| open Pack |
| open Value |
| |
| module V128Op = |
| struct |
| open Ast.V128Op |
| open V128Vec |
| open V128 |
| |
| let testop (op : testop) = |
| let f = match op with |
| | I8x16 AllTrue -> V128.I8x16.all_true |
| | I16x8 AllTrue -> V128.I16x8.all_true |
| | I32x4 AllTrue -> V128.I32x4.all_true |
| | I64x2 AllTrue -> V128.I64x2.all_true |
| | _ -> . |
| in fun v -> f (of_vec 1 v) |
| |
| let unop (op : unop) = |
| let f = match op with |
| | I8x16 Neg -> V128.I8x16.neg |
| | I8x16 Abs -> V128.I8x16.abs |
| | I8x16 Popcnt -> V128.I8x16.popcnt |
| | I16x8 Neg -> V128.I16x8.neg |
| | I16x8 Abs -> V128.I16x8.abs |
| | I32x4 Abs -> V128.I32x4.abs |
| | I32x4 Neg -> V128.I32x4.neg |
| | I64x2 Abs -> V128.I64x2.abs |
| | I64x2 Neg -> V128.I64x2.neg |
| | F32x4 Abs -> V128.F32x4.abs |
| | F32x4 Neg -> V128.F32x4.neg |
| | F32x4 Sqrt -> V128.F32x4.sqrt |
| | F32x4 Ceil -> V128.F32x4.ceil |
| | F32x4 Floor -> V128.F32x4.floor |
| | F32x4 Trunc -> V128.F32x4.trunc |
| | F32x4 Nearest -> V128.F32x4.nearest |
| | F64x2 Abs -> V128.F64x2.abs |
| | F64x2 Neg -> V128.F64x2.neg |
| | F64x2 Sqrt -> V128.F64x2.sqrt |
| | F64x2 Ceil -> V128.F64x2.ceil |
| | F64x2 Floor -> V128.F64x2.floor |
| | F64x2 Trunc -> V128.F64x2.trunc |
| | F64x2 Nearest -> V128.F64x2.nearest |
| | _ -> assert false |
| in fun v -> to_vec (f (of_vec 1 v)) |
| |
| let binop (op : binop) = |
| let f = match op with |
| | I8x16 Swizzle -> V128.V8x16.swizzle |
| | I8x16 (Shuffle is) -> V128.V8x16.shuffle (List.map I8.to_int_u is) |
| | I8x16 (Narrow S) -> V128.I8x16_convert.narrow_s |
| | I8x16 (Narrow U) -> V128.I8x16_convert.narrow_u |
| | I8x16 Add -> V128.I8x16.add |
| | I8x16 (AddSat S) -> V128.I8x16.add_sat_s |
| | I8x16 (AddSat U) -> V128.I8x16.add_sat_u |
| | I8x16 Sub -> V128.I8x16.sub |
| | I8x16 (SubSat S) -> V128.I8x16.sub_sat_s |
| | I8x16 (SubSat U) -> V128.I8x16.sub_sat_u |
| | I8x16 (Min S) -> V128.I8x16.min_s |
| | I8x16 (Min U) -> V128.I8x16.min_u |
| | I8x16 (Max S) -> V128.I8x16.max_s |
| | I8x16 (Max U) -> V128.I8x16.max_u |
| | I8x16 AvgrU -> V128.I8x16.avgr_u |
| | I8x16 RelaxedSwizzle -> V128.V8x16.swizzle |
| | I16x8 (Narrow S) -> V128.I16x8_convert.narrow_s |
| | I16x8 (Narrow U) -> V128.I16x8_convert.narrow_u |
| | I16x8 Add -> V128.I16x8.add |
| | I16x8 (AddSat S) -> V128.I16x8.add_sat_s |
| | I16x8 (AddSat U) -> V128.I16x8.add_sat_u |
| | I16x8 Sub -> V128.I16x8.sub |
| | I16x8 (SubSat S) -> V128.I16x8.sub_sat_s |
| | I16x8 (SubSat U) -> V128.I16x8.sub_sat_u |
| | I16x8 Mul -> V128.I16x8.mul |
| | I16x8 (Min S) -> V128.I16x8.min_s |
| | I16x8 (Min U) -> V128.I16x8.min_u |
| | I16x8 (Max S) -> V128.I16x8.max_s |
| | I16x8 (Max U) -> V128.I16x8.max_u |
| | I16x8 AvgrU -> V128.I16x8.avgr_u |
| | I16x8 (ExtMul (Low, S)) -> V128.I16x8_convert.extmul_low_s |
| | I16x8 (ExtMul (High, S)) -> V128.I16x8_convert.extmul_high_s |
| | I16x8 (ExtMul (Low, U)) -> V128.I16x8_convert.extmul_low_u |
| | I16x8 (ExtMul (High, U)) -> V128.I16x8_convert.extmul_high_u |
| | I16x8 Q15MulRSatS -> V128.I16x8.q15mulr_sat_s |
| | I16x8 RelaxedQ15MulRS -> V128.I16x8.q15mulr_sat_s |
| | I16x8 RelaxedDot -> V128.I16x8_convert.dot_s |
| | I32x4 Add -> V128.I32x4.add |
| | I32x4 Sub -> V128.I32x4.sub |
| | I32x4 (Min S) -> V128.I32x4.min_s |
| | I32x4 (Min U) -> V128.I32x4.min_u |
| | I32x4 (Max S) -> V128.I32x4.max_s |
| | I32x4 (Max U) -> V128.I32x4.max_u |
| | I32x4 Mul -> V128.I32x4.mul |
| | I32x4 (ExtMul (Low, S)) -> V128.I32x4_convert.extmul_low_s |
| | I32x4 (ExtMul (High, S)) -> V128.I32x4_convert.extmul_high_s |
| | I32x4 (ExtMul (Low, U)) -> V128.I32x4_convert.extmul_low_u |
| | I32x4 (ExtMul (High, U)) -> V128.I32x4_convert.extmul_high_u |
| | I64x2 Add -> V128.I64x2.add |
| | I64x2 Sub -> V128.I64x2.sub |
| | I64x2 Mul -> V128.I64x2.mul |
| | I32x4 DotS -> V128.I32x4_convert.dot_s |
| | I64x2 (ExtMul (Low, S)) -> V128.I64x2_convert.extmul_low_s |
| | I64x2 (ExtMul (High, S)) -> V128.I64x2_convert.extmul_high_s |
| | I64x2 (ExtMul (Low, U)) -> V128.I64x2_convert.extmul_low_u |
| | I64x2 (ExtMul (High, U)) -> V128.I64x2_convert.extmul_high_u |
| | F32x4 Add -> V128.F32x4.add |
| | F32x4 Sub -> V128.F32x4.sub |
| | F32x4 Mul -> V128.F32x4.mul |
| | F32x4 Div -> V128.F32x4.div |
| | F32x4 Min -> V128.F32x4.min |
| | F32x4 Max -> V128.F32x4.max |
| | F32x4 Pmin -> V128.F32x4.pmin |
| | F32x4 Pmax -> V128.F32x4.pmax |
| | F32x4 RelaxedMin -> V128.F32x4.min |
| | F32x4 RelaxedMax -> V128.F32x4.max |
| | F64x2 Add -> V128.F64x2.add |
| | F64x2 Sub -> V128.F64x2.sub |
| | F64x2 Mul -> V128.F64x2.mul |
| | F64x2 Div -> V128.F64x2.div |
| | F64x2 Min -> V128.F64x2.min |
| | F64x2 Max -> V128.F64x2.max |
| | F64x2 Pmin -> V128.F64x2.pmin |
| | F64x2 Pmax -> V128.F64x2.pmax |
| | F64x2 RelaxedMin -> V128.F64x2.min |
| | F64x2 RelaxedMax -> V128.F64x2.max |
| | _ -> assert false |
| in fun v1 v2 -> to_vec (f (of_vec 1 v1) (of_vec 2 v2)) |
| |
| let ternop (op : ternop) = |
| let f = match op with |
| | F32x4 RelaxedMadd -> V128.F32x4.fma |
| | F32x4 RelaxedNmadd -> V128.F32x4.fnma |
| | F64x2 RelaxedMadd -> V128.F64x2.fma |
| | F64x2 RelaxedNmadd -> V128.F64x2.fnma |
| | I8x16 RelaxedLaneselect -> V128.V1x128.bitselect |
| | I16x8 RelaxedLaneselect -> V128.V1x128.bitselect |
| | I32x4 RelaxedLaneselect -> V128.V1x128.bitselect |
| | I64x2 RelaxedLaneselect -> V128.V1x128.bitselect |
| | I32x4 RelaxedDotAddS -> V128.I32x4_convert.dot_add_s |
| | _ -> assert false |
| in fun v1 v2 v3 -> to_vec (f (of_vec 1 v1) (of_vec 2 v2) (of_vec 3 v3)) |
| |
| let relop (op : relop) = |
| let f = match op with |
| | I8x16 Eq -> V128.I8x16.eq |
| | I8x16 Ne -> V128.I8x16.ne |
| | I8x16 (Lt S) -> V128.I8x16.lt_s |
| | I8x16 (Lt U) -> V128.I8x16.lt_u |
| | I8x16 (Le S) -> V128.I8x16.le_s |
| | I8x16 (Le U) -> V128.I8x16.le_u |
| | I8x16 (Gt S) -> V128.I8x16.gt_s |
| | I8x16 (Gt U) -> V128.I8x16.gt_u |
| | I8x16 (Ge S) -> V128.I8x16.ge_s |
| | I8x16 (Ge U) -> V128.I8x16.ge_u |
| | I16x8 Eq -> V128.I16x8.eq |
| | I16x8 Ne -> V128.I16x8.ne |
| | I16x8 (Lt S) -> V128.I16x8.lt_s |
| | I16x8 (Lt U) -> V128.I16x8.lt_u |
| | I16x8 (Le S) -> V128.I16x8.le_s |
| | I16x8 (Le U) -> V128.I16x8.le_u |
| | I16x8 (Gt S) -> V128.I16x8.gt_s |
| | I16x8 (Gt U) -> V128.I16x8.gt_u |
| | I16x8 (Ge S) -> V128.I16x8.ge_s |
| | I16x8 (Ge U) -> V128.I16x8.ge_u |
| | I32x4 Eq -> V128.I32x4.eq |
| | I32x4 Ne -> V128.I32x4.ne |
| | I32x4 (Lt S) -> V128.I32x4.lt_s |
| | I32x4 (Lt U) -> V128.I32x4.lt_u |
| | I32x4 (Le S) -> V128.I32x4.le_s |
| | I32x4 (Le U) -> V128.I32x4.le_u |
| | I32x4 (Gt S) -> V128.I32x4.gt_s |
| | I32x4 (Gt U) -> V128.I32x4.gt_u |
| | I32x4 (Ge S) -> V128.I32x4.ge_s |
| | I32x4 (Ge U) -> V128.I32x4.ge_u |
| | I64x2 Eq -> V128.I64x2.eq |
| | I64x2 Ne -> V128.I64x2.ne |
| | I64x2 (Lt S) -> V128.I64x2.lt_s |
| | I64x2 (Le S) -> V128.I64x2.le_s |
| | I64x2 (Gt S) -> V128.I64x2.gt_s |
| | I64x2 (Ge S) -> V128.I64x2.ge_s |
| | F32x4 Eq -> V128.F32x4.eq |
| | F32x4 Ne -> V128.F32x4.ne |
| | F32x4 Lt -> V128.F32x4.lt |
| | F32x4 Le -> V128.F32x4.le |
| | F32x4 Gt -> V128.F32x4.gt |
| | F32x4 Ge -> V128.F32x4.ge |
| | F64x2 Eq -> V128.F64x2.eq |
| | F64x2 Ne -> V128.F64x2.ne |
| | F64x2 Lt -> V128.F64x2.lt |
| | F64x2 Le -> V128.F64x2.le |
| | F64x2 Gt -> V128.F64x2.gt |
| | F64x2 Ge -> V128.F64x2.ge |
| | _ -> assert false |
| in fun v1 v2 -> to_vec (f (of_vec 1 v1) (of_vec 2 v2)) |
| |
| let cvtop (op : cvtop) = |
| let f = match op with |
| | I16x8 (Extend (Low, S)) -> V128.I16x8_convert.extend_low_s |
| | I16x8 (Extend (High, S)) -> V128.I16x8_convert.extend_high_s |
| | I16x8 (Extend (Low, U)) -> V128.I16x8_convert.extend_low_u |
| | I16x8 (Extend (High, U)) -> V128.I16x8_convert.extend_high_u |
| | I16x8 (ExtAddPairwise S) -> V128.I16x8_convert.extadd_pairwise_s |
| | I16x8 (ExtAddPairwise U) -> V128.I16x8_convert.extadd_pairwise_u |
| | I32x4 (Extend (Low, S)) -> V128.I32x4_convert.extend_low_s |
| | I32x4 (Extend (High, S)) -> V128.I32x4_convert.extend_high_s |
| | I32x4 (Extend (Low, U)) -> V128.I32x4_convert.extend_low_u |
| | I32x4 (Extend (High, U)) -> V128.I32x4_convert.extend_high_u |
| | I32x4 (TruncSatF32x4 S) -> V128.I32x4_convert.trunc_sat_f32x4_s |
| | I32x4 (TruncSatF32x4 U) -> V128.I32x4_convert.trunc_sat_f32x4_u |
| | I32x4 (TruncSatZeroF64x2 S) -> V128.I32x4_convert.trunc_sat_f64x2_s_zero |
| | I32x4 (TruncSatZeroF64x2 U) -> V128.I32x4_convert.trunc_sat_f64x2_u_zero |
| | I32x4 (RelaxedTruncF32x4 S) -> V128.I32x4_convert.trunc_sat_f32x4_s |
| | I32x4 (RelaxedTruncF32x4 U) -> V128.I32x4_convert.trunc_sat_f32x4_u |
| | I32x4 (RelaxedTruncZeroF64x2 S) -> V128.I32x4_convert.trunc_sat_f64x2_s_zero |
| | I32x4 (RelaxedTruncZeroF64x2 U) -> V128.I32x4_convert.trunc_sat_f64x2_u_zero |
| | I32x4 (ExtAddPairwise S) -> V128.I32x4_convert.extadd_pairwise_s |
| | I32x4 (ExtAddPairwise U) -> V128.I32x4_convert.extadd_pairwise_u |
| | I64x2 (Extend (Low, S)) -> V128.I64x2_convert.extend_low_s |
| | I64x2 (Extend (High, S)) -> V128.I64x2_convert.extend_high_s |
| | I64x2 (Extend (Low, U)) -> V128.I64x2_convert.extend_low_u |
| | I64x2 (Extend (High, U)) -> V128.I64x2_convert.extend_high_u |
| | F32x4 (ConvertI32x4 S) -> V128.F32x4_convert.convert_i32x4_s |
| | F32x4 (ConvertI32x4 U)-> V128.F32x4_convert.convert_i32x4_u |
| | F32x4 DemoteZeroF64x2 -> V128.F32x4_convert.demote_f64x2_zero |
| | F64x2 PromoteLowF32x4 -> V128.F64x2_convert.promote_low_f32x4 |
| | F64x2 (ConvertI32x4 S) -> V128.F64x2_convert.convert_i32x4_s |
| | F64x2 (ConvertI32x4 U) -> V128.F64x2_convert.convert_i32x4_u |
| | _ -> assert false |
| in fun v -> to_vec (f (of_vec 1 v)) |
| |
| let shiftop (op : shiftop) = |
| let f = match op with |
| | I8x16 Shl -> V128.I8x16.shl |
| | I8x16 (Shr S) -> V128.I8x16.shr_s |
| | I8x16 (Shr U) -> V128.I8x16.shr_u |
| | I16x8 Shl -> V128.I16x8.shl |
| | I16x8 (Shr S) -> V128.I16x8.shr_s |
| | I16x8 (Shr U) -> V128.I16x8.shr_u |
| | I32x4 Shl -> V128.I32x4.shl |
| | I32x4 (Shr S) -> V128.I32x4.shr_s |
| | I32x4 (Shr U) -> V128.I32x4.shr_u |
| | I64x2 Shl -> V128.I64x2.shl |
| | I64x2 (Shr S) -> V128.I64x2.shr_s |
| | I64x2 (Shr U) -> V128.I64x2.shr_u |
| | _ -> . |
| in fun v n -> to_vec (f (of_vec 1 v) (I32Num.of_num 2 n)) |
| |
| let bitmaskop (op : bitmaskop) v = |
| let f = match op with |
| | I8x16 Bitmask -> V128.I8x16.bitmask |
| | I16x8 Bitmask -> V128.I16x8.bitmask |
| | I32x4 Bitmask -> V128.I32x4.bitmask |
| | I64x2 Bitmask -> V128.I64x2.bitmask |
| | _ -> . |
| in I32 (f (of_vec 1 v)) |
| |
| let vtestop (op : vtestop) = |
| let f = match op with |
| | AnyTrue -> V128.I8x16.any_true |
| in fun v -> f (of_vec 1 v) |
| |
| let vunop (op : vunop) = |
| let f = match op with |
| | Not -> V128.V1x128.not_ |
| in fun v -> to_vec (f (of_vec 1 v)) |
| |
| let vbinop (op : vbinop) = |
| let f = match op with |
| | And -> V128.V1x128.and_ |
| | Or -> V128.V1x128.or_ |
| | Xor -> V128.V1x128.xor |
| | AndNot -> V128.V1x128.andnot |
| in fun v1 v2 -> to_vec (f (of_vec 1 v1) (of_vec 2 v2)) |
| |
| let vternop (op : vternop) = |
| let f = match op with |
| | Bitselect -> V128.V1x128.bitselect |
| in fun v1 v2 v3 -> to_vec (f (of_vec 1 v1) (of_vec 2 v2) (of_vec 3 v3)) |
| end |
| |
| module V128CvtOp = |
| struct |
| open Ast.V128Op |
| open V128Vec |
| open V128 |
| |
| let splatop (op : splatop) v = |
| let i = |
| match op with |
| | I8x16 Splat -> V128.I8x16.splat (Convert.I8_.wrap_i32 (I32Num.of_num 1 v)) |
| | I16x8 Splat -> V128.I16x8.splat (Convert.I16_.wrap_i32 (I32Num.of_num 1 v)) |
| | I32x4 Splat -> V128.I32x4.splat (I32Num.of_num 1 v) |
| | I64x2 Splat -> V128.I64x2.splat (I64Num.of_num 1 v) |
| | F32x4 Splat -> V128.F32x4.splat (F32Num.of_num 1 v) |
| | F64x2 Splat -> V128.F64x2.splat (F64Num.of_num 1 v) |
| in to_vec i |
| |
| let extractop (op : extractop) v = |
| let v128 = of_vec 1 v in |
| match op with |
| | I8x16 (Extract (i, S)) -> I32 (Convert.I32_.extend_i8_s (V128.I8x16.extract_lane (I8.to_int_u i) v128)) |
| | I8x16 (Extract (i, U)) -> I32 (Convert.I32_.extend_i8_u (V128.I8x16.extract_lane (I8.to_int_u i) v128)) |
| | I16x8 (Extract (i, S)) -> I32 (Convert.I32_.extend_i16_s (V128.I16x8.extract_lane (I8.to_int_u i) v128)) |
| | I16x8 (Extract (i, U)) -> I32 (Convert.I32_.extend_i16_u (V128.I16x8.extract_lane (I8.to_int_u i) v128)) |
| | I32x4 (Extract (i, ())) -> I32 (V128.I32x4.extract_lane (I8.to_int_u i) v128) |
| | I64x2 (Extract (i, ())) -> I64 (V128.I64x2.extract_lane (I8.to_int_u i) v128) |
| | F32x4 (Extract (i, ())) -> F32 (V128.F32x4.extract_lane (I8.to_int_u i) v128) |
| | F64x2 (Extract (i, ())) -> F64 (V128.F64x2.extract_lane (I8.to_int_u i) v128) |
| |
| let replaceop (op : replaceop) v (n : num) = |
| let v128 = of_vec 1 v in |
| let v128' = match op with |
| | I8x16 (Replace i) -> V128.I8x16.replace_lane (I8.to_int_u i) v128 (Convert.I8_.wrap_i32 (I32Num.of_num 1 n)) |
| | I16x8 (Replace i) -> V128.I16x8.replace_lane (I8.to_int_u i) v128 (Convert.I16_.wrap_i32 (I32Num.of_num 1 n)) |
| | I32x4 (Replace i) -> V128.I32x4.replace_lane (I8.to_int_u i) v128 (I32Num.of_num 1 n) |
| | I64x2 (Replace i) -> V128.I64x2.replace_lane (I8.to_int_u i) v128 (I64Num.of_num 1 n) |
| | F32x4 (Replace i) -> V128.F32x4.replace_lane (I8.to_int_u i) v128 (F32Num.of_num 1 n) |
| | F64x2 (Replace i) -> V128.F64x2.replace_lane (I8.to_int_u i) v128 (F64Num.of_num 1 n) |
| in to_vec v128' |
| end |
| |
| (* Dispatch *) |
| |
| let op v128 = function |
| | V128 x -> v128 x |
| |
| let eval_vtestop = op V128Op.testop |
| let eval_vunop = op V128Op.unop |
| let eval_vbinop = op V128Op.binop |
| let eval_vternop = op V128Op.ternop |
| let eval_vrelop = op V128Op.relop |
| let eval_vcvtop = op V128Op.cvtop |
| let eval_vshiftop = op V128Op.shiftop |
| let eval_vbitmaskop = op V128Op.bitmaskop |
| let eval_vvtestop = op V128Op.vtestop |
| let eval_vvunop = op V128Op.vunop |
| let eval_vvbinop = op V128Op.vbinop |
| let eval_vvternop = op V128Op.vternop |
| let eval_vsplatop = op V128CvtOp.splatop |
| let eval_vextractop = op V128CvtOp.extractop |
| let eval_vreplaceop = op V128CvtOp.replaceop |