// (c) Copyright 2008-2009, Grazvydas "notaz" Ignotas
// Free for non-commercial use.
-#define CONTEXT_REG 7
+#define CONTEXT_REG 11
// XXX: tcache_ptr type for SVP and SH2 compilers differs..
#define EMIT_PTR(ptr, x) \
#define A_R10M (1 << 10)
#define A_R11M (1 << 11)
#define A_R14M (1 << 14)
+#define A_R15M (1 << 15)
#define A_COND_AL 0xe
#define A_COND_EQ 0x0
#define A_OP_ADD 0x4
#define A_OP_ADC 0x5
#define A_OP_SBC 0x6
+#define A_OP_RSC 0x7
#define A_OP_TST 0x8
#define A_OP_TEQ 0x9
#define A_OP_CMP 0xa
+#define A_OP_CMN 0xa
#define A_OP_ORR 0xc
#define A_OP_MOV 0xd
#define A_OP_BIC 0xe
#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))
#define EOP_MOV_IMM(rd, ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_MOV,0, 0,rd,ror2,imm8)
+#define EOP_MVN_IMM(rd, ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_MVN,0, 0,rd,ror2,imm8)
#define EOP_ORR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ORR,0,rn,rd,ror2,imm8)
#define EOP_EOR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_EOR,0,rn,rd,ror2,imm8)
#define EOP_ADD_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ADD,0,rn,rd,ror2,imm8)
#define EOP_RSB_IMM_C(cond,rd,rn,ror2,imm8) EOP_C_DOP_IMM(cond,A_OP_RSB,0,rn,rd,ror2,imm8)
#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)
+#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)
#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)
#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)
#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)
#define EOP_XXM(cond,p,u,s,w,l,rn,list) \
EMIT(((cond)<<28) | (1<<27) | ((p)<<24) | ((u)<<23) | ((s)<<22) | ((w)<<21) | ((l)<<20) | ((rn)<<16) | (list))
-#define EOP_STMFD_ST(list) EOP_XXM(A_COND_AL,1,0,0,1,0,13,list)
-#define EOP_LDMFD_ST(list) EOP_XXM(A_COND_AL,0,1,0,1,1,13,list)
+#define EOP_STMIA(rb,list) EOP_XXM(A_COND_AL,0,1,0,0,0,rb,list)
+#define EOP_LDMIA(rb,list) EOP_XXM(A_COND_AL,0,1,0,0,1,rb,list)
+
+#define EOP_STMFD_SP(list) EOP_XXM(A_COND_AL,1,0,0,1,0,13,list)
+#define EOP_LDMFD_SP(list) EOP_XXM(A_COND_AL,0,1,0,1,1,13,list)
/* branches */
#define EOP_C_BX(cond,rm) \
#define EOP_BX(rm) EOP_C_BX(A_COND_AL,rm)
+#define EOP_C_B_PTR(ptr,cond,l,signed_immed_24) \
+ EMIT_PTR(ptr, ((cond)<<28) | 0x0a000000 | ((l)<<24) | (signed_immed_24))
+
#define EOP_C_B(cond,l,signed_immed_24) \
- EMIT(((cond)<<28) | 0x0a000000 | ((l)<<24) | (signed_immed_24))
+ EOP_C_B_PTR(tcache_ptr,cond,l,signed_immed_24)
#define EOP_B( signed_immed_24) EOP_C_B(A_COND_AL,0,signed_immed_24)
#define EOP_BL(signed_immed_24) EOP_C_B(A_COND_AL,1,signed_immed_24)
#define EOP_C_SMULL(cond,s,rdhi,rdlo,rs,rm) \
EMIT(((cond)<<28) | 0x00c00000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm))
+#define EOP_C_SMLAL(cond,s,rdhi,rdlo,rs,rm) \
+ EMIT(((cond)<<28) | 0x00e00000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm))
+
#define EOP_MUL(rd,rm,rs) EOP_C_MUL(A_COND_AL,0,rd,rs,rm) // note: rd != rm
#define EOP_C_MRS(cond,rd) \
#define EOP_MSR_REG(rm) EOP_C_MSR_REG(A_COND_AL,rm)
-static void emith_op_imm(int cond, int s, int op, int r, unsigned int imm)
+static void emith_op_imm2(int cond, int s, int op, int rd, int rn, unsigned int imm)
{
- int ror2, rd = r, rn = r;
+ int ror2;
u32 v;
- if (op == A_OP_MOV)
+ if (op == A_OP_MOV) {
rn = 0;
- else if (imm == 0)
+ if (~imm < 0x100) {
+ imm = ~imm;
+ op = A_OP_MVN;
+ }
+ } else if (imm == 0)
return;
for (v = imm, ror2 = 0; v != 0 || op == A_OP_MOV; v >>= 8, ror2 -= 8/2) {
if (op == A_OP_MOV) {
op = A_OP_ORR;
- rn = r;
+ rn = rd;
}
}
}
+#define emith_op_imm(cond, s, op, r, imm) \
+ emith_op_imm2(cond, s, op, r, r, imm)
+
// test op
#define emith_top_imm(cond, op, r, imm) { \
u32 ror2, v; \
return (u32 *)tcache_ptr - start_ptr;
}
+#define JMP_POS(ptr) \
+ ptr = tcache_ptr; \
+ tcache_ptr += sizeof(u32)
+
+#define JMP_EMIT(cond, ptr) { \
+ int val = (u32 *)tcache_ptr - (u32 *)(ptr) - 2; \
+ EOP_C_B_PTR(ptr, cond, 0, val & 0xffffff); \
+}
+
+#define EMITH_JMP_START(cond) { \
+ void *cond_ptr; \
+ JMP_POS(cond_ptr)
+
+#define EMITH_JMP_END(cond) \
+ JMP_EMIT(cond, cond_ptr); \
+}
// fake "simple" or "short" jump - using cond insns instead
#define EMITH_SJMP_START(cond) \
#define emith_move_r_r(d, s) \
EOP_MOV_REG_SIMPLE(d, s)
+#define emith_mvn_r_r(d, s) \
+ EOP_MVN_REG(A_COND_AL,0,d,s,A_AM1_LSL,0)
+
#define emith_or_r_r_r_lsl(d, s1, s2, lslimm) \
EOP_ORR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
#define emith_eor_r_r_r_lsl(d, s1, s2, lslimm) \
EOP_EOR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm)
+#define emith_eor_r_r_r_lsr(d, s1, s2, lsrimm) \
+ EOP_EOR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSR,lsrimm)
+
+#define emith_or_r_r_lsl(d, s, lslimm) \
+ emith_or_r_r_r_lsl(d, d, s, lslimm)
+
+#define emith_eor_r_r_lsr(d, s, lsrimm) \
+ emith_eor_r_r_r_lsr(d, d, s, lsrimm)
+
#define emith_or_r_r_r(d, s1, s2) \
emith_or_r_r_r_lsl(d, s1, s2, 0)
#define emith_sub_r_r(d, s) \
EOP_SUB_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
+#define emith_adc_r_r(d, s) \
+ EOP_ADC_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
+
#define emith_and_r_r(d, s) \
EOP_AND_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0)
#define emith_sbcf_r_r(d, s) \
EOP_SBC_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
+#define emith_eorf_r_r(d, s) \
+ EOP_EOR_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0)
+
#define emith_move_r_imm(r, imm) \
emith_op_imm(A_COND_AL, 0, A_OP_MOV, r, imm)
#define emith_bic_r_imm(r, imm) \
emith_op_imm(A_COND_AL, 0, A_OP_BIC, r, imm)
+#define emith_and_r_imm(r, imm) \
+ emith_op_imm(A_COND_AL, 0, A_OP_AND, r, imm)
+
#define emith_or_r_imm(r, imm) \
emith_op_imm(A_COND_AL, 0, A_OP_ORR, r, imm)
+#define emith_eor_r_imm(r, imm) \
+ emith_op_imm(A_COND_AL, 0, A_OP_EOR, r, imm)
+
// note: only use 8bit imm for these
#define emith_tst_r_imm(r, imm) \
emith_top_imm(A_COND_AL, A_OP_TST, r, imm)
-#define emith_cmp_r_imm(r, imm) \
- emith_top_imm(A_COND_AL, A_OP_CMP, r, imm)
+#define emith_cmp_r_imm(r, imm) { \
+ u32 op = A_OP_CMP, imm_ = imm; \
+ if (~imm_ < 0x100) { \
+ imm_ = ~imm_; \
+ op = A_OP_CMN; \
+ } \
+ emith_top_imm(A_COND_AL, op, r, imm); \
+}
#define emith_subf_r_imm(r, imm) \
emith_op_imm(A_COND_AL, 1, A_OP_SUB, r, imm)
+#define emith_move_r_imm_c(cond, r, imm) \
+ emith_op_imm(cond, 0, A_OP_MOV, r, imm)
+
#define emith_add_r_imm_c(cond, r, imm) \
emith_op_imm(cond, 0, A_OP_ADD, r, imm)
#define emith_or_r_imm_c(cond, r, imm) \
emith_op_imm(cond, 0, A_OP_ORR, r, imm)
+#define emith_eor_r_imm_c(cond, r, imm) \
+ emith_op_imm(cond, 0, A_OP_EOR, r, imm)
+
#define emith_bic_r_imm_c(cond, r, imm) \
emith_op_imm(cond, 0, A_OP_BIC, r, imm)
+#define emith_move_r_imm_s8(r, imm) { \
+ if ((imm) & 0x80) \
+ EOP_MVN_IMM(r, 0, ((imm) ^ 0xff)); \
+ else \
+ EOP_MOV_IMM(r, 0, imm); \
+}
+
+#define emith_and_r_r_imm(d, s, imm) \
+ emith_op_imm2(A_COND_AL, 0, A_OP_AND, d, s, imm)
+
+#define emith_neg_r_r(d, s) \
+ EOP_RSB_IMM(d, s, 0, 0)
+
#define emith_lsl(d, s, cnt) \
EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSL,cnt)
#define emith_lsr(d, s, cnt) \
EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSR,cnt)
+#define emith_asr(d, s, cnt) \
+ EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_ASR,cnt)
+
#define emith_ror(d, s, cnt) \
EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_ROR,cnt)
+#define emith_rol(d, s, cnt) \
+ EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_ROR,32-(cnt)); \
+
#define emith_lslf(d, s, cnt) \
EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_LSL,cnt)
#define emith_rorcf(d) \
EOP_MOV_REG(A_COND_AL,1,d,d,A_AM1_ROR,0) /* ROR #0 -> RRX */
+#define emith_negcf_r_r(d, s) \
+ EOP_C_DOP_IMM(A_COND_AL,A_OP_RSC,1,s,d,0,0)
+
#define emith_mul(d, s1, s2) { \
if ((d) != (s1)) /* rd != rm limitation */ \
EOP_MUL(d, s1, s2); \
#define emith_mul_s64(dlo, dhi, s1, s2) \
EOP_C_SMULL(A_COND_AL,0,dhi,dlo,s1,s2)
+#define emith_mula_s64(dlo, dhi, s1, s2) \
+ EOP_C_SMLAL(A_COND_AL,0,dhi,dlo,s1,s2)
+
// misc
#define emith_ctx_read(r, offs) \
EOP_LDR_IMM(r, CONTEXT_REG, offs)
#define emith_ctx_write(r, offs) \
EOP_STR_IMM(r, CONTEXT_REG, offs)
-#define emith_clear_msb(d, s, count) { \
+#define emith_ctx_do_multiple(op, r, offs, count, tmpr) do { \
+ int v_, r_ = r, c_ = count, b_ = CONTEXT_REG; \
+ for (v_ = 0; c_; c_--, r_++) \
+ v_ |= 1 << r_; \
+ if ((offs) != 0) { \
+ EOP_ADD_IMM(tmpr,CONTEXT_REG,30/2,(offs)>>2);\
+ b_ = tmpr; \
+ } \
+ op(b_,v_); \
+} while(0)
+
+#define emith_ctx_read_multiple(r, offs, count, tmpr) \
+ emith_ctx_do_multiple(EOP_LDMIA, r, offs, count, tmpr)
+
+#define emith_ctx_write_multiple(r, offs, count, tmpr) \
+ emith_ctx_do_multiple(EOP_STMIA, r, offs, count, tmpr)
+
+#define emith_clear_msb_c(cond, d, s, count) { \
u32 t; \
if ((count) <= 8) { \
t = (count) - 8; \
t = (0xff << t) & 0xff; \
EOP_BIC_IMM(d,s,8/2,t); \
+ EOP_C_DOP_IMM(cond,A_OP_BIC,0,s,d,8/2,t); \
} else if ((count) >= 24) { \
t = (count) - 24; \
t = 0xff >> t; \
EOP_AND_IMM(d,s,0,t); \
+ EOP_C_DOP_IMM(cond,A_OP_AND,0,s,d,0,t); \
} else { \
- EOP_MOV_REG_LSL(d,s,count); \
- EOP_MOV_REG_LSR(d,d,count); \
+ EOP_MOV_REG(cond,0,d,s,A_AM1_LSL,count); \
+ EOP_MOV_REG(cond,0,d,d,A_AM1_LSR,count); \
} \
}
+#define emith_clear_msb(d, s, count) \
+ emith_clear_msb_c(A_COND_AL, d, s, count)
+
#define emith_sext(d, s, bits) { \
EOP_MOV_REG_LSL(d,s,32 - (bits)); \
EOP_MOV_REG_ASR(d,d,32 - (bits)); \
}
-// put bit0 of r0 to carry
-#define emith_set_carry(r0) \
- EOP_TST_REG(A_COND_AL,r0,r0,A_AM1_LSR,1) /* shift out to carry */ \
-
-// put bit0 of r0 to carry (for subtraction, inverted on ARM)
-#define emith_set_carry_sub(r0) { \
- int t = rcache_get_tmp(); \
- EOP_EOR_IMM(t,r0,0,1); /* invert */ \
- EOP_MOV_REG(A_COND_AL,1,t,t,A_AM1_LSR,1); /* shift out to carry */ \
- rcache_free_tmp(t); \
-}
-
#define host_arg2reg(rd, arg) \
rd = arg
#define emith_jump(target) \
emith_jump_cond(A_COND_AL, target)
+#define emith_jump_reg(r) \
+ EOP_BX(r)
+
/* SH2 drc specific */
+#define emith_sh2_drc_entry() \
+ EOP_STMFD_SP(A_R4M|A_R5M|A_R6M|A_R7M|A_R8M|A_R9M|A_R10M|A_R11M|A_R14M)
+
+#define emith_sh2_drc_exit() \
+ EOP_LDMFD_SP(A_R4M|A_R5M|A_R6M|A_R7M|A_R8M|A_R9M|A_R10M|A_R11M|A_R15M)
+
#define emith_sh2_test_t() { \
int r = rcache_get_reg(SHR_SR, RC_GR_READ); \
EOP_TST_IMM(r, 0, 1); \
#define emith_sh2_dtbf_loop() { \
int cr, rn; \
- tmp = rcache_get_tmp(); \
+ int tmp_ = rcache_get_tmp(); \
cr = rcache_get_reg(SHR_SR, RC_GR_RMW); \
rn = rcache_get_reg((op >> 8) & 0x0f, RC_GR_RMW); \
emith_sub_r_imm(rn, 1); /* sub rn, #1 */ \
emith_bic_r_imm(cr, 1); /* bic cr, #1 */ \
emith_sub_r_imm(cr, (cycles+1) << 12); /* sub cr, #(cycles+1)<<12 */ \
cycles = 0; \
- emith_asrf(tmp, cr, 2+12); /* movs tmp, cr, asr #2+12 */ \
- EOP_MOV_IMM_C(A_COND_MI,tmp,0,0); /* movmi tmp, #0 */ \
+ emith_asrf(tmp_, cr, 2+12); /* movs tmp_, cr, asr #2+12 */\
+ EOP_MOV_IMM_C(A_COND_MI,tmp_,0,0); /* movmi tmp_, #0 */ \
emith_lsl(cr, cr, 20); /* mov cr, cr, lsl #20 */ \
emith_lsr(cr, cr, 20); /* mov cr, cr, lsr #20 */ \
- emith_subf_r_r(rn, tmp); /* subs rn, tmp */ \
- EOP_RSB_IMM_C(A_COND_LS,tmp,rn,0,0); /* rsbls tmp, rn, #0 */ \
- EOP_ORR_REG(A_COND_LS,0,cr,cr,tmp,A_AM1_LSL,12+2); /* orrls cr,tmp,lsl #12+2 */\
+ emith_subf_r_r(rn, tmp_); /* subs rn, tmp_ */ \
+ EOP_RSB_IMM_C(A_COND_LS,tmp_,rn,0,0); /* rsbls tmp_, rn, #0 */ \
+ EOP_ORR_REG(A_COND_LS,0,cr,cr,tmp_,A_AM1_LSL,12+2); /* orrls cr,tmp_,lsl #12+2 */\
EOP_ORR_IMM_C(A_COND_LS,cr,cr,0,1); /* orrls cr, #1 */ \
EOP_MOV_IMM_C(A_COND_LS,rn,0,0); /* movls rn, #0 */ \
- rcache_free_tmp(tmp); \
+ rcache_free_tmp(tmp_); \
}
#define emith_write_sr(srcr) { \
emith_bic_r_imm_c(A_COND_CC, srr, 1); \
} \
}
+
+#define emith_tpop_carry(sr, is_sub) { \
+ if (is_sub) \
+ emith_eor_r_imm(sr, 1); \
+ emith_lsrf(sr, sr, 1); \
+}
+
+#define emith_tpush_carry(sr, is_sub) { \
+ emith_adc_r_r(sr, sr); \
+ if (is_sub) \
+ emith_eor_r_imm(sr, 1); \
+}
+
+/*
+ * if Q
+ * t = carry(Rn += Rm)
+ * else
+ * t = carry(Rn -= Rm)
+ * T ^= t
+ */
+#define emith_sh2_div1_step(rn, rm, sr) { \
+ void *jmp0, *jmp1; \
+ emith_tst_r_imm(sr, Q); /* if (Q ^ M) */ \
+ JMP_POS(jmp0); /* beq do_sub */ \
+ emith_addf_r_r(rn, rm); \
+ emith_eor_r_imm_c(A_COND_CS, sr, T); \
+ JMP_POS(jmp1); /* b done */ \
+ JMP_EMIT(A_COND_EQ, jmp0); /* do_sub: */ \
+ emith_subf_r_r(rn, rm); \
+ emith_eor_r_imm_c(A_COND_CC, sr, T); \
+ JMP_EMIT(A_COND_AL, jmp1); /* done: */ \
+}
+
notaz.gp2x.de / picodrive.git / blobdiff