From: notaz <notasas@gmail.com>
Date: Fri, 26 Jul 2013 22:23:56 +0000 (+0300)
Subject: rework sh2 sync, again..
X-Git-Tag: v1.85~36
X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=picodrive.git;a=commitdiff_plain;h=19886062f1a36f70b1f01d58f3fa1b79162defac

rework sh2 sync, again..

also some new debug and poll code
VF seems to be ok at least..
---

diff --git a/Makefile b/Makefile
index 4df1162..72896e0 100644
--- a/Makefile
+++ b/Makefile
@@ -6,6 +6,7 @@ CFLAGS += -Iplatform/linux/
 ifndef DEBUG
 CFLAGS += -O2 -DNDEBUG
 endif
+#CFLAGS += -DEVT_LOG
 #CFLAGS += -DDRC_CMP
 #drc_debug = 4
 #profile = 1
diff --git a/cpu/sh2/sh2.h b/cpu/sh2/sh2.h
index 92774d0..42de630 100644
--- a/cpu/sh2/sh2.h
+++ b/cpu/sh2/sh2.h
@@ -39,6 +39,15 @@ typedef struct SH2_
 	void		*p_rom;
 	unsigned int	pdb_io_csum[2];
 
+#define SH2_STATE_RUN   (1 << 0)	// to prevent recursion
+#define SH2_STATE_SLEEP (1 << 1)
+#define SH2_STATE_CPOLL (1 << 2)	// polling comm regs
+#define SH2_STATE_VPOLL (1 << 3)	// polling VDP
+	unsigned int	state;
+	unsigned int	poll_addr;
+	int		poll_cycles;
+	int		poll_cnt;
+
 	// interpreter stuff
 	int		icount;		// cycles left in current timeslice
 	unsigned int	ea;
diff --git a/pico/32x/32x.c b/pico/32x/32x.c
index 52b73cc..339a852 100644
--- a/pico/32x/32x.c
+++ b/pico/32x/32x.c
@@ -12,6 +12,8 @@
 struct Pico32x Pico32x;
 SH2 sh2s[2];
 
+#define SH2_IDLE_STATES (SH2_STATE_CPOLL|SH2_STATE_VPOLL|SH2_STATE_SLEEP)
+
 static int REGPARM(2) sh2_irq_cb(SH2 *sh2, int level)
 {
   if (sh2->pending_irl > sh2->pending_int_irq) {
@@ -28,11 +30,15 @@ static int REGPARM(2) sh2_irq_cb(SH2 *sh2, int level)
 }
 
 // if !nested_call, must sync CPUs before calling this
-void p32x_update_irls(int nested_call)
+void p32x_update_irls(SH2 *active_sh2)
 {
   int irqs, mlvl = 0, slvl = 0;
+  int m68k_cycles = 0;
   int mrun, srun;
 
+  if (active_sh2 != NULL)
+    m68k_cycles = sh2_cycles_done_m68k(active_sh2);
+
   // msh2
   irqs = (Pico32x.sh2irqs | Pico32x.sh2irqi[0]) & ((Pico32x.sh2irq_mask[0] << 3) | P32XI_VRES);
   while ((irqs >>= 1))
@@ -45,9 +51,14 @@ void p32x_update_irls(int nested_call)
     slvl++;
   slvl *= 2;
 
-  mrun = sh2_irl_irq(&msh2, mlvl, nested_call);
-  srun = sh2_irl_irq(&ssh2, slvl, nested_call);
-  p32x_poll_event(mrun | (srun << 1), 0);
+  mrun = sh2_irl_irq(&msh2, mlvl, active_sh2 != NULL);
+  if (mrun)
+    p32x_sh2_poll_event(&msh2, SH2_IDLE_STATES, m68k_cycles);
+
+  srun = sh2_irl_irq(&ssh2, slvl, active_sh2 != NULL);
+  if (srun)
+    p32x_sh2_poll_event(&ssh2, SH2_IDLE_STATES, m68k_cycles);
+
   elprintf(EL_32X, "update_irls: m %d/%d, s %d/%d", mlvl, mrun, slvl, srun);
 }
 
@@ -158,8 +169,9 @@ void PicoReset32x(void)
 {
   if (PicoAHW & PAHW_32X) {
     Pico32x.sh2irqs |= P32XI_VRES;
-    p32x_update_irls(0);
-    p32x_poll_event(3, 0);
+    p32x_update_irls(NULL);
+    p32x_sh2_poll_event(&msh2, SH2_IDLE_STATES, 0);
+    p32x_sh2_poll_event(&ssh2, SH2_IDLE_STATES, 0);
     p32x_timers_recalc();
   }
 }
@@ -204,10 +216,19 @@ static void p32x_start_blank(void)
   }
 
   Pico32x.sh2irqs |= P32XI_VINT;
-  p32x_update_irls(0);
-  p32x_poll_event(3, 1);
+  p32x_update_irls(NULL);
+  p32x_sh2_poll_event(&msh2, SH2_STATE_VPOLL, 0);
+  p32x_sh2_poll_event(&ssh2, SH2_STATE_VPOLL, 0);
 }
 
+// compare cycles, handling overflows
+// check if a > b
+#define CYCLES_GT(a, b) \
+  ((int)((a) - (b)) > 0)
+// check if a >= b
+#define CYCLES_GE(a, b) \
+  ((int)((a) - (b)) >= 0)
+
 /* events */
 static void pwm_irq_event(unsigned int now)
 {
@@ -215,13 +236,14 @@ static void pwm_irq_event(unsigned int now)
   p32x_pwm_schedule(now);
 
   Pico32x.sh2irqs |= P32XI_PWM;
-  p32x_update_irls(0);
+  p32x_update_irls(NULL);
 }
 
 static void fillend_event(unsigned int now)
 {
   Pico32x.vdp_regs[0x0a/2] &= ~P32XV_nFEN;
-  p32x_poll_event(3, 1);
+  p32x_sh2_poll_event(&msh2, SH2_STATE_VPOLL, now);
+  p32x_sh2_poll_event(&ssh2, SH2_STATE_VPOLL, now);
 }
 
 typedef void (event_cb)(unsigned int now);
@@ -233,18 +255,32 @@ static event_cb *event_cbs[] = {
   [P32X_EVENT_FILLEND]  = fillend_event,
 };
 
-// schedule event at some time (in m68k clocks)
-void p32x_event_schedule(enum p32x_event event, unsigned int now, int after)
+// schedule event at some time 'after', in m68k clocks
+void p32x_event_schedule(unsigned int now, enum p32x_event event, int after)
 {
-  unsigned int when = (now + after) | 1;
+  unsigned int when;
+
+  when = (now + after) | 1;
 
   elprintf(EL_32X, "new event #%u %u->%u", event, now, when);
   event_times[event] = when;
 
-  if (event_time_next == 0 || (int)(event_time_next - now) > after)
+  if (event_time_next == 0 || CYCLES_GT(event_time_next, when))
     event_time_next = when;
 }
 
+void p32x_event_schedule_sh2(SH2 *sh2, enum p32x_event event, int after)
+{
+  unsigned int now = sh2_cycles_done_m68k(sh2);
+  int left_to_next;
+
+  p32x_event_schedule(now, event, after);
+
+  left_to_next = (event_time_next - now) * 3;
+  if (sh2_cycles_left(sh2) > left_to_next)
+    sh2_end_run(sh2, left_to_next);
+}
+
 static void run_events(unsigned int until)
 {
   int oldest, oldest_diff, time;
@@ -283,13 +319,61 @@ static void run_events(unsigned int until)
     elprintf(EL_32X, "next event #%d at %u", oldest, event_time_next);
 }
 
-// compare cycles, handling overflows
-// check if a > b
-#define CYCLES_GT(a, b) \
-  ((int)((a) - (b)) > 0)
-// check if a >= b
-#define CYCLES_GE(a, b) \
-  ((int)((a) - (b)) >= 0)
+static inline void run_sh2(SH2 *sh2, int m68k_cycles)
+{
+  int cycles, done;
+
+  pevt_log_sh2_o(sh2, EVT_RUN_START);
+  sh2->state |= SH2_STATE_RUN;
+  cycles = C_M68K_TO_SH2(*sh2, m68k_cycles);
+  elprintf(EL_32X, "%csh2 +run %u %d",
+    sh2->is_slave?'s':'m', sh2->m68krcycles_done, cycles);
+
+  done = sh2_execute(sh2, cycles);
+
+  sh2->m68krcycles_done += C_SH2_TO_M68K(*sh2, done);
+  sh2->state &= ~SH2_STATE_RUN;
+  pevt_log_sh2_o(sh2, EVT_RUN_END);
+  elprintf(EL_32X, "%csh2 -run %u %d",
+    sh2->is_slave?'s':'m', sh2->m68krcycles_done, done);
+}
+
+// sync other sh2 to this one
+// note: recursive call
+void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target)
+{
+  SH2 *osh2 = &sh2s[sh2->is_slave ^ 1];
+  int left_to_event;
+  int m68k_cycles;
+
+  if (osh2->state & SH2_STATE_RUN)
+    return;
+
+  m68k_cycles = m68k_target - osh2->m68krcycles_done;
+  if (m68k_cycles < 200)
+    return;
+
+  if (osh2->state & SH2_IDLE_STATES) {
+    osh2->m68krcycles_done = m68k_target;
+    return;
+  }
+
+  elprintf(EL_32X, "%csh2 sync to %u %d",
+    osh2->is_slave?'s':'m', m68k_target, m68k_cycles);
+
+  run_sh2(osh2, m68k_cycles);
+
+  // there might be new event to schedule current sh2 to
+  if (event_time_next) {
+    left_to_event = event_time_next - m68k_target;
+    left_to_event *= 3;
+    if (sh2_cycles_left(sh2) > left_to_event) {
+      if (left_to_event < 1)
+        left_to_event = 1;
+      sh2_end_run(sh2, left_to_event);
+    }
+  }
+}
 
 #define sync_sh2s_normal p32x_sync_sh2s
 //#define sync_sh2s_lockstep p32x_sync_sh2s
@@ -298,7 +382,7 @@ static void run_events(unsigned int until)
 void sync_sh2s_normal(unsigned int m68k_target)
 {
   unsigned int now, target, timer_cycles;
-  int cycles, done;
+  int cycles;
 
   elprintf(EL_32X, "sh2 sync to %u", m68k_target);
 
@@ -327,42 +411,50 @@ void sync_sh2s_normal(unsigned int m68k_target)
         target - msh2.m68krcycles_done, target - ssh2.m68krcycles_done,
         m68k_target - now, Pico32x.emu_flags);
 
-      if (Pico32x.emu_flags & (P32XF_SSH2POLL|P32XF_SSH2VPOLL)) {
-        ssh2.m68krcycles_done = target;
-      }
-      else {
+      if (!(ssh2.state & SH2_IDLE_STATES)) {
         cycles = target - ssh2.m68krcycles_done;
         if (cycles > 0) {
-          done = sh2_execute(&ssh2, C_M68K_TO_SH2(ssh2, cycles));
-          ssh2.m68krcycles_done += C_SH2_TO_M68K(ssh2, done);
+          run_sh2(&ssh2, cycles);
 
           if (event_time_next && CYCLES_GT(target, event_time_next))
             target = event_time_next;
         }
       }
 
-      if (Pico32x.emu_flags & (P32XF_MSH2POLL|P32XF_MSH2VPOLL)) {
-        msh2.m68krcycles_done = target;
-      }
-      else {
+      if (!(msh2.state & SH2_IDLE_STATES)) {
         cycles = target - msh2.m68krcycles_done;
         if (cycles > 0) {
-          done = sh2_execute(&msh2, C_M68K_TO_SH2(msh2, cycles));
-          msh2.m68krcycles_done += C_SH2_TO_M68K(msh2, done);
+          run_sh2(&msh2, cycles);
 
           if (event_time_next && CYCLES_GT(target, event_time_next))
             target = event_time_next;
         }
       }
 
-      now = msh2.m68krcycles_done;
-      if (CYCLES_GT(now, ssh2.m68krcycles_done))
-        now = ssh2.m68krcycles_done;
+      now = target;
+      if (!(msh2.state & SH2_IDLE_STATES)) {
+        if (CYCLES_GT(now, msh2.m68krcycles_done))
+          now = msh2.m68krcycles_done;
+      }
+      if (!(ssh2.state & SH2_IDLE_STATES)) {
+        if (CYCLES_GT(now, ssh2.m68krcycles_done))
+          now = ssh2.m68krcycles_done;
+      }
     }
 
     p32x_timers_do(now - timer_cycles);
     timer_cycles = now;
   }
+
+  // advance idle CPUs
+  if (msh2.state & SH2_IDLE_STATES) {
+    if (CYCLES_GT(m68k_target, msh2.m68krcycles_done))
+      msh2.m68krcycles_done = m68k_target;
+  }
+  if (ssh2.state & SH2_IDLE_STATES) {
+    if (CYCLES_GT(m68k_target, ssh2.m68krcycles_done))
+      ssh2.m68krcycles_done = m68k_target;
+  }
 }
 
 #define STEP_68K 24
@@ -383,8 +475,8 @@ void sync_sh2s_lockstep(unsigned int m68k_target)
 
 #define CPUS_RUN(m68k_cycles,s68k_cycles) do { \
   SekRunM68k(m68k_cycles); \
-  if (Pico32x.emu_flags & P32XF_68KPOLL) \
-    p32x_sync_sh2s(SekCycleCntT + SekCycleCnt); \
+  if (Pico32x.emu_flags & (P32XF_68KCPOLL|P32XF_68KVPOLL)) \
+    p32x_sync_sh2s(SekCyclesDoneT2()); \
 } while (0)
 
 #define PICO_32X
@@ -396,13 +488,15 @@ void PicoFrame32x(void)
   if ((Pico32x.vdp_regs[0] & P32XV_Mx) != 0) // no forced blanking
     Pico32x.vdp_regs[0x0a/2] &= ~P32XV_PEN; // no palette access
 
-  p32x_poll_event(3, 1);
+  p32x_sh2_poll_event(&msh2, SH2_STATE_VPOLL, 0);
+  p32x_sh2_poll_event(&ssh2, SH2_STATE_VPOLL, 0);
 
   PicoFrameStart();
   PicoFrameHints();
   sh2_drc_frame();
 
-  elprintf(EL_32X, "poll: %02x", Pico32x.emu_flags);
+  elprintf(EL_32X, "poll: %02x %02x %02x",
+    Pico32x.emu_flags & 3, msh2.state, ssh2.state);
 }
 
 // calculate multipliers against 68k clock (7670442)
@@ -428,9 +522,9 @@ void Pico32xStateLoaded(int is_early)
     return;
   }
 
+  SekCycleCnt = 0;
   sh2s[0].m68krcycles_done = sh2s[1].m68krcycles_done = SekCycleCntT;
-  p32x_update_irls(0);
-  p32x_poll_event(3, 0);
+  p32x_update_irls(NULL);
   p32x_timers_recalc();
   run_events(SekCycleCntT);
 }
diff --git a/pico/32x/memory.c b/pico/32x/memory.c
index 2ec7563..db5d045 100644
--- a/pico/32x/memory.c
+++ b/pico/32x/memory.c
@@ -42,13 +42,6 @@
 #include "../memory.h"
 #include "../../cpu/sh2/compiler.h"
 
-#if 0
-#undef ash2_end_run
-#undef SekEndRun
-#define ash2_end_run(x)
-#define SekEndRun(x)
-#endif
-
 static const char str_mars[] = "MARS";
 
 void *p32x_bios_g, *p32x_bios_m, *p32x_bios_s;
@@ -57,64 +50,96 @@ struct Pico32xMem *Pico32xMem;
 static void bank_switch(int b);
 
 // poll detection
-#define POLL_THRESHOLD 6
+#define POLL_THRESHOLD 3
 
-struct poll_det {
-  u32 addr, cycles, cyc_max;
-  int cnt, flag;
-};
-static struct poll_det m68k_poll, sh2_poll[2];
+static struct {
+  u32 addr, cycles;
+  int cnt;
+} m68k_poll;
 
-static int p32x_poll_detect(struct poll_det *pd, u32 a, u32 cycles, int is_vdp)
+static int m68k_poll_detect(u32 a, u32 cycles, u32 flags)
 {
-  int ret = 0, flag = pd->flag;
-
-  if (is_vdp)
-    flag <<= 3;
-
-  if (a - 2 <= pd->addr && pd->addr <= a + 2 && cycles - pd->cycles <= pd->cyc_max) {
-    pd->cnt++;
-    if (pd->cnt > POLL_THRESHOLD) {
-      if (!(Pico32x.emu_flags & flag)) {
-        elprintf(EL_32X, "%s poll addr %08x, cyc %u",
-          flag & (P32XF_68KPOLL|P32XF_68KVPOLL) ? "m68k" :
-          (flag & (P32XF_MSH2POLL|P32XF_MSH2VPOLL) ? "msh2" : "ssh2"), a, cycles - pd->cycles);
+  int ret = 0;
+
+  if (a - 2 <= m68k_poll.addr && m68k_poll.addr <= a + 2
+    && cycles - m68k_poll.cycles <= 64)
+  {
+    if (m68k_poll.cnt++ > POLL_THRESHOLD) {
+      if (!(Pico32x.emu_flags & flags)) {
+        elprintf(EL_32X, "m68k poll addr %08x, cyc %u",
+          a, cycles - m68k_poll.cycles);
         ret = 1;
       }
-      Pico32x.emu_flags |= flag;
+      Pico32x.emu_flags |= flags;
     }
   }
   else {
-    pd->cnt = 0;
-    pd->addr = a;
+    m68k_poll.cnt = 0;
+    m68k_poll.addr = a;
   }
-  pd->cycles = cycles;
+  m68k_poll.cycles = cycles;
 
   return ret;
 }
 
-static int p32x_poll_undetect(struct poll_det *pd, int is_vdp)
+void p32x_m68k_poll_event(u32 flags)
+{
+  if (Pico32x.emu_flags & flags) {
+    elprintf(EL_32X, "m68k poll %02x -> %02x", Pico32x.emu_flags,
+      Pico32x.emu_flags & ~flags);
+    Pico32x.emu_flags &= ~flags;
+    SekSetStop(0);
+  }
+  m68k_poll.addr = m68k_poll.cnt = 0;
+}
+
+static void sh2_poll_detect(SH2 *sh2, u32 a, u32 flags)
 {
-  int ret = 0, flag = pd->flag;
-  if (is_vdp)
-    flag <<= 3; // VDP only
+  int cycles_left = sh2_cycles_left(sh2);
+
+  if (a == sh2->poll_addr && sh2->poll_cycles - cycles_left <= 10) {
+    if (sh2->poll_cnt++ > 3) {
+      if (!(sh2->state & flags))
+        elprintf(EL_32X, "%csh2 state: %02x->%02x", sh2->is_slave?'s':'m',
+          sh2->state, sh2->state | flags);
+
+      sh2->state |= flags;
+      sh2_end_run(sh2, 1);
+      pevt_log_sh2(sh2, EVT_POLL_START);
+      return;
+    }
+  }
   else
-    flag |= flag << 3; // both
-  if (Pico32x.emu_flags & flag) {
-    elprintf(EL_32X, "poll %02x -> %02x", Pico32x.emu_flags, Pico32x.emu_flags & ~flag);
-    ret = 1;
+    sh2->poll_cnt = 0;
+  sh2->poll_addr = a;
+  sh2->poll_cycles = cycles_left;
+}
+
+void p32x_sh2_poll_event(SH2 *sh2, u32 flags, u32 m68k_cycles)
+{
+  if (sh2->state & flags) {
+    elprintf(EL_32X, "%csh2 state: %02x->%02x", sh2->is_slave?'s':'m',
+      sh2->state, sh2->state & ~flags);
+
+    if (sh2->m68krcycles_done < m68k_cycles)
+      sh2->m68krcycles_done = m68k_cycles;
+
+    pevt_log_sh2_o(sh2, EVT_POLL_END);
   }
-  Pico32x.emu_flags &= ~flag;
-  pd->addr = pd->cnt = 0;
-  return ret;
+
+  sh2->state &= ~flags;
+  sh2->poll_addr = sh2->poll_cycles = sh2->poll_cnt = 0;
 }
 
-void p32x_poll_event(int cpu_mask, int is_vdp)
+static void sh2s_sync_on_read(SH2 *sh2)
 {
-  if (cpu_mask & 1)
-    p32x_poll_undetect(&sh2_poll[0], is_vdp);
-  if (cpu_mask & 2)
-    p32x_poll_undetect(&sh2_poll[1], is_vdp);
+  int cycles;
+  if (sh2->poll_cnt != 0)
+    return;
+
+  cycles = sh2_cycles_done(sh2);
+  if (cycles > 600)
+    p32x_sync_other_sh2(sh2, sh2->m68krcycles_done + cycles / 3);
 }
 
 // SH2 faking
@@ -165,8 +190,7 @@ static void dma_68k2sh2_do(void)
     elprintf(EL_32X|EL_ANOMALY, "tcr0 and dreq len differ: %d != %d", dmac0->tcr0, *dreqlen);
 
   // HACK: assume bus is busy and SH2 is halted
-  // XXX: use different mechanism for this, not poll det
-  Pico32x.emu_flags |= P32XF_MSH2POLL; // id ? P32XF_SSH2POLL : P32XF_MSH2POLL;
+  msh2.state |= SH2_STATE_SLEEP;
 
   for (i = 0; i < Pico32x.dmac_ptr && dmac0->tcr0 > 0; i++) {
     elprintf(EL_32X, "dmaw [%08x] %04x, left %d", dmac0->dar0, Pico32x.dmac_fifo[i], *dreqlen);
@@ -182,7 +206,7 @@ static void dma_68k2sh2_do(void)
     Pico32x.regs[6 / 2] &= ~P32XS_68S; // transfer complete
   if (dmac0->tcr0 == 0) {
     dmac0->chcr0 |= 2; // DMA has ended normally
-    p32x_poll_undetect(&sh2_poll[0], 0);
+    p32x_sh2_poll_event(&sh2s[0], SH2_STATE_SLEEP, SekCyclesDoneT());
   }
 }
 
@@ -211,7 +235,7 @@ static u32 p32x_reg_read16(u32 a)
       p32x_sync_sh2s(cycles);
     if (Pico32x.comm_dirty_sh2 & comreg)
       Pico32x.comm_dirty_sh2 &= ~comreg;
-    else if (p32x_poll_detect(&m68k_poll, a, cycles, 0)) {
+    else if (m68k_poll_detect(a, cycles, P32XF_68KCPOLL)) {
       SekSetStop(1);
       SekEndTimeslice(16);
     }
@@ -255,12 +279,12 @@ static void p32x_reg_write8(u32 a, u32 d)
       if ((d & 1) && !(Pico32x.sh2irqi[0] & P32XI_CMD)) {
         p32x_sync_sh2s(SekCyclesDoneT());
         Pico32x.sh2irqi[0] |= P32XI_CMD;
-        p32x_update_irls(0);
+        p32x_update_irls(NULL);
       }
       if ((d & 2) && !(Pico32x.sh2irqi[1] & P32XI_CMD)) {
         p32x_sync_sh2s(SekCyclesDoneT());
         Pico32x.sh2irqi[1] |= P32XI_CMD;
-        p32x_update_irls(0);
+        p32x_update_irls(NULL);
       }
       return;
     case 5: // bank
@@ -285,14 +309,14 @@ static void p32x_reg_write8(u32 a, u32 d)
     
     if (r8[a ^ 1] == d)
       return;
-    
+
     comreg = 1 << (a & 0x0f) / 2;
     if (Pico32x.comm_dirty_68k & comreg)
       p32x_sync_sh2s(cycles);
 
     r8[a ^ 1] = d;
-    p32x_poll_undetect(&sh2_poll[0], 0);
-    p32x_poll_undetect(&sh2_poll[1], 0);
+    p32x_sh2_poll_event(&sh2s[0], SH2_STATE_CPOLL, cycles);
+    p32x_sh2_poll_event(&sh2s[1], SH2_STATE_CPOLL, cycles);
     Pico32x.comm_dirty_68k |= comreg;
 
     if (cycles - (int)msh2.m68krcycles_done > 120)
@@ -351,8 +375,8 @@ static void p32x_reg_write16(u32 a, u32 d)
       p32x_sync_sh2s(cycles);
 
     r[a / 2] = d;
-    p32x_poll_undetect(&sh2_poll[0], 0);
-    p32x_poll_undetect(&sh2_poll[1], 0);
+    p32x_sh2_poll_event(&sh2s[0], SH2_STATE_CPOLL, cycles);
+    p32x_sh2_poll_event(&sh2s[1], SH2_STATE_CPOLL, cycles);
     Pico32x.comm_dirty_68k |= comreg;
 
     if (cycles - (int)msh2.m68krcycles_done > 120)
@@ -382,9 +406,6 @@ static void p32x_vdp_write8(u32 a, u32 d)
   u16 *r = Pico32x.vdp_regs;
   a &= 0x0f;
 
-  // for FEN checks between writes
-  sh2_poll[0].cnt = 0;
-
   // TODO: verify what's writeable
   switch (a) {
     case 0x01:
@@ -412,7 +433,7 @@ static void p32x_vdp_write8(u32 a, u32 d)
   }
 }
 
-static void p32x_vdp_write16(u32 a, u32 d, u32 cycles)
+static void p32x_vdp_write16(u32 a, u32 d, SH2 *sh2)
 {
   a &= 0x0e;
   if (a == 6) { // fill start
@@ -430,9 +451,10 @@ static void p32x_vdp_write16(u32 a, u32 d, u32 cycles)
     }
     Pico32x.vdp_regs[0x06 / 2] = a;
     Pico32x.vdp_regs[0x08 / 2] = d;
-    if (cycles > 0) {
+    if (sh2 != NULL && len > 4) {
       Pico32x.vdp_regs[0x0a / 2] |= P32XV_nFEN;
-      p32x_event_schedule(P32X_EVENT_FILLEND, cycles, len);
+      // supposedly takes 3 bus/6 sh2 cycles? or 3 sh2 cycles?
+      p32x_event_schedule_sh2(sh2, P32X_EVENT_FILLEND, 3 + len);
     }
     return;
   }
@@ -452,8 +474,8 @@ static u32 p32x_sh2reg_read16(u32 a, int cpuid)
     case 0x00: // adapter/irq ctl
       return (r[0] & P32XS_FM) | Pico32x.sh2_regs[0] | Pico32x.sh2irq_mask[cpuid];
     case 0x04: // H count (often as comm too)
-      if (p32x_poll_detect(&sh2_poll[cpuid], a, ash2_cycles_done(&sh2s[cpuid]), 0))
-        ash2_end_run(&sh2s[cpuid], 8);
+      sh2_poll_detect(&sh2s[cpuid], a, SH2_STATE_CPOLL);
+      sh2s_sync_on_read(&sh2s[cpuid]);
       return Pico32x.sh2_regs[4 / 2];
     case 0x10: // DREQ len
       return r[a / 2];
@@ -467,12 +489,12 @@ static u32 p32x_sh2reg_read16(u32 a, int cpuid)
     int comreg = 1 << (a & 0x0f) / 2;
     if (Pico32x.comm_dirty_68k & comreg)
       Pico32x.comm_dirty_68k &= ~comreg;
-    else if (p32x_poll_detect(&sh2_poll[cpuid], a, ash2_cycles_done(&sh2s[cpuid]), 0))
-      ash2_end_run(&sh2s[cpuid], 8);
+    else
+      sh2_poll_detect(&sh2s[cpuid], a, SH2_STATE_CPOLL);
+    sh2s_sync_on_read(&sh2s[cpuid]);
     return r[a / 2];
   }
   if ((a & 0x30) == 0x30) {
-    sh2_poll[cpuid].cnt = 0;
     return p32x_pwm_read16(a);
   }
 
@@ -482,22 +504,32 @@ static u32 p32x_sh2reg_read16(u32 a, int cpuid)
 static void p32x_sh2reg_write8(u32 a, u32 d, int cpuid)
 {
   a &= 0xff;
+
+  sh2s[cpuid].poll_addr = 0;
+
   switch (a) {
     case 0: // FM
       Pico32x.regs[0] &= ~P32XS_FM;
       Pico32x.regs[0] |= (d << 8) & P32XS_FM;
       return;
-    case 1: // 
+    case 1: // HEN/irq masks
+      if ((d ^ Pico32x.sh2_regs[0]) & 0x80)
+        elprintf(EL_ANOMALY|EL_32X, "HEN");
       Pico32x.sh2irq_mask[cpuid] = d & 0x8f;
       Pico32x.sh2_regs[0] &= ~0x80;
       Pico32x.sh2_regs[0] |= d & 0x80;
       if (d & 1)
-        p32x_pwm_schedule(sh2s[cpuid].m68krcycles_done); // XXX: timing?
-      p32x_update_irls(1);
+        p32x_pwm_schedule_sh2(&sh2s[cpuid]);
+      p32x_update_irls(&sh2s[cpuid]);
       return;
     case 5: // H count
-      Pico32x.sh2_regs[4 / 2] = d & 0xff;
-      p32x_poll_undetect(&sh2_poll[cpuid ^ 1], 0);
+      d &= 0xff;
+      if (Pico32x.sh2_regs[4 / 2] != d) {
+        Pico32x.sh2_regs[4 / 2] = d;
+        p32x_sh2_poll_event(&sh2s[cpuid ^ 1], SH2_STATE_CPOLL,
+          sh2_cycles_done_m68k(&sh2s[cpuid]));
+        sh2_end_run(&sh2s[cpuid], 4);
+      }
       return;
   }
 
@@ -508,9 +540,9 @@ static void p32x_sh2reg_write8(u32 a, u32 d, int cpuid)
       return;
 
     r8[a ^ 1] = d;
-    if (p32x_poll_undetect(&m68k_poll, 0))
-      SekSetStop(0);
-    p32x_poll_undetect(&sh2_poll[cpuid ^ 1], 0);
+    p32x_m68k_poll_event(P32XF_68KCPOLL);
+    p32x_sh2_poll_event(&sh2s[cpuid ^ 1], SH2_STATE_CPOLL,
+      sh2_cycles_done_m68k(&sh2s[cpuid]));
     comreg = 1 << (a & 0x0f) / 2;
     Pico32x.comm_dirty_sh2 |= comreg;
     return;
@@ -521,6 +553,8 @@ static void p32x_sh2reg_write16(u32 a, u32 d, int cpuid)
 {
   a &= 0xfe;
 
+  sh2s[cpuid].poll_addr = 0;
+
   // comm
   if ((a & 0x30) == 0x20) {
     int comreg;
@@ -528,9 +562,9 @@ static void p32x_sh2reg_write16(u32 a, u32 d, int cpuid)
       return;
 
     Pico32x.regs[a / 2] = d;
-    if (p32x_poll_undetect(&m68k_poll, 0))
-      SekSetStop(0);
-    p32x_poll_undetect(&sh2_poll[cpuid ^ 1], 0);
+    p32x_m68k_poll_event(P32XF_68KCPOLL);
+    p32x_sh2_poll_event(&sh2s[cpuid ^ 1], SH2_STATE_CPOLL,
+      sh2_cycles_done_m68k(&sh2s[cpuid]));
     comreg = 1 << (a & 0x0f) / 2;
     Pico32x.comm_dirty_sh2 |= comreg;
     return;
@@ -553,7 +587,7 @@ static void p32x_sh2reg_write16(u32 a, u32 d, int cpuid)
     case 0x1c:
       Pico32x.sh2irqs &= ~P32XI_PWM;
       if (!(Pico32x.emu_flags & P32XF_PWM_PEND))
-        p32x_pwm_schedule(sh2s[cpuid].m68krcycles_done); // timing?
+        p32x_pwm_schedule_sh2(&sh2s[cpuid]);
       goto irls;
   }
 
@@ -561,7 +595,7 @@ static void p32x_sh2reg_write16(u32 a, u32 d, int cpuid)
   return;
 
 irls:
-  p32x_update_irls(1);
+  p32x_update_irls(&sh2s[cpuid]);
 }
 
 // ------------------------------------------------------------------
@@ -693,7 +727,7 @@ static void sh2_peripheral_write32(u32 a, u32 d, int id)
     dmac0->tcr0 &= 0xffffff;
 
     // HACK: assume 68k starts writing soon and end the timeslice
-    ash2_end_run(&sh2s[id], 16);
+    sh2_end_run(&sh2s[id], 16);
 
     // DREQ is only sent after first 4 words are written.
     // we do multiple of 4 words to avoid messing up alignment
@@ -829,7 +863,7 @@ static void PicoWrite16_32x_on(u32 a, u32 d)
   }
 
   if ((a & 0xfff0) == 0x5180) { // a15180
-    p32x_vdp_write16(a, d, 0); // FIXME?
+    p32x_vdp_write16(a, d, NULL); // FIXME?
     return;
   }
 
@@ -1014,8 +1048,7 @@ static u32 sh2_read8_cs0(u32 a, int id)
 
   if ((a & 0x3ff00) == 0x4100) {
     d = p32x_vdp_read16(a);
-    if (p32x_poll_detect(&sh2_poll[id], a, ash2_cycles_done(&sh2s[id]), 1))
-      ash2_end_run(&sh2s[id], 8);
+    sh2_poll_detect(&sh2s[id], a, SH2_STATE_VPOLL);
     goto out_16to8;
   }
 
@@ -1069,8 +1102,7 @@ static u32 sh2_read16_cs0(u32 a, int id)
 
   if ((a & 0x3ff00) == 0x4100) {
     d = p32x_vdp_read16(a);
-    if (p32x_poll_detect(&sh2_poll[id], a, ash2_cycles_done(&sh2s[id]), 1))
-      ash2_end_run(&sh2s[id], 8);
+    sh2_poll_detect(&sh2s[id], a, SH2_STATE_VPOLL);
     goto out;
   }
 
@@ -1116,6 +1148,7 @@ static int REGPARM(3) sh2_write8_cs0(u32 a, u32 d, int id)
     id ? 's' : 'm', a, d & 0xff, sh2_pc(id));
 
   if ((a & 0x3ff00) == 0x4100) {
+    sh2s[id].poll_addr = 0;
     p32x_vdp_write8(a, d);
     return 0;
   }
@@ -1185,8 +1218,8 @@ static int REGPARM(3) sh2_write16_cs0(u32 a, u32 d, int id)
       id ? 's' : 'm', a, d & 0xffff, sh2_pc(id));
 
   if ((a & 0x3ff00) == 0x4100) {
-    sh2_poll[id].cnt = 0; // for poll before VDP accesses
-    p32x_vdp_write16(a, d, sh2s[id].m68krcycles_done);
+    sh2s[id].poll_addr = 0;
+    p32x_vdp_write16(a, d, &sh2s[id]);
     return 0;
   }
 
@@ -1596,14 +1629,6 @@ void PicoMemSetup32x(void)
   msh2.write8_tab  = ssh2.write8_tab  = (const void **)(void *)sh2_write8_map;
   msh2.write16_tab = ssh2.write16_tab = (const void **)(void *)sh2_write16_map;
 
-  // setup poll detector
-  m68k_poll.flag = P32XF_68KPOLL;
-  m68k_poll.cyc_max = 64;
-  sh2_poll[0].flag = P32XF_MSH2POLL;
-  sh2_poll[0].cyc_max = 21;
-  sh2_poll[1].flag = P32XF_SSH2POLL;
-  sh2_poll[1].cyc_max = 16;
-
   sh2_drc_mem_setup(&msh2);
   sh2_drc_mem_setup(&ssh2);
 }
@@ -1615,6 +1640,13 @@ void Pico32xMemStateLoaded(void)
   memset(Pico32xMem->pwm, 0, sizeof(Pico32xMem->pwm));
   Pico32x.dirty_pal = 1;
 
+  Pico32x.emu_flags &= ~(P32XF_68KCPOLL | P32XF_68KVPOLL);
+  memset(&m68k_poll, 0, sizeof(m68k_poll));
+  msh2.state = 0;
+  msh2.poll_addr = msh2.poll_cycles = msh2.poll_cnt = 0;
+  ssh2.state = 0;
+  ssh2.poll_addr = ssh2.poll_cycles = ssh2.poll_cnt = 0;
+
   sh2_drc_flush_all();
 }
 
diff --git a/pico/32x/pwm.c b/pico/32x/pwm.c
index e3385aa..36261d7 100644
--- a/pico/32x/pwm.c
+++ b/pico/32x/pwm.c
@@ -86,21 +86,35 @@ void p32x_timers_do(unsigned int cycles)
   }
 }
 
-void p32x_pwm_schedule(unsigned int now)
+static int p32x_pwm_schedule_(void)
 {
   int tm;
 
   if (Pico32x.emu_flags & P32XF_PWM_PEND)
-    return; // already scheduled
+    return 0; // already scheduled
   if (Pico32x.sh2irqs & P32XI_PWM)
-    return; // previous not acked
+    return 0; // previous not acked
   if (!((Pico32x.sh2irq_mask[0] | Pico32x.sh2irq_mask[1]) & 1))
-    return; // masked by everyone
+    return 0; // masked by everyone
 
+  Pico32x.emu_flags |= P32XF_PWM_PEND;
   tm = (Pico32x.regs[0x30 / 2] & 0x0f00) >> 8;
   tm = ((tm - 1) & 0x0f) + 1;
-  p32x_event_schedule(P32X_EVENT_PWM, now, pwm_cycles * tm / 3);
-  Pico32x.emu_flags |= P32XF_PWM_PEND;
+  return pwm_cycles * tm / 3;
+}
+
+void p32x_pwm_schedule(unsigned int now)
+{
+  int after = p32x_pwm_schedule_();
+  if (after != 0)
+    p32x_event_schedule(now, P32X_EVENT_PWM, after);
+}
+
+void p32x_pwm_schedule_sh2(SH2 *sh2)
+{
+  int after = p32x_pwm_schedule_();
+  if (after != 0)
+    p32x_event_schedule_sh2(sh2, P32X_EVENT_PWM, after);
 }
 
 unsigned int p32x_pwm_read16(unsigned int a)
diff --git a/pico/debug.c b/pico/debug.c
index 2370bfa..e579296 100644
--- a/pico/debug.c
+++ b/pico/debug.c
@@ -402,3 +402,137 @@ void PDebugCPUStep(void)
     SekStepM68k();
 }
 
+#ifdef EVT_LOG
+static struct evt_t {
+  unsigned int cycles;
+  short cpu;
+  short evt;
+} *evts;
+static int first_frame;
+static int evt_alloc;
+static int evt_cnt;
+
+void pevt_log(unsigned int cycles, enum evt_cpu c, enum evt e)
+{
+  if (first_frame == 0)
+    first_frame = Pico.m.frame_count;
+  if (evt_alloc == evt_cnt) {
+    evt_alloc = evt_alloc * 2 + 16 * 1024;
+    evts = realloc(evts, evt_alloc * sizeof(evts[0]));
+  }
+  evts[evt_cnt].cycles = cycles;
+  evts[evt_cnt].cpu = c;
+  evts[evt_cnt].evt = e;
+  evt_cnt++;
+}
+
+static int evt_cmp(const void *p1, const void *p2)
+{
+  const struct evt_t *e1 = p1, *e2 = p2;
+  int ret = (int)(e1->cycles - e2->cycles);
+  if (ret)
+    return ret;
+  if (e1->evt == EVT_RUN_END || e1->evt == EVT_POLL_END)
+    return -1;
+  if (e1->evt == EVT_RUN_START || e1->evt == EVT_POLL_START)
+    return 1;
+  if (e2->evt == EVT_RUN_END || e2->evt == EVT_POLL_END)
+    return 1;
+  if (e1->evt == EVT_RUN_START || e1->evt == EVT_POLL_START)
+    return -1;
+  return 0;
+}
+
+void pevt_dump(void)
+{
+  static const char *evt_names[EVT_CNT] = {
+    "x", "x", "+run", "-run", "+poll", "-poll",
+  };
+  char evt_print[EVT_CPU_CNT][EVT_CNT] = {{0,}};
+  unsigned int start_cycles[EVT_CPU_CNT] = {0,};
+  unsigned int run_cycles[EVT_CPU_CNT] = {0,};
+  unsigned int frame_cycles[EVT_CPU_CNT] = {0,};
+  unsigned int frame_resched[EVT_CPU_CNT] = {0,};
+  unsigned int cycles = 0;
+  int frame = first_frame - 1;
+  int line = 0;
+  int cpu_mask = 0;
+  int dirty = 0;
+  int i;
+
+  qsort(evts, evt_cnt, sizeof(evts[0]), evt_cmp);
+
+  for (i = 0; i < evt_cnt; i++) {
+    int c = evts[i].cpu, e = evts[i].evt;
+    int ei, ci;
+
+    if (cycles != evts[i].cycles || (cpu_mask & (1 << c))
+        || e == EVT_FRAME_START || e == EVT_NEXT_LINE)
+    {
+      if (dirty) {
+        printf("%u:%03u:%u ", frame, line, cycles);
+        for (ci = 0; ci < EVT_CPU_CNT; ci++) {
+          int found = 0;
+          for (ei = 0; ei < EVT_CNT; ei++) {
+            if (evt_print[ci][ei]) {
+              if (ei == EVT_RUN_END) {
+                printf("%8s%4d", evt_names[ei], run_cycles[ci]);
+                run_cycles[ci] = 0;
+              }
+              else
+                printf("%8s    ", evt_names[ei]);
+              found = 1;
+            }
+          }
+          if (!found)
+            printf("%12s", "");
+        }
+        printf("\n");
+        memset(evt_print, 0, sizeof(evt_print));
+        cpu_mask = 0;
+        dirty = 0;
+      }
+      cycles = evts[i].cycles;
+    }
+
+    switch (e) {
+    case EVT_FRAME_START:
+      frame++;
+      line = 0;
+      printf("%u:%03u:%u ", frame, line, cycles);
+      for (ci = 0; ci < EVT_CPU_CNT; ci++) {
+        printf("%12u", frame_cycles[ci]);
+        frame_cycles[ci] = 0;
+      }
+      printf("\n");
+      printf("%u:%03u:%u ", frame, line, cycles);
+      for (ci = 0; ci < EVT_CPU_CNT; ci++) {
+        printf("%12u", frame_resched[ci]);
+        frame_resched[ci] = 0;
+      }
+      printf("\n");
+      break;
+    case EVT_NEXT_LINE:
+      line++;
+      printf("%u:%03u:%u\n", frame, line, cycles);
+      break;
+    case EVT_RUN_START:
+      start_cycles[c] = cycles;
+      goto default_;
+    case EVT_RUN_END:
+      run_cycles[c] += cycles - start_cycles[c];
+      frame_cycles[c] += cycles - start_cycles[c];
+      frame_resched[c]++;
+      goto default_;
+    default_:
+    default:
+      evt_print[c][e] = 1;
+      cpu_mask |= 1 << c;
+      dirty = 1;
+      break;
+    }
+  }
+}
+#endif
+
+// vim:shiftwidth=2:ts=2:expandtab
diff --git a/pico/pico.c b/pico/pico.c
index 9a1faca..80ffb90 100644
--- a/pico/pico.c
+++ b/pico/pico.c
@@ -54,6 +54,7 @@ void PicoExit(void)
 
   if (SRam.data)
     free(SRam.data);
+  pevt_dump();
 }
 
 void PicoPower(void)
diff --git a/pico/pico_cmn.c b/pico/pico_cmn.c
index 12f649f..9099c85 100644
--- a/pico/pico_cmn.c
+++ b/pico/pico_cmn.c
@@ -30,9 +30,12 @@ static __inline void SekRunM68k(int cyc)
 {
   int cyc_do;
   pprof_start(m68k);
+  pevt_log_m68k_o(EVT_RUN_START);
 
   SekCycleAim+=cyc;
-  if ((cyc_do=SekCycleAim-SekCycleCnt) <= 0) return;
+  if ((cyc_do=SekCycleAim-SekCycleCnt) <= 0)
+    goto out;
+
 #if defined(EMU_CORE_DEBUG)
   // this means we do run-compare
   SekCycleCnt+=CM_compareRun(cyc_do, 0);
@@ -46,6 +49,8 @@ static __inline void SekRunM68k(int cyc)
   SekCycleCnt+=fm68k_emulate(cyc_do+1, 0, 0);
 #endif
 
+out:
+  pevt_log_m68k_o(EVT_RUN_END);
   pprof_end(m68k);
 }
 
@@ -55,6 +60,7 @@ static int PicoFrameHints(void)
   int lines, y, lines_vis = 224, line_sample, skip, vcnt_wrap;
   int hint; // Hint counter
 
+  pevt_log_m68k_o(EVT_FRAME_START);
   pv->v_counter = Pico.m.scanline = 0;
 
   if ((PicoOpt&POPT_ALT_RENDERER) && !PicoSkipFrame && (pv->reg[1]&0x40)) { // fast rend., display enabled
@@ -144,7 +150,7 @@ static int PicoFrameHints(void)
       if (ym2612.dacen && PsndDacLine <= y)
         PsndDoDAC(y);
 #ifdef PICO_32X
-      p32x_sync_sh2s(SekCycleCntT + SekCycleCnt);
+      p32x_sync_sh2s(SekCyclesDoneT2());
 #endif
       PsndGetSamples(y);
     }
@@ -158,6 +164,7 @@ static int PicoFrameHints(void)
 #else
     if (PicoLineHook) PicoLineHook();
 #endif
+    pevt_log_m68k_o(EVT_NEXT_LINE);
   }
 
   if (!skip)
@@ -213,7 +220,7 @@ static int PicoFrameHints(void)
   }
 
 #ifdef PICO_32X
-  p32x_sync_sh2s(SekCycleCntT + SekCycleCnt);
+  p32x_sync_sh2s(SekCyclesDoneT2());
   p32x_start_blank();
 #endif
 
@@ -235,6 +242,7 @@ static int PicoFrameHints(void)
 #else
   if (PicoLineHook) PicoLineHook();
 #endif
+  pevt_log_m68k_o(EVT_NEXT_LINE);
 
   lines = scanlines_total;
   vcnt_wrap = Pico.m.pal ? 0x103 : 0xEB; // based on Gens, TODO: verify
@@ -262,6 +270,7 @@ static int PicoFrameHints(void)
 #else
     if (PicoLineHook) PicoLineHook();
 #endif
+    pevt_log_m68k_o(EVT_NEXT_LINE);
   }
 
   // sync z80
@@ -271,7 +280,7 @@ static int PicoFrameHints(void)
     PsndDoDAC(lines-1);
 
 #ifdef PICO_32X
-  p32x_sync_sh2s(SekCycleCntT + SekCycleCnt);
+  p32x_sync_sh2s(SekCyclesDoneT2());
 #endif
   timers_cycle();
 
diff --git a/pico/pico_int.h b/pico/pico_int.h
index a654c5a..bfdf73a 100644
--- a/pico/pico_int.h
+++ b/pico/pico_int.h
@@ -146,6 +146,7 @@ extern unsigned int SekCycleCntT; // total cycle counter, updated once per frame
 #define SekCyclesBurn(c)  SekCycleCnt+=c
 #define SekCyclesDone()  (SekCycleAim-SekCyclesLeft)    // number of cycles done in this frame (can be checked anywhere)
 #define SekCyclesDoneT() (SekCycleCntT+SekCyclesDone()) // total nuber of cycles done for this rom
+#define SekCyclesDoneT2() (SekCycleCntT + SekCycleCnt)  // same as above but not from memhandlers
 
 #define SekEndRun(after) { \
 	SekCycleCnt -= SekCyclesLeft - (after); \
@@ -237,27 +238,31 @@ extern SH2 sh2s[2];
 #define ssh2 sh2s[1]
 
 #ifndef DRC_SH2
-# define ash2_end_run(sh2, after) do { \
-  if ((sh2)->icount > (after)) { \
+# define sh2_end_run(sh2, after_) do { \
+  if ((sh2)->icount > (after_)) { \
     (sh2)->cycles_timeslice -= (sh2)->icount; \
-    (sh2)->icount = after; \
+    (sh2)->icount = after_; \
   } \
 } while (0)
-# define ash2_cycles_done(sh2) ((sh2)->cycles_timeslice - (sh2)->icount)
+# define sh2_cycles_left(sh2) (sh2)->icount
 # define sh2_pc(c)    (c) ? ssh2.ppc : msh2.ppc
 #else
-# define ash2_end_run(sh2, after) do { \
-  int left = (sh2)->sr >> 12; \
-  if (left > (after)) { \
-    (sh2)->cycles_timeslice -= left; \
+# define sh2_end_run(sh2, after_) do { \
+  int left_ = (signed int)(sh2)->sr >> 12; \
+  if (left_ > (after_)) { \
+    (sh2)->cycles_timeslice -= left_; \
     (sh2)->sr &= 0xfff; \
-    (sh2)->sr |= (after) << 12; \
+    (sh2)->sr |= (after_) << 12; \
   } \
 } while (0)
-# define ash2_cycles_done(sh2) ((sh2)->cycles_timeslice - ((sh2)->sr >> 12))
+# define sh2_cycles_left(sh2) ((signed int)(sh2)->sr >> 12)
 # define sh2_pc(c)    (c) ? ssh2.pc : msh2.pc
 #endif
 
+#define sh2_cycles_done(sh2) ((int)(sh2)->cycles_timeslice - sh2_cycles_left(sh2))
+#define sh2_cycles_done_m68k(sh2) \
+  ((sh2)->m68krcycles_done + (sh2_cycles_done(sh2) / 3))
+
 #define sh2_reg(c, x) (c) ? ssh2.r[x] : msh2.r[x]
 #define sh2_gbr(c)    (c) ? ssh2.gbr : msh2.gbr
 #define sh2_vbr(c)    (c) ? ssh2.vbr : msh2.vbr
@@ -463,12 +468,8 @@ typedef struct
 #define P32XP_FULL  (1<<15) // PWM
 #define P32XP_EMPTY (1<<14)
 
-#define P32XF_68KPOLL   (1 << 0)
-#define P32XF_MSH2POLL  (1 << 1)
-#define P32XF_SSH2POLL  (1 << 2)
-#define P32XF_68KVPOLL  (1 << 3)
-#define P32XF_MSH2VPOLL (1 << 4)
-#define P32XF_SSH2VPOLL (1 << 5)
+#define P32XF_68KCPOLL  (1 << 0)
+#define P32XF_68KVPOLL  (1 << 1)
 #define P32XF_PWM_PEND  (1 << 6)
 
 #define P32XI_VRES (1 << 14/2) // IRL/2
@@ -737,9 +738,11 @@ void PicoUnload32x(void);
 void PicoFrame32x(void);
 void Pico32xStateLoaded(int is_early);
 void p32x_sync_sh2s(unsigned int m68k_target);
-void p32x_update_irls(int nested_call);
+void p32x_sync_other_sh2(SH2 *sh2, unsigned int m68k_target);
+void p32x_update_irls(SH2 *active_sh2);
 void p32x_reset_sh2s(void);
-void p32x_event_schedule(enum p32x_event event, unsigned int now, int after);
+void p32x_event_schedule(unsigned int now, enum p32x_event event, int after);
+void p32x_event_schedule_sh2(SH2 *sh2, enum p32x_event event, int after);
 
 // 32x/memory.c
 struct Pico32xMem *Pico32xMem;
@@ -750,7 +753,8 @@ void PicoWrite16_32x(unsigned int a, unsigned int d);
 void PicoMemSetup32x(void);
 void Pico32xSwapDRAM(int b);
 void Pico32xMemStateLoaded(void);
-void p32x_poll_event(int cpu_mask, int is_vdp);
+void p32x_m68k_poll_event(unsigned int flags);
+void p32x_sh2_poll_event(SH2 *sh2, unsigned int flags, unsigned int m68k_cycles);
 
 // 32x/draw.c
 void FinalizeLine32xRGB555(int sh, int line);
@@ -772,6 +776,7 @@ void p32x_pwm_update(int *buf32, int length, int stereo);
 void p32x_timers_do(unsigned int cycles);
 void p32x_timers_recalc(void);
 void p32x_pwm_schedule(unsigned int now);
+void p32x_pwm_schedule_sh2(SH2 *sh2);
 #else
 #define Pico32xInit()
 #define PicoPower32x()
@@ -848,6 +853,45 @@ do { \
 #define pprof_end_sub(...)
 #endif
 
+#ifdef EVT_LOG
+enum evt {
+  EVT_FRAME_START,
+  EVT_NEXT_LINE,
+  EVT_RUN_START,
+  EVT_RUN_END,
+  EVT_POLL_START,
+  EVT_POLL_END,
+  EVT_CNT
+};
+
+enum evt_cpu {
+  EVT_M68K,
+  EVT_S68K,
+  EVT_MSH2,
+  EVT_SSH2,
+  EVT_CPU_CNT
+};
+
+void pevt_log(unsigned int cycles, enum evt_cpu c, enum evt e);
+void pevt_dump(void);
+
+#define pevt_log_m68k(e) \
+  pevt_log(SekCyclesDoneT(), EVT_M68K, e)
+#define pevt_log_m68k_o(e) \
+  pevt_log(SekCyclesDoneT2(), EVT_M68K, e)
+#define pevt_log_sh2(sh2, e) \
+  pevt_log(sh2_cycles_done_m68k(sh2), EVT_MSH2 + (sh2)->is_slave, e)
+#define pevt_log_sh2_o(sh2, e) \
+  pevt_log((sh2)->m68krcycles_done, EVT_MSH2 + (sh2)->is_slave, e)
+#else
+#define pevt_log(c, e)
+#define pevt_log_m68k(e)
+#define pevt_log_m68k_o(e)
+#define pevt_log_sh2(sh2, e)
+#define pevt_log_sh2_o(sh2, e)
+#define pevt_dump()
+#endif
+
 // misc
 #ifdef _MSC_VER
 #define cdprintf