bpf, arm64: use more scalable stadd over ldxr / stxr loop in xadd
[linux/fpc-iii.git] / arch / arm64 / net / bpf_jit.h
blob76606e87233f38c8b59e64d2c0cc1f5cffc3dd7a
1 /*
2 * BPF JIT compiler for ARM64
4 * Copyright (C) 2014-2016 Zi Shen Lim <zlim.lnx@gmail.com>
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program. If not, see <http://www.gnu.org/licenses/>.
18 #ifndef _BPF_JIT_H
19 #define _BPF_JIT_H
21 #include <asm/insn.h>
23 /* 5-bit Register Operand */
24 #define A64_R(x) AARCH64_INSN_REG_##x
25 #define A64_FP AARCH64_INSN_REG_FP
26 #define A64_LR AARCH64_INSN_REG_LR
27 #define A64_ZR AARCH64_INSN_REG_ZR
28 #define A64_SP AARCH64_INSN_REG_SP
30 #define A64_VARIANT(sf) \
31 ((sf) ? AARCH64_INSN_VARIANT_64BIT : AARCH64_INSN_VARIANT_32BIT)
33 /* Compare & branch (immediate) */
34 #define A64_COMP_BRANCH(sf, Rt, offset, type) \
35 aarch64_insn_gen_comp_branch_imm(0, offset, Rt, A64_VARIANT(sf), \
36 AARCH64_INSN_BRANCH_COMP_##type)
37 #define A64_CBZ(sf, Rt, imm19) A64_COMP_BRANCH(sf, Rt, (imm19) << 2, ZERO)
38 #define A64_CBNZ(sf, Rt, imm19) A64_COMP_BRANCH(sf, Rt, (imm19) << 2, NONZERO)
40 /* Conditional branch (immediate) */
41 #define A64_COND_BRANCH(cond, offset) \
42 aarch64_insn_gen_cond_branch_imm(0, offset, cond)
43 #define A64_COND_EQ AARCH64_INSN_COND_EQ /* == */
44 #define A64_COND_NE AARCH64_INSN_COND_NE /* != */
45 #define A64_COND_CS AARCH64_INSN_COND_CS /* unsigned >= */
46 #define A64_COND_HI AARCH64_INSN_COND_HI /* unsigned > */
47 #define A64_COND_LS AARCH64_INSN_COND_LS /* unsigned <= */
48 #define A64_COND_CC AARCH64_INSN_COND_CC /* unsigned < */
49 #define A64_COND_GE AARCH64_INSN_COND_GE /* signed >= */
50 #define A64_COND_GT AARCH64_INSN_COND_GT /* signed > */
51 #define A64_COND_LE AARCH64_INSN_COND_LE /* signed <= */
52 #define A64_COND_LT AARCH64_INSN_COND_LT /* signed < */
53 #define A64_B_(cond, imm19) A64_COND_BRANCH(cond, (imm19) << 2)
55 /* Unconditional branch (immediate) */
56 #define A64_BRANCH(offset, type) aarch64_insn_gen_branch_imm(0, offset, \
57 AARCH64_INSN_BRANCH_##type)
58 #define A64_B(imm26) A64_BRANCH((imm26) << 2, NOLINK)
59 #define A64_BL(imm26) A64_BRANCH((imm26) << 2, LINK)
61 /* Unconditional branch (register) */
62 #define A64_BR(Rn) aarch64_insn_gen_branch_reg(Rn, AARCH64_INSN_BRANCH_NOLINK)
63 #define A64_BLR(Rn) aarch64_insn_gen_branch_reg(Rn, AARCH64_INSN_BRANCH_LINK)
64 #define A64_RET(Rn) aarch64_insn_gen_branch_reg(Rn, AARCH64_INSN_BRANCH_RETURN)
66 /* Load/store register (register offset) */
67 #define A64_LS_REG(Rt, Rn, Rm, size, type) \
68 aarch64_insn_gen_load_store_reg(Rt, Rn, Rm, \
69 AARCH64_INSN_SIZE_##size, \
70 AARCH64_INSN_LDST_##type##_REG_OFFSET)
71 #define A64_STRB(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 8, STORE)
72 #define A64_LDRB(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 8, LOAD)
73 #define A64_STRH(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 16, STORE)
74 #define A64_LDRH(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 16, LOAD)
75 #define A64_STR32(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 32, STORE)
76 #define A64_LDR32(Wt, Xn, Xm) A64_LS_REG(Wt, Xn, Xm, 32, LOAD)
77 #define A64_STR64(Xt, Xn, Xm) A64_LS_REG(Xt, Xn, Xm, 64, STORE)
78 #define A64_LDR64(Xt, Xn, Xm) A64_LS_REG(Xt, Xn, Xm, 64, LOAD)
80 /* Load/store register pair */
81 #define A64_LS_PAIR(Rt, Rt2, Rn, offset, ls, type) \
82 aarch64_insn_gen_load_store_pair(Rt, Rt2, Rn, offset, \
83 AARCH64_INSN_VARIANT_64BIT, \
84 AARCH64_INSN_LDST_##ls##_PAIR_##type)
85 /* Rn -= 16; Rn[0] = Rt; Rn[8] = Rt2; */
86 #define A64_PUSH(Rt, Rt2, Rn) A64_LS_PAIR(Rt, Rt2, Rn, -16, STORE, PRE_INDEX)
87 /* Rt = Rn[0]; Rt2 = Rn[8]; Rn += 16; */
88 #define A64_POP(Rt, Rt2, Rn) A64_LS_PAIR(Rt, Rt2, Rn, 16, LOAD, POST_INDEX)
90 /* Load/store exclusive */
91 #define A64_SIZE(sf) \
92 ((sf) ? AARCH64_INSN_SIZE_64 : AARCH64_INSN_SIZE_32)
93 #define A64_LSX(sf, Rt, Rn, Rs, type) \
94 aarch64_insn_gen_load_store_ex(Rt, Rn, Rs, A64_SIZE(sf), \
95 AARCH64_INSN_LDST_##type)
96 /* Rt = [Rn]; (atomic) */
97 #define A64_LDXR(sf, Rt, Rn) \
98 A64_LSX(sf, Rt, Rn, A64_ZR, LOAD_EX)
99 /* [Rn] = Rt; (atomic) Rs = [state] */
100 #define A64_STXR(sf, Rt, Rn, Rs) \
101 A64_LSX(sf, Rt, Rn, Rs, STORE_EX)
103 /* LSE atomics */
104 #define A64_STADD(sf, Rn, Rs) \
105 aarch64_insn_gen_stadd(Rn, Rs, A64_SIZE(sf))
107 /* Add/subtract (immediate) */
108 #define A64_ADDSUB_IMM(sf, Rd, Rn, imm12, type) \
109 aarch64_insn_gen_add_sub_imm(Rd, Rn, imm12, \
110 A64_VARIANT(sf), AARCH64_INSN_ADSB_##type)
111 /* Rd = Rn OP imm12 */
112 #define A64_ADD_I(sf, Rd, Rn, imm12) A64_ADDSUB_IMM(sf, Rd, Rn, imm12, ADD)
113 #define A64_SUB_I(sf, Rd, Rn, imm12) A64_ADDSUB_IMM(sf, Rd, Rn, imm12, SUB)
114 /* Rd = Rn */
115 #define A64_MOV(sf, Rd, Rn) A64_ADD_I(sf, Rd, Rn, 0)
117 /* Bitfield move */
118 #define A64_BITFIELD(sf, Rd, Rn, immr, imms, type) \
119 aarch64_insn_gen_bitfield(Rd, Rn, immr, imms, \
120 A64_VARIANT(sf), AARCH64_INSN_BITFIELD_MOVE_##type)
121 /* Signed, with sign replication to left and zeros to right */
122 #define A64_SBFM(sf, Rd, Rn, ir, is) A64_BITFIELD(sf, Rd, Rn, ir, is, SIGNED)
123 /* Unsigned, with zeros to left and right */
124 #define A64_UBFM(sf, Rd, Rn, ir, is) A64_BITFIELD(sf, Rd, Rn, ir, is, UNSIGNED)
126 /* Rd = Rn << shift */
127 #define A64_LSL(sf, Rd, Rn, shift) ({ \
128 int sz = (sf) ? 64 : 32; \
129 A64_UBFM(sf, Rd, Rn, (unsigned)-(shift) % sz, sz - 1 - (shift)); \
131 /* Rd = Rn >> shift */
132 #define A64_LSR(sf, Rd, Rn, shift) A64_UBFM(sf, Rd, Rn, shift, (sf) ? 63 : 31)
133 /* Rd = Rn >> shift; signed */
134 #define A64_ASR(sf, Rd, Rn, shift) A64_SBFM(sf, Rd, Rn, shift, (sf) ? 63 : 31)
136 /* Zero extend */
137 #define A64_UXTH(sf, Rd, Rn) A64_UBFM(sf, Rd, Rn, 0, 15)
138 #define A64_UXTW(sf, Rd, Rn) A64_UBFM(sf, Rd, Rn, 0, 31)
140 /* Move wide (immediate) */
141 #define A64_MOVEW(sf, Rd, imm16, shift, type) \
142 aarch64_insn_gen_movewide(Rd, imm16, shift, \
143 A64_VARIANT(sf), AARCH64_INSN_MOVEWIDE_##type)
144 /* Rd = Zeros (for MOVZ);
145 * Rd |= imm16 << shift (where shift is {0, 16, 32, 48});
146 * Rd = ~Rd; (for MOVN); */
147 #define A64_MOVN(sf, Rd, imm16, shift) A64_MOVEW(sf, Rd, imm16, shift, INVERSE)
148 #define A64_MOVZ(sf, Rd, imm16, shift) A64_MOVEW(sf, Rd, imm16, shift, ZERO)
149 #define A64_MOVK(sf, Rd, imm16, shift) A64_MOVEW(sf, Rd, imm16, shift, KEEP)
151 /* Add/subtract (shifted register) */
152 #define A64_ADDSUB_SREG(sf, Rd, Rn, Rm, type) \
153 aarch64_insn_gen_add_sub_shifted_reg(Rd, Rn, Rm, 0, \
154 A64_VARIANT(sf), AARCH64_INSN_ADSB_##type)
155 /* Rd = Rn OP Rm */
156 #define A64_ADD(sf, Rd, Rn, Rm) A64_ADDSUB_SREG(sf, Rd, Rn, Rm, ADD)
157 #define A64_SUB(sf, Rd, Rn, Rm) A64_ADDSUB_SREG(sf, Rd, Rn, Rm, SUB)
158 #define A64_SUBS(sf, Rd, Rn, Rm) A64_ADDSUB_SREG(sf, Rd, Rn, Rm, SUB_SETFLAGS)
159 /* Rd = -Rm */
160 #define A64_NEG(sf, Rd, Rm) A64_SUB(sf, Rd, A64_ZR, Rm)
161 /* Rn - Rm; set condition flags */
162 #define A64_CMP(sf, Rn, Rm) A64_SUBS(sf, A64_ZR, Rn, Rm)
164 /* Data-processing (1 source) */
165 #define A64_DATA1(sf, Rd, Rn, type) aarch64_insn_gen_data1(Rd, Rn, \
166 A64_VARIANT(sf), AARCH64_INSN_DATA1_##type)
167 /* Rd = BSWAPx(Rn) */
168 #define A64_REV16(sf, Rd, Rn) A64_DATA1(sf, Rd, Rn, REVERSE_16)
169 #define A64_REV32(sf, Rd, Rn) A64_DATA1(sf, Rd, Rn, REVERSE_32)
170 #define A64_REV64(Rd, Rn) A64_DATA1(1, Rd, Rn, REVERSE_64)
172 /* Data-processing (2 source) */
173 /* Rd = Rn OP Rm */
174 #define A64_DATA2(sf, Rd, Rn, Rm, type) aarch64_insn_gen_data2(Rd, Rn, Rm, \
175 A64_VARIANT(sf), AARCH64_INSN_DATA2_##type)
176 #define A64_UDIV(sf, Rd, Rn, Rm) A64_DATA2(sf, Rd, Rn, Rm, UDIV)
177 #define A64_LSLV(sf, Rd, Rn, Rm) A64_DATA2(sf, Rd, Rn, Rm, LSLV)
178 #define A64_LSRV(sf, Rd, Rn, Rm) A64_DATA2(sf, Rd, Rn, Rm, LSRV)
179 #define A64_ASRV(sf, Rd, Rn, Rm) A64_DATA2(sf, Rd, Rn, Rm, ASRV)
181 /* Data-processing (3 source) */
182 /* Rd = Ra + Rn * Rm */
183 #define A64_MADD(sf, Rd, Ra, Rn, Rm) aarch64_insn_gen_data3(Rd, Ra, Rn, Rm, \
184 A64_VARIANT(sf), AARCH64_INSN_DATA3_MADD)
185 /* Rd = Rn * Rm */
186 #define A64_MUL(sf, Rd, Rn, Rm) A64_MADD(sf, Rd, A64_ZR, Rn, Rm)
188 /* Logical (shifted register) */
189 #define A64_LOGIC_SREG(sf, Rd, Rn, Rm, type) \
190 aarch64_insn_gen_logical_shifted_reg(Rd, Rn, Rm, 0, \
191 A64_VARIANT(sf), AARCH64_INSN_LOGIC_##type)
192 /* Rd = Rn OP Rm */
193 #define A64_AND(sf, Rd, Rn, Rm) A64_LOGIC_SREG(sf, Rd, Rn, Rm, AND)
194 #define A64_ORR(sf, Rd, Rn, Rm) A64_LOGIC_SREG(sf, Rd, Rn, Rm, ORR)
195 #define A64_EOR(sf, Rd, Rn, Rm) A64_LOGIC_SREG(sf, Rd, Rn, Rm, EOR)
196 #define A64_ANDS(sf, Rd, Rn, Rm) A64_LOGIC_SREG(sf, Rd, Rn, Rm, AND_SETFLAGS)
197 /* Rn & Rm; set condition flags */
198 #define A64_TST(sf, Rn, Rm) A64_ANDS(sf, A64_ZR, Rn, Rm)
200 #endif /* _BPF_JIT_H */