, i32 7)
  %r = and i32 %z, 192
  ret i32 %r
}

define i33 @fshr_known_bits(i33 %x, i33 %y) {
; CHECK-LABEL: @fshr_known_bits(
; CHECK-NEXT:    ret i33 128
;
  %x2 = or i33 %x, 1 ; lo bit set
  %y2 = lshr i33 %y, 1 ; hi bit set
  %z = call i33 @llvm.fshr.i33(i33 %x2, i33 %y2, i33 26)
  %r = and i33 %z, 192
  ret i33 %r
}

; This case fails to simplify due to multiple uses.

define i33 @fshr_multi_use(i33 %a) {
; CHECK-LABEL: @fshr_multi_use(
; CHECK-NEXT:    [[B:%.*]] = call i33 @llvm.fshl.i33(i33 [[A:%.*]], i33 [[A]], i33 32)
; CHECK-NEXT:    [[C:%.*]] = lshr i33 [[B]], 23
; CHECK-NEXT:    [[D:%.*]] = xor i33 [[C]], [[B]]
; CHECK-NEXT:    [[E:%.*]] = and i33 [[D]], 31
; CHECK-NEXT:    ret i33 [[E]]
;
  %b = tail call i33 @llvm.fshr.i33(i33 %a, i33 %a, i33 1)
  %c = lshr i33 %b, 23
  %d = xor i33 %c, %b
  %e = and i33 %d, 31
  ret i33 %e
}

; This demonstrates the same simplification working if the fshr intrinsic
; is expanded into shifts and or.

define i33 @expanded_fshr_multi_use(i33 %a) {
; CHECK-LABEL: @expanded_fshr_multi_use(
; CHECK-NEXT:    [[B:%.*]] = call i33 @llvm.fshl.i33(i33 [[A:%.*]], i33 [[A]], i33 32)
; CHECK-NEXT:    [[C:%.*]] = lshr i33 [[B]], 23
; CHECK-NEXT:    [[D:%.*]] = xor i33 [[C]], [[B]]
; CHECK-NEXT:    [[E:%.*]] = and i33 [[D]], 31
; CHECK-NEXT:    ret i33 [[E]]
;
  %t = lshr i33 %a, 1
  %t2 = shl i33 %a, 32
  %b = or i33 %t, %t2
  %c = lshr i33 %b, 23
  %d = xor i33 %c, %b
  %e = and i33 %d, 31
  ret i33 %e
}

; Special-case: rotate a 16-bit value left/right by 8-bits is bswap.

define i16 @fshl_bswap(i16 %x) {
; CHECK-LABEL: @fshl_bswap(
; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.bswap.i16(i16 [[X:%.*]])
; CHECK-NEXT:    ret i16 [[R]]
;
  %r = call i16 @llvm.fshl.i16(i16 %x, i16 %x, i16 8)
  ret i16 %r
}

define i16 @fshr_bswap(i16 %x) {
; CHECK-LABEL: @fshr_bswap(
; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.bswap.i16(i16 [[X:%.*]])
; CHECK-NEXT:    ret i16 [[R]]
;
  %r = call i16 @llvm.fshr.i16(i16 %x, i16 %x, i16 8)
  ret i16 %r
}

define <3 x i16> @fshl_bswap_vector(<3 x i16> %x) {
; CHECK-LABEL: @fshl_bswap_vector(
; CHECK-NEXT:    [[R:%.*]] = call <3 x i16> @llvm.bswap.v3i16(<3 x i16> [[X:%.*]])
; CHECK-NEXT:    ret <3 x i16> [[R]]
;
  %r = call <3 x i16> @llvm.fshl.v3i16(<3 x i16> %x, <3 x i16> %x, <3 x i16> <i16 8, i16 8, i16 8>)
  ret <3 x i16> %r
}

; Negative test

define i16 @fshl_bswap_wrong_op(i16 %x, i16 %y) {
; CHECK-LABEL: @fshl_bswap_wrong_op(
; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[Y:%.*]], i16 8)
; CHECK-NEXT:    ret i16 [[R]]
;
  %r = call i16 @llvm.fshl.i16(i16 %x, i16 %y, i16 8)
  ret i16 %r
}

; Negative test

define i16 @fshr_bswap_wrong_amount(i16 %x) {
; CHECK-LABEL: @fshr_bswap_wrong_amount(
; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 12)
; CHECK-NEXT:    ret i16 [[R]]
;
  %r = call i16 @llvm.fshr.i16(i16 %x, i16 %x, i16 4)
  ret i16 %r
}

; Negative test

define i32 @fshl_bswap_wrong_width(i32 %x) {
; CHECK-LABEL: @fshl_bswap_wrong_width(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 8)
; CHECK-NEXT:    ret i32 [[R]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 8)
  ret i32 %r
}

define i32 @fshl_mask_args_same1(i32 %a) {
; CHECK-LABEL: @fshl_mask_args_same1(
; CHECK-NEXT:    [[T2:%.*]] = lshr i32 [[A:%.*]], 16
; CHECK-NEXT:    ret i32 [[T2]]
;
  %t1 = and i32 %a, 4294901760 ; 0xffff0000
  %t2 = call i32 @llvm.fshl.i32(i32 %t1, i32 %t1, i32 16)
  ret i32 %t2
}

define i32 @fshl_mask_args_same2(i32 %a) {
; CHECK-LABEL: @fshl_mask_args_same2(
; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i32 [[A:%.*]] to i16
; CHECK-NEXT:    [[REV:%.*]] = shl i16 [[TRUNC]], 8
; CHECK-NEXT:    [[T2:%.*]] = zext i16 [[REV]] to i32
; CHECK-NEXT:    ret i32 [[T2]]
;
  %t1 = and i32 %a, 255
  %t2 = call i32 @llvm.fshl.i32(i32 %t1, i32 %t1, i32 8)
  ret i32 %t2
}

define i32 @fshl_mask_args_same3(i32 %a) {
; CHECK-LABEL: @fshl_mask_args_same3(
; CHECK-NEXT:    [[REV:%.*]] = shl i32 [[A:%.*]], 24
; CHECK-NEXT:    ret i32 [[REV]]
;
  %t1 = and i32 %a, 255
  %t2 = call i32 @llvm.fshl.i32(i32 %t1, i32 %t1, i32 24)
  ret i32 %t2
}

define i32 @fshl_mask_args_different(i32 %a) {
; CHECK-LABEL: @fshl_mask_args_different(
; CHECK-NEXT:    [[T1:%.*]] = lshr i32 [[A:%.*]], 15
; CHECK-NEXT:    [[T3:%.*]] = and i32 [[T1]], 130560
; CHECK-NEXT:    ret i32 [[T3]]
;
  %t2 = and i32 %a, 4294901760 ; 0xfffff00f
  %t1 = and i32 %a, 4278190080 ; 0xff00f00f
  %t3 = call i32 @llvm.fshl.i32(i32 %t2, i32 %t1, i32 17)
  ret i32 %t3
}

define i32 @fsh_andconst_rotate(i32 %a) {
; CHECK-LABEL: @fsh_andconst_rotate(
; CHECK-NEXT:    [[T2:%.*]] = lshr i32 [[A:%.*]], 16
; CHECK-NEXT:    ret i32 [[T2]]
;
  %t1 = and i32 %a, 4294901760 ; 0xffff0000
  %t2 = call i32 @llvm.fshl.i32(i32 %t1, i32 %t1, i32 16)
  ret i32 %t2
}

define i32 @fsh_orconst_rotate(i32 %a) {
; CHECK-LABEL: @fsh_orconst_rotate(
; CHECK-NEXT:    [[T2:%.*]] = call i32 @llvm.fshl.i32(i32 [[A:%.*]], i32 -268435456, i32 4)
; CHECK-NEXT:    ret i32 [[T2]]
;
  %t1 = or i32 %a, 4026531840 ; 0xf0000000
  %t2 = call i32 @llvm.fshl.i32(i32 %t1, i32 %t1, i32 4)
  ret i32 %t2
}

define i32 @fsh_rotate_5(i8 %x, i32 %y) {
; CHECK-LABEL: @fsh_rotate_5(
; CHECK-NEXT:    [[T1:%.*]] = zext i8 [[X:%.*]] to i32
; CHECK-NEXT:    [[OR1:%.*]] = or i32 [[Y:%.*]], [[T1]]
; CHECK-NEXT:    [[OR2:%.*]] = call i32 @llvm.fshl.i32(i32 [[OR1]], i32 [[OR1]], i32 5)
; CHECK-NEXT:    ret i32 [[OR2]]
;

  %t1 = zext i8 %x to i32
  %or1 = or i32 %t1, %y
  %shr = lshr i32 %or1, 27
  %shl = shl i32 %or1, 5
  %or2 = or i32 %shr, %shl
  ret i32 %or2
}

define i32 @fsh_rotate_18(i8 %x, i32 %y) {
; CHECK-LABEL: @fsh_rotate_18(
; CHECK-NEXT:    [[T1:%.*]] = zext i8 [[X:%.*]] to i32
; CHECK-NEXT:    [[OR1:%.*]] = or i32 [[Y:%.*]], [[T1]]
; CHECK-NEXT:    [[OR2:%.*]] = call i32 @llvm.fshl.i32(i32 [[OR1]], i32 [[OR1]], i32 18)
; CHECK-NEXT:    ret i32 [[OR2]]
;

  %t1 = zext i8 %x to i32
  %or1 = or i32 %t1, %y
  %shr = lshr i32 %or1, 14
  %shl = shl i32 %or1, 18
  %or2 = or i32 %shr, %shl
  ret i32 %or2
}

define i32 @fsh_load_rotate_12(ptr %data) {
; CHECK-LABEL: @fsh_load_rotate_12(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[DATA:%.*]], align 1
; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP0]] to i32
; CHECK-NEXT:    [[SHL:%.*]] = shl nuw i32 [[CONV]], 24
; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 1
; CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
; CHECK-NEXT:    [[CONV2:%.*]] = zext i8 [[TMP1]] to i32
; CHECK-NEXT:    [[SHL3:%.*]] = shl nuw nsw i32 [[CONV2]], 16
; CHECK-NEXT:    [[OR:%.*]] = or disjoint i32 [[SHL3]], [[SHL]]
; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 2
; CHECK-NEXT:    [[TMP2:%.*]] = load i8, ptr [[ARRAYIDX4]], align 1
; CHECK-NEXT:    [[CONV5:%.*]] = zext i8 [[TMP2]] to i32
; CHECK-NEXT:    [[SHL6:%.*]] = shl nuw nsw i32 [[CONV5]], 8
; CHECK-NEXT:    [[OR7:%.*]] = or disjoint i32 [[OR]], [[SHL6]]
; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 3
; CHECK-NEXT:    [[TMP3:%.*]] = load i8, ptr [[ARRAYIDX8]], align 1
; CHECK-NEXT:    [[CONV9:%.*]] = zext i8 [[TMP3]] to i32
; CHECK-NEXT:    [[OR10:%.*]] = or disjoint i32 [[OR7]], [[CONV9]]
; CHECK-NEXT:    [[OR15:%.*]] = call i32 @llvm.fshl.i32(i32 [[OR10]], i32 [[OR10]], i32 12)
; CHECK-NEXT:    ret i32 [[OR15]]
;

entry:
  %0 = load i8, ptr %data
  %conv = zext i8 %0 to i32
  %shl = shl nuw i32 %conv, 24
  %arrayidx1 = getelementptr inbounds i8, ptr %data, i64 1
  %1 = load i8, ptr %arrayidx1
  %conv2 = zext i8 %1 to i32
  %shl3 = shl nuw nsw i32 %conv2, 16
  %or = or i32 %shl3, %shl
  %arrayidx4 = getelementptr inbounds i8, ptr %data, i64 2
  %2 = load i8, ptr %arrayidx4
  %conv5 = zext i8 %2 to i32
  %shl6 = shl nuw nsw i32 %conv5, 8
  %or7 = or i32 %or, %shl6
  %arrayidx8 = getelementptr inbounds i8, ptr %data, i64 3
  %3 = load i8, ptr %arrayidx8
  %conv9 = zext i8 %3 to i32
  %or10 = or i32 %or7, %conv9
  %shr = lshr i32 %or10, 20
  %shl7 = shl i32 %or10, 12
  %or15 = or i32 %shr, %shl7
  ret i32 %or15
}

define i32 @fsh_load_rotate_25(ptr %data) {
; CHECK-LABEL: @fsh_load_rotate_25(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[DATA:%.*]], align 1
; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP0]] to i32
; CHECK-NEXT:    [[SHL:%.*]] = shl nuw i32 [[CONV]], 24
; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 1
; CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
; CHECK-NEXT:    [[CONV2:%.*]] = zext i8 [[TMP1]] to i32
; CHECK-NEXT:    [[SHL3:%.*]] = shl nuw nsw i32 [[CONV2]], 16
; CHECK-NEXT:    [[OR:%.*]] = or disjoint i32 [[SHL3]], [[SHL]]
; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 2
; CHECK-NEXT:    [[TMP2:%.*]] = load i8, ptr [[ARRAYIDX4]], align 1
; CHECK-NEXT:    [[CONV5:%.*]] = zext i8 [[TMP2]] to i32
; CHECK-NEXT:    [[SHL6:%.*]] = shl nuw nsw i32 [[CONV5]], 8
; CHECK-NEXT:    [[OR7:%.*]] = or disjoint i32 [[OR]], [[SHL6]]
; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds nuw i8, ptr [[DATA]], i64 3
; CHECK-NEXT:    [[TMP3:%.*]] = load i8, ptr [[ARRAYIDX8]], align 1
; CHECK-NEXT:    [[CONV9:%.*]] = zext i8 [[TMP3]] to i32
; CHECK-NEXT:    [[OR10:%.*]] = or disjoint i32 [[OR7]], [[CONV9]]
; CHECK-NEXT:    [[OR15:%.*]] = call i32 @llvm.fshl.i32(i32 [[OR10]], i32 [[OR10]], i32 25)
; CHECK-NEXT:    ret i32 [[OR15]]
;

entry:
  %0 = load i8, ptr %data
  %conv = zext i8 %0 to i32
  %shl = shl nuw i32 %conv, 24
  %arrayidx1 = getelementptr inbounds i8, ptr %data, i64 1
  %1 = load i8, ptr %arrayidx1
  %conv2 = zext i8 %1 to i32
  %shl3 = shl nuw nsw i32 %conv2, 16
  %or = or i32 %shl3, %shl
  %arrayidx4 = getelementptr inbounds i8, ptr %data, i64 2
  %2 = load i8, ptr %arrayidx4
  %conv5 = zext i8 %2 to i32
  %shl6 = shl nuw nsw i32 %conv5, 8
  %or7 = or i32 %or, %shl6
  %arrayidx8 = getelementptr inbounds i8, ptr %data, i64 3
  %3 = load i8, ptr %arrayidx8
  %conv9 = zext i8 %3 to i32
  %or10 = or i32 %or7, %conv9
  %shr = lshr i32 %or10, 7
  %shl7 = shl i32 %or10, 25
  %or15 = or i32 %shr, %shl7
  ret i32 %or15
}

define <2 x i31> @fshr_mask_args_same_vector(<2 x i31> %a) {
; CHECK-LABEL: @fshr_mask_args_same_vector(
; CHECK-NEXT:    [[T3:%.*]] = shl <2 x i31> [[A:%.*]], splat (i31 10)
; CHECK-NEXT:    ret <2 x i31> [[T3]]
;
  %t1 = and <2 x i31> %a, <i31 1000, i31 1000>
  %t2 = and <2 x i31> %a, <i31 6442450943, i31 6442450943>
  %t3 = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %t2, <2 x i31> %t1, <2 x i31> <i31 10, i31 10>)
  ret <2 x i31> %t3
}

define <2 x i32> @fshr_mask_args_same_vector2(<2 x i32> %a, <2 x i32> %b) {
; CHECK-LABEL: @fshr_mask_args_same_vector2(
; CHECK-NEXT:    [[T1:%.*]] = and <2 x i32> [[A:%.*]], <i32 1000000, i32 100000>
; CHECK-NEXT:    [[T3:%.*]] = lshr exact <2 x i32> [[T1]], splat (i32 3)
; CHECK-NEXT:    ret <2 x i32> [[T3]]
;
  %t1 = and <2 x i32> %a, <i32 1000000, i32 100000>
  %t2 = and <2 x i32> %a, <i32 6442450943, i32 6442450943>
  %t3 = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %t1, <2 x i32> %t1, <2 x i32> <i32 3, i32 3>)
  ret <2 x i32> %t3
}

define <2 x i31> @fshr_mask_args_same_vector3_different_but_still_prunable(<2 x i31> %a) {
; CHECK-LABEL: @fshr_mask_args_same_vector3_different_but_still_prunable(
; CHECK-NEXT:    [[T1:%.*]] = and <2 x i31> [[A:%.*]], splat (i31 1000)
; CHECK-NEXT:    [[T3:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[A]], <2 x i31> [[T1]], <2 x i31> <i31 10, i31 3>)
; CHECK-NEXT:    ret <2 x i31> [[T3]]
;
  %t1 = and <2 x i31> %a, <i31 1000, i31 1000>
  %t2 = and <2 x i31> %a, <i31 6442450943, i31 6442450943>
  %t3 = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %t2, <2 x i31> %t1, <2 x i31> <i31 10, i31 3>)
  ret <2 x i31> %t3
}

define <2 x i32> @fsh_unary_shuffle_ops(<2 x i32> %x, <2 x i32> %y, <2 x i32> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops(
; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[Y:%.*]], <2 x i32> [[Z:%.*]])
; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %c = shufflevector <2 x i32> %z, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c)
  ret <2 x i32> %r
}

define <3 x i16> @fsh_unary_shuffle_ops_widening(<2 x i16> %x, <2 x i16> %y, <2 x i16> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_widening(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i16> [[X:%.*]], <2 x i16> poison, <3 x i32> <i32 1, i32 0, i32 1>
; CHECK-NEXT:    call void @use_v3(<3 x i16> [[A]])
; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> [[X]], <2 x i16> [[Y:%.*]], <2 x i16> [[Z:%.*]])
; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i16> [[TMP1]], <2 x i16> poison, <3 x i32> <i32 1, i32 0, i32 1>
; CHECK-NEXT:    ret <3 x i16> [[R]]
;
  %a = shufflevector <2 x i16> %x, <2 x i16> poison, <3 x i32> <i32 1, i32 0, i32 1>
  call void @use_v3(<3 x i16> %a)
  %b = shufflevector <2 x i16> %y, <2 x i16> poison, <3 x i32> <i32 1, i32 0, i32 1>
  %c = shufflevector <2 x i16> %z, <2 x i16> poison, <3 x i32> <i32 1, i32 0, i32 1>
  %r = call <3 x i16> @llvm.fshl.v3i16(<3 x i16> %a, <3 x i16> %b, <3 x i16> %c)
  ret <3 x i16> %r
}

define <2 x i31> @fsh_unary_shuffle_ops_narrowing(<3 x i31> %x, <3 x i31> %y, <3 x i31> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_narrowing(
; CHECK-NEXT:    [[B:%.*]] = shufflevector <3 x i31> [[Y:%.*]], <3 x i31> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    call void @use_v2(<2 x i31> [[B]])
; CHECK-NEXT:    [[TMP1:%.*]] = call <3 x i31> @llvm.fshl.v3i31(<3 x i31> [[X:%.*]], <3 x i31> [[Y]], <3 x i31> [[Z:%.*]])
; CHECK-NEXT:    [[R:%.*]] = shufflevector <3 x i31> [[TMP1]], <3 x i31> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %a = shufflevector <3 x i31> %x, <3 x i31> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <3 x i31> %y, <3 x i31> poison, <2 x i32> <i32 1, i32 0>
  call void @use_v2(<2 x i31> %b)
  %c = shufflevector <3 x i31> %z, <3 x i31> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %a, <2 x i31> %b, <2 x i31> %c)
  ret <2 x i31> %r
}

define <2 x i32> @fsh_unary_shuffle_ops_1_const(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_1_const(
; CHECK-NEXT:    [[Y:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> <i32 2, i32 1>, <2 x i32> [[X:%.*]], <2 x i32> [[Y1:%.*]])
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i32> [[Y]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    ret <2 x i32> [[B]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr(<2 x i32> <i32 1, i32 2>, <2 x i32> %a, <2 x i32> %b)
  ret <2 x i32> %r
}

define <2 x i32> @fsh_unary_shuffle_ops_2_const(<2 x i32> %x) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_2_const(
; CHECK-NEXT:    [[X:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> <i32 2, i32 1>, <2 x i32> <i32 2, i32 1>, <2 x i32> [[X1:%.*]])
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i32> [[X]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    ret <2 x i32> [[A]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr(<2 x i32> <i32 1, i32 2>, <2 x i32> <i32 1, i32 2>, <2 x i32> %a)
  ret <2 x i32> %r
}

define <vscale x 2 x i32> @fsh_unary_shuffle_ops_1_const_scalable(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_1_const_scalable(
; CHECK-NEXT:    [[Y:%.*]] = call <vscale x 2 x i32> @llvm.fshr.nxv2i32(<vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> [[X:%.*]], <vscale x 2 x i32> [[Y1:%.*]])
; CHECK-NEXT:    [[B:%.*]] = shufflevector <vscale x 2 x i32> [[Y]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
; CHECK-NEXT:    ret <vscale x 2 x i32> [[B]]
;
  %a = shufflevector <vscale x 2 x i32> %x, <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
  %b = shufflevector <vscale x 2 x i32> %y, <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
  %r = call <vscale x 2 x i32> @llvm.fshr(<vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> %a, <vscale x 2 x i32> %b)
  ret <vscale x 2 x i32> %r
}

define <vscale x 2 x i32> @fsh_unary_shuffle_ops_2_const_scalable(<vscale x 2 x i32> %x) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_2_const_scalable(
; CHECK-NEXT:    [[X:%.*]] = call <vscale x 2 x i32> @llvm.fshr.nxv2i32(<vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> [[X1:%.*]])
; CHECK-NEXT:    [[A:%.*]] = shufflevector <vscale x 2 x i32> [[X]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
; CHECK-NEXT:    ret <vscale x 2 x i32> [[A]]
;
  %a = shufflevector <vscale x 2 x i32> %x, <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
  %r = call <vscale x 2 x i32> @llvm.fshr(<vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> splat (i32 42), <vscale x 2 x i32> %a)
  ret <vscale x 2 x i32> %r
}

define <3 x i32> @fsh_unary_shuffle_ops_widening_1_const(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_widening_1_const(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i32> [[X:%.*]], <2 x i32> poison, <3 x i32> <i32 1, i32 0, i32 poison>
; CHECK-NEXT:    call void @use_v3(<3 x i32> [[A]])
; CHECK-NEXT:    [[Y:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> splat (i32 42), <2 x i32> [[X]], <2 x i32> [[Y1:%.*]])
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i32> [[Y]], <2 x i32> poison, <3 x i32> <i32 1, i32 0, i32 poison>
; CHECK-NEXT:    ret <3 x i32> [[B]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <3 x i32> <i32 1, i32 0, i32 poison>
  call void @use_v3(<3 x i32> %a)
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <3 x i32> <i32 1, i32 0, i32 poison>
  %r = call <3 x i32> @llvm.fshr(<3 x i32> splat (i32 42), <3 x i32> %a, <3 x i32> %b)
  ret <3 x i32> %r
}

; negative test - must have 3 shuffles

define <2 x i32> @fsh_unary_shuffle_ops_unshuffled(<2 x i32> %x, <2 x i32> %y, <2 x i32> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_unshuffled(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i32> [[X:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i32> [[Y:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[A]], <2 x i32> [[B]], <2 x i32> [[Z:%.*]])
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %z)
  ret <2 x i32> %r
}

; negative test - must have identical masks

define <2 x i32> @fsh_unary_shuffle_ops_wrong_mask(<2 x i32> %x, <2 x i32> %y, <2 x i32> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_wrong_mask(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i32> [[X:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i32> [[Y:%.*]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT:    [[C:%.*]] = shufflevector <2 x i32> [[Z:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[A]], <2 x i32> [[B]], <2 x i32> [[C]])
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %a = shufflevector <2 x i32> %x, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> <i32 0, i32 0>
  %c = shufflevector <2 x i32> %z, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c)
  ret <2 x i32> %r
}

; negative test - too many uses

define <2 x i31> @fsh_unary_shuffle_ops_uses(<2 x i31> %x, <2 x i31> %y, <2 x i31> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_uses(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <2 x i31> [[X:%.*]], <2 x i31> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    call void @use_v2(<2 x i31> [[A]])
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i31> [[Y:%.*]], <2 x i31> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    call void @use_v2(<2 x i31> [[B]])
; CHECK-NEXT:    [[C:%.*]] = shufflevector <2 x i31> [[Z:%.*]], <2 x i31> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    call void @use_v2(<2 x i31> [[C]])
; CHECK-NEXT:    [[R:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[A]], <2 x i31> [[B]], <2 x i31> [[C]])
; CHECK-NEXT:    ret <2 x i31> [[R]]
;
  %a = shufflevector <2 x i31> %x, <2 x i31> poison, <2 x i32> <i32 1, i32 0>
  call void @use_v2(<2 x i31> %a)
  %b = shufflevector <2 x i31> %y, <2 x i31> poison, <2 x i32> <i32 1, i32 0>
  call void @use_v2(<2 x i31> %b)
  %c = shufflevector <2 x i31> %z, <2 x i31> poison, <2 x i32> <i32 1, i32 0>
  call void @use_v2(<2 x i31> %c)
  %r = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %a, <2 x i31> %b, <2 x i31> %c)
  ret <2 x i31> %r
}

; negative test - all source ops must have the same type

define <2 x i32> @fsh_unary_shuffle_ops_partial_widening(<3 x i32> %x, <2 x i32> %y, <2 x i32> %z) {
; CHECK-LABEL: @fsh_unary_shuffle_ops_partial_widening(
; CHECK-NEXT:    [[A:%.*]] = shufflevector <3 x i32> [[X:%.*]], <3 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[B:%.*]] = shufflevector <2 x i32> [[Y:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[C:%.*]] = shufflevector <2 x i32> [[Z:%.*]], <2 x i32> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[A]], <2 x i32> [[B]], <2 x i32> [[C]])
; CHECK-NEXT:    ret <2 x i32> [[R]]
;
  %a = shufflevector <3 x i32> %x, <3 x i32> poison, <2 x i32> <i32 1, i32 0>
  %b = shufflevector <2 x i32> %y, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %c = shufflevector <2 x i32> %z, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
  %r = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c)
  ret <2 x i32> %r
}

define <2 x i32> @fshr_vec_zero_elem(<2 x i32> %x, <2 x i32> %y) {
; CHECK-LABEL: @fshr_vec_zero_elem(
; CHECK-NEXT:    [[FSH:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[Y:%.*]], <2 x i32> <i32 2, i32 0>)
; CHECK-NEXT:    ret <2 x i32> [[FSH]]
;
  %fsh = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i32> <i32 2, i32 0>)
  ret <2 x i32> %fsh
}

define i16 @fshl_i16_shl(i16 %x, i16 %y) {
; CHECK-LABEL: @fshl_i16_shl(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = and i16 [[Y:%.*]], 15
; CHECK-NEXT:    [[RES:%.*]] = shl i16 [[X:%.*]], [[TMP0]]
; CHECK-NEXT:    ret i16 [[RES]]
;
entry:
  %res = call i16 @llvm.fshl.i16(i16 %x, i16 0, i16 %y)
  ret i16 %res
}

define i32 @fshl_i32_shl(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_i32_shl(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = and i32 [[Y:%.*]], 31
; CHECK-NEXT:    [[RES:%.*]] = shl i32 [[X:%.*]], [[TMP0]]
; CHECK-NEXT:    ret i32 [[RES]]
;
entry:
  %res = call i32 @llvm.fshl.i32(i32 %x, i32 0, i32 %y)
  ret i32 %res
}

define <2 x i16> @fshl_vi16_shl(<2 x i16> %x, <2 x i16> %y) {
; CHECK-LABEL: @fshl_vi16_shl(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = and <2 x i16> [[Y:%.*]], splat (i16 15)
; CHECK-NEXT:    [[RES:%.*]] = shl <2 x i16> [[X:%.*]], [[TMP0]]
; CHECK-NEXT:    ret <2 x i16> [[RES]]
;
entry:
  %res = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> %x, <2 x i16> zeroinitializer, <2 x i16> %y)
  ret <2 x i16> %res
}

define i32 @fshr_i32_shl_negative_test(i32 %x, i32 %y) {
; CHECK-LABEL: @fshr_i32_shl_negative_test(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 0, i32 [[Y:%.*]])
; CHECK-NEXT:    ret i32 [[RES]]
;
entry:
  %res = call i32 @llvm.fshr.i32(i32 %x, i32 0, i32 %y)
  ret i32 %res
}

define <2 x i31> @fshl_vi31_shl_negative_test(<2 x i31> %x, <2 x i31> %y) {
; CHECK-LABEL: @fshl_vi31_shl_negative_test(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[RES:%.*]] = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> [[X:%.*]], <2 x i31> zeroinitializer, <2 x i31> [[Y:%.*]])
; CHECK-NEXT:    ret <2 x i31> [[RES]]
;
entry:
  %res = call <2 x i31> @llvm.fshl.v2i31(<2 x i31> %x, <2 x i31> zeroinitializer, <2 x i31>  %y)
  ret <2 x i31>  %res
}

;; Issue #124387 Range attribute no longer holds after operands changed.
define i8 @fshl_range_trunc(i1 %x) {
; CHECK-LABEL: @fshl_range_trunc(
; CHECK-NEXT:    [[ZEXT:%.*]] = zext i1 [[X:%.*]] to i32
; CHECK-NEXT:    [[OR:%.*]] = or disjoint i32 [[ZEXT]], 126
; CHECK-NEXT:    [[FSHL:%.*]] = call i32 @llvm.fshl.i32(i32 [[OR]], i32 -2, i32 1)
; CHECK-NEXT:    [[TR:%.*]] = trunc nuw i32 [[FSHL]] to i8
; CHECK-NEXT:    ret i8 [[TR]]
;
  %zext = zext i1 %x to i32
  %or = or disjoint i32 %zext, -2
  %fshl = call range(i32 -4, 2) i32 @llvm.fshl.i32(i32 %or, i32 %or, i32 1)
  %tr = trunc nsw i32 %fshl to i8
  ret i8 %tr
}

;; Issue #138334 negative rotate amounts can be folded into the opposite direction
define i32 @fshl_neg_amount(i32 %x, i32 %y) {
; CHECK-LABEL: @fshl_neg_amount(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %n = sub i32 0, %y
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 %n)
  ret i32 %r
}

define i32 @fshr_neg_amount(i32 %x, i32 %y) {
; CHECK-LABEL: @fshr_neg_amount(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %n = sub i32 0, %y
  %r = call i32 @llvm.fshr.i32(i32 %x, i32 %x, i32 %n)
  ret i32 %r
}

;; negative test, funnel shift is not a rotate

define i32 @fshl_neg_amount_non_rotate(i32 %x, i32 %y, i32 %z) {
; CHECK-LABEL: @fshl_neg_amount_non_rotate(
; CHECK-NEXT:    [[N:%.*]] = sub i32 0, [[Y:%.*]]
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Z:%.*]], i32 [[N]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %n = sub i32 0, %y
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %z, i32 %n)
  ret i32 %r
}

define i32 @fshr_neg_amount_non_rotate(i32 %x, i32 %y, i32 %z) {
; CHECK-LABEL: @fshr_neg_amount_non_rotate(
; CHECK-NEXT:    [[N:%.*]] = sub i32 0, [[Y:%.*]]
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 [[Z:%.*]], i32 [[N]])
; CHECK-NEXT:    ret i32 [[R]]
;
  %n = sub i32 0, %y
  %r = call i32 @llvm.fshr.i32(i32 %x, i32 %z, i32 %n)
  ret i32 %r
}

;; negative test, bitwidth is not a power of two

define i31 @fshl_neg_amount_non_power_two(i31 %x, i31 %y) {
; CHECK-LABEL: @fshl_neg_amount_non_power_two(
; CHECK-NEXT:    [[N:%.*]] = sub i31 0, [[Y:%.*]]
; CHECK-NEXT:    [[R:%.*]] = call i31 @llvm.fshl.i31(i31 [[X:%.*]], i31 [[X]], i31 [[N]])
; CHECK-NEXT:    ret i31 [[R]]
;
  %n = sub i31 0, %y
  %r = call i31 @llvm.fshl.i31(i31 %x, i31 %x, i31 %n)
  ret i31 %r
}

define i31 @fshr_neg_amount_non_power_two(i31 %x, i31 %y) {
; CHECK-LABEL: @fshr_neg_amount_non_power_two(
; CHECK-NEXT:    [[N:%.*]] = sub i31 0, [[Y:%.*]]
; CHECK-NEXT:    [[R:%.*]] = call i31 @llvm.fshr.i31(i31 [[X:%.*]], i31 [[X]], i31 [[N]])
; CHECK-NEXT:    ret i31 [[R]]
;
  %n = sub i31 0, %y
  %r = call i31 @llvm.fshr.i31(i31 %x, i31 %x, i31 %n)
  ret i31 %r
}

define i32 @rot_const_consecutive(i32 %x) {
; CHECK-LABEL: @rot_const_consecutive(
; CHECK-NEXT:    [[R2:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 8)
; CHECK-NEXT:    ret i32 [[R2]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 13)
  %r2 = call i32 @llvm.fshl.i32(i32 %r, i32 %r, i32 27)
  ret i32 %r2
}

define i32 @rot_const_consecutive_multi_use(i32 %x) {
; CHECK-LABEL: @rot_const_consecutive_multi_use(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 7)
; CHECK-NEXT:    [[R3:%.*]] = call i32 @llvm.fshl.i32(i32 [[X]], i32 [[X]], i32 11)
; CHECK-NEXT:    [[R2:%.*]] = and i32 [[R]], [[R3]]
; CHECK-NEXT:    ret i32 [[R2]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 7)
  %r2 = call i32 @llvm.fshl.i32(i32 %r, i32 %r, i32 4)
  %and = and i32 %r, %r2
  ret i32 %and
}

define i32 @rot_const_consecutive_cancel_out(i32 %x) {
; CHECK-LABEL: @rot_const_consecutive_cancel_out(
; CHECK-NEXT:    ret i32 [[X:%.*]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 7)
  %r2 = call i32 @llvm.fshl.i32(i32 %r, i32 %r, i32 25)
  ret i32 %r2
}

;; negative test, consecutive rotates only fold if shift amounts are const

define i32 @rot_nonconst_shift(i32 %x, i32 %amt) {
; CHECK-LABEL: @rot_nonconst_shift(
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 7)
; CHECK-NEXT:    [[R2:%.*]] = call i32 @llvm.fshl.i32(i32 [[R]], i32 [[R]], i32 [[AMT:%.*]])
; CHECK-NEXT:    ret i32 [[R2]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 7)
  %r2 = call i32 @llvm.fshl.i32(i32 %r, i32 %r, i32 %amt)
  ret i32 %r2
}

;; negative test, 1st funnel shift isn't a rotate.

define i32 @fsh_rot(i32 %x, i32 %y) {
; CHECK-LABEL: @fsh_rot(
; CHECK-NEXT:    [[FSH:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[Y:%.*]], i32 7)
; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[FSH]], i32 [[FSH]], i32 4)
; CHECK-NEXT:    ret i32 [[R]]
;
  %fsh = call i32 @llvm.fshl.i32(i32 %x, i32 %y, i32 7)
  %r = call i32 @llvm.fshl.i32(i32 %fsh, i32 %fsh, i32 4)
  ret i32 %r
}

;; negative test, 2nd funnel shift isn't a rotate.

define i32 @rot_fsh(i32 %x, i32 %y) {
; CHECK-LABEL: @rot_fsh(
; CHECK-NEXT:    [[Y:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 7)
; CHECK-NEXT:    [[R2:%.*]] = call i32 @llvm.fshl.i32(i32 [[Y]], i32 [[R:%.*]], i32 4)
; CHECK-NEXT:    ret i32 [[R2]]
;
  %r = call i32 @llvm.fshl.i32(i32 %x, i32 %x, i32 7)
  %r2 = call i32 @llvm.fshl.i32(i32 %r, i32 %y, i32 4)
  ret i32 %r2
}