drc: support ms ABI
[picodrive.git] / cpu / drc / emit_arm.c
1 /*
2  * Basic macros to emit ARM instructions and some utils
3  * Copyright (C) 2008,2009,2010 notaz
4  *
5  * This work is licensed under the terms of MAME license.
6  * See COPYING file in the top-level directory.
7  */
8 #define CONTEXT_REG 11
9 #define RET_REG     0
10
11 // XXX: tcache_ptr type for SVP and SH2 compilers differs..
12 #define EMIT_PTR(ptr, x) \
13         do { \
14                 *(u32 *)ptr = x; \
15                 ptr = (void *)((u8 *)ptr + sizeof(u32)); \
16                 COUNT_OP; \
17         } while (0)
18
19 #define EMIT(x) EMIT_PTR(tcache_ptr, x)
20
21 #define A_R4M  (1 << 4)
22 #define A_R5M  (1 << 5)
23 #define A_R6M  (1 << 6)
24 #define A_R7M  (1 << 7)
25 #define A_R8M  (1 << 8)
26 #define A_R9M  (1 << 9)
27 #define A_R10M (1 << 10)
28 #define A_R11M (1 << 11)
29 #define A_R12M (1 << 12)
30 #define A_R14M (1 << 14)
31 #define A_R15M (1 << 15)
32
33 #define A_COND_AL 0xe
34 #define A_COND_EQ 0x0
35 #define A_COND_NE 0x1
36 #define A_COND_HS 0x2
37 #define A_COND_LO 0x3
38 #define A_COND_MI 0x4
39 #define A_COND_PL 0x5
40 #define A_COND_VS 0x6
41 #define A_COND_VC 0x7
42 #define A_COND_HI 0x8
43 #define A_COND_LS 0x9
44 #define A_COND_GE 0xa
45 #define A_COND_LT 0xb
46 #define A_COND_GT 0xc
47 #define A_COND_LE 0xd
48 #define A_COND_CS A_COND_HS
49 #define A_COND_CC A_COND_LO
50
51 /* unified conditions */
52 #define DCOND_EQ A_COND_EQ
53 #define DCOND_NE A_COND_NE
54 #define DCOND_MI A_COND_MI
55 #define DCOND_PL A_COND_PL
56 #define DCOND_HI A_COND_HI
57 #define DCOND_HS A_COND_HS
58 #define DCOND_LO A_COND_LO
59 #define DCOND_GE A_COND_GE
60 #define DCOND_GT A_COND_GT
61 #define DCOND_LT A_COND_LT
62 #define DCOND_LS A_COND_LS
63 #define DCOND_LE A_COND_LE
64 #define DCOND_VS A_COND_VS
65 #define DCOND_VC A_COND_VC
66
67 /* addressing mode 1 */
68 #define A_AM1_LSL 0
69 #define A_AM1_LSR 1
70 #define A_AM1_ASR 2
71 #define A_AM1_ROR 3
72
73 #define A_AM1_IMM(ror2,imm8)                  (((ror2)<<8) | (imm8) | 0x02000000)
74 #define A_AM1_REG_XIMM(shift_imm,shift_op,rm) (((shift_imm)<<7) | ((shift_op)<<5) | (rm))
75 #define A_AM1_REG_XREG(rs,shift_op,rm)        (((rs)<<8) | ((shift_op)<<5) | 0x10 | (rm))
76
77 /* data processing op */
78 #define A_OP_AND 0x0
79 #define A_OP_EOR 0x1
80 #define A_OP_SUB 0x2
81 #define A_OP_RSB 0x3
82 #define A_OP_ADD 0x4
83 #define A_OP_ADC 0x5
84 #define A_OP_SBC 0x6
85 #define A_OP_RSC 0x7
86 #define A_OP_TST 0x8
87 #define A_OP_TEQ 0x9
88 #define A_OP_CMP 0xa
89 #define A_OP_CMN 0xa
90 #define A_OP_ORR 0xc
91 #define A_OP_MOV 0xd
92 #define A_OP_BIC 0xe
93 #define A_OP_MVN 0xf
94
95 #define EOP_C_DOP_X(cond,op,s,rn,rd,shifter_op) \
96         EMIT(((cond)<<28) | ((op)<< 21) | ((s)<<20) | ((rn)<<16) | ((rd)<<12) | (shifter_op))
97
98 #define EOP_C_DOP_IMM(     cond,op,s,rn,rd,ror2,imm8)             EOP_C_DOP_X(cond,op,s,rn,rd,A_AM1_IMM(ror2,imm8))
99 #define EOP_C_DOP_REG_XIMM(cond,op,s,rn,rd,shift_imm,shift_op,rm) EOP_C_DOP_X(cond,op,s,rn,rd,A_AM1_REG_XIMM(shift_imm,shift_op,rm))
100 #define EOP_C_DOP_REG_XREG(cond,op,s,rn,rd,rs,       shift_op,rm) EOP_C_DOP_X(cond,op,s,rn,rd,A_AM1_REG_XREG(rs,       shift_op,rm))
101
102 #define EOP_MOV_IMM(rd,   ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_MOV,0, 0,rd,ror2,imm8)
103 #define EOP_MVN_IMM(rd,   ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_MVN,0, 0,rd,ror2,imm8)
104 #define EOP_ORR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ORR,0,rn,rd,ror2,imm8)
105 #define EOP_EOR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_EOR,0,rn,rd,ror2,imm8)
106 #define EOP_ADD_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ADD,0,rn,rd,ror2,imm8)
107 #define EOP_BIC_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_BIC,0,rn,rd,ror2,imm8)
108 #define EOP_AND_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_AND,0,rn,rd,ror2,imm8)
109 #define EOP_SUB_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_SUB,0,rn,rd,ror2,imm8)
110 #define EOP_TST_IMM(   rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_TST,1,rn, 0,ror2,imm8)
111 #define EOP_CMP_IMM(   rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_CMP,1,rn, 0,ror2,imm8)
112 #define EOP_RSB_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_RSB,0,rn,rd,ror2,imm8)
113
114 #define EOP_MOV_IMM_C(cond,rd,   ror2,imm8) EOP_C_DOP_IMM(cond,A_OP_MOV,0, 0,rd,ror2,imm8)
115 #define EOP_ORR_IMM_C(cond,rd,rn,ror2,imm8) EOP_C_DOP_IMM(cond,A_OP_ORR,0,rn,rd,ror2,imm8)
116 #define EOP_RSB_IMM_C(cond,rd,rn,ror2,imm8) EOP_C_DOP_IMM(cond,A_OP_RSB,0,rn,rd,ror2,imm8)
117
118 #define EOP_MOV_REG(cond,s,rd,   rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_MOV,s, 0,rd,shift_imm,shift_op,rm)
119 #define EOP_MVN_REG(cond,s,rd,   rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_MVN,s, 0,rd,shift_imm,shift_op,rm)
120 #define EOP_ORR_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ORR,s,rn,rd,shift_imm,shift_op,rm)
121 #define EOP_ADD_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ADD,s,rn,rd,shift_imm,shift_op,rm)
122 #define EOP_ADC_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ADC,s,rn,rd,shift_imm,shift_op,rm)
123 #define EOP_SUB_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_SUB,s,rn,rd,shift_imm,shift_op,rm)
124 #define EOP_SBC_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_SBC,s,rn,rd,shift_imm,shift_op,rm)
125 #define EOP_AND_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_AND,s,rn,rd,shift_imm,shift_op,rm)
126 #define EOP_EOR_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_EOR,s,rn,rd,shift_imm,shift_op,rm)
127 #define EOP_CMP_REG(cond,     rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_CMP,1,rn, 0,shift_imm,shift_op,rm)
128 #define EOP_TST_REG(cond,     rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_TST,1,rn, 0,shift_imm,shift_op,rm)
129 #define EOP_TEQ_REG(cond,     rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_TEQ,1,rn, 0,shift_imm,shift_op,rm)
130
131 #define EOP_MOV_REG2(s,rd,   rm,shift_op,rs) EOP_C_DOP_REG_XREG(A_COND_AL,A_OP_MOV,s, 0,rd,rs,shift_op,rm)
132 #define EOP_ADD_REG2(s,rd,rn,rm,shift_op,rs) EOP_C_DOP_REG_XREG(A_COND_AL,A_OP_ADD,s,rn,rd,rs,shift_op,rm)
133 #define EOP_SUB_REG2(s,rd,rn,rm,shift_op,rs) EOP_C_DOP_REG_XREG(A_COND_AL,A_OP_SUB,s,rn,rd,rs,shift_op,rm)
134
135 #define EOP_MOV_REG_SIMPLE(rd,rm)           EOP_MOV_REG(A_COND_AL,0,rd,rm,A_AM1_LSL,0)
136 #define EOP_MOV_REG_LSL(rd,   rm,shift_imm) EOP_MOV_REG(A_COND_AL,0,rd,rm,A_AM1_LSL,shift_imm)
137 #define EOP_MOV_REG_LSR(rd,   rm,shift_imm) EOP_MOV_REG(A_COND_AL,0,rd,rm,A_AM1_LSR,shift_imm)
138 #define EOP_MOV_REG_ASR(rd,   rm,shift_imm) EOP_MOV_REG(A_COND_AL,0,rd,rm,A_AM1_ASR,shift_imm)
139 #define EOP_MOV_REG_ROR(rd,   rm,shift_imm) EOP_MOV_REG(A_COND_AL,0,rd,rm,A_AM1_ROR,shift_imm)
140
141 #define EOP_ORR_REG_SIMPLE(rd,rm)           EOP_ORR_REG(A_COND_AL,0,rd,rd,rm,A_AM1_LSL,0)
142 #define EOP_ORR_REG_LSL(rd,rn,rm,shift_imm) EOP_ORR_REG(A_COND_AL,0,rd,rn,rm,A_AM1_LSL,shift_imm)
143 #define EOP_ORR_REG_LSR(rd,rn,rm,shift_imm) EOP_ORR_REG(A_COND_AL,0,rd,rn,rm,A_AM1_LSR,shift_imm)
144 #define EOP_ORR_REG_ASR(rd,rn,rm,shift_imm) EOP_ORR_REG(A_COND_AL,0,rd,rn,rm,A_AM1_ASR,shift_imm)
145 #define EOP_ORR_REG_ROR(rd,rn,rm,shift_imm) EOP_ORR_REG(A_COND_AL,0,rd,rn,rm,A_AM1_ROR,shift_imm)
146
147 #define EOP_ADD_REG_SIMPLE(rd,rm)           EOP_ADD_REG(A_COND_AL,0,rd,rd,rm,A_AM1_LSL,0)
148 #define EOP_ADD_REG_LSL(rd,rn,rm,shift_imm) EOP_ADD_REG(A_COND_AL,0,rd,rn,rm,A_AM1_LSL,shift_imm)
149 #define EOP_ADD_REG_LSR(rd,rn,rm,shift_imm) EOP_ADD_REG(A_COND_AL,0,rd,rn,rm,A_AM1_LSR,shift_imm)
150
151 #define EOP_TST_REG_SIMPLE(rn,rm)           EOP_TST_REG(A_COND_AL,  rn,   0,A_AM1_LSL,rm)
152
153 #define EOP_MOV_REG2_LSL(rd,   rm,rs)       EOP_MOV_REG2(0,rd,   rm,A_AM1_LSL,rs)
154 #define EOP_MOV_REG2_ROR(rd,   rm,rs)       EOP_MOV_REG2(0,rd,   rm,A_AM1_ROR,rs)
155 #define EOP_ADD_REG2_LSL(rd,rn,rm,rs)       EOP_ADD_REG2(0,rd,rn,rm,A_AM1_LSL,rs)
156 #define EOP_SUB_REG2_LSL(rd,rn,rm,rs)       EOP_SUB_REG2(0,rd,rn,rm,A_AM1_LSL,rs)
157
158 /* addressing mode 2 */
159 #define EOP_C_AM2_IMM(cond,u,b,l,rn,rd,offset_12) \
160         EMIT(((cond)<<28) | 0x05000000 | ((u)<<23) | ((b)<<22) | ((l)<<20) | ((rn)<<16) | ((rd)<<12) | (offset_12))
161
162 #define EOP_C_AM2_REG(cond,u,b,l,rn,rd,shift_imm,shift_op,rm) \
163         EMIT(((cond)<<28) | 0x07000000 | ((u)<<23) | ((b)<<22) | ((l)<<20) | ((rn)<<16) | ((rd)<<12) | \
164                 ((shift_imm)<<7) | ((shift_op)<<5) | (rm))
165
166 /* addressing mode 3 */
167 #define EOP_C_AM3(cond,u,r,l,rn,rd,s,h,immed_reg) \
168         EMIT(((cond)<<28) | 0x01000090 | ((u)<<23) | ((r)<<22) | ((l)<<20) | ((rn)<<16) | ((rd)<<12) | \
169                         ((s)<<6) | ((h)<<5) | (immed_reg))
170
171 #define EOP_C_AM3_IMM(cond,u,l,rn,rd,s,h,offset_8) EOP_C_AM3(cond,u,1,l,rn,rd,s,h,(((offset_8)&0xf0)<<4)|((offset_8)&0xf))
172
173 #define EOP_C_AM3_REG(cond,u,l,rn,rd,s,h,rm)       EOP_C_AM3(cond,u,0,l,rn,rd,s,h,rm)
174
175 /* ldr and str */
176 #define EOP_LDR_IMM2(cond,rd,rn,offset_12)  EOP_C_AM2_IMM(cond,1,0,1,rn,rd,offset_12)
177 #define EOP_LDRB_IMM2(cond,rd,rn,offset_12) EOP_C_AM2_IMM(cond,1,1,1,rn,rd,offset_12)
178
179 #define EOP_LDR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,1,0,1,rn,rd,offset_12)
180 #define EOP_LDR_NEGIMM(rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,0,0,1,rn,rd,offset_12)
181 #define EOP_LDR_SIMPLE(rd,rn)           EOP_C_AM2_IMM(A_COND_AL,1,0,1,rn,rd,0)
182 #define EOP_STR_IMM(   rd,rn,offset_12) EOP_C_AM2_IMM(A_COND_AL,1,0,0,rn,rd,offset_12)
183 #define EOP_STR_SIMPLE(rd,rn)           EOP_C_AM2_IMM(A_COND_AL,1,0,0,rn,rd,0)
184
185 #define EOP_LDR_REG_LSL(cond,rd,rn,rm,shift_imm) EOP_C_AM2_REG(cond,1,0,1,rn,rd,shift_imm,A_AM1_LSL,rm)
186
187 #define EOP_LDRH_IMM2(cond,rd,rn,offset_8)  EOP_C_AM3_IMM(cond,1,1,rn,rd,0,1,offset_8)
188
189 #define EOP_LDRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,1,1,rn,rd,0,1,offset_8)
190 #define EOP_LDRH_SIMPLE(rd,rn)           EOP_C_AM3_IMM(A_COND_AL,1,1,rn,rd,0,1,0)
191 #define EOP_LDRH_REG(   rd,rn,rm)        EOP_C_AM3_REG(A_COND_AL,1,1,rn,rd,0,1,rm)
192 #define EOP_STRH_IMM(   rd,rn,offset_8)  EOP_C_AM3_IMM(A_COND_AL,1,0,rn,rd,0,1,offset_8)
193 #define EOP_STRH_SIMPLE(rd,rn)           EOP_C_AM3_IMM(A_COND_AL,1,0,rn,rd,0,1,0)
194 #define EOP_STRH_REG(   rd,rn,rm)        EOP_C_AM3_REG(A_COND_AL,1,0,rn,rd,0,1,rm)
195
196 /* ldm and stm */
197 #define EOP_XXM(cond,p,u,s,w,l,rn,list) \
198         EMIT(((cond)<<28) | (1<<27) | ((p)<<24) | ((u)<<23) | ((s)<<22) | ((w)<<21) | ((l)<<20) | ((rn)<<16) | (list))
199
200 #define EOP_STMIA(rb,list) EOP_XXM(A_COND_AL,0,1,0,0,0,rb,list)
201 #define EOP_LDMIA(rb,list) EOP_XXM(A_COND_AL,0,1,0,0,1,rb,list)
202
203 #define EOP_STMFD_SP(list) EOP_XXM(A_COND_AL,1,0,0,1,0,13,list)
204 #define EOP_LDMFD_SP(list) EOP_XXM(A_COND_AL,0,1,0,1,1,13,list)
205
206 /* branches */
207 #define EOP_C_BX(cond,rm) \
208         EMIT(((cond)<<28) | 0x012fff10 | (rm))
209
210 #define EOP_C_B_PTR(ptr,cond,l,signed_immed_24) \
211         EMIT_PTR(ptr, ((cond)<<28) | 0x0a000000 | ((l)<<24) | (signed_immed_24))
212
213 #define EOP_C_B(cond,l,signed_immed_24) \
214         EOP_C_B_PTR(tcache_ptr,cond,l,signed_immed_24)
215
216 #define EOP_B( signed_immed_24) EOP_C_B(A_COND_AL,0,signed_immed_24)
217 #define EOP_BL(signed_immed_24) EOP_C_B(A_COND_AL,1,signed_immed_24)
218
219 /* misc */
220 #define EOP_C_MUL(cond,s,rd,rs,rm) \
221         EMIT(((cond)<<28) | ((s)<<20) | ((rd)<<16) | ((rs)<<8) | 0x90 | (rm))
222
223 #define EOP_C_UMULL(cond,s,rdhi,rdlo,rs,rm) \
224         EMIT(((cond)<<28) | 0x00800000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm))
225
226 #define EOP_C_SMULL(cond,s,rdhi,rdlo,rs,rm) \
227         EMIT(((cond)<<28) | 0x00c00000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm))
228
229 #define EOP_C_SMLAL(cond,s,rdhi,rdlo,rs,rm) \
230         EMIT(((cond)<<28) | 0x00e00000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm))
231
232 #define EOP_MUL(rd,rm,rs) EOP_C_MUL(A_COND_AL,0,rd,rs,rm) // note: rd != rm
233
234 #define EOP_C_MRS(cond,rd) \
235         EMIT(((cond)<<28) | 0x010f0000 | ((rd)<<12))
236
237 #define EOP_C_MSR_IMM(cond,ror2,imm) \
238         EMIT(((cond)<<28) | 0x0328f000 | ((ror2)<<8) | (imm)) // cpsr_f
239
240 #define EOP_C_MSR_REG(cond,rm) \
241         EMIT(((cond)<<28) | 0x0128f000 | (rm)) // cpsr_f
242
243 #define EOP_MRS(rd)           EOP_C_MRS(A_COND_AL,rd)
244 #define EOP_MSR_IMM(ror2,imm) EOP_C_MSR_IMM(A_COND_AL,ror2,imm)
245 #define EOP_MSR_REG(rm)       EOP_C_MSR_REG(A_COND_AL,rm)
246
247 #define EOP_MOVW(rd,imm) \
248         EMIT(0xe3000000 | ((rd)<<12) | ((imm)&0xfff) | (((imm)<<4)&0xf0000))
249
250 #define EOP_MOVT(rd,imm) \
251         EMIT(0xe3400000 | ((rd)<<12) | (((imm)>>16)&0xfff) | (((imm)>>12)&0xf0000))
252
253 // XXX: AND, RSB, *C, will break if 1 insn is not enough
254 static void emith_op_imm2(int cond, int s, int op, int rd, int rn, unsigned int imm)
255 {
256         int ror2;
257         u32 v;
258
259         switch (op) {
260         case A_OP_MOV:
261                 rn = 0;
262                 if (~imm < 0x10000) {
263                         imm = ~imm;
264                         op = A_OP_MVN;
265                 }
266 #ifdef HAVE_ARMV7
267                 for (v = imm, ror2 = 0; v && !(v & 3); v >>= 2)
268                         ror2--;
269                 if (v >> 8) {
270                         /* 2+ insns needed - prefer movw/movt */
271                         if (op == A_OP_MVN)
272                                 imm = ~imm;
273                         EOP_MOVW(rd, imm);
274                         if (imm & 0xffff0000)
275                                 EOP_MOVT(rd, imm);
276                         return;
277                 }
278 #endif
279                 break;
280
281         case A_OP_EOR:
282         case A_OP_SUB:
283         case A_OP_ADD:
284         case A_OP_ORR:
285         case A_OP_BIC:
286                 if (s == 0 && imm == 0)
287                         return;
288                 break;
289         }
290
291         for (v = imm, ror2 = 0; ; ror2 -= 8/2) {
292                 /* shift down to get 'best' rot2 */
293                 for (; v && !(v & 3); v >>= 2)
294                         ror2--;
295
296                 EOP_C_DOP_IMM(cond, op, s, rn, rd, ror2 & 0x0f, v & 0xff);
297
298                 v >>= 8;
299                 if (v == 0)
300                         break;
301                 if (op == A_OP_MOV)
302                         op = A_OP_ORR;
303                 if (op == A_OP_MVN)
304                         op = A_OP_BIC;
305                 rn = rd;
306         }
307 }
308
309 #define emith_op_imm(cond, s, op, r, imm) \
310         emith_op_imm2(cond, s, op, r, r, imm)
311
312 // test op
313 #define emith_top_imm(cond, op, r, imm) do { \
314         u32 ror2, v; \
315         for (ror2 = 0, v = imm; v && !(v & 3); v >>= 2) \
316                 ror2--; \
317         EOP_C_DOP_IMM(cond, op, 1, r, 0, ror2 & 0x0f, v & 0xff); \
318 } while (0)
319
320 #define is_offset_24(val) \
321         ((val) >= (int)0xff000000 && (val) <= 0x00ffffff)
322
323 static int emith_xbranch(int cond, void *target, int is_call)
324 {
325         int val = (u32 *)target - (u32 *)tcache_ptr - 2;
326         int direct = is_offset_24(val);
327         u32 *start_ptr = (u32 *)tcache_ptr;
328
329         if (direct)
330         {
331                 EOP_C_B(cond,is_call,val & 0xffffff);           // b, bl target
332         }
333         else
334         {
335 #ifdef __EPOC32__
336 //              elprintf(EL_SVP, "emitting indirect jmp %08x->%08x", tcache_ptr, target);
337                 if (is_call)
338                         EOP_ADD_IMM(14,15,0,8);                 // add lr,pc,#8
339                 EOP_C_AM2_IMM(cond,1,0,1,15,15,0);              // ldrcc pc,[pc]
340                 EOP_MOV_REG_SIMPLE(15,15);                      // mov pc, pc
341                 EMIT((u32)target);
342 #else
343                 // should never happen
344                 elprintf(EL_STATUS|EL_SVP|EL_ANOMALY, "indirect jmp %08x->%08x", target, tcache_ptr);
345                 exit(1);
346 #endif
347         }
348
349         return (u32 *)tcache_ptr - start_ptr;
350 }
351
352 #define JMP_POS(ptr) \
353         ptr = tcache_ptr; \
354         tcache_ptr += sizeof(u32)
355
356 #define JMP_EMIT(cond, ptr) { \
357         u32 val_ = (u32 *)tcache_ptr - (u32 *)(ptr) - 2; \
358         EOP_C_B_PTR(ptr, cond, 0, val_ & 0xffffff); \
359 }
360
361 #define EMITH_JMP_START(cond) { \
362         void *cond_ptr; \
363         JMP_POS(cond_ptr)
364
365 #define EMITH_JMP_END(cond) \
366         JMP_EMIT(cond, cond_ptr); \
367 }
368
369 // fake "simple" or "short" jump - using cond insns instead
370 #define EMITH_NOTHING1(cond) \
371         (void)(cond)
372
373 #define EMITH_SJMP_DECL_()
374 #define EMITH_SJMP_START_(cond) EMITH_NOTHING1(cond)
375 #define EMITH_SJMP_END_(cond)   EMITH_NOTHING1(cond)
376 #define EMITH_SJMP_START(cond)  EMITH_NOTHING1(cond)
377 #define EMITH_SJMP_END(cond)    EMITH_NOTHING1(cond)
378 #define EMITH_SJMP3_START(cond) EMITH_NOTHING1(cond)
379 #define EMITH_SJMP3_MID(cond)   EMITH_NOTHING1(cond)
380 #define EMITH_SJMP3_END()
381
382 #define emith_move_r_r(d, s) \
383         EOP_MOV_REG_SIMPLE(d, s)
384
385 #define emith_move_r_r_ptr(d, s) \
386         emith_move_r_r(d, s)
387
388 #define emith_mvn_r_r(d, s) \
389         EOP_MVN_REG(A_COND_AL,0,d,s,A_AM1_LSL,0)
390
391 #define emith_add_r_r_r_lsl(d, s1, s2, lslimm) \
392         EOP_ADD_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
393
394 #define emith_or_r_r_r_lsl(d, s1, s2, lslimm) \
395         EOP_ORR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
396
397 #define emith_eor_r_r_r_lsl(d, s1, s2, lslimm) \
398         EOP_EOR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
399
400 #define emith_eor_r_r_r_lsr(d, s1, s2, lsrimm) \
401         EOP_EOR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSR,lsrimm)
402
403 #define emith_or_r_r_lsl(d, s, lslimm) \
404         emith_or_r_r_r_lsl(d, d, s, lslimm)
405
406 #define emith_eor_r_r_lsr(d, s, lsrimm) \
407         emith_eor_r_r_r_lsr(d, d, s, lsrimm)
408
409 #define emith_add_r_r_r(d, s1, s2) \
410         emith_add_r_r_r_lsl(d, s1, s2, 0)
411
412 #define emith_or_r_r_r(d, s1, s2) \
413         emith_or_r_r_r_lsl(d, s1, s2, 0)
414
415 #define emith_eor_r_r_r(d, s1, s2) \
416         emith_eor_r_r_r_lsl(d, s1, s2, 0)
417
418 #define emith_add_r_r(d, s) \
419         emith_add_r_r_r(d, d, s)
420
421 #define emith_sub_r_r(d, s) \
422         EOP_SUB_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
423
424 #define emith_adc_r_r(d, s) \
425         EOP_ADC_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
426
427 #define emith_and_r_r(d, s) \
428         EOP_AND_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
429
430 #define emith_or_r_r(d, s) \
431         emith_or_r_r_r(d, d, s)
432
433 #define emith_eor_r_r(d, s) \
434         emith_eor_r_r_r(d, d, s)
435
436 #define emith_tst_r_r(d, s) \
437         EOP_TST_REG(A_COND_AL,d,s,A_AM1_LSL,0)
438
439 #define emith_tst_r_r_ptr(d, s) \
440         emith_tst_r_r(d, s)
441
442 #define emith_teq_r_r(d, s) \
443         EOP_TEQ_REG(A_COND_AL,d,s,A_AM1_LSL,0)
444
445 #define emith_cmp_r_r(d, s) \
446         EOP_CMP_REG(A_COND_AL,d,s,A_AM1_LSL,0)
447
448 #define emith_addf_r_r(d, s) \
449         EOP_ADD_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
450
451 #define emith_subf_r_r(d, s) \
452         EOP_SUB_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
453
454 #define emith_adcf_r_r(d, s) \
455         EOP_ADC_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
456
457 #define emith_sbcf_r_r(d, s) \
458         EOP_SBC_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
459
460 #define emith_eorf_r_r(d, s) \
461         EOP_EOR_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
462
463 #define emith_move_r_imm(r, imm) \
464         emith_op_imm(A_COND_AL, 0, A_OP_MOV, r, imm)
465
466 #define emith_add_r_imm(r, imm) \
467         emith_op_imm(A_COND_AL, 0, A_OP_ADD, r, imm)
468
469 #define emith_adc_r_imm(r, imm) \
470         emith_op_imm(A_COND_AL, 0, A_OP_ADC, r, imm)
471
472 #define emith_sub_r_imm(r, imm) \
473         emith_op_imm(A_COND_AL, 0, A_OP_SUB, r, imm)
474
475 #define emith_bic_r_imm(r, imm) \
476         emith_op_imm(A_COND_AL, 0, A_OP_BIC, r, imm)
477
478 #define emith_and_r_imm(r, imm) \
479         emith_op_imm(A_COND_AL, 0, A_OP_AND, r, imm)
480
481 #define emith_or_r_imm(r, imm) \
482         emith_op_imm(A_COND_AL, 0, A_OP_ORR, r, imm)
483
484 #define emith_eor_r_imm(r, imm) \
485         emith_op_imm(A_COND_AL, 0, A_OP_EOR, r, imm)
486
487 // note: only use 8bit imm for these
488 #define emith_tst_r_imm(r, imm) \
489         emith_top_imm(A_COND_AL, A_OP_TST, r, imm)
490
491 #define emith_cmp_r_imm(r, imm) { \
492         u32 op = A_OP_CMP, imm_ = imm; \
493         if (~imm_ < 0x100) { \
494                 imm_ = ~imm_; \
495                 op = A_OP_CMN; \
496         } \
497         emith_top_imm(A_COND_AL, op, r, imm); \
498 }
499
500 #define emith_subf_r_imm(r, imm) \
501         emith_op_imm(A_COND_AL, 1, A_OP_SUB, r, imm)
502
503 #define emith_move_r_imm_c(cond, r, imm) \
504         emith_op_imm(cond, 0, A_OP_MOV, r, imm)
505
506 #define emith_add_r_imm_c(cond, r, imm) \
507         emith_op_imm(cond, 0, A_OP_ADD, r, imm)
508
509 #define emith_sub_r_imm_c(cond, r, imm) \
510         emith_op_imm(cond, 0, A_OP_SUB, r, imm)
511
512 #define emith_or_r_imm_c(cond, r, imm) \
513         emith_op_imm(cond, 0, A_OP_ORR, r, imm)
514
515 #define emith_eor_r_imm_c(cond, r, imm) \
516         emith_op_imm(cond, 0, A_OP_EOR, r, imm)
517
518 #define emith_bic_r_imm_c(cond, r, imm) \
519         emith_op_imm(cond, 0, A_OP_BIC, r, imm)
520
521 #define emith_move_r_imm_s8(r, imm) { \
522         if ((imm) & 0x80) \
523                 EOP_MVN_IMM(r, 0, ((imm) ^ 0xff)); \
524         else \
525                 EOP_MOV_IMM(r, 0, imm); \
526 }
527
528 #define emith_and_r_r_imm(d, s, imm) \
529         emith_op_imm2(A_COND_AL, 0, A_OP_AND, d, s, imm)
530
531 #define emith_add_r_r_imm(d, s, imm) \
532         emith_op_imm2(A_COND_AL, 0, A_OP_ADD, d, s, imm)
533
534 #define emith_add_r_r_ptr_imm(d, s, imm) \
535         emith_add_r_r_imm(d, s, imm)
536
537 #define emith_sub_r_r_imm(d, s, imm) \
538         emith_op_imm2(A_COND_AL, 0, A_OP_SUB, d, s, imm)
539
540 #define emith_neg_r_r(d, s) \
541         EOP_RSB_IMM(d, s, 0, 0)
542
543 #define emith_lsl(d, s, cnt) \
544         EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSL,cnt)
545
546 #define emith_lsr(d, s, cnt) \
547         EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSR,cnt)
548
549 #define emith_asr(d, s, cnt) \
550         EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_ASR,cnt)
551
552 #define emith_ror_c(cond, d, s, cnt) \
553         EOP_MOV_REG(cond,0,d,s,A_AM1_ROR,cnt)
554
555 #define emith_ror(d, s, cnt) \
556         emith_ror_c(A_COND_AL, d, s, cnt)
557
558 #define emith_rol(d, s, cnt) \
559         EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_ROR,32-(cnt)); \
560
561 #define emith_lslf(d, s, cnt) \
562         EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_LSL,cnt)
563
564 #define emith_lsrf(d, s, cnt) \
565         EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_LSR,cnt)
566
567 #define emith_asrf(d, s, cnt) \
568         EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_ASR,cnt)
569
570 // note: only C flag updated correctly
571 #define emith_rolf(d, s, cnt) { \
572         EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_ROR,32-(cnt)); \
573         /* we don't have ROL so we shift to get the right carry */ \
574         EOP_TST_REG(A_COND_AL,d,d,A_AM1_LSR,1); \
575 }
576
577 #define emith_rorf(d, s, cnt) \
578         EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_ROR,cnt)
579
580 #define emith_rolcf(d) \
581         emith_adcf_r_r(d, d)
582
583 #define emith_rorcf(d) \
584         EOP_MOV_REG(A_COND_AL,1,d,d,A_AM1_ROR,0) /* ROR #0 -> RRX */
585
586 #define emith_negcf_r_r(d, s) \
587         EOP_C_DOP_IMM(A_COND_AL,A_OP_RSC,1,s,d,0,0)
588
589 #define emith_mul(d, s1, s2) { \
590         if ((d) != (s1)) /* rd != rm limitation */ \
591                 EOP_MUL(d, s1, s2); \
592         else \
593                 EOP_MUL(d, s2, s1); \
594 }
595
596 #define emith_mul_u64(dlo, dhi, s1, s2) \
597         EOP_C_UMULL(A_COND_AL,0,dhi,dlo,s1,s2)
598
599 #define emith_mul_s64(dlo, dhi, s1, s2) \
600         EOP_C_SMULL(A_COND_AL,0,dhi,dlo,s1,s2)
601
602 #define emith_mula_s64(dlo, dhi, s1, s2) \
603         EOP_C_SMLAL(A_COND_AL,0,dhi,dlo,s1,s2)
604
605 // misc
606 #define emith_read_r_r_offs_c(cond, r, rs, offs) \
607         EOP_LDR_IMM2(cond, r, rs, offs)
608
609 #define emith_read8_r_r_offs_c(cond, r, rs, offs) \
610         EOP_LDRB_IMM2(cond, r, rs, offs)
611
612 #define emith_read16_r_r_offs_c(cond, r, rs, offs) \
613         EOP_LDRH_IMM2(cond, r, rs, offs)
614
615 #define emith_read_r_r_offs(r, rs, offs) \
616         emith_read_r_r_offs_c(A_COND_AL, r, rs, offs)
617
618 #define emith_read8_r_r_offs(r, rs, offs) \
619         emith_read8_r_r_offs_c(A_COND_AL, r, rs, offs)
620
621 #define emith_read16_r_r_offs(r, rs, offs) \
622         emith_read16_r_r_offs_c(A_COND_AL, r, rs, offs)
623
624 #define emith_ctx_read(r, offs) \
625         emith_read_r_r_offs(r, CONTEXT_REG, offs)
626
627 #define emith_ctx_read_ptr(r, offs) \
628         emith_ctx_read(r, offs)
629
630 #define emith_ctx_write(r, offs) \
631         EOP_STR_IMM(r, CONTEXT_REG, offs)
632
633 #define emith_ctx_do_multiple(op, r, offs, count, tmpr) do { \
634         int v_, r_ = r, c_ = count, b_ = CONTEXT_REG;        \
635         for (v_ = 0; c_; c_--, r_++)                         \
636                 v_ |= 1 << r_;                               \
637         if ((offs) != 0) {                                   \
638                 EOP_ADD_IMM(tmpr,CONTEXT_REG,30/2,(offs)>>2);\
639                 b_ = tmpr;                                   \
640         }                                                    \
641         op(b_,v_);                                           \
642 } while(0)
643
644 #define emith_ctx_read_multiple(r, offs, count, tmpr) \
645         emith_ctx_do_multiple(EOP_LDMIA, r, offs, count, tmpr)
646
647 #define emith_ctx_write_multiple(r, offs, count, tmpr) \
648         emith_ctx_do_multiple(EOP_STMIA, r, offs, count, tmpr)
649
650 #define emith_clear_msb_c(cond, d, s, count) { \
651         u32 t; \
652         if ((count) <= 8) { \
653                 t = (count) - 8; \
654                 t = (0xff << t) & 0xff; \
655                 EOP_BIC_IMM(d,s,8/2,t); \
656                 EOP_C_DOP_IMM(cond,A_OP_BIC,0,s,d,8/2,t); \
657         } else if ((count) >= 24) { \
658                 t = (count) - 24; \
659                 t = 0xff >> t; \
660                 EOP_AND_IMM(d,s,0,t); \
661                 EOP_C_DOP_IMM(cond,A_OP_AND,0,s,d,0,t); \
662         } else { \
663                 EOP_MOV_REG(cond,0,d,s,A_AM1_LSL,count); \
664                 EOP_MOV_REG(cond,0,d,d,A_AM1_LSR,count); \
665         } \
666 }
667
668 #define emith_clear_msb(d, s, count) \
669         emith_clear_msb_c(A_COND_AL, d, s, count)
670
671 #define emith_sext(d, s, bits) { \
672         EOP_MOV_REG_LSL(d,s,32 - (bits)); \
673         EOP_MOV_REG_ASR(d,d,32 - (bits)); \
674 }
675
676 #define emith_do_caller_regs(mask, func) { \
677         u32 _reg_mask = (mask) & 0x500f; \
678         if (_reg_mask) { \
679                 if (__builtin_parity(_reg_mask) == 1) \
680                         _reg_mask |= 0x10; /* eabi align */ \
681                 func(_reg_mask); \
682         } \
683 }
684
685 #define emith_save_caller_regs(mask) \
686         emith_do_caller_regs(mask, EOP_STMFD_SP)
687
688 #define emith_restore_caller_regs(mask) \
689         emith_do_caller_regs(mask, EOP_LDMFD_SP)
690
691 // upto 4 args
692 #define emith_pass_arg_r(arg, reg) \
693         EOP_MOV_REG_SIMPLE(arg, reg)
694
695 #define emith_pass_arg_imm(arg, imm) \
696         emith_move_r_imm(arg, imm)
697
698 #define emith_jump(target) \
699         emith_jump_cond(A_COND_AL, target)
700
701 #define emith_jump_patchable(target) \
702         emith_jump(target)
703
704 #define emith_jump_cond(cond, target) \
705         emith_xbranch(cond, target, 0)
706
707 #define emith_jump_cond_patchable(cond, target) \
708         emith_jump_cond(cond, target)
709
710 #define emith_jump_patch(ptr, target) do { \
711         u32 *ptr_ = ptr; \
712         u32 val_ = (u32 *)(target) - ptr_ - 2; \
713         *ptr_ = (*ptr_ & 0xff000000) | (val_ & 0x00ffffff); \
714 } while (0)
715
716 #define emith_jump_at(ptr, target) { \
717         u32 val_ = (u32 *)(target) - (u32 *)(ptr) - 2; \
718         EOP_C_B_PTR(ptr, A_COND_AL, 0, val_ & 0xffffff); \
719 }
720
721 #define emith_jump_reg_c(cond, r) \
722         EOP_C_BX(cond, r)
723
724 #define emith_jump_reg(r) \
725         emith_jump_reg_c(A_COND_AL, r)
726
727 #define emith_jump_ctx_c(cond, offs) \
728         EOP_LDR_IMM2(cond,15,CONTEXT_REG,offs)
729
730 #define emith_jump_ctx(offs) \
731         emith_jump_ctx_c(A_COND_AL, offs)
732
733 #define emith_call_cond(cond, target) \
734         emith_xbranch(cond, target, 1)
735
736 #define emith_call(target) \
737         emith_call_cond(A_COND_AL, target)
738
739 #define emith_call_ctx(offs) { \
740         emith_move_r_r(14, 15); \
741         emith_jump_ctx(offs); \
742 }
743
744 #define emith_ret_c(cond) \
745         emith_jump_reg_c(cond, 14)
746
747 #define emith_ret() \
748         emith_ret_c(A_COND_AL)
749
750 #define emith_ret_to_ctx(offs) \
751         emith_ctx_write(14, offs)
752
753 #define emith_push_ret() \
754         EOP_STMFD_SP(A_R14M)
755
756 #define emith_pop_and_ret() \
757         EOP_LDMFD_SP(A_R15M)
758
759 #define host_instructions_updated(base, end) \
760         cache_flush_d_inval_i(base, end)
761
762 #define host_arg2reg(rd, arg) \
763         rd = arg
764
765 /* SH2 drc specific */
766 /* pushes r12 for eabi alignment */
767 #define emith_sh2_drc_entry() \
768         EOP_STMFD_SP(A_R4M|A_R5M|A_R6M|A_R7M|A_R8M|A_R9M|A_R10M|A_R11M|A_R12M|A_R14M)
769
770 #define emith_sh2_drc_exit() \
771         EOP_LDMFD_SP(A_R4M|A_R5M|A_R6M|A_R7M|A_R8M|A_R9M|A_R10M|A_R11M|A_R12M|A_R15M)
772
773 #define emith_sh2_wcall(a, tab) { \
774         emith_lsr(12, a, SH2_WRITE_SHIFT); \
775         EOP_LDR_REG_LSL(A_COND_AL,12,tab,12,2); \
776         emith_move_r_r(2, CONTEXT_REG); \
777         emith_jump_reg(12); \
778 }
779
780 #define emith_sh2_dtbf_loop() { \
781         int cr, rn;                                                          \
782         int tmp_ = rcache_get_tmp();                                         \
783         cr = rcache_get_reg(SHR_SR, RC_GR_RMW);                              \
784         rn = rcache_get_reg((op >> 8) & 0x0f, RC_GR_RMW);                    \
785         emith_sub_r_imm(rn, 1);                /* sub rn, #1 */              \
786         emith_bic_r_imm(cr, 1);                /* bic cr, #1 */              \
787         emith_sub_r_imm(cr, (cycles+1) << 12); /* sub cr, #(cycles+1)<<12 */ \
788         cycles = 0;                                                          \
789         emith_asrf(tmp_, cr, 2+12);            /* movs tmp_, cr, asr #2+12 */\
790         EOP_MOV_IMM_C(A_COND_MI,tmp_,0,0);     /* movmi tmp_, #0 */          \
791         emith_lsl(cr, cr, 20);                 /* mov cr, cr, lsl #20 */     \
792         emith_lsr(cr, cr, 20);                 /* mov cr, cr, lsr #20 */     \
793         emith_subf_r_r(rn, tmp_);              /* subs rn, tmp_ */           \
794         EOP_RSB_IMM_C(A_COND_LS,tmp_,rn,0,0);  /* rsbls tmp_, rn, #0 */      \
795         EOP_ORR_REG(A_COND_LS,0,cr,cr,tmp_,A_AM1_LSL,12+2); /* orrls cr,tmp_,lsl #12+2 */\
796         EOP_ORR_IMM_C(A_COND_LS,cr,cr,0,1);    /* orrls cr, #1 */            \
797         EOP_MOV_IMM_C(A_COND_LS,rn,0,0);       /* movls rn, #0 */            \
798         rcache_free_tmp(tmp_);                                               \
799 }
800
801 #define emith_write_sr(sr, srcr) { \
802         emith_lsr(sr, sr, 10); \
803         emith_or_r_r_r_lsl(sr, sr, srcr, 22); \
804         emith_ror(sr, sr, 22); \
805 }
806
807 #define emith_carry_to_t(srr, is_sub) { \
808         if (is_sub) { /* has inverted C on ARM */ \
809                 emith_or_r_imm_c(A_COND_CC, srr, 1); \
810                 emith_bic_r_imm_c(A_COND_CS, srr, 1); \
811         } else { \
812                 emith_or_r_imm_c(A_COND_CS, srr, 1); \
813                 emith_bic_r_imm_c(A_COND_CC, srr, 1); \
814         } \
815 }
816
817 #define emith_tpop_carry(sr, is_sub) {  \
818         if (is_sub)                     \
819                 emith_eor_r_imm(sr, 1); \
820         emith_lsrf(sr, sr, 1);          \
821 }
822
823 #define emith_tpush_carry(sr, is_sub) { \
824         emith_adc_r_r(sr, sr);          \
825         if (is_sub)                     \
826                 emith_eor_r_imm(sr, 1); \
827 }
828
829 /*
830  * if Q
831  *   t = carry(Rn += Rm)
832  * else
833  *   t = carry(Rn -= Rm)
834  * T ^= t
835  */
836 #define emith_sh2_div1_step(rn, rm, sr) {         \
837         void *jmp0, *jmp1;                        \
838         emith_tst_r_imm(sr, Q);  /* if (Q ^ M) */ \
839         JMP_POS(jmp0);           /* beq do_sub */ \
840         emith_addf_r_r(rn, rm);                   \
841         emith_eor_r_imm_c(A_COND_CS, sr, T);      \
842         JMP_POS(jmp1);           /* b done */     \
843         JMP_EMIT(A_COND_EQ, jmp0); /* do_sub: */  \
844         emith_subf_r_r(rn, rm);                   \
845         emith_eor_r_imm_c(A_COND_CC, sr, T);      \
846         JMP_EMIT(A_COND_AL, jmp1); /* done: */    \
847 }
848