drc: debug improvements

diff --git a/cpu/drc/emit_arm.c b/cpu/drc/emit_arm.c
index efee0f3..2ac8e32 100644 (file)
--- a/cpu/drc/emit_arm.c
+++ b/cpu/drc/emit_arm.c
@@ -639,6 +639,21 @@ static int emith_xbranch(int cond, void *target, int is_call)
        EOP_MOV_REG_ASR(d,d,32 - (bits)); \
 }
 
+#define emith_do_caller_regs(mask, func) { \
+       u32 _reg_mask = (mask) & 0x500f; \
+       if (_reg_mask) { \
+               if (__builtin_parity(_reg_mask) == 1) \
+                       _reg_mask |= 0x10; /* eabi align */ \
+               func(_reg_mask); \
+       } \
+}
+
+#define emith_save_caller_regs(mask) \
+       emith_do_caller_regs(mask, EOP_STMFD_SP)
+
+#define emith_restore_caller_regs(mask) \
+       emith_do_caller_regs(mask, EOP_LDMFD_SP)
+
 // upto 4 args
 #define emith_pass_arg_r(arg, reg) \
        EOP_MOV_REG_SIMPLE(arg, reg)

diff --git a/cpu/sh2/compiler.c b/cpu/sh2/compiler.c
index bdf2cea..46acb52 100644 (file)
--- a/cpu/sh2/compiler.c
+++ b/cpu/sh2/compiler.c
@@ -572,7 +572,7 @@ static void *dr_prepare_ext_branch(u32 pc, int is_slave, int tcache_id)
       break;
   cnt = i + 1;
   if (cnt >= block_link_pool_max_counts[tcache_id]) {
-    dbg(1, "bl overflow for tcache %d\n", tcache_id);
+    dbg(1, "bl overflow for tcache %d", tcache_id);
     return NULL;
   }
   bl += cnt;
@@ -994,6 +994,18 @@ static void rcache_unlock_all(void)
     reg_temp[i].flags &= ~HRF_LOCKED;
 }
 
+static inline u32 rcache_used_hreg_mask(void)
+{
+  u32 mask = 0;
+  int i;
+
+  for (i = 0; i < ARRAY_SIZE(reg_temp); i++)
+    if (reg_temp[i].type != HR_FREE)
+      mask |= 1 << reg_temp[i].hreg;
+
+  return mask;
+}
+
 static void rcache_clean(void)
 {
   int i;
@@ -1477,16 +1489,18 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
       emit_move_r_imm32(SHR_PC, pc);
       sr = rcache_get_reg(SHR_SR, RC_GR_RMW);
       FLUSH_CYCLES(sr);
-      // rcache_clean(); // FIXME
-      rcache_flush();
+      rcache_clean();
+
+      tmp = rcache_used_hreg_mask();
+      emith_save_caller_regs(tmp);
       emit_do_static_regs(1, 0);
       emith_pass_arg_r(0, CONTEXT_REG);
       emith_call(do_sh2_cmp);
+      emith_restore_caller_regs(tmp);
     }
 #endif
 
     pc += 2;
-    cycles += opd->cycles;
 
     if (skip_op > 0) {
       skip_op--;
@@ -2517,6 +2531,8 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id)
 end_op:
     rcache_unlock_all();
 
+    cycles += opd->cycles;
+
     if (op_flags[i+1] & OF_DELAY_OP) {
       do_host_disasm(tcache_id);
       continue;
@@ -3083,9 +3099,6 @@ int sh2_drc_init(SH2 *sh2)
     for (i = 1; i < ARRAY_SIZE(tcache_bases); i++)
       tcache_bases[i] = tcache_ptrs[i] = tcache_bases[i - 1] + tcache_sizes[i - 1];
 
-    // tmp
-    PicoOpt |= POPT_DIS_VDP_FIFO;
-
 #if (DRC_DEBUG & 4)
     for (i = 0; i < ARRAY_SIZE(block_tables); i++)
       tcache_dsm_ptrs[i] = tcache_bases[i];

diff --git a/cpu/sh2/mame/sh2pico.c b/cpu/sh2/mame/sh2pico.c
index 1a70d43..10d16ba 100644 (file)
--- a/cpu/sh2/mame/sh2pico.c
+++ b/cpu/sh2/mame/sh2pico.c
@@ -72,8 +72,13 @@ static unsigned int op_refs[0x10000];
 int sh2_execute(SH2 *sh2, int cycles)
 {
 #ifdef DRC_CMP
-       unsigned int base_pc = 0, end_pc = 0;
-       unsigned char op_flags[BLOCK_INSN_LIMIT];
+       static unsigned int base_pc_[2] = { 0, 0 };
+       static unsigned int end_pc_[2] = { 0, 0 };
+       static unsigned char op_flags_[2][BLOCK_INSN_LIMIT];
+       unsigned int *base_pc = &base_pc_[sh2->is_slave];
+       unsigned int *end_pc = &end_pc_[sh2->is_slave];
+       unsigned char *op_flags = op_flags_[sh2->is_slave];
+       unsigned int pc_expect = sh2->pc;
 #endif
        UINT32 opcode;
 
@@ -88,20 +93,23 @@ int sh2_execute(SH2 *sh2, int cycles)
        {
 #ifdef DRC_CMP
                if (!sh2->delay) {
-                       if (sh2->pc < base_pc || sh2->pc >= end_pc) {
-                               base_pc = sh2->pc;
-                               scan_block(base_pc, sh2->is_slave,
-                                       op_flags, &end_pc, NULL);
+                       if (sh2->pc < *base_pc || sh2->pc >= *end_pc) {
+                               *base_pc = sh2->pc;
+                               scan_block(*base_pc, sh2->is_slave,
+                                       op_flags, end_pc, NULL);
                        }
-                       if ((op_flags[(sh2->pc - base_pc) / 2]
-                               & OF_BTARGET) || sh2->pc == base_pc)
+                       if ((op_flags[(sh2->pc - *base_pc) / 2]
+                               & OF_BTARGET) || sh2->pc == *base_pc
+                               || pc_expect != sh2->pc) // branched
                        {
+                               pc_expect = sh2->pc;
                                if (sh2->icount < 0)
                                        break;
                        }
 
                        do_sh2_trace(sh2, sh2->icount);
                }
+               pc_expect += 2;
 #endif
 
                if (sh2->delay)

diff --git a/cpu/sh2/sh2.c b/cpu/sh2/sh2.c
index 1f034cd..fbf0c6d 100644 (file)
--- a/cpu/sh2/sh2.c
+++ b/cpu/sh2/sh2.c
@@ -129,25 +129,50 @@ void sh2_unpack(SH2 *sh2, const unsigned char *buff)
 #include <stdio.h>
 #include <stdlib.h>
 #include <pico/memory.h>
+#undef _USE_CZ80 // HACK
+#include <pico/pico_int.h>
 #include <pico/debug.h>
 
 static SH2 sh2ref[2];
-static int current_slave = -1;
 static unsigned int mem_val;
 static FILE *f;
 
+enum ctl_byte {
+       CTL_MASTERSLAVE = 0x80,
+       CTL_EA = 0x82,
+       CTL_EAVAL = 0x83,
+       CTL_M68KPC = 0x84,
+       CTL_CYCLES = 0x85,
+};
+
 #define SH2MAP_ADDR2OFFS_R(a) \
   ((((a) >> 25) & 3) | (((a) >> 27) & 0x1c))
 
 static unsigned int local_read32(SH2 *sh2, u32 a)
 {
        const sh2_memmap *sh2_map = sh2->read16_map;
+       u16 *pd;
        uptr p;
 
        sh2_map += SH2MAP_ADDR2OFFS_R(a);
        p = sh2_map->addr;
        if (!map_flag_set(p)) {
-               u16 *pd = (u16 *)((p << 1) + ((a & sh2_map->mask) & ~3));
+               pd = (u16 *)((p << 1) + ((a & sh2_map->mask) & ~1));
+               return (pd[0] << 16) | pd[1];
+       }
+
+       if ((a & 0xfffff000) == 0xc0000000) {
+               // data array
+               pd = (u16 *)Pico32xMem->data_array[sh2->is_slave]
+                       + (a & 0xfff) / 2;
+               return (pd[0] << 16) | pd[1];
+       }
+       if ((a & 0xdfffffc0) == 0x4000) {
+               pd = &Pico32x.regs[(a & 0x3f) / 2];
+               return (pd[0] << 16) | pd[1];
+       }
+       if ((a & 0xdffffe00) == 0x4200) {
+               pd = &Pico32xMem->pal[(a & 0x1ff) / 2];
                return (pd[0] << 16) | pd[1];
        }
 
@@ -162,6 +187,8 @@ static void write_uint(unsigned char ctl, unsigned int v)
 
 void do_sh2_trace(SH2 *current, int cycles)
 {
+       static int current_slave = -1;
+       static u32 current_m68k_pc;
        SH2 *sh2o = &sh2ref[current->is_slave];
        u32 *regs_a = (void *)current;
        u32 *regs_o = (void *)sh2o;
@@ -172,9 +199,14 @@ void do_sh2_trace(SH2 *current, int cycles)
        if (f == NULL)
                f = fopen("tracelog", "wb");
 
+       if (SekPc != current_m68k_pc) {
+               current_m68k_pc = SekPc;
+               write_uint(CTL_M68KPC, current_m68k_pc);
+       }
+
        if (current->is_slave != current_slave) {
                current_slave = current->is_slave;
-               v = 0x80 | current->is_slave;
+               v = CTL_MASTERSLAVE | current->is_slave;
                fwrite(&v, 1, 1, f);
        }
 
@@ -188,15 +220,15 @@ void do_sh2_trace(SH2 *current, int cycles)
        }
 
        if (current->ea != sh2o->ea) {
-               write_uint(0x82, current->ea);
+               write_uint(CTL_EA, current->ea);
                sh2o->ea = current->ea;
        }
        val = local_read32(current, current->ea);
        if (mem_val != val) {
-               write_uint(0x83, val);
+               write_uint(CTL_EAVAL, val);
                mem_val = val;
        }
-       write_uint(0x84, cycles);
+       write_uint(CTL_CYCLES, cycles);
 }
 
 static const char *regnames[] = {
@@ -208,6 +240,19 @@ static const char *regnames[] = {
        "gbr", "vbr", "mach","macl",
 };
 
+static void dump_regs(SH2 *sh2)
+{
+       char csh2;
+       int i;
+
+       csh2 = sh2->is_slave ? 's' : 'm';
+       for (i = 0; i < 16/2; i++)
+               printf("%csh2 r%d: %08x r%02d: %08x\n", csh2,
+                       i, sh2->r[i], i+8, sh2->r[i+8]);
+       printf("%csh2 PC: %08x  ,   %08x\n", csh2, sh2->pc, sh2->ppc);
+       printf("%csh2 SR:      %03x  PR: %08x\n", csh2, sh2->sr, sh2->pr);
+}
+
 void do_sh2_cmp(SH2 *current)
 {
        static int current_slave;
@@ -221,31 +266,39 @@ void do_sh2_cmp(SH2 *current)
        int bad = 0;
        int cycles;
        int i, ret;
-       char csh2;
 
-       if (f == NULL)
+       if (f == NULL) {
                f = fopen("tracelog", "rb");
+               sh2ref[1].is_slave = 1;
+       }
 
        while (1) {
                ret = fread(&code, 1, 1, f);
                if (ret <= 0)
                        break;
-               if (code == 0x84) {
+               if (code == CTL_CYCLES) {
                        fread(&cycles_o, 1, 4, f);
                        break;
                }
 
                switch (code) {
-               case 0x80:
-               case 0x81:
+               case CTL_MASTERSLAVE:
+               case CTL_MASTERSLAVE + 1:
                        current_slave = code & 1;
                        break;
-               case 0x82:
+               case CTL_EA:
                        fread(&sh2o->ea, 4, 1, f);
                        break;
-               case 0x83:
+               case CTL_EAVAL:
                        fread(&current_val, 4, 1, f);
                        break;
+               case CTL_M68KPC:
+                       fread(&val, 4, 1, f);
+                       if (SekPc != val) {
+                               printf("m68k: %08x %08x\n", SekPc, val);
+                               bad = 1;
+                       }
+                       break;
                default:
                        if (code < offsetof(SH2, read8_map) / 4)
                                fread(regs_o + code, 4, 1, f);
@@ -304,12 +357,9 @@ void do_sh2_cmp(SH2 *current)
 
 end:
        printf("--\n");
-       csh2 = current->is_slave ? 's' : 'm';
-       for (i = 0; i < 16/2; i++)
-               printf("%csh2 r%d: %08x r%02d: %08x\n", csh2,
-                       i, sh2o->r[i], i+8, sh2o->r[i+8]);
-       printf("%csh2 PC: %08x  ,   %08x\n", csh2, sh2o->pc, sh2o->ppc);
-       printf("%csh2 SR:      %03x  PR: %08x\n", csh2, sh2o->sr, sh2o->pr);
+       dump_regs(sh2o);
+       if (current->is_slave != current_slave)
+               dump_regs(&sh2ref[current->is_slave ^ 1]);
        PDebugDumpMem();
        exit(1);
 }

diff --git a/pico/pico.c b/pico/pico.c
index 20648f0..9a1faca 100644 (file)
--- a/pico/pico.c
+++ b/pico/pico.c
@@ -145,6 +145,10 @@ int PicoReset(void)
   if (Pico.romsize <= 0)\r
     return 1;\r
 \r
+#ifdef DRC_CMP\r
+  PicoOpt |= POPT_DIS_VDP_FIFO|POPT_DIS_IDLE_DET;\r
+#endif\r
+\r
   /* must call now, so that banking is reset, and correct vectors get fetched */\r
   if (PicoResetHook)\r
     PicoResetHook();\r

diff --git a/pico/pico_int.h b/pico/pico_int.h
index 420e82c..a654c5a 100644 (file)
--- a/pico/pico_int.h
+++ b/pico/pico_int.h
@@ -244,6 +244,7 @@ extern SH2 sh2s[2];
   } \\r
 } while (0)\r
 # define ash2_cycles_done(sh2) ((sh2)->cycles_timeslice - (sh2)->icount)\r
+# define sh2_pc(c)    (c) ? ssh2.ppc : msh2.ppc\r
 #else\r
 # define ash2_end_run(sh2, after) do { \\r
   int left = (sh2)->sr >> 12; \\r
@@ -254,10 +255,9 @@ extern SH2 sh2s[2];
   } \\r
 } while (0)\r
 # define ash2_cycles_done(sh2) ((sh2)->cycles_timeslice - ((sh2)->sr >> 12))\r
+# define sh2_pc(c)    (c) ? ssh2.pc : msh2.pc\r
 #endif\r
 \r
-//#define sh2_pc(c)     (c) ? ssh2.ppc : msh2.ppc\r
-#define sh2_pc(c)     (c) ? ssh2.pc : msh2.pc\r
 #define sh2_reg(c, x) (c) ? ssh2.r[x] : msh2.r[x]\r
 #define sh2_gbr(c)    (c) ? ssh2.gbr : msh2.gbr\r
 #define sh2_vbr(c)    (c) ? ssh2.vbr : msh2.vbr\r