[WebAssembly] SIMD Bitwise binary arithmetic

[tlively]

Summary

AND, OR, and XOR. This CL depends on D51113.

[aheejin]

lib/Target/WebAssembly/WebAssemblyInstrSIMD.td

187

Can we add a few newlines to make it more readable? Otherwise LGTM

[tlively]

Add whitespace for readability

lib/Target/WebAssembly/WebAssemblyInstrSIMD.td

 multiclass SIMDBinaryInt<SDNode node, string name, bits<32> baseInst> {
   defm "" : SIMDBinaryIntNoI64x2<node, name, baseInst>;
   defm "" : SIMDBinary<v2i64, "i64x2", node, name, !add(baseInst, 3)>;
 }
 multiclass SIMDBinaryFP<SDNode node, string name, bits<32> baseInst> {
   defm "" : SIMDBinary<v4f32, "f32x4", node, name, baseInst>;
   defm "" : SIMDBinary<v2f64, "f64x2", node, name, !add(baseInst, 1)>;
 }
+multiclass SIMDBitwise<SDNode node, string name, bits<32> simdop> {
+  defm "" : SIMDBinary<v16i8, "v128", node, name, simdop>;
+  defm "" : SIMDBinary<v8i16, "v128", node, name, simdop>;
+  defm "" : SIMDBinary<v4i32, "v128", node, name, simdop>;
+  defm "" : SIMDBinary<v2i64, "v128", node, name, simdop>;
+}
 
 let Defs = [ARGUMENTS] in {
 defm "" : ConstVec<v16i8,
                    (ins vec_i8imm_op:$i0, vec_i8imm_op:$i1,
                         vec_i8imm_op:$i2, vec_i8imm_op:$i3,
                         vec_i8imm_op:$i4, vec_i8imm_op:$i5,
                         vec_i8imm_op:$i6, vec_i8imm_op:$i7,
                         vec_i8imm_op:$i8, vec_i8imm_op:$i9,
defm "" : ConstVec<v4f32,
                         f32imm_op:$i2, f32imm_op:$i3),
                    (build_vector (f32 fpimm:$i0), (f32 fpimm:$i1),
                                  (f32 fpimm:$i2), (f32 fpimm:$i3)),
                    "$i0, $i1, $i2, $i3">;
 defm "" : ConstVec<v2f64,
                   (ins f64imm_op:$i0, f64imm_op:$i1),
                   (build_vector (f64 fpimm:$i0), (f64 fpimm:$i1)),
                   "$i0, $i1">;
+
 defm "" : ExtractLaneExtended<"_s", 9>;
 defm "" : ExtractLaneExtended<"_u", 10>;
 defm "" : ExtractLane<v4i32, "i32x4", LaneIdx4, I32, 13>;
 defm "" : ExtractLane<v2i64, "i64x2", LaneIdx2, I64, 14>;
 defm "" : ExtractLane<v4f32, "f32x4", LaneIdx4, F32, 15>;
 defm "" : ExtractLane<v2f64, "f64x2", LaneIdx2, F64, 16>;
+
 defm "" : ReplaceLane<v16i8, "i8x16", LaneIdx16, I32, i32, 17>;
 defm "" : ReplaceLane<v8i16, "i16x8", LaneIdx8, I32, i32, 18>;
 defm "" : ReplaceLane<v4i32, "i32x4", LaneIdx4, I32, i32, 19>;
 defm "" : ReplaceLane<v2i64, "i64x2", LaneIdx2, I64, i64, 20>;
 defm "" : ReplaceLane<v4f32, "f32x4", LaneIdx4, F32, f32, 21>;
 defm "" : ReplaceLane<v2f64, "f64x2", LaneIdx2, F64, f64, 22>;
+
 defm "" : Splat<v16i8, "i8x16", I32, splat16, 3>;
 defm "" : Splat<v8i16, "i16x8", I32, splat8, 4>;
 defm "" : Splat<v4i32, "i32x4", I32, splat4, 5>;
 defm "" : Splat<v2i64, "i64x2", I64, splat2, 6>;
 defm "" : Splat<v4f32, "f32x4", F32, splat4, 7>;
 defm "" : Splat<v2f64, "f64x2", F64, splat2, 8>;
+
 let isCommutable = 1 in {
 defm ADD : SIMDBinaryInt<add, "add", 24>;
 defm ADD : SIMDBinaryFP<fadd, "add", 122>;
 defm MUL : SIMDBinaryIntNoI64x2<mul, "mul", 32>;
 defm MUL : SIMDBinaryFP<fmul, "mul", 128>;
 } // isCommutable = 1
+
 defm SUB : SIMDBinaryInt<sub, "sub", 28>;
 defm SUB : SIMDBinaryFP<fsub, "sub", 124>;
 defm DIV : SIMDBinaryFP<fdiv, "div", 126>;
+
+let isCommutable = 1 in {
+defm AND : SIMDBitwise<and, "and", 59>;
+defm OR : SIMDBitwise<or, "or", 60>;
+defm XOR : SIMDBitwise<xor, "xor", 61>;
+} // isCommutable = 1
+
 } // Defs = [ARGUMENTS]
 
 // follow convention of making implicit expansions unsigned
 def : Pat<(i32 (vector_extract (v16i8 V128:$vec), (i32 LaneIdx16:$idx))),
           (EXTRACT_LANE_v16i8_u V128:$vec, (i32 LaneIdx16:$idx))>;
 def : Pat<(i32 (vector_extract (v8i16 V128:$vec), (i32 LaneIdx8:$idx))),
           (EXTRACT_LANE_v8i16_u V128:$vec, (i32 LaneIdx8:$idx))>;
 

test/CodeGen/WebAssembly/simd-arith.ll

 ; SIMD128: .result v128{{$}}
 ; SIMD128: i8x16.mul $push0=, $0, $1 # encoding: [0xfd,0x20]{{$}}
 ; SIMD128: return $pop0 #
 define <16 x i8> @mul_v16i8(<16 x i8> %x, <16 x i8> %y) {
   %a = mul <16 x i8> %x, %y
   ret <16 x i8> %a
 }
 
+; CHECK-LABEL: and_v16i8
+; NO-SIMD128-NOT: i8x16
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.and $push0=, $0, $1 # encoding: [0xfd,0x3b]{{$}}
+; SIMD128: return $pop0 #
+define <16 x i8> @and_v16i8(<16 x i8> %x, <16 x i8> %y) {
+  %a = and <16 x i8> %x, %y
+  ret <16 x i8> %a
+}
+
+; CHECK-LABEL: or_v16i8
+; NO-SIMD128-NOT: i8x16
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.or $push0=, $0, $1 # encoding: [0xfd,0x3c]{{$}}
+; SIMD128: return $pop0 #
+define <16 x i8> @or_v16i8(<16 x i8> %x, <16 x i8> %y) {
+  %a = or <16 x i8> %x, %y
+  ret <16 x i8> %a
+}
+
+; CHECK-LABEL: xor_v16i8
+; NO-SIMD128-NOT: i8x16
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.xor $push0=, $0, $1 # encoding: [0xfd,0x3d]{{$}}
+; SIMD128: return $pop0 #
+define <16 x i8> @xor_v16i8(<16 x i8> %x, <16 x i8> %y) {
+  %a = xor <16 x i8> %x, %y
+  ret <16 x i8> %a
+}
+
 ; ==============================================================================
 ; 8 x i16
 ; ==============================================================================
 ; CHECK-LABEL: add_v8i16
 ; NO-SIMD128-NOT: i16x8
 ; SIMD128: .param v128, v128{{$}}
 ; SIMD128: .result v128{{$}}
 ; SIMD128: i16x8.add $push0=, $0, $1 # encoding: [0xfd,0x19]{{$}}

 ; SIMD128: .result v128{{$}}
 ; SIMD128: i16x8.mul $push0=, $0, $1 # encoding: [0xfd,0x21]{{$}}
 ; SIMD128: return $pop0 #
 define <8 x i16> @mul_v8i16(<8 x i16> %x, <8 x i16> %y) {
   %a = mul <8 x i16> %x, %y
   ret <8 x i16> %a
 }
 
+; CHECK-LABEL: and_v8i16
+; NO-SIMD128-NOT: i16x8
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.and $push0=, $0, $1 # encoding: [0xfd,0x3b]{{$}}
+; SIMD128: return $pop0 #
+define <8 x i16> @and_v8i16(<8 x i16> %x, <8 x i16> %y) {
+  %a = and <8 x i16> %x, %y
+  ret <8 x i16> %a
+}
+
+; CHECK-LABEL: or_v8i16
+; NO-SIMD128-NOT: i16x8
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.or $push0=, $0, $1 # encoding: [0xfd,0x3c]{{$}}
+; SIMD128: return $pop0 #
+define <8 x i16> @or_v8i16(<8 x i16> %x, <8 x i16> %y) {
+  %a = or <8 x i16> %x, %y
+  ret <8 x i16> %a
+}
+
+; CHECK-LABEL: xor_v8i16
+; NO-SIMD128-NOT: i16x8
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.xor $push0=, $0, $1 # encoding: [0xfd,0x3d]{{$}}
+; SIMD128: return $pop0 #
+define <8 x i16> @xor_v8i16(<8 x i16> %x, <8 x i16> %y) {
+  %a = xor <8 x i16> %x, %y
+  ret <8 x i16> %a
+}
+
 ; ==============================================================================
 ; 4 x i32
 ; ==============================================================================
 ; CHECK-LABEL: add_v4i32
 ; NO-SIMD128-NOT: i32x4
 ; SIMD128: .param v128, v128{{$}}
 ; SIMD128: .result v128{{$}}
 ; SIMD128: i32x4.add $push0=, $0, $1 # encoding: [0xfd,0x1a]{{$}}

 ; SIMD128: .result v128{{$}}
 ; SIMD128: i32x4.mul $push0=, $0, $1 # encoding: [0xfd,0x22]{{$}}
 ; SIMD128: return $pop0 #
 define <4 x i32> @mul_v4i32(<4 x i32> %x, <4 x i32> %y) {
   %a = mul <4 x i32> %x, %y
   ret <4 x i32> %a
 }
 
+; CHECK-LABEL: and_v4i32
+; NO-SIMD128-NOT: i32x4
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.and $push0=, $0, $1 # encoding: [0xfd,0x3b]{{$}}
+; SIMD128: return $pop0 #
+define <4 x i32> @and_v4i32(<4 x i32> %x, <4 x i32> %y) {
+  %a = and <4 x i32> %x, %y
+  ret <4 x i32> %a
+}
+
+; CHECK-LABEL: or_v4i32
+; NO-SIMD128-NOT: i32x4
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.or $push0=, $0, $1 # encoding: [0xfd,0x3c]{{$}}
+; SIMD128: return $pop0 #
+define <4 x i32> @or_v4i32(<4 x i32> %x, <4 x i32> %y) {
+  %a = or <4 x i32> %x, %y
+  ret <4 x i32> %a
+}
+
+; CHECK-LABEL: xor_v4i32
+; NO-SIMD128-NOT: i32x4
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.xor $push0=, $0, $1 # encoding: [0xfd,0x3d]{{$}}
+; SIMD128: return $pop0 #
+define <4 x i32> @xor_v4i32(<4 x i32> %x, <4 x i32> %y) {
+  %a = xor <4 x i32> %x, %y
+  ret <4 x i32> %a
+}
+
 ; ==============================================================================
 ; 2 x i64
 ; ==============================================================================
 ; CHECK-LABEL: add_v2i64
 ; NO-SIMD128-NOT: i64x2
 ; SIMD128-VM-NOT: i64x2
 ; SIMD128: .param v128, v128{{$}}
 ; SIMD128: .result v128{{$}}

 ; SIMD128-NOT: i64x2.mul
 ; SIMD128: i64x2.extract_lane
 ; SIMD128: i64.mul
 define <2 x i64> @mul_v2i64(<2 x i64> %x, <2 x i64> %y) {
   %a = mul <2 x i64> %x, %y
   ret <2 x i64> %a
 }
 
+; CHECK-LABEL: and_v2i64
+; NO-SIMD128-NOT: i64x2
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.and $push0=, $0, $1 # encoding: [0xfd,0x3b]{{$}}
+; SIMD128: return $pop0 #
+define <2 x i64> @and_v2i64(<2 x i64> %x, <2 x i64> %y) {
+  %a = and <2 x i64> %x, %y
+  ret <2 x i64> %a
+}
+
+; CHECK-LABEL: or_v2i64
+; NO-SIMD128-NOT: i64x2
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.or $push0=, $0, $1 # encoding: [0xfd,0x3c]{{$}}
+; SIMD128: return $pop0 #
+define <2 x i64> @or_v2i64(<2 x i64> %x, <2 x i64> %y) {
+  %a = or <2 x i64> %x, %y
+  ret <2 x i64> %a
+}
+
+; CHECK-LABEL: xor_v2i64
+; NO-SIMD128-NOT: i64x2
+; SIMD128: .param v128, v128{{$}}
+; SIMD128: .result v128{{$}}
+; SIMD128: v128.xor $push0=, $0, $1 # encoding: [0xfd,0x3d]{{$}}
+; SIMD128: return $pop0 #
+define <2 x i64> @xor_v2i64(<2 x i64> %x, <2 x i64> %y) {
+  %a = xor <2 x i64> %x, %y
+  ret <2 x i64> %a
+}
+
 ; ==============================================================================
 ; 4 x float
 ; ==============================================================================
 ; CHECK-LABEL: add_v4f32
 ; NO-SIMD128-NOT: f32x4
 ; SIMD128: .param v128, v128{{$}}
 ; SIMD128: .result v128{{$}}
 ; SIMD128: f32x4.add $push0=, $0, $1 # encoding: [0xfd,0x7a]{{$}}