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