1 /*
  2  * Copyright (c) 2020, 2025, Oracle and/or its affiliates. All rights reserved.
  3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  4  *
  5  * This code is free software; you can redistribute it and/or modify it
  6  * under the terms of the GNU General Public License version 2 only, as
  7  * published by the Free Software Foundation.
  8  *
  9  * This code is distributed in the hope that it will be useful, but WITHOUT
 10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 12  * version 2 for more details (a copy is included in the LICENSE file that
 13  * accompanied this code).
 14  *
 15  * You should have received a copy of the GNU General Public License version
 16  * 2 along with this work; if not, write to the Free Software Foundation,
 17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 18  *
 19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 20  * or visit www.oracle.com if you need additional information or have any
 21  * questions.
 22  *
 23  */
 24 
 25 #ifndef CPU_AARCH64_C2_MACROASSEMBLER_AARCH64_HPP
 26 #define CPU_AARCH64_C2_MACROASSEMBLER_AARCH64_HPP
 27 
 28 // C2_MacroAssembler contains high-level macros for C2
 29 
 30  private:
 31   // Return true if the phase output is in the scratch emit size mode.
 32   virtual bool in_scratch_emit_size() override;
 33 
 34   void neon_reduce_logical_helper(int opc, bool sf, Register Rd, Register Rn, Register Rm,
 35                                   enum shift_kind kind = Assembler::LSL, unsigned shift = 0);
 36 
 37   void select_from_two_vectors_neon(FloatRegister dst, FloatRegister src1,
 38                                     FloatRegister src2, FloatRegister index,
 39                                     FloatRegister tmp, unsigned vector_length_in_bytes);
 40 
 41   void select_from_two_vectors_sve(FloatRegister dst, FloatRegister src1,
 42                                    FloatRegister src2, FloatRegister index,
 43                                    FloatRegister tmp, SIMD_RegVariant T,
 44                                    unsigned vector_length_in_bytes);
 45 
 46  public:
 47   // jdk.internal.util.ArraysSupport.vectorizedHashCode
 48   address arrays_hashcode(Register ary, Register cnt, Register result, FloatRegister vdata0,
 49                           FloatRegister vdata1, FloatRegister vdata2, FloatRegister vdata3,
 50                           FloatRegister vmul0, FloatRegister vmul1, FloatRegister vmul2,
 51                           FloatRegister vmul3, FloatRegister vpow, FloatRegister vpowm,
 52                           BasicType eltype);
 53 
 54   // Code used by cmpFastLock and cmpFastUnlock mach instructions in .ad file.
 55   void fast_lock(Register object, Register box, Register t1, Register t2, Register t3);
 56   void fast_unlock(Register object, Register box, Register t1, Register t2, Register t3);
 57 
 58   void string_compare(Register str1, Register str2,
 59                       Register cnt1, Register cnt2, Register result,
 60                       Register tmp1, Register tmp2, FloatRegister vtmp1,
 61                       FloatRegister vtmp2, FloatRegister vtmp3,
 62                       PRegister pgtmp1, PRegister pgtmp2, int ae);
 63 
 64   void string_indexof(Register str1, Register str2,
 65                       Register cnt1, Register cnt2,
 66                       Register tmp1, Register tmp2,
 67                       Register tmp3, Register tmp4,
 68                       Register tmp5, Register tmp6,
 69                       int int_cnt1, Register result, int ae);
 70 
 71   void string_indexof_char(Register str1, Register cnt1,
 72                            Register ch, Register result,
 73                            Register tmp1, Register tmp2, Register tmp3);
 74 
 75   void stringL_indexof_char(Register str1, Register cnt1,
 76                             Register ch, Register result,
 77                             Register tmp1, Register tmp2, Register tmp3);
 78 
 79   void string_indexof_char_sve(Register str1, Register cnt1,
 80                                Register ch, Register result,
 81                                FloatRegister ztmp1, FloatRegister ztmp2,
 82                                PRegister pgtmp, PRegister ptmp, bool isL);
 83 
 84   // Compress the least significant bit of each byte to the rightmost and clear
 85   // the higher garbage bits.
 86   void bytemask_compress(Register dst);
 87 
 88   // Pack the value of each mask element in "src" into a long value in "dst", at most the
 89   // first 64 lane elements. The input "src" is a vector of boolean represented as bytes
 90   // with 0x00/0x01 as element values. Each lane value from "src" is packed into one bit in
 91   // "dst".
 92   void sve_vmask_tolong(Register dst, FloatRegister src, FloatRegister vtmp, int lane_cnt);
 93 
 94   void sve2_vmask_tolong(Register dst, FloatRegister src, FloatRegister vtmp1,
 95                          FloatRegister vtmp2, int lane_cnt);
 96 
 97   // Unpack the mask, a long value in "src", into vector register "dst" with boolean type.
 98   // Each bit in "src" is unpacked into one byte lane in "dst". Note that "dst" can support
 99   // at most 64 lanes.
100   void sve_vmask_fromlong(FloatRegister dst, Register src, FloatRegister vtmp, int lane_cnt);
101 
102   // SIMD&FP comparison
103   void neon_compare(FloatRegister dst, BasicType bt, FloatRegister src1,
104                     FloatRegister src2, Condition cond, bool isQ);
105 
106   void neon_compare_zero(FloatRegister dst, BasicType bt, FloatRegister src,
107                          Condition cond, bool isQ);
108 
109   void sve_compare(PRegister pd, BasicType bt, PRegister pg,
110                    FloatRegister zn, FloatRegister zm, Condition cond);
111 
112   void sve_vmask_lasttrue(Register dst, BasicType bt, PRegister src, PRegister ptmp);
113 
114   // Vector cast
115   void neon_vector_extend(FloatRegister dst, BasicType dst_bt, unsigned dst_vlen_in_bytes,
116                           FloatRegister src, BasicType src_bt, bool is_unsigned = false);
117 
118   void neon_vector_narrow(FloatRegister dst, BasicType dst_bt,
119                           FloatRegister src, BasicType src_bt, unsigned src_vlen_in_bytes);
120 
121   void sve_vector_extend(FloatRegister dst, SIMD_RegVariant dst_size,
122                          FloatRegister src, SIMD_RegVariant src_size, bool is_unsigned = false);
123 
124   void sve_vector_narrow(FloatRegister dst, SIMD_RegVariant dst_size,
125                          FloatRegister src, SIMD_RegVariant src_size, FloatRegister tmp);
126 
127   void sve_vmaskcast_extend(PRegister dst, PRegister src,
128                             uint dst_element_length_in_bytes, uint src_element_lenght_in_bytes);
129 
130   void sve_vmaskcast_narrow(PRegister dst, PRegister src, PRegister ptmp,
131                             uint dst_element_length_in_bytes, uint src_element_lenght_in_bytes);
132 
133   // Vector reduction
134   void neon_reduce_add_integral(Register dst, BasicType bt,
135                                 Register isrc, FloatRegister vsrc,
136                                 unsigned vector_length_in_bytes, FloatRegister vtmp);
137 
138   void neon_reduce_mul_integral(Register dst, BasicType bt,
139                                 Register isrc, FloatRegister vsrc,
140                                 unsigned vector_length_in_bytes,
141                                 FloatRegister vtmp1, FloatRegister vtmp2);
142 
143   void neon_reduce_mul_fp(FloatRegister dst, BasicType bt,
144                           FloatRegister fsrc, FloatRegister vsrc,
145                           unsigned vector_length_in_bytes, FloatRegister vtmp);
146 
147   void neon_reduce_logical(int opc, Register dst, BasicType bt, Register isrc,
148                            FloatRegister vsrc, unsigned vector_length_in_bytes);
149 
150   void neon_reduce_minmax_integral(int opc, Register dst, BasicType bt,
151                                    Register isrc, FloatRegister vsrc,
152                                    unsigned vector_length_in_bytes, FloatRegister vtmp);
153 
154   void sve_reduce_integral(int opc, Register dst, BasicType bt, Register src1,
155                            FloatRegister src2, PRegister pg, FloatRegister tmp);
156 
157   // Set elements of the dst predicate to true for lanes in the range of
158   // [0, lane_cnt), or to false otherwise. The input "lane_cnt" should be
159   // smaller than or equal to the supported max vector length of the basic
160   // type. Clobbers: rscratch1 and the rFlagsReg.
161   void sve_gen_mask_imm(PRegister dst, BasicType bt, uint32_t lane_cnt);
162 
163   // Extract a scalar element from an sve vector at position 'idx'.
164   // The input elements in src are expected to be of integral type.
165   void sve_extract_integral(Register dst, BasicType bt, FloatRegister src,
166                             int idx, FloatRegister vtmp);
167 
168   // java.lang.Math::round intrinsics
169   void vector_round_neon(FloatRegister dst, FloatRegister src, FloatRegister tmp1,
170                          FloatRegister tmp2, FloatRegister tmp3,
171                          SIMD_Arrangement T);
172   void vector_round_sve(FloatRegister dst, FloatRegister src, FloatRegister tmp1,
173                         FloatRegister tmp2, PRegister pgtmp,
174                         SIMD_RegVariant T);
175 
176   // Pack active elements of src, under the control of mask, into the
177   // lowest-numbered elements of dst. Any remaining elements of dst will
178   // be filled with zero.
179   void sve_compress_byte(FloatRegister dst, FloatRegister src, PRegister mask,
180                          FloatRegister vtmp1, FloatRegister vtmp2, FloatRegister vtmp3,
181                          PRegister ptmp, PRegister pgtmp, unsigned vector_length_in_bytes);
182 
183   void sve_compress_short(FloatRegister dst, FloatRegister src, PRegister mask,
184                           FloatRegister vzr, FloatRegister vtmp,
185                           PRegister pgtmp, unsigned vector_length_in_bytes);
186 
187   void neon_reverse_bits(FloatRegister dst, FloatRegister src, BasicType bt, bool isQ);
188 
189   void neon_reverse_bytes(FloatRegister dst, FloatRegister src, BasicType bt, bool isQ);
190 
191   void neon_rearrange_hsd(FloatRegister dst, FloatRegister src, FloatRegister shuffle,
192                           FloatRegister tmp, BasicType bt, bool isQ);
193   // java.lang.Math::signum intrinsics
194   void vector_signum_neon(FloatRegister dst, FloatRegister src, FloatRegister zero,
195                           FloatRegister one, SIMD_Arrangement T);
196 
197   void vector_signum_sve(FloatRegister dst, FloatRegister src, FloatRegister zero,
198                          FloatRegister one, FloatRegister vtmp, PRegister pgtmp, SIMD_RegVariant T);
199 
200   void verify_int_in_range(uint idx, const TypeInt* t, Register val, Register tmp);
201   void verify_long_in_range(uint idx, const TypeLong* t, Register val, Register tmp);
202 
203   void reconstruct_frame_pointer(Register rtmp);
204 
205   // Select from a table of two vectors
206   void select_from_two_vectors(FloatRegister dst, FloatRegister src1, FloatRegister src2,
207                                FloatRegister index, FloatRegister tmp, BasicType bt,
208                                unsigned vector_length_in_bytes);
209 
210   void vector_expand_neon(FloatRegister dst, FloatRegister src, FloatRegister mask,
211                           FloatRegister tmp1, FloatRegister tmp2, BasicType bt,
212                           int vector_length_in_bytes);
213   void vector_expand_sve(FloatRegister dst, FloatRegister src, PRegister pg,
214                          FloatRegister tmp1, FloatRegister tmp2, BasicType bt,
215                          int vector_length_in_bytes);
216 #endif // CPU_AARCH64_C2_MACROASSEMBLER_AARCH64_HPP