dev-release: try to make it work
[pcsx_rearmed.git] / libpcsxcore / lightrec / plugin.c
index 2a46942..3e11265 100644 (file)
@@ -1,4 +1,3 @@
-#include <lightrec.h>
 #include <errno.h>
 #include <stdbool.h>
 #include <stdio.h>
@@ -10,6 +9,7 @@
 #include <sys/mman.h>
 #endif
 
+#include "lightrec.h"
 #include "../cdrom.h"
 #include "../gpu.h"
 #include "../gte.h"
@@ -19,7 +19,8 @@
 #include "../psxmem.h"
 #include "../r3000a.h"
 #include "../psxinterpreter.h"
-#include "../new_dynarec/events.h"
+#include "../psxhle.h"
+#include "../psxevents.h"
 
 #include "../frontend/main.h"
 
 #      define unlikely(x)     (x)
 #endif
 
+#ifndef LIGHTREC_PROG_NAME
+#  ifdef __linux__
+#    define LIGHTREC_PROG_NAME "/proc/self/exe"
+#  else
+#    define LIGHTREC_PROG_NAME "retroarch.exe"
+#  endif
+#endif
+
+#ifndef GPUSTATUS_POLLING_THRESHOLD
+#  define GPUSTATUS_POLLING_THRESHOLD 0
+#endif
+
 psxRegisters psxRegs;
 Rcnt rcnts[4];
 
@@ -59,11 +72,18 @@ void* code_buffer;
 
 static struct lightrec_state *lightrec_state;
 
-static char *name = "retroarch.exe";
-
 static bool use_lightrec_interpreter;
-static bool use_pcsx_interpreter;
 static bool block_stepping;
+//static bool use_pcsx_interpreter;
+#define use_pcsx_interpreter 0
+static bool ram_disabled;
+static bool lightrec_debug, lightrec_very_debug;
+static u32 lightrec_begin_cycles;
+
+extern u32 lightrec_hacks;
+
+static void lightrec_plugin_apply_config();
+extern void lightrec_code_inv(void *ptr, uint32_t len);
 
 enum my_cp2_opcodes {
        OP_CP2_RTPS             = 0x01,
@@ -91,7 +111,6 @@ enum my_cp2_opcodes {
 };
 
 static void (*cp2_ops[])(struct psxCP2Regs *) = {
-       [OP_CP2_RTPS] = gteRTPS,
        [OP_CP2_RTPS] = gteRTPS,
        [OP_CP2_NCLIP] = gteNCLIP,
        [OP_CP2_OP] = gteOP,
@@ -150,7 +169,7 @@ static void lightrec_tansition_to_pcsx(struct lightrec_state *state)
 
 static void lightrec_tansition_from_pcsx(struct lightrec_state *state)
 {
-       s32 cycles_left = next_interupt - psxRegs.cycle;
+       s32 cycles_left = psxRegs.next_interupt - psxRegs.cycle;
 
        if (block_stepping || cycles_left <= 0 || has_interrupt())
                lightrec_set_exit_flags(state, LIGHTREC_EXIT_CHECK_INTERRUPT);
@@ -160,7 +179,7 @@ static void lightrec_tansition_from_pcsx(struct lightrec_state *state)
 }
 
 static void hw_write_byte(struct lightrec_state *state,
-                         u32 op, void *host, u32 mem, u8 val)
+                         u32 op, void *host, u32 mem, u32 val)
 {
        lightrec_tansition_to_pcsx(state);
 
@@ -170,7 +189,7 @@ static void hw_write_byte(struct lightrec_state *state,
 }
 
 static void hw_write_half(struct lightrec_state *state,
-                         u32 op, void *host, u32 mem, u16 val)
+                         u32 op, void *host, u32 mem, u32 val)
 {
        lightrec_tansition_to_pcsx(state);
 
@@ -219,12 +238,30 @@ static u16 hw_read_half(struct lightrec_state *state,
 static u32 hw_read_word(struct lightrec_state *state,
                        u32 op, void *host, u32 mem)
 {
-       u32 val;
+       static u32 old_cycle, oldold_cycle, old_gpusr;
+       u32 val, diff;
 
        lightrec_tansition_to_pcsx(state);
 
        val = psxHwRead32(mem);
 
+       if (GPUSTATUS_POLLING_THRESHOLD > 0 && mem == 0x1f801814) {
+               diff = psxRegs.cycle - old_cycle;
+
+               if (diff > 0
+                   && diff < GPUSTATUS_POLLING_THRESHOLD
+                   && diff == old_cycle - oldold_cycle) {
+                       while (psxRegs.next_interupt > psxRegs.cycle && val == old_gpusr) {
+                               psxRegs.cycle += diff;
+                               val = psxHwRead32(mem);
+                       }
+               }
+
+               oldold_cycle = old_cycle;
+               old_cycle = psxRegs.cycle;
+               old_gpusr = val;
+       }
+
        lightrec_tansition_from_pcsx(state);
 
        return val;
@@ -328,10 +365,20 @@ static void lightrec_enable_ram(struct lightrec_state *state, bool enable)
                memcpy(psxM, cache_buf, sizeof(cache_buf));
        else
                memcpy(cache_buf, psxM, sizeof(cache_buf));
+
+       ram_disabled = !enable;
 }
 
 static bool lightrec_can_hw_direct(u32 kaddr, bool is_write, u8 size)
 {
+       if (is_write && size != 32) {
+               // force32 so must go through handlers
+               if (0x1f801000 <= kaddr && kaddr < 0x1f801024)
+                       return false;
+               if ((kaddr & 0x1fffff80) == 0x1f801080) // dma
+                       return false;
+       }
+
        switch (size) {
        case 8:
                switch (kaddr) {
@@ -406,29 +453,11 @@ static bool lightrec_can_hw_direct(u32 kaddr, bool is_write, u8 size)
        }
 }
 
-#if defined(HW_DOL) || defined(HW_RVL)
-static void lightrec_code_inv(void *ptr, uint32_t len)
-{
-       extern void DCFlushRange(void *ptr, u32 len);
-       extern void ICInvalidateRange(void *ptr, u32 len);
-
-       DCFlushRange(ptr, len);
-       ICInvalidateRange(ptr, len);
-}
-#elif defined(HW_WUP)
-static void lightrec_code_inv(void *ptr, uint32_t len)
-{
-       wiiu_clear_cache(ptr, (void *)((uintptr_t)ptr + len));
-}
-#endif
-
 static const struct lightrec_ops lightrec_ops = {
        .cop2_op = cop2_op,
        .enable_ram = lightrec_enable_ram,
        .hw_direct = lightrec_can_hw_direct,
-#if defined(HW_DOL) || defined(HW_RVL) || defined(HW_WUP)
-       .code_inv = lightrec_code_inv,
-#endif
+       .code_inv = LIGHTREC_CODE_INV ? lightrec_code_inv : NULL,
 };
 
 static int lightrec_plugin_init(void)
@@ -463,7 +492,17 @@ static int lightrec_plugin_init(void)
 
        use_lightrec_interpreter = !!getenv("LIGHTREC_INTERPRETER");
 
-       lightrec_state = lightrec_init(name,
+#ifdef LIGHTREC_DEBUG
+       char *cycles = getenv("LIGHTREC_BEGIN_CYCLES");
+
+       lightrec_very_debug = !!getenv("LIGHTREC_VERY_DEBUG");
+       lightrec_debug = lightrec_very_debug || !!getenv("LIGHTREC_DEBUG");
+
+       if (cycles)
+               lightrec_begin_cycles = (unsigned int) strtol(cycles, NULL, 0);
+#endif
+
+       lightrec_state = lightrec_init(LIGHTREC_PROG_NAME,
                        lightrec_map, ARRAY_SIZE(lightrec_map),
                        &lightrec_ops);
 
@@ -476,17 +515,123 @@ static int lightrec_plugin_init(void)
 #ifndef _WIN32
        signal(SIGPIPE, exit);
 #endif
+       lightrec_plugin_apply_config();
        return 0;
 }
 
+static u32 do_calculate_hash(const void *buffer, u32 count, u32 needle, bool le)
+{
+       unsigned int i;
+       const u32 *data = (const u32 *) buffer;
+       u32 hash = needle;
+
+       count /= 4;
+       for(i = 0; i < count; ++i) {
+               hash += le ? LE32TOH(data[i]) : data[i];
+               hash += (hash << 10);
+               hash ^= (hash >> 6);
+       }
+
+       hash += (hash << 3);
+       hash ^= (hash >> 11);
+       hash += (hash << 15);
+
+       return hash;
+}
+
+static u32 hash_calculate_le(const void *buffer, u32 count)
+{
+       return do_calculate_hash(buffer, count, 0xffffffff, true);
+}
+
+u32 hash_calculate(const void *buffer, u32 count)
+{
+       return do_calculate_hash(buffer, count, 0xffffffff, false);
+}
+
+static u32 hash_calculate_ram(const void *buffer, u32 ram_size)
+{
+       u32 hash;
+
+       if (ram_disabled)
+               hash = hash_calculate_le(cache_buf, sizeof(cache_buf));
+       else
+               hash = hash_calculate_le(buffer, sizeof(cache_buf));
+
+       return do_calculate_hash(buffer + sizeof(cache_buf),
+                                ram_size - sizeof(cache_buf),
+                                hash, true);
+}
+
+static const char * const mips_regs[] = {
+       "zero",
+       "at",
+       "v0", "v1",
+       "a0", "a1", "a2", "a3",
+       "t0", "t1", "t2", "t3", "t4", "t5", "t6", "t7",
+       "s0", "s1", "s2", "s3", "s4", "s5", "s6", "s7",
+       "t8", "t9",
+       "k0", "k1",
+       "gp", "sp", "fp", "ra",
+       "lo", "hi",
+};
+
+static void print_for_big_ass_debugger(void)
+{
+       struct lightrec_registers *regs;
+       unsigned int i;
+
+       regs = lightrec_get_registers(lightrec_state);
+
+       printf("CYCLE 0x%08x PC 0x%08x", psxRegs.cycle, psxRegs.pc);
+
+       if (lightrec_very_debug)
+               printf(" RAM 0x%08x SCRATCH 0x%08x HW 0x%08x",
+                               hash_calculate_ram(psxM, 0x200000),
+                               hash_calculate_le(psxH, 0x400),
+                               hash_calculate_le(psxH + 0x1000, 0x2000));
+
+       printf(" CP0 0x%08x CP2D 0x%08x CP2C 0x%08x INT 0x%04x INTCYCLE 0x%08x GPU 0x%08x",
+                       hash_calculate(regs->cp0, sizeof(regs->cp0)),
+                       hash_calculate(regs->cp2d, sizeof(regs->cp2d)),
+                       hash_calculate(regs->cp2c, sizeof(regs->cp2c)),
+                       psxRegs.interrupt,
+                       hash_calculate(psxRegs.intCycle, sizeof(psxRegs.intCycle)),
+                       LE32TOH(HW_GPU_STATUS));
+
+       if (lightrec_very_debug) {
+               for (i = 0; i < 32; i++)
+                       printf(" CP2D%u 0x%08x", i, regs->cp2d[i]);
+               for (i = 0; i < 32; i++)
+                       printf(" CP2C%u 0x%08x", i, regs->cp2c[i]);
+       }
+
+       if (lightrec_very_debug)
+               for (i = 0; i < 34; i++)
+                       printf(" %s 0x%08x", mips_regs[i], regs->gpr[i]);
+       else
+               printf(" GPR 0x%08x",
+                      hash_calculate(regs->gpr, sizeof(regs->gpr)));
+       printf("\n");
+
+       fflush(stdout);
+}
+
+static void lightrec_plugin_sync_regs_to_pcsx(bool need_cp2);
+static void lightrec_plugin_sync_regs_from_pcsx(bool need_cp2);
+
 static void lightrec_plugin_execute_internal(bool block_only)
 {
        struct lightrec_registers *regs;
        u32 flags, cycles_pcsx;
+       u32 old_pc = psxRegs.pc;
 
        regs = lightrec_get_registers(lightrec_state);
        gen_interupt((psxCP0Regs *)regs->cp0);
-       cycles_pcsx = next_interupt - psxRegs.cycle;
+       if (!block_only && psxRegs.stop)
+               return;
+
+       cycles_pcsx = psxRegs.next_interupt - psxRegs.cycle;
        assert((s32)cycles_pcsx > 0);
 
        // step during early boot so that 0x80030000 fastboot hack works
@@ -495,7 +640,7 @@ static void lightrec_plugin_execute_internal(bool block_only)
                cycles_pcsx = 0;
 
        if (use_pcsx_interpreter) {
-               intExecuteBlock(0);
+               psxInt.ExecuteBlock(&psxRegs, 0);
        } else {
                u32 cycles_lightrec = cycles_pcsx * 1024;
                if (unlikely(use_lightrec_interpreter)) {
@@ -514,11 +659,30 @@ static void lightrec_plugin_execute_internal(bool block_only)
                if (flags & LIGHTREC_EXIT_SEGFAULT) {
                        fprintf(stderr, "Exiting at cycle 0x%08x\n",
                                psxRegs.cycle);
+                       if (lightrec_debug)
+                               print_for_big_ass_debugger();
                        exit(1);
                }
 
                if (flags & LIGHTREC_EXIT_SYSCALL)
-                       psxException(0x20, 0, (psxCP0Regs *)regs->cp0);
+                       psxException(R3000E_Syscall << 2, 0, (psxCP0Regs *)regs->cp0);
+               if (flags & LIGHTREC_EXIT_BREAK)
+                       psxException(R3000E_Bp << 2, 0, (psxCP0Regs *)regs->cp0);
+               else if (flags & LIGHTREC_EXIT_UNKNOWN_OP) {
+                       u32 op = intFakeFetch(psxRegs.pc);
+                       u32 hlec = op & 0x03ffffff;
+                       if ((op >> 26) == 0x3b && hlec < ARRAY_SIZE(psxHLEt) && Config.HLE) {
+                               lightrec_plugin_sync_regs_to_pcsx(0);
+                               psxHLEt[hlec]();
+                               lightrec_plugin_sync_regs_from_pcsx(0);
+                       }
+                       else
+                               psxException(R3000E_RI << 2, 0, (psxCP0Regs *)regs->cp0);
+               }
+       }
+
+       if (lightrec_debug && psxRegs.cycle >= lightrec_begin_cycles && psxRegs.pc != old_pc) {
+               print_for_big_ass_debugger();
        }
 
        if ((regs->cp0[13] & regs->cp0[12] & 0x300) && (regs->cp0[12] & 0x1)) {
@@ -528,31 +692,28 @@ static void lightrec_plugin_execute_internal(bool block_only)
        }
 }
 
-static void lightrec_plugin_execute(void)
+static void lightrec_plugin_execute(psxRegisters *regs)
 {
-       extern int stop;
-
-       while (!stop)
-               lightrec_plugin_execute_internal(false);
+       while (!regs->stop)
+               lightrec_plugin_execute_internal(lightrec_very_debug);
 }
 
-static void lightrec_plugin_execute_block(enum blockExecCaller caller)
+static void lightrec_plugin_execute_block(psxRegisters *regs,
+       enum blockExecCaller caller)
 {
        lightrec_plugin_execute_internal(true);
 }
 
 static void lightrec_plugin_clear(u32 addr, u32 size)
 {
-       if (addr == 0 && size == UINT32_MAX)
+       if ((addr == 0 && size == UINT32_MAX)
+           || (lightrec_hacks & LIGHTREC_OPT_INV_DMA_ONLY))
                lightrec_invalidate_all(lightrec_state);
        else
                /* size * 4: PCSX uses DMA units */
                lightrec_invalidate(lightrec_state, addr, size * 4);
 }
 
-static void lightrec_plugin_sync_regs_to_pcsx(void);
-static void lightrec_plugin_sync_regs_from_pcsx(void);
-
 static void lightrec_plugin_notify(enum R3000Anote note, void *data)
 {
        switch (note)
@@ -562,21 +723,34 @@ static void lightrec_plugin_notify(enum R3000Anote note, void *data)
                /* not used, lightrec calls lightrec_enable_ram() instead */
                break;
        case R3000ACPU_NOTIFY_BEFORE_SAVE:
-               lightrec_plugin_sync_regs_to_pcsx();
+               /* non-null 'data' means this is HLE related sync */
+               lightrec_plugin_sync_regs_to_pcsx(data == NULL);
                break;
        case R3000ACPU_NOTIFY_AFTER_LOAD:
-               lightrec_plugin_sync_regs_from_pcsx();
+               lightrec_plugin_sync_regs_from_pcsx(data == NULL);
+               if (data == NULL)
+                       lightrec_invalidate_all(lightrec_state);
                break;
        }
 }
 
 static void lightrec_plugin_apply_config()
 {
+       static u32 cycles_per_op_old;
        u32 cycle_mult = Config.cycle_multiplier_override && Config.cycle_multiplier == CYCLE_MULT_DEFAULT
                ? Config.cycle_multiplier_override : Config.cycle_multiplier;
-       assert(cycle_mult);
+       u32 cycles_per_op = cycle_mult * 1024 / 100;
+       assert(cycles_per_op);
+
+       if (cycles_per_op_old && cycles_per_op_old != cycles_per_op) {
+               SysPrintf("lightrec: reinit block cache for cycles_per_op %.2f\n",
+                       cycles_per_op / 1024.f);
+       }
+       cycles_per_op_old = cycles_per_op;
+       lightrec_set_cycles_per_opcode(lightrec_state, cycles_per_op);
 
-       lightrec_set_cycles_per_opcode(lightrec_state, cycle_mult * 1024 / 100);
+       lightrec_set_unsafe_opt_flags(lightrec_state, lightrec_hacks);
+       intApplyConfig();
 }
 
 static void lightrec_plugin_shutdown(void)
@@ -608,26 +782,26 @@ static void lightrec_plugin_reset(void)
        regs->cp0[15] = 0x00000002; // PRevID = Revision ID, same as R3000A
 }
 
-static void lightrec_plugin_sync_regs_from_pcsx(void)
+static void lightrec_plugin_sync_regs_from_pcsx(bool need_cp2)
 {
        struct lightrec_registers *regs;
 
        regs = lightrec_get_registers(lightrec_state);
-       memcpy(regs->cp2d, &psxRegs.CP2, sizeof(regs->cp2d) + sizeof(regs->cp2c));
-       memcpy(regs->cp0, &psxRegs.CP0, sizeof(regs->cp0));
        memcpy(regs->gpr, &psxRegs.GPR, sizeof(regs->gpr));
-
-       lightrec_invalidate_all(lightrec_state);
+       memcpy(regs->cp0, &psxRegs.CP0, sizeof(regs->cp0));
+       if (need_cp2)
+               memcpy(regs->cp2d, &psxRegs.CP2, sizeof(regs->cp2d) + sizeof(regs->cp2c));
 }
 
-static void lightrec_plugin_sync_regs_to_pcsx(void)
+static void lightrec_plugin_sync_regs_to_pcsx(bool need_cp2)
 {
        struct lightrec_registers *regs;
 
        regs = lightrec_get_registers(lightrec_state);
-       memcpy(&psxRegs.CP2, regs->cp2d, sizeof(regs->cp2d) + sizeof(regs->cp2c));
-       memcpy(&psxRegs.CP0, regs->cp0, sizeof(regs->cp0));
        memcpy(&psxRegs.GPR, regs->gpr, sizeof(regs->gpr));
+       memcpy(&psxRegs.CP0, regs->cp0, sizeof(regs->cp0));
+       if (need_cp2)
+               memcpy(&psxRegs.CP2, regs->cp2d, sizeof(regs->cp2d) + sizeof(regs->cp2c));
 }
 
 R3000Acpu psxRec =