32x drc functional on ARM, random adjustments
[picodrive.git] / pico / carthw / svp / compiler.c
index 4fad3d7..c26e8e3 100644 (file)
@@ -7,7 +7,13 @@
 #include "../../../cpu/drc/cmn.h"
 #include "compiler.h"
 
-#define u32 unsigned int
+// FIXME: asm has these hardcoded
+#define SSP_BLOCKTAB_ENTS       (0x5090/2)
+#define SSP_BLOCKTAB_IRAM_ONE   (0x800/2) // table entries
+#define SSP_BLOCKTAB_IRAM_ENTS  (15*SSP_BLOCKTAB_IRAM_ONE)
+
+static u32 **ssp_block_table; // [0x5090/2];
+static u32 **ssp_block_table_iram; // [15][0x800/2];
 
 static u32 *tcache_ptr = NULL;
 
@@ -23,13 +29,14 @@ extern ssp1601_t *ssp;
 #define SSP_FLAG_N (1<<0xf)
 
 #ifndef ARM
-#define DUMP_BLOCK 0x0c9a
+//#define DUMP_BLOCK 0x0c9a
 void ssp_drc_next(void){}
 void ssp_drc_next_patch(void){}
 void ssp_drc_end(void){}
 #endif
 
-#include "gen_arm.c"
+#define COUNT_OP
+#include "../../../cpu/drc/emit_arm.c"
 
 // -----------------------------------------------------
 
@@ -279,11 +286,11 @@ static void tr_flush_dirty_prs(void)
        int i, ror = 0, reg;
        int dirty = dirty_regb >> 8;
        if ((dirty&7) == 7) {
-               emit_mov_const(A_COND_AL, 8, known_regs.r[0]|(known_regs.r[1]<<8)|(known_regs.r[2]<<16));
+               emith_move_r_imm(8, known_regs.r[0]|(known_regs.r[1]<<8)|(known_regs.r[2]<<16));
                dirty &= ~7;
        }
        if ((dirty&0x70) == 0x70) {
-               emit_mov_const(A_COND_AL, 9, known_regs.r[4]|(known_regs.r[5]<<8)|(known_regs.r[6]<<16));
+               emith_move_r_imm(9, known_regs.r[4]|(known_regs.r[5]<<8)|(known_regs.r[6]<<16));
                dirty &= ~0x70;
        }
        /* r0-r7 */
@@ -342,14 +349,14 @@ static void tr_make_dirty_ST(void)
 static void tr_mov16(int r, int val)
 {
        if (hostreg_r[r] != val) {
-               emit_mov_const(A_COND_AL, r, val);
+               emith_move_r_imm(r, val);
                hostreg_r[r] = val;
        }
 }
 
 static void tr_mov16_cond(int cond, int r, int val)
 {
-       emit_mov_const(cond, r, val);
+       emith_op_imm(cond, A_OP_MOV, r, val);
        hostreg_r[r] = -1;
 }
 
@@ -361,7 +368,7 @@ static void tr_flush_dirty_pmcrs(void)
 
        if (dirty_regb & KRREG_PMC) {
                val = known_regs.pmc.v;
-               emit_mov_const(A_COND_AL, 1, val);
+               emith_move_r_imm(1, val);
                EOP_STR_IMM(1,7,0x400+SSP_PMC*4);
 
                if (known_regs.emu_status & (SSP_PMC_SET|SSP_PMC_HAVE_ADDR)) {
@@ -374,14 +381,14 @@ static void tr_flush_dirty_pmcrs(void)
                if (dirty_regb & (1 << (20+i))) {
                        if (val != known_regs.pmac_read[i]) {
                                val = known_regs.pmac_read[i];
-                               emit_mov_const(A_COND_AL, 1, val);
+                               emith_move_r_imm(1, val);
                        }
                        EOP_STR_IMM(1,7,0x454+i*4); // pmac_read
                }
                if (dirty_regb & (1 << (25+i))) {
                        if (val != known_regs.pmac_write[i]) {
                                val = known_regs.pmac_write[i];
-                               emit_mov_const(A_COND_AL, 1, val);
+                               emith_move_r_imm(1, val);
                        }
                        EOP_STR_IMM(1,7,0x46c+i*4); // pmac_write
                }
@@ -786,7 +793,7 @@ static void tr_PMX_to_r0(int reg)
                if      ((mode & 0xfff0) == 0x0800)
                {
                        EOP_LDR_IMM(1,7,0x488);         // rom_ptr
-                       emit_mov_const(A_COND_AL, 0, (pmcv&0xfffff)<<1);
+                       emith_move_r_imm(0, (pmcv&0xfffff)<<1);
                        EOP_LDRH_REG(0,1,0);            // ldrh r0, [r1, r0]
                        known_regs.pmac_read[reg] += 1;
                }
@@ -794,7 +801,7 @@ static void tr_PMX_to_r0(int reg)
                {
                        int inc = get_inc(mode);
                        EOP_LDR_IMM(1,7,0x490);         // dram_ptr
-                       emit_mov_const(A_COND_AL, 0, (pmcv&0xffff)<<1);
+                       emith_move_r_imm(0, (pmcv&0xffff)<<1);
                        EOP_LDRH_REG(0,1,0);            // ldrh r0, [r1, r0]
                        if (reg == 4 && (pmcv == 0x187f03 || pmcv == 0x187f04)) // wait loop detection
                        {
@@ -829,7 +836,7 @@ static void tr_PMX_to_r0(int reg)
        tr_flush_dirty_ST();
        //tr_flush_dirty_pmcrs();
        tr_mov16(0, reg);
-       emit_call(A_COND_AL, ssp_pm_read);
+       emith_call(ssp_pm_read);
        hostreg_clear();
 }
 
@@ -1028,7 +1035,7 @@ static void tr_r0_to_PMX(int reg)
                        int inc = get_inc(mode);
                        if (mode & 0x0400) tr_unhandled();
                        EOP_LDR_IMM(1,7,0x490);         // dram_ptr
-                       emit_mov_const(A_COND_AL, 2, addr<<1);
+                       emith_move_r_imm(2, addr << 1);
                        EOP_STRH_REG(0,1,2);            // strh r0, [r1, r2]
                        known_regs.pmac_write[reg] += inc;
                }
@@ -1036,7 +1043,7 @@ static void tr_r0_to_PMX(int reg)
                {
                        if (mode & 0x0400) tr_unhandled();
                        EOP_LDR_IMM(1,7,0x490);         // dram_ptr
-                       emit_mov_const(A_COND_AL, 2, addr<<1);
+                       emith_move_r_imm(2, addr << 1);
                        EOP_STRH_REG(0,1,2);            // strh r0, [r1, r2]
                        known_regs.pmac_write[reg] += (addr&1) ? 31 : 1;
                }
@@ -1044,7 +1051,7 @@ static void tr_r0_to_PMX(int reg)
                {
                        int inc = get_inc(mode);
                        EOP_LDR_IMM(1,7,0x48c);         // iram_ptr
-                       emit_mov_const(A_COND_AL, 2, (addr&0x3ff)<<1);
+                       emith_move_r_imm(2, (addr&0x3ff) << 1);
                        EOP_STRH_REG(0,1,2);            // strh r0, [r1, r2]
                        EOP_MOV_IMM(1,0,1);
                        EOP_STR_IMM(1,7,0x494);         // iram_dirty
@@ -1070,7 +1077,7 @@ static void tr_r0_to_PMX(int reg)
        tr_flush_dirty_ST();
        //tr_flush_dirty_pmcrs();
        tr_mov16(1, reg);
-       emit_call(A_COND_AL, ssp_pm_write);
+       emith_call(ssp_pm_write);
        hostreg_clear();
 }
 
@@ -1111,7 +1118,7 @@ static void tr_r0_to_PMC(int const_val)
        {
                tr_flush_dirty_ST();
                if (known_regb & KRREG_PMC) {
-                       emit_mov_const(A_COND_AL, 1, known_regs.pmc.v);
+                       emith_move_r_imm(1, known_regs.pmc.v);
                        EOP_STR_IMM(1,7,0x400+SSP_PMC*4);
                        known_regb &= ~KRREG_PMC;
                        dirty_regb &= ~KRREG_PMC;
@@ -1660,7 +1667,7 @@ static void emit_block_prologue(void)
        // check if there are enough cycles..
        // note: r0 must contain PC of current block
        EOP_CMP_IMM(11,0,0);                    // cmp r11, #0
-       emit_jump(A_COND_LE, ssp_drc_end);
+       emith_jump_cond(A_COND_LE, ssp_drc_end);
 }
 
 /* cond:
@@ -1674,35 +1681,41 @@ static void emit_block_epilogue(int cycles, int cond, int pc, int end_pc)
 
        if (cond < 0 || (end_pc >= 0x400 && pc < 0x400)) {
                // indirect jump, or rom -> iram jump, must use dispatcher
-               emit_jump(A_COND_AL, ssp_drc_next);
+               emith_jump(ssp_drc_next);
        }
        else if (cond == A_COND_AL) {
-               u32 *target = (pc < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][pc] : ssp_block_table[pc];
+               u32 *target = (pc < 0x400) ?
+                       ssp_block_table_iram[ssp->drc.iram_context * SSP_BLOCKTAB_IRAM_ONE + pc] :
+                       ssp_block_table[pc];
                if (target != NULL)
-                       emit_jump(A_COND_AL, target);
+                       emith_jump(target);
                else {
-                       int ops = emit_jump(A_COND_AL, ssp_drc_next);
+                       int ops = emith_jump(ssp_drc_next);
                        // cause the next block to be emitted over jump instruction
                        tcache_ptr -= ops;
                }
        }
        else {
-               u32 *target1 = (pc     < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][pc] : ssp_block_table[pc];
-               u32 *target2 = (end_pc < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][end_pc] : ssp_block_table[end_pc];
+               u32 *target1 = (pc     < 0x400) ?
+                       ssp_block_table_iram[ssp->drc.iram_context * SSP_BLOCKTAB_IRAM_ONE + pc] :
+                       ssp_block_table[pc];
+               u32 *target2 = (end_pc < 0x400) ?
+                       ssp_block_table_iram[ssp->drc.iram_context * SSP_BLOCKTAB_IRAM_ONE + end_pc] :
+                       ssp_block_table[end_pc];
                if (target1 != NULL)
-                    emit_jump(cond, target1);
+                    emith_jump_cond(cond, target1);
                if (target2 != NULL)
-                    emit_jump(tr_neg_cond(cond), target2); // neg_cond, to be able to swap jumps if needed
+                    emith_jump_cond(tr_neg_cond(cond), target2); // neg_cond, to be able to swap jumps if needed
 #ifndef __EPOC32__
                // emit patchable branches
                if (target1 == NULL)
-                       emit_call(cond, ssp_drc_next_patch);
+                       emith_call_cond(cond, ssp_drc_next_patch);
                if (target2 == NULL)
-                       emit_call(tr_neg_cond(cond), ssp_drc_next_patch);
+                       emith_call_cond(tr_neg_cond(cond), ssp_drc_next_patch);
 #else
                // won't patch indirect jumps
                if (target1 == NULL || target2 == NULL)
-                       emit_jump(A_COND_AL, ssp_drc_next);
+                       emith_jump(ssp_drc_next);
 #endif
        }
 }
@@ -1746,7 +1759,7 @@ void *ssp_translate_block(int pc)
        if (ccount >= 100) {
                end_cond = A_COND_AL;
                jump_pc = pc;
-               emit_mov_const(A_COND_AL, 0, pc);
+               emith_move_r_imm(0, pc);
        }
 
        tr_flush_dirty_prs();
@@ -1754,7 +1767,7 @@ void *ssp_translate_block(int pc)
        tr_flush_dirty_pmcrs();
        emit_block_epilogue(ccount, end_cond, jump_pc, pc);
 
-       if (tcache_ptr - tcache > SSP_TCACHE_SIZE/4) {
+       if (tcache_ptr - (u32 *)tcache > DRC_TCACHE_SIZE/4) {
                elprintf(EL_ANOMALY|EL_STATUS|EL_SVP, "tcache overflow!\n");
                fflush(stdout);
                exit(1);
@@ -1775,7 +1788,9 @@ void *ssp_translate_block(int pc)
        exit(0);
 #endif
 
-       handle_caches();
+#ifdef ARM
+       cache_flush_d_inval_i(tcache, tcache_ptr);
+#endif
 
        return block_start;
 }
@@ -1790,14 +1805,30 @@ static void ssp1601_state_load(void)
        ssp->drc.iram_context = 0;
 }
 
+void ssp1601_dyn_exit(void)
+{
+       free(ssp_block_table);
+       free(ssp_block_table_iram);
+       ssp_block_table = ssp_block_table_iram = NULL;
+
+       drc_cmn_cleanup();
+}
+
 int ssp1601_dyn_startup(void)
 {
        drc_cmn_init();
 
-       memset(tcache, 0, SSP_TCACHE_SIZE);
-       memset(ssp_block_table, 0, sizeof(ssp_block_table));
-       memset(ssp_block_table_iram, 0, sizeof(ssp_block_table_iram));
-       tcache_ptr = tcache;
+       ssp_block_table = calloc(sizeof(ssp_block_table[0]), SSP_BLOCKTAB_ENTS);
+       if (ssp_block_table == NULL)
+               return -1;
+       ssp_block_table_iram = calloc(sizeof(ssp_block_table_iram[0]), SSP_BLOCKTAB_IRAM_ENTS);
+       if (ssp_block_table_iram == NULL) {
+               free(ssp_block_table);
+               return -1;
+       }
+
+       memset(tcache, 0, DRC_TCACHE_SIZE);
+       tcache_ptr = (void *)tcache;
 
        PicoLoadStateHook = ssp1601_state_load;
 
@@ -1806,12 +1837,12 @@ int ssp1601_dyn_startup(void)
        // hle'd blocks
        ssp_block_table[0x800/2] = (void *) ssp_hle_800;
        ssp_block_table[0x902/2] = (void *) ssp_hle_902;
-       ssp_block_table_iram[ 7][0x030/2] = (void *) ssp_hle_07_030;
-       ssp_block_table_iram[ 7][0x036/2] = (void *) ssp_hle_07_036;
-       ssp_block_table_iram[ 7][0x6d6/2] = (void *) ssp_hle_07_6d6;
-       ssp_block_table_iram[11][0x12c/2] = (void *) ssp_hle_11_12c;
-       ssp_block_table_iram[11][0x384/2] = (void *) ssp_hle_11_384;
-       ssp_block_table_iram[11][0x38a/2] = (void *) ssp_hle_11_38a;
+       ssp_block_table_iram[ 7 * SSP_BLOCKTAB_IRAM_ONE + 0x030/2] = (void *) ssp_hle_07_030;
+       ssp_block_table_iram[ 7 * SSP_BLOCKTAB_IRAM_ONE + 0x036/2] = (void *) ssp_hle_07_036;
+       ssp_block_table_iram[ 7 * SSP_BLOCKTAB_IRAM_ONE + 0x6d6/2] = (void *) ssp_hle_07_6d6;
+       ssp_block_table_iram[11 * SSP_BLOCKTAB_IRAM_ONE + 0x12c/2] = (void *) ssp_hle_11_12c;
+       ssp_block_table_iram[11 * SSP_BLOCKTAB_IRAM_ONE + 0x384/2] = (void *) ssp_hle_11_384;
+       ssp_block_table_iram[11 * SSP_BLOCKTAB_IRAM_ONE + 0x38a/2] = (void *) ssp_hle_11_38a;
 #endif
 
        return 0;