1 /*
  2  * Copyright (c) 2020, 2026, Oracle and/or its affiliates. All rights reserved.
  3  * Copyright (c) 2020, 2022, Huawei Technologies Co., Ltd. All rights reserved.
  4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  5  *
  6  * This code is free software; you can redistribute it and/or modify it
  7  * under the terms of the GNU General Public License version 2 only, as
  8  * published by the Free Software Foundation.
  9  *
 10  * This code is distributed in the hope that it will be useful, but WITHOUT
 11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 13  * version 2 for more details (a copy is included in the LICENSE file that
 14  * accompanied this code).
 15  *
 16  * You should have received a copy of the GNU General Public License version
 17  * 2 along with this work; if not, write to the Free Software Foundation,
 18  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 19  *
 20  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 21  * or visit www.oracle.com if you need additional information or have any
 22  * questions.
 23  *
 24  */
 25 
 26 #ifndef CPU_RISCV_C2_MACROASSEMBLER_RISCV_HPP
 27 #define CPU_RISCV_C2_MACROASSEMBLER_RISCV_HPP
 28 
 29 // C2_MacroAssembler contains high-level macros for C2
 30 
 31  private:
 32   // Return true if the phase output is in the scratch emit size mode.
 33   virtual bool in_scratch_emit_size() override;
 34 
 35   void element_compare(Register r1, Register r2,
 36                        Register result, Register cnt,
 37                        Register tmp1, Register tmp2,
 38                        VectorRegister vr1, VectorRegister vr2,
 39                        VectorRegister vrs,
 40                        bool is_latin, Label& DONE, Assembler::LMUL lmul);
 41 
 42   void string_compare_long_same_encoding(Register result, Register str1, Register str2,
 43                                   const bool isLL, Register cnt1, Register cnt2,
 44                                   Register tmp1, Register tmp2, Register tmp3,
 45                                   const int STUB_THRESHOLD, Label *STUB, Label *SHORT_STRING, Label *DONE);
 46   void string_compare_long_different_encoding(Register result, Register str1, Register str2,
 47                                   bool isLU, Register cnt1, Register cnt2,
 48                                   Register tmp1, Register tmp2, Register tmp3,
 49                                   const int STUB_THRESHOLD, Label *STUB, Label *DONE);
 50 
 51  public:
 52   // Code used by cmpFastLock and cmpFastUnlock mach instructions in .ad file.
 53   void fast_lock(Register object, Register box,
 54                  Register tmp1, Register tmp2, Register tmp3, Register tmp4);
 55   void fast_unlock(Register object, Register box,
 56                    Register tmp1, Register tmp2, Register tmp3);
 57 
 58   void string_compare(Register str1, Register str2,
 59                       Register cnt1, Register cnt2, Register result,
 60                       Register tmp1, Register tmp2, Register tmp3,
 61                       int ae);
 62 
 63   void string_indexof_char_short(Register str1, Register cnt1,
 64                                  Register ch, Register result,
 65                                  bool isL);
 66 
 67   void string_indexof_char(Register str1, Register cnt1,
 68                            Register ch, Register result,
 69                            Register tmp1, Register tmp2,
 70                            Register tmp3, Register tmp4,
 71                            bool isL);
 72 
 73   void string_indexof(Register str1, Register str2,
 74                       Register cnt1, Register cnt2,
 75                       Register tmp1, Register tmp2,
 76                       Register tmp3, Register tmp4,
 77                       Register tmp5, Register tmp6,
 78                       Register result, int ae);
 79 
 80   void string_indexof_linearscan(Register haystack, Register needle,
 81                                  Register haystack_len, Register needle_len,
 82                                  Register tmp1, Register tmp2,
 83                                  Register tmp3, Register tmp4,
 84                                  int needle_con_cnt, Register result, int ae);
 85 
 86   void arrays_equals(Register r1, Register r2,
 87                      Register tmp1, Register tmp2, Register tmp3,
 88                      Register result, int elem_size);
 89 
 90   void arrays_hashcode(Register ary, Register cnt, Register result,
 91                        Register tmp1, Register tmp2,
 92                        Register tmp3, Register tmp4,
 93                        Register tmp5, Register tmp6,
 94                        BasicType eltype);
 95   int arrays_hashcode_elsize(BasicType eltype);
 96   void arrays_hashcode_elload(Register dst, Address src, BasicType eltype);
 97 
 98   void arrays_hashcode_v(Register ary, Register cnt, Register result,
 99                          Register tmp1, Register tmp2, Register tmp3,
100                          BasicType eltype);
101   void arrays_hashcode_elload_v(VectorRegister vdst, VectorRegister vtmp,
102                                 Register src, BasicType eltype);
103 
104   void string_equals(Register r1, Register r2,
105                      Register result, Register cnt1);
106 
107   // refer to conditional_branches and float_conditional_branches
108   static const int bool_test_bits = 3;
109   static const int unsigned_branch_mask = 1 << bool_test_bits;
110   static const int double_branch_mask = 1 << bool_test_bits;
111 
112   // cmp
113   void cmp_branch(int cmpFlag,
114                   Register op1, Register op2,
115                   Label& label, bool is_far = false);
116 
117   void float_cmp_branch(int cmpFlag,
118                         FloatRegister op1, FloatRegister op2,
119                         Label& label, bool is_far = false);
120 
121   void enc_cmpUEqNeLeGt_imm0_branch(int cmpFlag, Register op,
122                                     Label& L, bool is_far = false);
123 
124   void enc_cmpEqNe_imm0_branch(int cmpFlag, Register op,
125                                Label& L, bool is_far = false);
126 
127   void enc_cmove(int cmpFlag,
128                  Register op1, Register op2,
129                  Register dst, Register src);
130 
131   void enc_cmove_cmp_fp(int cmpFlag,
132                         FloatRegister op1, FloatRegister op2,
133                         Register dst, Register src, bool is_single);
134 
135   void enc_cmove_fp_cmp(int cmpFlag, Register op1, Register op2,
136                         FloatRegister dst, FloatRegister src, bool is_single);
137 
138   void enc_cmove_fp_cmp_fp(int cmpFlag, FloatRegister op1, FloatRegister op2,
139                            FloatRegister dst, FloatRegister src,
140                            bool cmp_single, bool cmov_single);
141 
142   void spill(Register r, bool is64, int offset) {
143     is64 ? sd(r, Address(sp, offset))
144          : sw(r, Address(sp, offset));
145   }
146 
147   void spill(FloatRegister f, bool is64, int offset) {
148     is64 ? fsd(f, Address(sp, offset))
149          : fsw(f, Address(sp, offset));
150   }
151 
152   void spill(VectorRegister v, int offset) {
153     add(t0, sp, offset);
154     vs1r_v(v, t0);
155   }
156 
157   void unspill(Register r, bool is64, int offset) {
158     is64 ? ld(r, Address(sp, offset))
159          : lw(r, Address(sp, offset));
160   }
161 
162   void unspillu(Register r, bool is64, int offset) {
163     is64 ? ld(r, Address(sp, offset))
164          : lwu(r, Address(sp, offset));
165   }
166 
167   void unspill(FloatRegister f, bool is64, int offset) {
168     is64 ? fld(f, Address(sp, offset))
169          : flw(f, Address(sp, offset));
170   }
171 
172   void unspill(VectorRegister v, int offset) {
173     add(t0, sp, offset);
174     vl1r_v(v, t0);
175   }
176 
177   void spill_copy_vector_stack_to_stack(int src_offset, int dst_offset, uint vector_length_in_bytes) {
178     assert(vector_length_in_bytes % 16 == 0, "unexpected vector reg size");
179     for (int i = 0; i < (int)vector_length_in_bytes / 8; i++) {
180       unspill(t0, true, src_offset + (i * 8));
181       spill(t0, true, dst_offset + (i * 8));
182     }
183   }
184 
185   enum class FLOAT_TYPE {
186     half_precision,
187     single_precision,
188     double_precision
189   };
190 
191   void minmax_fp(FloatRegister dst,
192                  FloatRegister src1, FloatRegister src2,
193                  FLOAT_TYPE ft, bool is_min);
194 
195   void round_double_mode(FloatRegister dst, FloatRegister src, int round_mode,
196                          Register tmp1, Register tmp2, Register tmp3);
197 
198   void signum_fp(FloatRegister dst, FloatRegister one, bool is_double);
199 
200   void float16_to_float(FloatRegister dst, Register src, Register tmp);
201   void float_to_float16(Register dst, FloatRegister src, FloatRegister ftmp, Register xtmp);
202 
203   void signum_fp_v(VectorRegister dst, VectorRegister one, BasicType bt, int vlen);
204 
205 
206   // intrinsic methods implemented by rvv instructions
207 
208   void java_round_float_v(VectorRegister dst, VectorRegister src, FloatRegister ftmp, BasicType bt, uint vector_length);
209   void java_round_double_v(VectorRegister dst, VectorRegister src, FloatRegister ftmp, BasicType bt, uint vector_length);
210 
211   void float16_to_float_v(VectorRegister dst, VectorRegister src, uint vector_length);
212   void float_to_float16_v(VectorRegister dst, VectorRegister src, VectorRegister vtmp, Register tmp, uint vector_length);
213 
214   void string_equals_v(Register r1, Register r2,
215                        Register result, Register cnt1);
216 
217   void arrays_equals_v(Register r1, Register r2,
218                        Register result, Register cnt1,
219                        int elem_size);
220 
221   void string_compare_v(Register str1, Register str2,
222                         Register cnt1, Register cnt2,
223                         Register result,
224                         Register tmp1, Register tmp2,
225                         int encForm);
226 
227   void clear_array_v(Register base, Register cnt);
228 
229   void byte_array_inflate_v(Register src, Register dst,
230                             Register len, Register tmp);
231 
232   void char_array_compress_v(Register src, Register dst,
233                             Register len, Register result,
234                             Register tmp);
235 
236   void encode_iso_array_v(Register src, Register dst,
237                           Register len, Register result,
238                           Register tmp, bool ascii);
239 
240   void count_positives_v(Register ary, Register len,
241                         Register result, Register tmp);
242 
243   void string_indexof_char_v(Register str1, Register cnt1,
244                             Register ch, Register result,
245                             Register tmp1, Register tmp2,
246                             bool isL);
247 
248   void minmax_fp_v(VectorRegister dst,
249                   VectorRegister src1, VectorRegister src2,
250                   BasicType bt, bool is_min, uint vector_length);
251 
252   void minmax_fp_masked_v(VectorRegister dst, VectorRegister src1, VectorRegister src2,
253                           VectorRegister vmask, VectorRegister tmp1, VectorRegister tmp2,
254                           BasicType bt, bool is_min, uint vector_length);
255 
256   void reduce_minmax_fp_v(FloatRegister dst,
257                           FloatRegister src1, VectorRegister src2,
258                           VectorRegister tmp1, VectorRegister tmp2,
259                           bool is_double, bool is_min, uint vector_length,
260                           VectorMask vm = Assembler::unmasked);
261 
262   void reduce_integral_v(Register dst, Register src1,
263                         VectorRegister src2, VectorRegister tmp,
264                         int opc, BasicType bt, uint vector_length,
265                         VectorMask vm = Assembler::unmasked);
266 
267   void reduce_mul_integral_v(Register dst, Register src1, VectorRegister src2,
268                              VectorRegister vtmp1, VectorRegister vtmp2, BasicType bt,
269                              uint vector_length, VectorMask vm = Assembler::unmasked);
270 
271   void vsetvli_helper(BasicType bt, uint vector_length, LMUL vlmul = Assembler::m1, Register tmp = t0);
272 
273   void compare_integral_v(VectorRegister dst, VectorRegister src1, VectorRegister src2, int cond,
274                           BasicType bt, uint vector_length, VectorMask vm = Assembler::unmasked);
275 
276   void compare_fp_v(VectorRegister dst, VectorRegister src1, VectorRegister src2, int cond,
277                     BasicType bt, uint vector_length, VectorMask vm = Assembler::unmasked);
278 
279   void spill_vmask(VectorRegister v, int offset);
280 
281   void unspill_vmask(VectorRegister v, int offset);
282 
283   void spill_copy_vmask_stack_to_stack(int src_offset, int dst_offset, uint vector_length_in_bytes) {
284     assert(vector_length_in_bytes % 4 == 0, "unexpected vector mask reg size");
285     for (int i = 0; i < (int)vector_length_in_bytes / 4; i++) {
286       unspill(t0, false, src_offset + (i * 4));
287       spill(t0, false, dst_offset + (i * 4));
288     }
289   }
290 
291   void integer_extend_v(VectorRegister dst, BasicType dst_bt, uint vector_length,
292                         VectorRegister src, BasicType src_bt, bool is_signed);
293 
294   void integer_narrow_v(VectorRegister dst, BasicType dst_bt, uint vector_length,
295                         VectorRegister src, BasicType src_bt);
296 
297   void vfcvt_rtz_x_f_v_safe(VectorRegister dst, VectorRegister src);
298 
299   void extract_v(Register dst, VectorRegister src,
300                  BasicType bt, int idx, VectorRegister vtmp);
301 
302   void extract_fp_v(FloatRegister dst, VectorRegister src,
303                     BasicType bt, int idx, VectorRegister vtmp);
304 
305   void slidedown_v(VectorRegister dst, VectorRegister src,
306                    uint32_t offset, Register tmp = t0);
307 
308 #endif // CPU_RISCV_C2_MACROASSEMBLER_RISCV_HPP