< prev index next >
src/hotspot/cpu/aarch64/aarch64.ad
Print this page
rev 61247 : Remove AArch64 SVE related code
rev 61248 : Remove AArch64 SVE related code
@@ -2104,11 +2104,11 @@
}
// Identify extra cases that we might want to provide match rules for vector nodes and
// other intrinsics guarded with vector length (vlen) and element type (bt).
const bool Matcher::match_rule_supported_vector(int opcode, int vlen, BasicType bt) {
- if (!match_rule_supported(opcode)) {
+ if (!match_rule_supported(opcode) || !vector_size_supported(bt, vlen)) {
return false;
}
// Special cases which require vector length
switch (opcode) {
@@ -2116,19 +2116,29 @@
if (vlen != 4) {
return false;
}
break;
}
+ case Op_VectorLoadShuffle:
+ case Op_VectorRearrange:
+ if (vlen < 4) {
+ return false;
+ }
+ break;
}
return true; // Per default match rules are supported.
}
const bool Matcher::has_predicated_vectors(void) {
return false;
}
+bool Matcher::supports_vector_variable_shifts(void) {
+ return true;
+}
+
const int Matcher::float_pressure(int default_pressure_threshold) {
return default_pressure_threshold;
}
int Matcher::regnum_to_fpu_offset(int regnum)
@@ -2171,19 +2181,30 @@
// Limits on vector size (number of elements) loaded into vector.
const int Matcher::max_vector_size(const BasicType bt) {
return vector_width_in_bytes(bt)/type2aelembytes(bt);
}
const int Matcher::min_vector_size(const BasicType bt) {
-// For the moment limit the vector size to 8 bytes
+ int max_size = max_vector_size(bt);
+ // Limit the vector size to 8 bytes
int size = 8 / type2aelembytes(bt);
+ if (bt == T_BYTE) {
+ // To support vector api shuffle/rearrange.
+ size = 4;
+ } else if (bt == T_BOOLEAN) {
+ // To support vector api load/store mask.
+ size = 2;
+ }
if (size < 2) size = 2;
- return size;
+ return MIN2(size,max_size);
}
// Vector ideal reg.
const uint Matcher::vector_ideal_reg(int len) {
switch(len) {
+ // For 16-bit/32-bit mask vector, reuse VecD.
+ case 2:
+ case 4:
case 8: return Op_VecD;
case 16: return Op_VecX;
}
ShouldNotReachHere();
return 0;
@@ -2822,10 +2843,16 @@
%}
// END Non-volatile memory access
// Vector loads and stores
+ enc_class aarch64_enc_ldrvH(vecD dst, memory mem) %{
+ FloatRegister dst_reg = as_FloatRegister($dst$$reg);
+ loadStore(C2_MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::H,
+ $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
+ %}
+
enc_class aarch64_enc_ldrvS(vecD dst, memory mem) %{
FloatRegister dst_reg = as_FloatRegister($dst$$reg);
loadStore(C2_MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::S,
$mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
%}
@@ -2840,10 +2867,16 @@
FloatRegister dst_reg = as_FloatRegister($dst$$reg);
loadStore(C2_MacroAssembler(&cbuf), &MacroAssembler::ldr, dst_reg, MacroAssembler::Q,
$mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
%}
+ enc_class aarch64_enc_strvH(vecD src, memory mem) %{
+ FloatRegister src_reg = as_FloatRegister($src$$reg);
+ loadStore(C2_MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, MacroAssembler::H,
+ $mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
+ %}
+
enc_class aarch64_enc_strvS(vecD src, memory mem) %{
FloatRegister src_reg = as_FloatRegister($src$$reg);
loadStore(C2_MacroAssembler(&cbuf), &MacroAssembler::str, src_reg, MacroAssembler::S,
$mem->opcode(), as_Register($mem$$base), $mem$$index, $mem$$scale, $mem$$disp);
%}
@@ -3924,10 +3957,30 @@
op_cost(0);
format %{ %}
interface(CONST_INTER);
%}
+operand immI_2()
+%{
+ predicate(n->get_int() == 2);
+ match(ConI);
+
+ op_cost(0);
+ format %{ %}
+ interface(CONST_INTER);
+%}
+
+operand immI_4()
+%{
+ predicate(n->get_int() == 4);
+ match(ConI);
+
+ op_cost(0);
+ format %{ %}
+ interface(CONST_INTER);
+%}
+
operand immI_8()
%{
predicate(n->get_int() == 8);
match(ConI);
@@ -10826,10 +10879,11 @@
ins_pipe(ialu_reg_shift);
%}
// BEGIN This section of the file is automatically generated. Do not edit --------------
+// This section is generated from aarch64_ad.m4
instruct regL_not_reg(iRegLNoSp dst,
iRegL src1, immL_M1 m1,
rFlagsReg cr) %{
match(Set dst (XorL src1 m1));
@@ -15982,10 +16036,11 @@
// ====================REDUCTION ARITHMETIC====================================
instruct reduce_add2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp, iRegINoSp tmp2)
%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
match(Set dst (AddReductionVI isrc vsrc));
ins_cost(INSN_COST);
effect(TEMP tmp, TEMP tmp2);
format %{ "umov $tmp, $vsrc, S, 0\n\t"
"umov $tmp2, $vsrc, S, 1\n\t"
@@ -16001,10 +16056,11 @@
ins_pipe(pipe_class_default);
%}
instruct reduce_add4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX vtmp, iRegINoSp itmp)
%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
match(Set dst (AddReductionVI isrc vsrc));
ins_cost(INSN_COST);
effect(TEMP vtmp, TEMP itmp);
format %{ "addv $vtmp, T4S, $vsrc\n\t"
"umov $itmp, $vtmp, S, 0\n\t"
@@ -16019,10 +16075,11 @@
ins_pipe(pipe_class_default);
%}
instruct reduce_mul2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
match(Set dst (MulReductionVI isrc vsrc));
ins_cost(INSN_COST);
effect(TEMP tmp, TEMP dst);
format %{ "umov $tmp, $vsrc, S, 0\n\t"
"mul $dst, $tmp, $isrc\n\t"
@@ -16038,10 +16095,11 @@
ins_pipe(pipe_class_default);
%}
instruct reduce_mul4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX vtmp, iRegINoSp itmp)
%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
match(Set dst (MulReductionVI isrc vsrc));
ins_cost(INSN_COST);
effect(TEMP vtmp, TEMP itmp, TEMP dst);
format %{ "ins $vtmp, D, $vsrc, 0, 1\n\t"
"mulv $vtmp, T2S, $vtmp, $vsrc\n\t"
@@ -17027,19 +17085,103 @@
ins_pipe(vsqrt_fp128);
%}
// --------------------------------- ABS --------------------------------------
+instruct vabs8B(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 ||
+ n->as_Vector()->length() == 8);
+ match(Set dst (AbsVB src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (8B)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical64);
+%}
+
+instruct vabs16B(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 16);
+ match(Set dst (AbsVB src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (16B)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical128);
+%}
+
+instruct vabs4S(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4);
+ match(Set dst (AbsVS src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (4H)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical64);
+%}
+
+instruct vabs8S(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 8);
+ match(Set dst (AbsVS src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (8H)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical128);
+%}
+
+instruct vabs2I(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2);
+ match(Set dst (AbsVI src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (2S)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical64);
+%}
+
+instruct vabs4I(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 4);
+ match(Set dst (AbsVI src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (4S)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical128);
+%}
+
+instruct vabs2L(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 2);
+ match(Set dst (AbsVL src));
+ ins_cost(INSN_COST);
+ format %{ "abs $dst, $src\t# vector (2D)" %}
+ ins_encode %{
+ __ absr(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(vlogical128);
+%}
+
instruct vabs2F(vecD dst, vecD src)
%{
predicate(n->as_Vector()->length() == 2);
match(Set dst (AbsVF src));
ins_cost(INSN_COST * 3);
format %{ "fabs $dst,$src\t# vector (2S)" %}
ins_encode %{
- __ fabs(as_FloatRegister($dst$$reg), __ T2S,
- as_FloatRegister($src$$reg));
+ __ fabs(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($src$$reg));
%}
ins_pipe(vunop_fp64);
%}
instruct vabs4F(vecX dst, vecX src)
@@ -17047,12 +17189,11 @@
predicate(n->as_Vector()->length() == 4);
match(Set dst (AbsVF src));
ins_cost(INSN_COST * 3);
format %{ "fabs $dst,$src\t# vector (4S)" %}
ins_encode %{
- __ fabs(as_FloatRegister($dst$$reg), __ T4S,
- as_FloatRegister($src$$reg));
+ __ fabs(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($src$$reg));
%}
ins_pipe(vunop_fp128);
%}
instruct vabs2D(vecX dst, vecX src)
@@ -17060,12 +17201,11 @@
predicate(n->as_Vector()->length() == 2);
match(Set dst (AbsVD src));
ins_cost(INSN_COST * 3);
format %{ "fabs $dst,$src\t# vector (2D)" %}
ins_encode %{
- __ fabs(as_FloatRegister($dst$$reg), __ T2D,
- as_FloatRegister($src$$reg));
+ __ fabs(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($src$$reg));
%}
ins_pipe(vunop_fp128);
%}
// --------------------------------- NEG --------------------------------------
@@ -17202,11 +17342,12 @@
ins_pipe(vlogical128);
%}
// ------------------------------ Shift ---------------------------------------
instruct vshiftcnt8B(vecD dst, iRegIorL2I cnt) %{
- predicate(n->as_Vector()->length_in_bytes() == 8);
+ predicate(n->as_Vector()->length_in_bytes() == 4 ||
+ n->as_Vector()->length_in_bytes() == 8);
match(Set dst (LShiftCntV cnt));
match(Set dst (RShiftCntV cnt));
format %{ "dup $dst, $cnt\t# shift count vector (8B)" %}
ins_encode %{
__ dup(as_FloatRegister($dst$$reg), __ T8B, as_Register($cnt$$reg));
@@ -18055,10 +18196,3443 @@
as_FloatRegister($dst$$reg));
%}
ins_pipe(pipe_class_default);
%}
+// BEGIN This section of the file is automatically generated. Do not edit --------------
+// This section is generated from aarch64_neon_ad.m4
+
+// ====================VECTOR INSTRUCTIONS==================================
+
+// ------------------------------ Load/store/reinterpret -----------------------
+
+// Load vector (16 bits)
+instruct loadV2(vecD dst, memory mem)
+%{
+ predicate(n->as_LoadVector()->memory_size() == 2);
+ match(Set dst (LoadVector mem));
+ ins_cost(4 * INSN_COST);
+ format %{ "ldrh $dst,$mem\t# vector (16 bits)" %}
+ ins_encode( aarch64_enc_ldrvH(dst, mem) );
+ ins_pipe(vload_reg_mem64);
+%}
+
+// Store Vector (16 bits)
+instruct storeV2(vecD src, memory mem)
+%{
+ predicate(n->as_StoreVector()->memory_size() == 2);
+ match(Set mem (StoreVector mem src));
+ ins_cost(4 * INSN_COST);
+ format %{ "strh $mem,$src\t# vector (16 bits)" %}
+ ins_encode( aarch64_enc_strvH(src, mem) );
+ ins_pipe(vstore_reg_mem64);
+%}
+
+instruct reinterpretD(vecD dst)
+%{
+ predicate(n->bottom_type()->is_vect()->length_in_bytes() == 8 &&
+ n->in(1)->bottom_type()->is_vect()->length_in_bytes() == 8);
+ match(Set dst (VectorReinterpret dst));
+ ins_cost(0);
+ format %{ " # reinterpret $dst" %}
+ ins_encode %{
+ // empty
+ %}
+ ins_pipe(pipe_class_empty);
+%}
+
+instruct reinterpretX(vecX dst)
+%{
+ predicate(n->bottom_type()->is_vect()->length_in_bytes() == 16 &&
+ n->in(1)->bottom_type()->is_vect()->length_in_bytes() == 16);
+ match(Set dst (VectorReinterpret dst));
+ ins_cost(0);
+ format %{ " # reinterpret $dst" %}
+ ins_encode %{
+ // empty
+ %}
+ ins_pipe(pipe_class_empty);
+%}
+
+instruct reinterpretD2X(vecX dst, vecD src)
+%{
+ predicate(n->bottom_type()->is_vect()->length_in_bytes() == 16 &&
+ n->in(1)->bottom_type()->is_vect()->length_in_bytes() == 8);
+ match(Set dst (VectorReinterpret src));
+ ins_cost(INSN_COST);
+ format %{ " # reinterpret $dst,$src" %}
+ ins_encode %{
+ // If register is the same, then move is not needed.
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg),
+ as_FloatRegister($src$$reg));
+ }
+ %}
+ ins_pipe(vlogical64);
+%}
+
+instruct reinterpretX2D(vecD dst, vecX src)
+%{
+ predicate(n->bottom_type()->is_vect()->length_in_bytes() == 8 &&
+ n->in(1)->bottom_type()->is_vect()->length_in_bytes() == 16);
+ match(Set dst (VectorReinterpret src));
+ ins_cost(INSN_COST);
+ format %{ " # reinterpret $dst,$src" %}
+ ins_encode %{
+ // If register is the same, then move is not needed.
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg),
+ as_FloatRegister($src$$reg));
+ }
+ %}
+ ins_pipe(vlogical64);
+%}
+
+// ------------------------------ Vector cast -------------------------------
+
+instruct vcvt4Bto4S(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorCastB2X src));
+ format %{ "sxtl $dst, T8H, $src, T8B\t# convert 4B to 4S vector" %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt8Bto8S(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorCastB2X src));
+ format %{ "sxtl $dst, T8H, $src, T8B\t# convert 8B to 8S vector" %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt4Sto4B(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorCastS2X src));
+ format %{ "xtn $dst, T8B, $src, T8H\t# convert 4S to 4B vector" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg), __ T8H);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt8Sto8B(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorCastS2X src));
+ format %{ "xtn $dst, T8B, $src, T8H\t# convert 8S to 8B vector" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg), __ T8H);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt4Sto4I(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorCastS2X src));
+ format %{ "sxtl $dst, T4S, $src, T4H\t# convert 4S to 4I vector" %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($src$$reg), __ T4H);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt4Ito4S(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorCastI2X src));
+ format %{ "xtn $dst, T4H, $src, T4S\t# convert 4I to 4S vector" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($src$$reg), __ T4S);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Ito2L(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorCastI2X src));
+ format %{ "sxtl $dst, T2D, $src, T2S\t# convert 2I to 2L vector" %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($src$$reg), __ T2S);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Lto2I(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorCastL2X src));
+ format %{ "xtn $dst, T2S, $src, T2D\t# convert 2L to 2I vector" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($src$$reg), __ T2D);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt4Bto4I(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorCastB2X src));
+ format %{ "sxtl $dst, T8H, $src, T8B\n\t"
+ "sxtl $dst, T4S, $dst, T4H\t# convert 4B to 4I vector"
+ %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ sxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcvt4Ito4B(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorCastI2X src));
+ format %{ "xtn $dst, T4H, $src, T4S\n\t"
+ "xtn $dst, T8B, $dst, T8H\t# convert 4I to 4B vector"
+ %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($src$$reg), __ T4S);
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg), __ T8H);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcvt4Bto4F(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastB2X src));
+ format %{ "sxtl $dst, T8H, $src, T8B\n\t"
+ "sxtl $dst, T4S, $dst, T4H\n\t"
+ "scvtfv T4S, $dst, $dst\t# convert 4B to 4F vector"
+ %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ sxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ __ scvtfv(__ T4S, as_FloatRegister($dst$$reg), as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcvt4Sto4F(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastS2X src));
+ format %{ "sxtl $dst, T4S, $src, T4H\n\t"
+ "scvtfv T4S, $dst, $dst\t# convert 4S to 4F vector"
+ %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($src$$reg), __ T4H);
+ __ scvtfv(__ T4S, as_FloatRegister($dst$$reg), as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcvt2Ito2D(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorCastI2X src));
+ format %{ "sxtl $dst, T2D, $src, T2S\n\t"
+ "scvtfv T2D, $dst, $dst\t# convert 2I to 2D vector"
+ %}
+ ins_encode %{
+ __ sxtl(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($src$$reg), __ T2S);
+ __ scvtfv(__ T2D, as_FloatRegister($dst$$reg), as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcvt2Ito2F(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastI2X src));
+ format %{ "scvtfv T2S, $dst, $src\t# convert 2I to 2F vector" %}
+ ins_encode %{
+ __ scvtfv(__ T2S, as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt4Ito4F(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastI2X src));
+ format %{ "scvtfv T4S, $dst, $src\t# convert 4I to 4F vector" %}
+ ins_encode %{
+ __ scvtfv(__ T4S, as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Lto2D(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorCastL2X src));
+ format %{ "scvtfv T2D, $dst, $src\t# convert 2L to 2D vector" %}
+ ins_encode %{
+ __ scvtfv(__ T2D, as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Fto2D(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorCastF2X src));
+ format %{ "fcvtl $dst, T2D, $src, T2S\t# convert 2F to 2D vector" %}
+ ins_encode %{
+ __ fcvtl(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($src$$reg), __ T2S);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Dto2F(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastD2X src));
+ format %{ "fcvtn $dst, T2S, $src, T2D\t# convert 2D to 2F vector" %}
+ ins_encode %{
+ __ fcvtn(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($src$$reg), __ T2D);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vcvt2Lto2F(vecD dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorCastL2X src));
+ format %{ "scvtfv T2D, $dst, $src\n\t"
+ "fcvtn $dst, T2S, $dst, T2D\t# convert 2L to 2F vector"
+ %}
+ ins_encode %{
+ __ scvtfv(__ T2D, as_FloatRegister($dst$$reg), as_FloatRegister($src$$reg));
+ __ fcvtn(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($dst$$reg), __ T2D);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+// ------------------------------ Reduction -------------------------------
+
+instruct reduce_add8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (AddReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "addv $tmp, T8B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "addw $dst, $dst, $isrc\n\t"
+ "sxtb $dst, $dst\t# add reduction8B"
+ %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($vsrc$$reg));
+ __ smov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ addw($dst$$Register, $dst$$Register, $isrc$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_add16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (AddReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "addv $tmp, T16B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "addw $dst, $dst, $isrc\n\t"
+ "sxtb $dst, $dst\t# add reduction16B"
+ %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($vsrc$$reg));
+ __ smov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ addw($dst$$Register, $dst$$Register, $isrc$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_add4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (AddReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "addv $tmp, T4H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "addw $dst, $dst, $isrc\n\t"
+ "sxth $dst, $dst\t# add reduction4S"
+ %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T4H, as_FloatRegister($vsrc$$reg));
+ __ smov($dst$$Register, as_FloatRegister($tmp$$reg), __ H, 0);
+ __ addw($dst$$Register, $dst$$Register, $isrc$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_add8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (AddReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "addv $tmp, T8H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "addw $dst, $dst, $isrc\n\t"
+ "sxth $dst, $dst\t# add reduction8S"
+ %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T8H, as_FloatRegister($vsrc$$reg));
+ __ smov($dst$$Register, as_FloatRegister($tmp$$reg), __ H, 0);
+ __ addw($dst$$Register, $dst$$Register, $isrc$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_add2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, vecX tmp)
+%{
+ match(Set dst (AddReductionVL isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "addpd $tmp, $vsrc\n\t"
+ "umov $dst, $tmp, D, 0\n\t"
+ "add $dst, $isrc, $dst\t# add reduction2L"
+ %}
+ ins_encode %{
+ __ addpd(as_FloatRegister($tmp$$reg), as_FloatRegister($vsrc$$reg));
+ __ umov($dst$$Register, as_FloatRegister($tmp$$reg), __ D, 0);
+ __ add($dst$$Register, $isrc$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_mul8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD vtmp1, vecD vtmp2, iRegINoSp itmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MulReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP vtmp1, TEMP vtmp2, TEMP itmp);
+ format %{ "ins $vtmp1, S, $vsrc, 0, 1\n\t"
+ "mulv $vtmp1, T8B, $vtmp1, $vsrc\n\t"
+ "ins $vtmp2, H, $vtmp1, 0, 1\n\t"
+ "mulv $vtmp2, T8B, $vtmp2, $vtmp1\n\t"
+ "umov $itmp, $vtmp2, B, 0\n\t"
+ "mulw $dst, $itmp, $isrc\n\t"
+ "sxtb $dst, $dst\n\t"
+ "umov $itmp, $vtmp2, B, 1\n\t"
+ "mulw $dst, $itmp, $dst\n\t"
+ "sxtb $dst, $dst\t# mul reduction8B"
+ %}
+ ins_encode %{
+ __ ins(as_FloatRegister($vtmp1$$reg), __ S,
+ as_FloatRegister($vsrc$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp1$$reg), __ T8B,
+ as_FloatRegister($vtmp1$$reg), as_FloatRegister($vsrc$$reg));
+ __ ins(as_FloatRegister($vtmp2$$reg), __ H,
+ as_FloatRegister($vtmp1$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp2$$reg), __ T8B,
+ as_FloatRegister($vtmp2$$reg), as_FloatRegister($vtmp1$$reg));
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ B, 0);
+ __ mulw($dst$$Register, $itmp$$Register, $isrc$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ B, 1);
+ __ mulw($dst$$Register, $itmp$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_mul16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX vtmp1, vecX vtmp2, iRegINoSp itmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MulReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP vtmp1, TEMP vtmp2, TEMP itmp);
+ format %{ "ins $vtmp1, D, $vsrc, 0, 1\n\t"
+ "mulv $vtmp1, T8B, $vtmp1, $vsrc\n\t"
+ "ins $vtmp2, S, $vtmp1, 0, 1\n\t"
+ "mulv $vtmp1, T8B, $vtmp2, $vtmp1\n\t"
+ "ins $vtmp2, H, $vtmp1, 0, 1\n\t"
+ "mulv $vtmp2, T8B, $vtmp2, $vtmp1\n\t"
+ "umov $itmp, $vtmp2, B, 0\n\t"
+ "mulw $dst, $itmp, $isrc\n\t"
+ "sxtb $dst, $dst\n\t"
+ "umov $itmp, $vtmp2, B, 1\n\t"
+ "mulw $dst, $itmp, $dst\n\t"
+ "sxtb $dst, $dst\t# mul reduction16B"
+ %}
+ ins_encode %{
+ __ ins(as_FloatRegister($vtmp1$$reg), __ D,
+ as_FloatRegister($vsrc$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp1$$reg), __ T8B,
+ as_FloatRegister($vtmp1$$reg), as_FloatRegister($vsrc$$reg));
+ __ ins(as_FloatRegister($vtmp2$$reg), __ S,
+ as_FloatRegister($vtmp1$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp1$$reg), __ T8B,
+ as_FloatRegister($vtmp2$$reg), as_FloatRegister($vtmp1$$reg));
+ __ ins(as_FloatRegister($vtmp2$$reg), __ H,
+ as_FloatRegister($vtmp1$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp2$$reg), __ T8B,
+ as_FloatRegister($vtmp2$$reg), as_FloatRegister($vtmp1$$reg));
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ B, 0);
+ __ mulw($dst$$Register, $itmp$$Register, $isrc$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ B, 1);
+ __ mulw($dst$$Register, $itmp$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_mul4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD vtmp, iRegINoSp itmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MulReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP vtmp, TEMP itmp);
+ format %{ "ins $vtmp, S, $vsrc, 0, 1\n\t"
+ "mulv $vtmp, T4H, $vtmp, $vsrc\n\t"
+ "umov $itmp, $vtmp, H, 0\n\t"
+ "mulw $dst, $itmp, $isrc\n\t"
+ "sxth $dst, $dst\n\t"
+ "umov $itmp, $vtmp, H, 1\n\t"
+ "mulw $dst, $itmp, $dst\n\t"
+ "sxth $dst, $dst\t# mul reduction4S"
+ %}
+ ins_encode %{
+ __ ins(as_FloatRegister($vtmp$$reg), __ S,
+ as_FloatRegister($vsrc$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp$$reg), __ T4H,
+ as_FloatRegister($vtmp$$reg), as_FloatRegister($vsrc$$reg));
+ __ umov($itmp$$Register, as_FloatRegister($vtmp$$reg), __ H, 0);
+ __ mulw($dst$$Register, $itmp$$Register, $isrc$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ __ umov($itmp$$Register, as_FloatRegister($vtmp$$reg), __ H, 1);
+ __ mulw($dst$$Register, $itmp$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_mul8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX vtmp1, vecX vtmp2, iRegINoSp itmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MulReductionVI isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP vtmp1, TEMP vtmp2, TEMP itmp);
+ format %{ "ins $vtmp1, D, $vsrc, 0, 1\n\t"
+ "mulv $vtmp1, T4H, $vtmp1, $vsrc\n\t"
+ "ins $vtmp2, S, $vtmp1, 0, 1\n\t"
+ "mulv $vtmp2, T4H, $vtmp2, $vtmp1\n\t"
+ "umov $itmp, $vtmp2, H, 0\n\t"
+ "mulw $dst, $itmp, $isrc\n\t"
+ "sxth $dst, $dst\n\t"
+ "umov $itmp, $vtmp2, H, 1\n\t"
+ "mulw $dst, $itmp, $dst\n\t"
+ "sxth $dst, $dst\t# mul reduction8S"
+ %}
+ ins_encode %{
+ __ ins(as_FloatRegister($vtmp1$$reg), __ D,
+ as_FloatRegister($vsrc$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp1$$reg), __ T4H,
+ as_FloatRegister($vtmp1$$reg), as_FloatRegister($vsrc$$reg));
+ __ ins(as_FloatRegister($vtmp2$$reg), __ S,
+ as_FloatRegister($vtmp1$$reg), 0, 1);
+ __ mulv(as_FloatRegister($vtmp2$$reg), __ T4H,
+ as_FloatRegister($vtmp2$$reg), as_FloatRegister($vtmp1$$reg));
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ H, 0);
+ __ mulw($dst$$Register, $itmp$$Register, $isrc$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ __ umov($itmp$$Register, as_FloatRegister($vtmp2$$reg), __ H, 1);
+ __ mulw($dst$$Register, $itmp$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_mul2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp)
+%{
+ match(Set dst (MulReductionVL isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "mul $dst, $isrc, $tmp\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "mul $dst, $dst, $tmp\t# mul reduction2L"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ mul($dst$$Register, $isrc$$Register, $tmp$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ mul($dst$$Register, $dst$$Register, $tmp$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "smaxv $tmp, T8B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction8B"
+ %}
+ ins_encode %{
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "smaxv $tmp, T16B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction16B"
+ %}
+ ins_encode %{
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "smaxv $tmp, T4H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction4S"
+ %}
+ ins_encode %{
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T4H, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ H, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "smaxv $tmp, T8H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction8S"
+ %}
+ ins_encode %{
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T8H, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ H, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "smaxv $tmp, T4S, $vsrc\n\t"
+ "umov $dst, $tmp, S, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction4I"
+ %}
+ ins_encode %{
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T4S, as_FloatRegister($vsrc$$reg));
+ __ umov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ S, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "sminv $tmp, T8B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction8B"
+ %}
+ ins_encode %{
+ __ sminv(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "sminv $tmp, T16B, $vsrc\n\t"
+ "smov $dst, $tmp, B, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction16B"
+ %}
+ ins_encode %{
+ __ sminv(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecD tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "sminv $tmp, T4H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction4S"
+ %}
+ ins_encode %{
+ __ sminv(as_FloatRegister($tmp$$reg), __ T4H, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ H, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "sminv $tmp, T8H, $vsrc\n\t"
+ "smov $dst, $tmp, H, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction8S"
+ %}
+ ins_encode %{
+ __ sminv(as_FloatRegister($tmp$$reg), __ T8H, as_FloatRegister($vsrc$$reg));
+ __ smov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ H, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "sminv $tmp, T4S, $vsrc\n\t"
+ "umov $dst, $tmp, S, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction4I"
+ %}
+ ins_encode %{
+ __ sminv(as_FloatRegister($tmp$$reg), __ T4S, as_FloatRegister($vsrc$$reg));
+ __ umov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ S, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "dup $tmp, T2D, $vsrc\n\t"
+ "smaxv $tmp, T4S, $tmp\n\t"
+ "umov $dst, $tmp, S, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc GT\t# max reduction2I"
+ %}
+ ins_encode %{
+ __ dup(as_FloatRegister($tmp$$reg), __ T2D, as_FloatRegister($vsrc$$reg));
+ __ smaxv(as_FloatRegister($tmp$$reg), __ T4S, as_FloatRegister($tmp$$reg));
+ __ umov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ S, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, vecX tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "dup $tmp, T2D, $vsrc\n\t"
+ "sminv $tmp, T4S, $tmp\n\t"
+ "umov $dst, $tmp, S, 0\n\t"
+ "cmpw $dst, $isrc\n\t"
+ "cselw $dst, $dst, $isrc LT\t# min reduction2I"
+ %}
+ ins_encode %{
+ __ dup(as_FloatRegister($tmp$$reg), __ T2D, as_FloatRegister($vsrc$$reg));
+ __ sminv(as_FloatRegister($tmp$$reg), __ T4S, as_FloatRegister($tmp$$reg));
+ __ umov(as_Register($dst$$reg), as_FloatRegister($tmp$$reg), __ S, 0);
+ __ cmpw(as_Register($dst$$reg), as_Register($isrc$$reg));
+ __ cselw(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($isrc$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_max2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (MaxReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "cmp $isrc,$tmp\n\t"
+ "csel $dst, $isrc, $tmp GT\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "cmp $dst, $tmp\n\t"
+ "csel $dst, $dst, $tmp GT\t# max reduction2L"
+ %}
+ ins_encode %{
+ __ umov(as_Register($tmp$$reg), as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ cmp(as_Register($isrc$$reg), as_Register($tmp$$reg));
+ __ csel(as_Register($dst$$reg), as_Register($isrc$$reg), as_Register($tmp$$reg), Assembler::GT);
+ __ umov(as_Register($tmp$$reg), as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ cmp(as_Register($dst$$reg), as_Register($tmp$$reg));
+ __ csel(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($tmp$$reg), Assembler::GT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_min2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp, rFlagsReg cr)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (MinReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp, KILL cr);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "cmp $isrc,$tmp\n\t"
+ "csel $dst, $isrc, $tmp LT\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "cmp $dst, $tmp\n\t"
+ "csel $dst, $dst, $tmp LT\t# min reduction2L"
+ %}
+ ins_encode %{
+ __ umov(as_Register($tmp$$reg), as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ cmp(as_Register($isrc$$reg), as_Register($tmp$$reg));
+ __ csel(as_Register($dst$$reg), as_Register($isrc$$reg), as_Register($tmp$$reg), Assembler::LT);
+ __ umov(as_Register($tmp$$reg), as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ cmp(as_Register($dst$$reg), as_Register($tmp$$reg));
+ __ csel(as_Register($dst$$reg), as_Register($dst$$reg), as_Register($tmp$$reg), Assembler::LT);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "andw $dst, $dst, $tmp\n\t"
+ "andw $dst, $dst, $dst, LSR #16\n\t"
+ "andw $dst, $dst, $dst, LSR #8\n\t"
+ "andw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# and reduction8B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ andw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ andw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "orrw $dst, $dst, $tmp\n\t"
+ "orrw $dst, $dst, $dst, LSR #16\n\t"
+ "orrw $dst, $dst, $dst, LSR #8\n\t"
+ "orrw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# orr reduction8B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ orrw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ orrw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor8B(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "eorw $dst, $dst, $tmp\n\t"
+ "eorw $dst, $dst, $dst, LSR #16\n\t"
+ "eorw $dst, $dst, $dst, LSR #8\n\t"
+ "eorw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# eor reduction8B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ eorw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ eorw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "andr $dst, $dst, $tmp\n\t"
+ "andr $dst, $dst, $dst, LSR #32\n\t"
+ "andw $dst, $dst, $dst, LSR #16\n\t"
+ "andw $dst, $dst, $dst, LSR #8\n\t"
+ "andw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# and reduction16B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ andr($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ andr($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ andw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "orr $dst, $dst, $tmp\n\t"
+ "orr $dst, $dst, $dst, LSR #32\n\t"
+ "orrw $dst, $dst, $dst, LSR #16\n\t"
+ "orrw $dst, $dst, $dst, LSR #8\n\t"
+ "orrw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# orr reduction16B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ orr ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ orr ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ orrw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor16B(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "eor $dst, $dst, $tmp\n\t"
+ "eor $dst, $dst, $dst, LSR #32\n\t"
+ "eorw $dst, $dst, $dst, LSR #16\n\t"
+ "eorw $dst, $dst, $dst, LSR #8\n\t"
+ "eorw $dst, $isrc, $dst\n\t"
+ "sxtb $dst, $dst\t# eor reduction16B"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ eor ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ eor ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 8);
+ __ eorw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxtb($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "andw $dst, $dst, $tmp\n\t"
+ "andw $dst, $dst, $dst, LSR #16\n\t"
+ "andw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# and reduction4S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ andw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ andw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "orrw $dst, $dst, $tmp\n\t"
+ "orrw $dst, $dst, $dst, LSR #16\n\t"
+ "orrw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# orr reduction4S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ orrw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ orrw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor4S(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "umov $dst, $vsrc, S, 1\n\t"
+ "eorw $dst, $dst, $tmp\n\t"
+ "eorw $dst, $dst, $dst, LSR #16\n\t"
+ "eorw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# eor reduction4S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ eorw($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ eorw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "andr $dst, $dst, $tmp\n\t"
+ "andr $dst, $dst, $dst, LSR #32\n\t"
+ "andw $dst, $dst, $dst, LSR #16\n\t"
+ "andw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# and reduction8S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ andr($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ andr($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ andw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ andw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "orr $dst, $dst, $tmp\n\t"
+ "orr $dst, $dst, $dst, LSR #32\n\t"
+ "orrw $dst, $dst, $dst, LSR #16\n\t"
+ "orrw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# orr reduction8S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ orr ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ orr ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ orrw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ orrw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor8S(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "eor $dst, $dst, $tmp\n\t"
+ "eor $dst, $dst, $dst, LSR #32\n\t"
+ "eorw $dst, $dst, $dst, LSR #16\n\t"
+ "eorw $dst, $isrc, $dst\n\t"
+ "sxth $dst, $dst\t# eor reduction8S"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ eor ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ eor ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ eorw($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 16);
+ __ eorw($dst$$Register, $isrc$$Register, $dst$$Register);
+ __ sxth($dst$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "andw $dst, $tmp, $isrc\n\t"
+ "umov $tmp, $vsrc, S, 1\n\t"
+ "andw $dst, $tmp, $dst\t# and reduction2I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ andw($dst$$Register, $tmp$$Register, $isrc$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ andw($dst$$Register, $tmp$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "orrw $dst, $tmp, $isrc\n\t"
+ "umov $tmp, $vsrc, S, 1\n\t"
+ "orrw $dst, $tmp, $dst\t# orr reduction2I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ orrw($dst$$Register, $tmp$$Register, $isrc$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ orrw($dst$$Register, $tmp$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor2I(iRegINoSp dst, iRegIorL2I isrc, vecD vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, S, 0\n\t"
+ "eorw $dst, $tmp, $isrc\n\t"
+ "umov $tmp, $vsrc, S, 1\n\t"
+ "eorw $dst, $tmp, $dst\t# eor reduction2I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 0);
+ __ eorw($dst$$Register, $tmp$$Register, $isrc$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ S, 1);
+ __ eorw($dst$$Register, $tmp$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "andr $dst, $dst, $tmp\n\t"
+ "andr $dst, $dst, $dst, LSR #32\n\t"
+ "andw $dst, $isrc, $dst\t# and reduction4I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ andr($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ andr($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ andw($dst$$Register, $isrc$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "orr $dst, $dst, $tmp\n\t"
+ "orr $dst, $dst, $dst, LSR #32\n\t"
+ "orrw $dst, $isrc, $dst\t# orr reduction4I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ orr ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ orr ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ orrw($dst$$Register, $isrc$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor4I(iRegINoSp dst, iRegIorL2I isrc, vecX vsrc, iRegINoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "umov $dst, $vsrc, D, 1\n\t"
+ "eor $dst, $dst, $tmp\n\t"
+ "eor $dst, $dst, $dst, LSR #32\n\t"
+ "eorw $dst, $isrc, $dst\t# eor reduction4I"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ umov($dst$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ eor ($dst$$Register, $dst$$Register, $tmp$$Register);
+ __ eor ($dst$$Register, $dst$$Register, $dst$$Register, Assembler::LSR, 32);
+ __ eorw($dst$$Register, $isrc$$Register, $dst$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_and2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (AndReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "andr $dst, $isrc, $tmp\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "andr $dst, $dst, $tmp\t# and reduction2L"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ andr($dst$$Register, $isrc$$Register, $tmp$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ andr($dst$$Register, $dst$$Register, $tmp$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_orr2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (OrReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "orr $dst, $isrc, $tmp\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "orr $dst, $dst, $tmp\t# orr reduction2L"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ orr ($dst$$Register, $isrc$$Register, $tmp$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ orr ($dst$$Register, $dst$$Register, $tmp$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct reduce_eor2L(iRegLNoSp dst, iRegL isrc, vecX vsrc, iRegLNoSp tmp)
+%{
+ predicate(n->in(2)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (XorReductionV isrc vsrc));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp);
+ format %{ "umov $tmp, $vsrc, D, 0\n\t"
+ "eor $dst, $isrc, $tmp\n\t"
+ "umov $tmp, $vsrc, D, 1\n\t"
+ "eor $dst, $dst, $tmp\t# eor reduction2L"
+ %}
+ ins_encode %{
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 0);
+ __ eor ($dst$$Register, $isrc$$Register, $tmp$$Register);
+ __ umov($tmp$$Register, as_FloatRegister($vsrc$$reg), __ D, 1);
+ __ eor ($dst$$Register, $dst$$Register, $tmp$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+// ------------------------------ Vector insert ---------------------------------
+
+instruct insert8B(vecD dst, vecD src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T8B, $src, $src\n\t"
+ "mov $dst, T8B, $idx, $val\t# insert into vector(8B)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T8B, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert16B(vecX dst, vecX src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "mov $dst, T16B, $idx, $val\t# insert into vector(16B)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T16B, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert4S(vecD dst, vecD src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T8B, $src, $src\n\t"
+ "mov $dst, T4H, $idx, $val\t# insert into vector(4S)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T4H, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert8S(vecX dst, vecX src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "mov $dst, T8H, $idx, $val\t# insert into vector(8S)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T8H, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert2I(vecD dst, vecD src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T8B, $src, $src\n\t"
+ "mov $dst, T2S, $idx, $val\t# insert into vector(2I)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T2S, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert4I(vecX dst, vecX src, iRegIorL2I val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "mov $dst, T4S, $idx, $val\t# insert into vector(4I)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T4S, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert2L(vecX dst, vecX src, iRegL val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "mov $dst, T2D, $idx, $val\t# insert into vector(2L)" %}
+ ins_encode %{
+ if (as_FloatRegister($dst$$reg) != as_FloatRegister($src$$reg)) {
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ }
+ __ mov(as_FloatRegister($dst$$reg), __ T2D, $idx$$constant, $val$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert2F(vecD dst, vecD src, vRegF val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst);
+ format %{ "orr $dst, T8B, $src, $src\n\t"
+ "ins $dst, S, $val, $idx, 0\t# insert into vector(2F)" %}
+ ins_encode %{
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ __ ins(as_FloatRegister($dst$$reg), __ S,
+ as_FloatRegister($val$$reg), $idx$$constant, 0);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert4F(vecX dst, vecX src, vRegF val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "ins $dst, S, $val, $idx, 0\t# insert into vector(4F)" %}
+ ins_encode %{
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ __ ins(as_FloatRegister($dst$$reg), __ S,
+ as_FloatRegister($val$$reg), $idx$$constant, 0);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct insert2D(vecX dst, vecX src, vRegD val, immI idx)
+%{
+ predicate(n->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorInsert (Binary src val) idx));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst);
+ format %{ "orr $dst, T16B, $src, $src\n\t"
+ "ins $dst, D, $val, $idx, 0\t# insert into vector(2D)" %}
+ ins_encode %{
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ __ ins(as_FloatRegister($dst$$reg), __ D,
+ as_FloatRegister($val$$reg), $idx$$constant, 0);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+// ------------------------------ Vector extract ---------------------------------
+
+instruct extract8B(iRegINoSp dst, vecD src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 8);
+ match(Set dst (ExtractB src idx));
+ ins_cost(INSN_COST);
+ format %{ "smov $dst, $src, B, $idx\t# extract from vector(8B)" %}
+ ins_encode %{
+ __ smov($dst$$Register, as_FloatRegister($src$$reg), __ B, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract16B(iRegINoSp dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 16);
+ match(Set dst (ExtractB src idx));
+ ins_cost(INSN_COST);
+ format %{ "smov $dst, $src, B, $idx\t# extract from vector(16B)" %}
+ ins_encode %{
+ __ smov($dst$$Register, as_FloatRegister($src$$reg), __ B, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract4S(iRegINoSp dst, vecD src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 4);
+ match(Set dst (ExtractS src idx));
+ ins_cost(INSN_COST);
+ format %{ "smov $dst, $src, H, $idx\t# extract from vector(4S)" %}
+ ins_encode %{
+ __ smov($dst$$Register, as_FloatRegister($src$$reg), __ H, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract8S(iRegINoSp dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 8);
+ match(Set dst (ExtractS src idx));
+ ins_cost(INSN_COST);
+ format %{ "smov $dst, $src, H, $idx\t# extract from vector(8S)" %}
+ ins_encode %{
+ __ smov($dst$$Register, as_FloatRegister($src$$reg), __ H, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract2I(iRegINoSp dst, vecD src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 2);
+ match(Set dst (ExtractI src idx));
+ ins_cost(INSN_COST);
+ format %{ "umov $dst, $src, S, $idx\t# extract from vector(2I)" %}
+ ins_encode %{
+ __ umov($dst$$Register, as_FloatRegister($src$$reg), __ S, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract4I(iRegINoSp dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 4);
+ match(Set dst (ExtractI src idx));
+ ins_cost(INSN_COST);
+ format %{ "umov $dst, $src, S, $idx\t# extract from vector(4I)" %}
+ ins_encode %{
+ __ umov($dst$$Register, as_FloatRegister($src$$reg), __ S, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract2L(iRegLNoSp dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 2);
+ match(Set dst (ExtractL src idx));
+ ins_cost(INSN_COST);
+ format %{ "umov $dst, $src, D, $idx\t# extract from vector(2L)" %}
+ ins_encode %{
+ __ umov($dst$$Register, as_FloatRegister($src$$reg), __ D, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract2F(vRegF dst, vecD src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 2);
+ match(Set dst (ExtractF src idx));
+ ins_cost(INSN_COST);
+ format %{ "ins $dst, S, $src, 0, $idx\t# extract from vector(2F)" %}
+ ins_encode %{
+ __ ins(as_FloatRegister($dst$$reg), __ S,
+ as_FloatRegister($src$$reg), 0, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract4F(vRegF dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 4);
+ match(Set dst (ExtractF src idx));
+ ins_cost(INSN_COST);
+ format %{ "ins $dst, S, $src, 0, $idx\t# extract from vector(4F)" %}
+ ins_encode %{
+ __ ins(as_FloatRegister($dst$$reg), __ S,
+ as_FloatRegister($src$$reg), 0, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct extract2D(vRegD dst, vecX src, immI idx)
+%{
+ predicate(n->in(1)->bottom_type()->is_vect()->length() == 2);
+ match(Set dst (ExtractD src idx));
+ ins_cost(INSN_COST);
+ format %{ "ins $dst, D, $src, 0, $idx\t# extract from vector(2D)" %}
+ ins_encode %{
+ __ ins(as_FloatRegister($dst$$reg), __ D,
+ as_FloatRegister($src$$reg), 0, $idx$$constant);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+// ------------------------------ Vector comparison ---------------------------------
+
+instruct vcmeq8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (8B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmeq16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (16B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmeq4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (4S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmeq8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (8S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmeq2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (2I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmeq4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (4I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmeq2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\t# vector cmp (2L)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmeq2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\t# vector cmp (2F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmeq4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\t# vector cmp (4F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmeq2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::eq &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\t# vector cmp (2D)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (8B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmgt16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (16B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (4S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmgt8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (8S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (2I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmgt4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (4I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src1, $src2\t# vector cmp (2L)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src1, $src2\t# vector cmp (2F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmgt4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src1, $src2\t# vector cmp (4F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmgt2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::gt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src1, $src2\t# vector cmp (2D)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (8B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmge16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (16B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (4S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmge8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (8S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (2I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmge4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (4I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src1, $src2\t# vector cmp (2L)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src1, $src2\t# vector cmp (2F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmge4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src1, $src2\t# vector cmp (4F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmge2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ge &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src1, $src2\t# vector cmp (2D)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmne8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (8B)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (16B)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (4S)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (8S)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (2I)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (4I)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmeq $dst, $src1, $src2\n\t# vector cmp (2L)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmeq(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\n\t# vector cmp (2F)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\n\t# vector cmp (4F)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmne2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::ne &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmeq $dst, $src1, $src2\n\t# vector cmp (2D)"
+ "not $dst, $dst\t" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmeq(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct vcmlt8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (8B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmlt16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (16B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmlt4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (4S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmlt8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (8S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmlt2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (2I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmlt4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (4I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmlt2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmgt $dst, $src2, $src1\t# vector cmp (2L)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmlt2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src2, $src1\t# vector cmp (2F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmlt4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src2, $src1\t# vector cmp (4F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmlt2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::lt &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmgt $dst, $src2, $src1\t# vector cmp (2D)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle8B(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (8B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmle16B(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (16B)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle4S(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (4S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmle8S(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (8S)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle2I(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (2I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmle4I(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (4I)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle2L(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "cmge $dst, $src2, $src1\t# vector cmp (2L)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ cmge(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle2F(vecD dst, vecD src1, vecD src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src2, $src1\t# vector cmp (2F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vcmle4F(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_FLOAT);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src2, $src1\t# vector cmp (4F)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vcmle2D(vecX dst, vecX src1, vecX src2, immI cond)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ n->as_VectorMaskCmp()->get_predicate() == BoolTest::le &&
+ n->in(1)->in(1)->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE);
+ match(Set dst (VectorMaskCmp (Binary src1 src2) cond));
+ format %{ "fcmge $dst, $src2, $src1\t# vector cmp (2D)" %}
+ ins_cost(INSN_COST);
+ ins_encode %{
+ __ fcmge(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+// ------------------------------ Vector mul -----------------------------------
+
+instruct vmul2L(vecX dst, vecX src1, vecX src2, iRegLNoSp tmp1, iRegLNoSp tmp2)
+%{
+ predicate(n->as_Vector()->length() == 2);
+ match(Set dst (MulVL src1 src2));
+ ins_cost(INSN_COST);
+ effect(TEMP tmp1, TEMP tmp2);
+ format %{ "umov $tmp1, $src1, D, 0\n\t"
+ "umov $tmp2, $src2, D, 0\n\t"
+ "mul $tmp2, $tmp2, $tmp1\n\t"
+ "mov $dst, T2D, 0, $tmp2\t# insert into vector(2L)\n\t"
+ "umov $tmp1, $src1, D, 1\n\t"
+ "umov $tmp2, $src2, D, 1\n\t"
+ "mul $tmp2, $tmp2, $tmp1\n\t"
+ "mov $dst, T2D, 1, $tmp2\t# insert into vector(2L)\n\t"
+ %}
+ ins_encode %{
+ __ umov($tmp1$$Register, as_FloatRegister($src1$$reg), __ D, 0);
+ __ umov($tmp2$$Register, as_FloatRegister($src2$$reg), __ D, 0);
+ __ mul(as_Register($tmp2$$reg), as_Register($tmp2$$reg), as_Register($tmp1$$reg));
+ __ mov(as_FloatRegister($dst$$reg), __ T2D, 0, $tmp2$$Register);
+ __ umov($tmp1$$Register, as_FloatRegister($src1$$reg), __ D, 1);
+ __ umov($tmp2$$Register, as_FloatRegister($src2$$reg), __ D, 1);
+ __ mul(as_Register($tmp2$$reg), as_Register($tmp2$$reg), as_Register($tmp1$$reg));
+ __ mov(as_FloatRegister($dst$$reg), __ T2D, 1, $tmp2$$Register);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+// --------------------------------- Vector not --------------------------------
+
+instruct vnot2I(vecD dst, vecD src, immI_M1 m1)
+%{
+ predicate(n->as_Vector()->length_in_bytes() == 8);
+ match(Set dst (XorV src (ReplicateB m1)));
+ match(Set dst (XorV src (ReplicateS m1)));
+ match(Set dst (XorV src (ReplicateI m1)));
+ ins_cost(INSN_COST);
+ format %{ "not $dst, $src\t# vector (8B)" %}
+ ins_encode %{
+ __ notr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vnot4I(vecX dst, vecX src, immI_M1 m1)
+%{
+ predicate(n->as_Vector()->length_in_bytes() == 16);
+ match(Set dst (XorV src (ReplicateB m1)));
+ match(Set dst (XorV src (ReplicateS m1)));
+ match(Set dst (XorV src (ReplicateI m1)));
+ ins_cost(INSN_COST);
+ format %{ "not $dst, $src\t# vector (16B)" %}
+ ins_encode %{
+ __ notr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct vnot2L(vecX dst, vecX src, immL_M1 m1)
+%{
+ predicate(n->as_Vector()->length_in_bytes() == 16);
+ match(Set dst (XorV src (ReplicateL m1)));
+ ins_cost(INSN_COST);
+ format %{ "not $dst, $src\t# vector (16B)" %}
+ ins_encode %{
+ __ notr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+// ------------------------------ Vector max/min -------------------------------
+
+instruct vmax8B(vecD dst, vecD src1, vecD src2)
+%{
+ predicate((n->as_Vector()->length() == 4 || n->as_Vector()->length() == 8) &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (8B)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmax16B(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 16 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (16B)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmax4S(vecD dst, vecD src1, vecD src2)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (4S)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmax8S(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (8S)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmax2I(vecD dst, vecD src1, vecD src2)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (2I)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmax4I(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "maxv $dst, $src1, $src2\t# vector (4I)" %}
+ ins_encode %{
+ __ maxv(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmin8B(vecD dst, vecD src1, vecD src2)
+%{
+ predicate((n->as_Vector()->length() == 4 || n->as_Vector()->length() == 8) &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (8B)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmin16B(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 16 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (16B)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmin4S(vecD dst, vecD src1, vecD src2)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (4S)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmin8S(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (8S)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmin2I(vecD dst, vecD src1, vecD src2)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (2I)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T2S,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop64);
+%}
+
+instruct vmin4I(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_INT);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ format %{ "minv $dst, $src1, $src2\t# vector (4I)" %}
+ ins_encode %{
+ __ minv(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($src1$$reg),
+ as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+
+instruct vmax2L(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (MaxV src1 src2));
+ ins_cost(INSN_COST);
+ effect(TEMP dst);
+ format %{ "cmgt $dst, $src1, $src2\t# vector (2L)\n\t"
+ "bsl $dst, $src1, $src2\t# vector (16B)" %}
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ bsl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+instruct vmin2L(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length() == 2 && n->bottom_type()->is_vect()->element_basic_type() == T_LONG);
+ match(Set dst (MinV src1 src2));
+ ins_cost(INSN_COST);
+ effect(TEMP dst);
+ format %{ "cmgt $dst, $src1, $src2\t# vector (2L)\n\t"
+ "bsl $dst, $src2, $src1\t# vector (16B)" %}
+ ins_encode %{
+ __ cmgt(as_FloatRegister($dst$$reg), __ T2D,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ bsl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vdop128);
+%}
+
+// --------------------------------- blend (bsl) ----------------------------
+
+instruct vbsl8B(vecD dst, vecD src1, vecD src2)
+%{
+ predicate(n->as_Vector()->length_in_bytes() == 8);
+ match(Set dst (VectorBlend (Binary src1 src2) dst));
+ ins_cost(INSN_COST);
+ format %{ "bsl $dst, $src2, $src1\t# vector (8B)" %}
+ ins_encode %{
+ __ bsl(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vlogical64);
+%}
+
+instruct vbsl16B(vecX dst, vecX src1, vecX src2)
+%{
+ predicate(n->as_Vector()->length_in_bytes() == 16);
+ match(Set dst (VectorBlend (Binary src1 src2) dst));
+ ins_cost(INSN_COST);
+ format %{ "bsl $dst, $src2, $src1\t# vector (16B)" %}
+ ins_encode %{
+ __ bsl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src2$$reg), as_FloatRegister($src1$$reg));
+ %}
+ ins_pipe(vlogical128);
+%}
+
+// --------------------------------- Load/store Mask ----------------------------
+
+instruct loadmask8B(vecD dst, vecD src )
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "negr $dst, $src\t# load mask (8B to 8B)" %}
+ ins_encode %{
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadmask16B(vecX dst, vecX src )
+%{
+ predicate(n->as_Vector()->length() == 16 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "negr $dst, $src\t# load mask (16B to 16B)" %}
+ ins_encode %{
+ __ negr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct storemask8B(vecD dst, vecD src , immI_1 size)
+%{
+ predicate(n->as_Vector()->length() == 8);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "negr $dst, $src\t# store mask (8B to 8B)" %}
+ ins_encode %{
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct storemask16B(vecX dst, vecX src , immI_1 size)
+%{
+ predicate(n->as_Vector()->length() == 16);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "negr $dst, $src\t# store mask (16B to 16B)" %}
+ ins_encode %{
+ __ negr(as_FloatRegister($dst$$reg), __ T16B, as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadmask4S(vecD dst, vecD src )
+%{
+ predicate(n->as_Vector()->length() == 4 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\n\t"
+ "negr $dst, $dst\t# load mask (4B to 4H)" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ negr(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct loadmask8S(vecX dst, vecD src )
+%{
+ predicate(n->as_Vector()->length() == 8 && n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\n\t"
+ "negr $dst, $dst\t# load mask (8B to 8H)" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ negr(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct storemask4S(vecD dst, vecD src , immI_2 size)
+%{
+ predicate(n->as_Vector()->length() == 4);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "xtn $dst, $src\n\t"
+ "negr $dst, $dst\t# store mask (4H to 4B)" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg), __ T8H);
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct storemask8S(vecD dst, vecX src , immI_2 size)
+%{
+ predicate(n->as_Vector()->length() == 8);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "xtn $dst, $src\n\t"
+ "negr $dst, $dst\t# store mask (8H to 8B)" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($src$$reg), __ T8H);
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct loadmask2I(vecD dst, vecD src )
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ (n->bottom_type()->is_vect()->element_basic_type() == T_INT ||
+ n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT));
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 2B to 2H\n\t"
+ "uxtl $dst, $dst\t# 2H to 2S\n\t"
+ "negr $dst, $dst\t# load mask (2B to 2S)" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ uxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ __ negr(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct loadmask4I(vecX dst, vecD src )
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ (n->bottom_type()->is_vect()->element_basic_type() == T_INT ||
+ n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT));
+ match(Set dst (VectorLoadMask src ));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 4B to 4H\n\t"
+ "uxtl $dst, $dst\t# 4H to 4S\n\t"
+ "negr $dst, $dst\t# load mask (4B to 4S)" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ uxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ __ negr(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct storemask2I(vecD dst, vecD src , immI_4 size)
+%{
+ predicate(n->as_Vector()->length() == 2);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "xtn $dst, $src\t# 2S to 2H\n\t"
+ "xtn $dst, $dst\t# 2H to 2B\n\t"
+ "negr $dst, $dst\t# store mask (2S to 2B)" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($src$$reg), __ T4S);
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg), __ T8H);
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct storemask4I(vecD dst, vecX src , immI_4 size)
+%{
+ predicate(n->as_Vector()->length() == 4);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "xtn $dst, $src\t# 4S to 4H\n\t"
+ "xtn $dst, $dst\t# 4H to 4B\n\t"
+ "negr $dst, $dst\t# store mask (4S to 4B)" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($src$$reg), __ T4S);
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg), __ T8H);
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct loadmask2L(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 2 &&
+ (n->bottom_type()->is_vect()->element_basic_type() == T_LONG ||
+ n->bottom_type()->is_vect()->element_basic_type() == T_DOUBLE));
+ match(Set dst (VectorLoadMask src));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 2B to 2S\n\t"
+ "uxtl $dst, $dst\t# 2S to 2I\n\t"
+ "uxtl $dst, $dst\t# 2I to 2L\n\t"
+ "neg $dst, $dst\t# load mask (2B to 2L)" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ uxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ __ uxtl(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($dst$$reg), __ T2S);
+ __ negr(as_FloatRegister($dst$$reg), __ T2D, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct storemask2L(vecD dst, vecX src, immI_8 size)
+%{
+ predicate(n->as_Vector()->length() == 2);
+ match(Set dst (VectorStoreMask src size));
+ ins_cost(INSN_COST);
+ format %{ "xtn $dst, $src\t# 2L to 2I\n\t"
+ "xtn $dst, $dst\t# 2I to 2S\n\t"
+ "xtn $dst, $dst\t# 2S to 2B\n\t"
+ "neg $dst, $dst\t# store mask (2L to 2B)" %}
+ ins_encode %{
+ __ xtn(as_FloatRegister($dst$$reg), __ T2S, as_FloatRegister($src$$reg), __ T2D);
+ __ xtn(as_FloatRegister($dst$$reg), __ T4H, as_FloatRegister($dst$$reg), __ T4S);
+ __ xtn(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg), __ T8H);
+ __ negr(as_FloatRegister($dst$$reg), __ T8B, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+//-------------------------------- LOAD_IOTA_INDICES----------------------------------
+
+instruct loadcon8B(vecD dst, immI0 src)
+%{
+ predicate((n->as_Vector()->length() == 2 || n->as_Vector()->length() == 4 ||
+ n->as_Vector()->length() == 8) &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadConst src));
+ ins_cost(INSN_COST);
+ format %{ "ldr $dst, CONSTANT_MEMORY\t# load iota indices" %}
+ ins_encode %{
+ __ lea(rscratch1, ExternalAddress(StubRoutines::aarch64::vector_iota_indices()));
+ __ ldrd(as_FloatRegister($dst$$reg), rscratch1);
+ %}
+ ins_pipe(pipe_class_memory);
+%}
+
+instruct loadcon16B(vecX dst, immI0 src)
+%{
+ predicate(n->as_Vector()->length() == 16 && n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadConst src));
+ ins_cost(INSN_COST);
+ format %{ "ldr $dst, CONSTANT_MEMORY\t# load iota indices" %}
+ ins_encode %{
+ __ lea(rscratch1, ExternalAddress(StubRoutines::aarch64::vector_iota_indices()));
+ __ ldrq(as_FloatRegister($dst$$reg), rscratch1);
+ %}
+ ins_pipe(pipe_class_memory);
+%}
+
+//-------------------------------- LOAD_SHUFFLE ----------------------------------
+
+instruct loadshuffle8B(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadShuffle src));
+ ins_cost(INSN_COST);
+ format %{ "mov $dst, $src\t# get 8B shuffle" %}
+ ins_encode %{
+ __ orr(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadshuffle16B(vecX dst, vecX src)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorLoadShuffle src));
+ ins_cost(INSN_COST);
+ format %{ "mov $dst, $src\t# get 16B shuffle" %}
+ ins_encode %{
+ __ orr(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), as_FloatRegister($src$$reg));
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadshuffle4S(vecD dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorLoadShuffle src));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 4B to 4H" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadshuffle8S(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorLoadShuffle src));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 8B to 8H" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ %}
+ ins_pipe(pipe_class_default);
+%}
+
+instruct loadshuffle4I(vecX dst, vecD src)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ (n->bottom_type()->is_vect()->element_basic_type() == T_INT ||
+ n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT));
+ match(Set dst (VectorLoadShuffle src));
+ ins_cost(INSN_COST);
+ format %{ "uxtl $dst, $src\t# 4B to 4H \n\t"
+ "uxtl $dst, $dst\t# 4H to 4S" %}
+ ins_encode %{
+ __ uxtl(as_FloatRegister($dst$$reg), __ T8H, as_FloatRegister($src$$reg), __ T8B);
+ __ uxtl(as_FloatRegister($dst$$reg), __ T4S, as_FloatRegister($dst$$reg), __ T4H);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+//-------------------------------- Rearrange -------------------------------------
+// Here is an example that rearranges a NEON vector with 4 ints:
+// Rearrange V1 int[a0, a1, a2, a3] to V2 int[a2, a3, a0, a1]
+// 1. Get the indices of V1 and store them as Vi byte[0, 1, 2, 3].
+// 2. Convert Vi byte[0, 1, 2, 3] to the indices of V2 and also store them as Vi byte[2, 3, 0, 1].
+// 3. Unsigned extend Long Vi from byte[2, 3, 0, 1] to int[2, 3, 0, 1].
+// 4. Multiply Vi int[2, 3, 0, 1] with constant int[0x04040404, 0x04040404, 0x04040404, 0x04040404]
+// and get tbl base Vm int[0x08080808, 0x0c0c0c0c, 0x00000000, 0x04040404].
+// 5. Add Vm with constant int[0x03020100, 0x03020100, 0x03020100, 0x03020100]
+// and get tbl index Vm int[0x0b0a0908, 0x0f0e0d0c, 0x03020100, 0x07060504]
+// 6. Use Vm as index register, and use V1 as table register.
+// Then get V2 as the result by tbl NEON instructions.
+// Notes:
+// Step 1 matches VectorLoadConst.
+// Step 3 matches VectorLoadShuffle.
+// Step 4, 5, 6 match VectorRearrange.
+// For VectorRearrange short/int, the reason why such complex calculation is
+// required is because NEON tbl supports bytes table only, so for short/int, we
+// need to lookup 2/4 bytes as a group. For VectorRearrange long, we use bsl
+// to implement rearrange.
+
+instruct rearrange8B(vecD dst, vecD src, vecD shuffle)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorRearrange src shuffle));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst);
+ format %{ "tbl $dst, {$dst}, $shuffle\t# rearrange 8B" %}
+ ins_encode %{
+ __ tbl(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), 1, as_FloatRegister($shuffle$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct rearrange16B(vecX dst, vecX src, vecX shuffle)
+%{
+ predicate(n->as_Vector()->length() == 16 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_BYTE);
+ match(Set dst (VectorRearrange src shuffle));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst);
+ format %{ "tbl $dst, {$dst}, $shuffle\t# rearrange 16B" %}
+ ins_encode %{
+ __ tbl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), 1, as_FloatRegister($shuffle$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct rearrange4S(vecD dst, vecD src, vecD shuffle, vecD tmp0, vecD tmp1)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorRearrange src shuffle));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp0, TEMP tmp1);
+ format %{ "mov $tmp0, CONSTANT\t# constant 0x0202020202020202\n\t"
+ "mov $tmp1, CONSTANT\t# constant 0x0100010001000100\n\t"
+ "mulv $dst, T4H, $shuffle, $tmp0\n\t"
+ "addv $dst, T8B, $dst, $tmp1\n\t"
+ "tbl $dst, {$src}, $dst\t# rearrange 4S" %}
+ ins_encode %{
+ __ mov(as_FloatRegister($tmp0$$reg), __ T8B, 0x02);
+ __ mov(as_FloatRegister($tmp1$$reg), __ T4H, 0x0100);
+ __ mulv(as_FloatRegister($dst$$reg), __ T4H,
+ as_FloatRegister($shuffle$$reg), as_FloatRegister($tmp0$$reg));
+ __ addv(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($dst$$reg), as_FloatRegister($tmp1$$reg));
+ __ tbl(as_FloatRegister($dst$$reg), __ T8B,
+ as_FloatRegister($src$$reg), 1, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct rearrange8S(vecX dst, vecX src, vecX shuffle, vecX tmp0, vecX tmp1)
+%{
+ predicate(n->as_Vector()->length() == 8 &&
+ n->bottom_type()->is_vect()->element_basic_type() == T_SHORT);
+ match(Set dst (VectorRearrange src shuffle));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp0, TEMP tmp1);
+ format %{ "mov $tmp0, CONSTANT\t# constant 0x0202020202020202\n\t"
+ "mov $tmp1, CONSTANT\t# constant 0x0100010001000100\n\t"
+ "mulv $dst, T8H, $shuffle, $tmp0\n\t"
+ "addv $dst, T16B, $dst, $tmp1\n\t"
+ "tbl $dst, {$src}, $dst\t# rearrange 8S" %}
+ ins_encode %{
+ __ mov(as_FloatRegister($tmp0$$reg), __ T16B, 0x02);
+ __ mov(as_FloatRegister($tmp1$$reg), __ T8H, 0x0100);
+ __ mulv(as_FloatRegister($dst$$reg), __ T8H,
+ as_FloatRegister($shuffle$$reg), as_FloatRegister($tmp0$$reg));
+ __ addv(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($dst$$reg), as_FloatRegister($tmp1$$reg));
+ __ tbl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), 1, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct rearrange4I(vecX dst, vecX src, vecX shuffle, vecX tmp0, vecX tmp1)
+%{
+ predicate(n->as_Vector()->length() == 4 &&
+ (n->bottom_type()->is_vect()->element_basic_type() == T_INT ||
+ n->bottom_type()->is_vect()->element_basic_type() == T_FLOAT));
+ match(Set dst (VectorRearrange src shuffle));
+ ins_cost(INSN_COST);
+ effect(TEMP_DEF dst, TEMP tmp0, TEMP tmp1);
+ format %{ "mov $tmp0, CONSTANT\t# constant 0x0404040404040404\n\t"
+ "mov $tmp1, CONSTANT\t# constant 0x0302010003020100\n\t"
+ "mulv $dst, T8H, $shuffle, $tmp0\n\t"
+ "addv $dst, T16B, $dst, $tmp1\n\t"
+ "tbl $dst, {$src}, $dst\t# rearrange 4I" %}
+ ins_encode %{
+ __ mov(as_FloatRegister($tmp0$$reg), __ T16B, 0x04);
+ __ mov(as_FloatRegister($tmp1$$reg), __ T4S, 0x03020100);
+ __ mulv(as_FloatRegister($dst$$reg), __ T4S,
+ as_FloatRegister($shuffle$$reg), as_FloatRegister($tmp0$$reg));
+ __ addv(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($dst$$reg), as_FloatRegister($tmp1$$reg));
+ __ tbl(as_FloatRegister($dst$$reg), __ T16B,
+ as_FloatRegister($src$$reg), 1, as_FloatRegister($dst$$reg));
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+//-------------------------------- Anytrue/alltrue -----------------------------
+
+instruct anytrue_in_mask8B(iRegINoSp dst, vecD src1, vecD src2, vecD tmp, rFlagsReg cr)
+%{
+ predicate(static_cast<const VectorTestNode*>(n)->get_predicate() == BoolTest::ne);
+ match(Set dst (VectorTest src1 src2 ));
+ ins_cost(INSN_COST);
+ effect(TEMP tmp, KILL cr);
+ format %{ "addv $tmp, T8B, $src1\t# src1 and src2 are the same\n\t"
+ "umov $dst, $tmp, B, 0\n\t"
+ "cmp $dst, 0\n\t"
+ "cset $dst" %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($src1$$reg));
+ __ umov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw($dst$$Register, zr);
+ __ csetw($dst$$Register, Assembler::NE);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct anytrue_in_mask16B(iRegINoSp dst, vecX src1, vecX src2, vecX tmp, rFlagsReg cr)
+%{
+ predicate(static_cast<const VectorTestNode*>(n)->get_predicate() == BoolTest::ne);
+ match(Set dst (VectorTest src1 src2 ));
+ ins_cost(INSN_COST);
+ effect(TEMP tmp, KILL cr);
+ format %{ "addv $tmp, T16B, $src1\t# src1 and src2 are the same\n\t"
+ "umov $dst, $tmp, B, 0\n\t"
+ "cmp $dst, 0\n\t"
+ "cset $dst" %}
+ ins_encode %{
+ __ addv(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($src1$$reg));
+ __ umov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw($dst$$Register, zr);
+ __ csetw($dst$$Register, Assembler::NE);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct alltrue_in_mask8B(iRegINoSp dst, vecD src1, vecD src2, vecD tmp, rFlagsReg cr)
+%{
+ predicate(static_cast<const VectorTestNode*>(n)->get_predicate() == BoolTest::overflow);
+ match(Set dst (VectorTest src1 src2 ));
+ ins_cost(INSN_COST);
+ effect(TEMP tmp, KILL cr);
+ format %{ "andr $tmp, T8B, $src1, $src2\t# src2 is maskAllTrue\n\t"
+ "notr $tmp, T8B, $tmp\n\t"
+ "addv $tmp, T8B, $tmp\n\t"
+ "umov $dst, $tmp, B, 0\n\t"
+ "cmp $dst, 0\n\t"
+ "cset $dst" %}
+ ins_encode %{
+ __ andr(as_FloatRegister($tmp$$reg), __ T8B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($tmp$$reg));
+ __ addv(as_FloatRegister($tmp$$reg), __ T8B, as_FloatRegister($tmp$$reg));
+ __ umov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw($dst$$Register, zr);
+ __ csetw($dst$$Register, Assembler::EQ);
+ %}
+ ins_pipe(pipe_slow);
+%}
+
+instruct alltrue_in_mask16B(iRegINoSp dst, vecX src1, vecX src2, vecX tmp, rFlagsReg cr)
+%{
+ predicate(static_cast<const VectorTestNode*>(n)->get_predicate() == BoolTest::overflow);
+ match(Set dst (VectorTest src1 src2 ));
+ ins_cost(INSN_COST);
+ effect(TEMP tmp, KILL cr);
+ format %{ "andr $tmp, T16B, $src1, $src2\t# src2 is maskAllTrue\n\t"
+ "notr $tmp, T16B, $tmp\n\t"
+ "addv $tmp, T16B, $tmp\n\t"
+ "umov $dst, $tmp, B, 0\n\t"
+ "cmp $dst, 0\n\t"
+ "cset $dst" %}
+ ins_encode %{
+ __ andr(as_FloatRegister($tmp$$reg), __ T16B,
+ as_FloatRegister($src1$$reg), as_FloatRegister($src2$$reg));
+ __ notr(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($tmp$$reg));
+ __ addv(as_FloatRegister($tmp$$reg), __ T16B, as_FloatRegister($tmp$$reg));
+ __ umov($dst$$Register, as_FloatRegister($tmp$$reg), __ B, 0);
+ __ cmpw($dst$$Register, zr);
+ __ csetw($dst$$Register, Assembler::EQ);
+ %}
+ ins_pipe(pipe_slow);
+%}
+// END This section of the file is automatically generated. Do not edit --------------
+
//----------PEEPHOLE RULES-----------------------------------------------------
// These must follow all instruction definitions as they use the names
// defined in the instructions definitions.
//
// peepmatch ( root_instr_name [preceding_instruction]* );
< prev index next >