imported winuaesrc1100b5.zip

diff --git a/blitter.c b/blitter.c
index 3bf706de50c96e484a8bfb10752b93553c036c3e..5a0c8c9ea7021b4d2dc55f1990223900bce94c22 100755 (executable)
--- a/blitter.c
+++ b/blitter.c
@@ -37,7 +37,6 @@ static int blitonedot,blitsign;
 static int blit_add;
 static int blit_modadda, blit_modaddb, blit_modaddc, blit_modaddd;
 static int blit_ch;
-int blit_singlechannel;
 
 #ifdef BLITTER_DEBUG
 static int blitter_dontdo;
@@ -65,61 +64,131 @@ static long blit_firstline_cycles;
 static long blit_first_cycle;
 static int blit_last_cycle, blit_dmacount, blit_dmacount2;
 static int blit_linecycles, blit_extracycles;
-static uae_u8 *blit_diag;
+static int *blit_diag;
 
 static uae_u16 ddat1, ddat2;
 static int ddat1use, ddat2use;
 
-static uae_u8 blit_cycle_diagram_finald[] =
-    { 0, 2, 0,4 };
-
-static uae_u8 blit_cycle_diagram[][10] =
+/*
+
+Confirmed blitter ínformation by Toni Wilen
+(order of channels or position of idle cycles are not confirmed)
+
+1=BLTCON0 channel mask
+2=total cycles per blitted word
+3=steals all cycles if BLTNASTY=1 (always if A-channel is enabled. this is illogical..)
+4=total cycles per blitted word in fillmode
+5=cycle diagram (first cycle)
+6=main cycle diagram (ABCD=channels,-=idle cycle,x=idle cycle but bus allocated)
+
+1 234 5    6
+
+F 4*4*ABC- ABCD
+E 4*4*ABC- ABCx
+D 3*4 AB-  ABD
+C 3*4 AB-  ABx
+B 3*3*AC-  ACD
+A 3*3*AC-  ACx
+9 2*3 A-   AD
+8 2*3 A-   Ax
+7 4 4 -BC- -BCD
+6 4 4 -BC- -BC-
+5 3 4 -B-  -BD
+4 3 4 -B-  -B-
+3 3 3 -C-  -CD
+2 3 3 -C-  -C-
+1 2 3 -D   -D
+0 2 3 --   --
+
+NOTES: (BLTNASTY=1)
+
+- Blitter ALWAYS needs free bus cycle, even if it is running an "idle" cycle.
+  Exception: possible extra fill mode idle cycle.is "real" idle cycle.
+- Fill mode may add one extra real idle cycle.(depends on channel mask)
+- All blits with channel A enabled use all available bus cycles
+  (stops CPU accesses to chip RAM if BLTNASTY=1)
+  Can someone explain this? Why does idle cycles need bus cycles?
+- idle cycles (no A-channel enabled) are not "used" by blitter, they are freely
+  available for CPU..
+
+BLTNASTY=0 makes things even more interesting..
+
+- even zero channel blits get slower if BLTNASTY=0 depending on the number of
+  active bitplanes. ALSO "2 cycle" blits with one real cycle and one idle cycle
+  have the exact same speed as zero channel blit in all situations -> only the
+  total number of cycles count, number of active channels does not matter.
+
+*/
+
+
+/* -1 = idle cycle and allocate bus */
+
+static int blit_cycle_diagram[][10] =
 {
     { 0, 2, 0,0 },             /* 0 */
-    { 0, 2, 4,0 },             /* 1 */
-    { 0, 2, 3,0 },             /* 2 */
+    { 0, 2, 0,4 },             /* 1 */
+    { 0, 3, 0,3,0 },           /* 2 */
     { 2, 3, 0,3,4, 3,0 },      /* 3 */
-    { 0, 3, 2,0,0 },           /* 4 */
+    { 0, 3, 0,2,0 },           /* 4 */
     { 2, 3, 0,2,4, 2,0 },      /* 5 */
-    { 0, 3, 2,3,0 },           /* 6 */
+    { 0, 4, 0,2,3,0 },         /* 6 */
     { 3, 4, 0,2,3,4, 2,3,0 },  /* 7 */
-    { 0, 2, 1,0 },             /* 8 */
+    { 0, 2, 1,-1 },            /* 8 */
     { 2, 2, 1,4, 1,0 },                /* 9 */
-    { 0, 2, 1,3 },             /* A */
+    { 0, 3, 1,3,-1 },          /* A */
     { 3, 3, 1,3,4, 1,3,0 },    /* B */
-    { 2, 3, 0,1,2, 1,2 },      /* C */
+    { 2, 3, 1,2,-1, 1,2 },     /* C */
     { 3, 3, 1,2,4, 1,2,0 },    /* D */
-    { 0, 3, 1,2,3 },           /* E */
+    { 0, 4, 1,2,3,-1 },                /* E */
     { 4, 4, 1,2,3,4, 1,2,3,0 } /* F */
 };
 
-/* fill mode always adds C-channel to cycle-diagram */
-/* Reflect - Sound Vision freezes without this */
-static uae_u8 blit_cycle_diagram_fill[][10] =
+/* 5 = fill mode idle cycle ("real" idle cycle) */
+
+static int blit_cycle_diagram_fill[][10] =
 {
-    { 0, 2, 0,0 },             /* 0 */
-    { 0, 3, 0,3,4 },           /* 1 */
-    { 0, 2, 3,0 },             /* 2 */
-    { 2, 3, 0,3,4, 3,0 },      /* 3 */
-    { 0, 3, 2,0,0 },           /* 4 */
-    { 3, 4, 0,2,0,4, 2,0,0 },  /* 5 */
-    { 0, 3, 2,3,0 },           /* 6 */
-    { 3, 4, 0,2,3,4, 2,3,0 },  /* 7 */
-    { 0, 2, 1,0 },             /* 8 */
-    { 3, 3, 1,0,4, 1,0,0},     /* 9 */
-    { 0, 2, 1,3 },             /* A */
+    { 0, 3, 0,5,0 },           /* 0 */
+    { 0, 3, 3,5,4 },           /* 1 */
+    { 0, 3, 0,3,0 },           /* 2 */
+    { 2, 3, 3,5,4, 3,0 },      /* 3 */
+    { 0, 4, 0,2,5,0 },         /* 4 */
+    { 3, 4, 0,2,5,4, 2,0,0 },  /* 5 */
+    { 0, 4, 2,3,5,0 },         /* 6 */
+    { 3, 4, 2,3,5,4, 2,3,0 },  /* 7 */
+    { 0, 3, 1,5,-1 },          /* 8 */
+    { 2, 3, 1,5,4, 1,0},       /* 9 */
+    { 0, 3, 1,3,5, },          /* A */
     { 3, 3, 1,3,4, 1,3,0 },    /* B */
-    { 2, 3, 0,1,2, 1,2 },      /* C */
-    { 4, 4, 1,2,0,4, 1,2,0,0 },        /* D */
-    { 0, 3, 1,2,3 },           /* E */
+    { 2, 4, 1,2,5,-1, 1,2 },   /* C */
+    { 3, 4, 1,2,5,4, 1,2,0 },  /* D */
+    { 0, 4, 1,2,3,-1 },                /* E */
     { 4, 4, 1,2,3,4, 1,2,3,0 } /* F */
 };
 
-static uae_u8 blit_cycle_diagram_line[] =
+/*
+
+    line draw takes 4 cycles (two 2-cycle blits combined)
+    it also have real idle cycles and only 2 dma fetches
+    (read from C, write to D, but see below)
+
+    Oddities:
+
+    - first word is written to address pointed by BLTDPT
+      but all following writes go to address pointed by BLTCPT!
+    - BLTDMOD is ignored by blitter (BLTCMOD is used)
+    - state of D-channel enable bit does not matter!
+    - disabling A-channel freezes the content of BPLAPT
+
+*/
+
+static int blit_cycle_diagram_line[] =
 {
-    0, 2, 0,4,0,4 /* total guess.. */
+    0, 4, 3,5,4,5  /* guessed */
 };
 
+static int blit_cycle_diagram_finald[] =
+    { 0, 2, 0,4 };
+
 void build_blitfilltable(void)
 {
     unsigned int d, fillmask;
@@ -179,10 +248,10 @@ STATIC_INLINE int channel_state (int cycles)
     return blit_diag[((cycles - blit_diag[0]) % blit_diag[1]) + 2];
 }
 
-int is_bitplane_dma (int hpos);
+extern int is_bitplane_dma (int hpos);
 STATIC_INLINE int canblit (int hpos)
 {
-    if ((cycle_line[hpos] == 0 || cycle_line[hpos] == CYCLE_NOCPU) && !is_bitplane_dma (hpos))
+    if (cycle_line[hpos] == 0 && !is_bitplane_dma (hpos))
        return 1;
     return 0;
 }
@@ -504,22 +573,23 @@ static void decide_blitter_line (int hpos)
     hpos++;
     if (dmaen (DMA_BLITTER)) {
        while (blit_last_hpos < hpos) {
-           int c = blit_cyclecounter % 2;
+           int c = channel_state (blit_cyclecounter);
            for (;;) {
-               if (blit_cyclecounter < 0) {
+
+               if (c == 5) {
                    blit_cyclecounter++;
                    break;
                }
 #if 1
-               if ((c == 0 || c == 1) && !canblit(blit_last_hpos))
+               if (c && !canblit(blit_last_hpos))
                    break;
 #endif
 #if 1
-               //if (c == 1)
+               if (c != 0)
                    cycle_line[blit_last_hpos] |= CYCLE_BLITTER;
 #endif
                blit_cyclecounter++;
-               if (c == 1) {
+               if (c == 4) {
                    /* believe it or not but try Cardamon or Cardamom without this.. */
                    if (ddat1use)
                        bltdpt = bltcpt;
@@ -742,11 +812,21 @@ void decide_blitter (int hpos)
            }
 #endif
            for (;;) {
-               if (c && !canblit (blit_last_hpos)) {
+               if (c == 5) { /* real idle cycle */
+                   blit_cyclecounter++;
+                   break;
+               }
+               
+               /* all cycles need free bus, even idle cycles (except fillmode idle) */
+               if (!canblit (blit_last_hpos)) {
                    blit_misscyclecounter++;
                    break;
                }
-               if (c == 4) {
+
+               if (c < 0) { /* no channel but bus still needs to be allocated.. */
+                   cycle_line[blit_last_hpos] |= CYCLE_BLITTER;
+                   blit_cyclecounter++;
+               } else if (c == 4) {
                    if (blitter_doddma ()) {
                        cycle_line[blit_last_hpos] |= CYCLE_BLITTER;
                        blit_cyclecounter++;
@@ -827,14 +907,10 @@ static void blit_bltset (int con)
     blitdesc = bltcon1 & 2;
     blit_ch = (bltcon0 & 0x0f00) >> 8;
 
-    blit_singlechannel = 0;
-    if (blit_ch == 0 || blit_ch == 1 || blit_ch == 2 || blit_ch == 4 || blit_ch == 8)
-       blit_singlechannel = 1;
     if (blitline) {
        if (blt_info.hblitsize != 2)
            write_log ("weird hblitsize in linemode: %d vsize=%d PC%=%x\n", blt_info.hblitsize, blt_info.vblitsize, m68k_getpc());
        blit_diag = blit_cycle_diagram_line;
-       blit_singlechannel = 1;
     } else {
        if (con & 2) {
            blitfc = !!(bltcon1 & 0x4);
@@ -888,7 +964,7 @@ void blit_modset (void)
     blit_modaddd = mult * blt_info.bltdmod;
 }
 
-void reset_blit        (int bltcon)
+void reset_blit (int bltcon)
 {
     if (bltstate == BLT_done)
        return;
@@ -897,7 +973,7 @@ void reset_blit     (int bltcon)
     blit_modset ();
 }
 
-void do_blitter        (int hpos)
+void do_blitter (int hpos)
 {
     int cycles;
 #ifdef BLITTER_DEBUG

diff --git a/compemu_raw_x86.c b/compemu_raw_x86.c
index 158db86c77ed05c4723f890fce6ef14ae993e3ef..aa331717560515faab14c050bb37646ceb333ec2 100755 (executable)
--- a/compemu_raw_x86.c
+++ b/compemu_raw_x86.c
@@ -2627,15 +2627,16 @@ static void raw_init_cpu(void)
   /* Have CMOV support? */
   have_cmov = c->x86_hwcap & (1 << 15);
 
+#if 0
   /* Can the host CPU suffer from partial register stalls? */
   have_rat_stall = (c->x86_vendor == X86_VENDOR_INTEL);
-#if 1
   /* It appears that partial register writes are a bad idea even on
         AMD K7 cores, even though they are not supposed to have the
         dreaded rat stall. Why? Anyway, that's why we lie about it ;-) */
   if (c->x86_processor == X86_PROCESSOR_ATHLON)
        have_rat_stall = 1;
 #endif
+  have_rat_stall = 1;
 
   /* Alignments */
   if (tune_alignment) {

diff --git a/custom.c b/custom.c
index 9bad294c0e7fe6b90ff41a5037ffb78a55fd36c3..48ba59679fbb0b47ba240898234a99e85824c31f 100755 (executable)
--- a/custom.c
+++ b/custom.c
@@ -3788,7 +3788,7 @@ STATIC_INLINE void do_sprites_1 (int num, int cycle, int hpos)
     if (cycle && !s->dmacycle)
        return; /* Superfrog intro flashing bee fix */
 
-    dma = hpos < plfstrt || diwstate != DIW_waiting_stop || !dmaen (DMA_BITPLANE);
+    dma = hpos < plfstrt || diwstate != DIW_waiting_stop;
     if (vpos == s->vstop || vpos == sprite_vblank_endline) {
        s->dmastate = 0;
        posctl = 1;
@@ -4962,8 +4962,8 @@ int REGPARAM2 custom_wput_1 (int hpos, uaecptr addr, uae_u32 value, int noget)
      case 0x1C6: if (hbstop != value) { hbstop = value; varsync (); } break;
      case 0x1C8: if (vtotal != value) { vtotal = value; varsync (); } break;
      case 0x1CA: if (vsstop != value) { vsstop = value; varsync (); } break;
-     case 0x1CC: if (vbstrt > value + 1 || vbstrt < value - 1) { vbstrt = value; varsync (); } break;
-     case 0x1CE: if (vbstop > value + 1 || vbstrt < value - 1) { vbstop = value; varsync (); } break;
+     case 0x1CC: if (vbstrt < value || vbstrt > value + 1) { vbstrt = value; varsync (); } break;
+     case 0x1CE: if (vbstop < value || vbstop > value + 1) { vbstop = value; varsync (); } break;
      case 0x1DE: if (hsstrt != value) { hsstrt = value; varsync (); } break;
      case 0x1E0: if (vsstrt != value) { vsstrt = value; varsync (); } break;
      case 0x1E2: if (hcenter != value) { hcenter = value; varsync (); } break;
@@ -5448,7 +5448,8 @@ STATIC_INLINE decide_fetch_ce (int hpos)
 
 STATIC_INLINE int dma_cycle(void)
 {
-    int hpos, cycles = 0, bnasty = 0;
+    int hpos, cycles = 0;
+    static int bnasty;
 
     for (;;) {
        int bpldma;
@@ -5460,19 +5461,21 @@ STATIC_INLINE int dma_cycle(void)
        decide_line (hpos);
        decide_fetch_ce (hpos);
        bpldma = is_bitplane_dma (hpos);
-       if (cycle_line[hpos] == 0 && !bpldma) {
-           if (bltstate == BLT_done || bnasty >= 3)
+       if (bltstate != BLT_done) {
+           if (!blitpri && bnasty >= 3 && !cycle_line[hpos] && !bpldma) {
+               bnasty = 0;
                break;
+           }
            decide_blitter (hpos);
-           if (cycle_line[hpos] == 0)
-               break;
-           if (!blitpri || blit_singlechannel)
-               bnasty++;
-       } else if (bpldma && (blit_singlechannel || !blitpri)) {
+           bnasty++;
+       } else if (bpldma || cycle_line[hpos]) {
            bnasty++;
        }
+       if (cycle_line[hpos] == 0 && !bpldma)
+           break;
        /* bus was allocated to dma channel, wait for next cycle.. */
     }
+    bnasty = 0;
     cycle_line[hpos] |= CYCLE_CPU;
     return cycles;
 }

diff --git a/gencpu.c b/gencpu.c
index 3181df4c192d430da9e59d829e612aa4765935f7..f3a0bc0e99c592f533b119acb935ca06feedd196 100755 (executable)
--- a/gencpu.c
+++ b/gencpu.c
@@ -1881,6 +1881,7 @@ static void gen_opcode (unsigned long int opcode)
            printf("\t\tSET_VFLG (1);\n");
            printf("\t\tif (dst < 0) SET_NFLG (1);\n");
        }
+       printf ("\t\tm68k_incpc (%d);\n", m68k_pc_offset);
        printf ("\t\tException (5, oldpc);\n");
        printf ("\t\tgoto %s;\n", endlabelstr);
        printf ("\t} else {\n");
@@ -1924,6 +1925,7 @@ static void gen_opcode (unsigned long int opcode)
            printf("\t\tSET_VFLG (1);\n");
            printf("\t\tSET_ZFLG (1);\n");
        }
+       printf ("\t\tm68k_incpc (%d);\n", m68k_pc_offset);
        printf ("\t\tException (5, oldpc);\n");
        printf ("\t\tgoto %s;\n", endlabelstr);
        printf ("\t} else {\n");

diff --git a/od-win32/ahidsound.c b/od-win32/ahidsound.c
index e49f3be98a0856d628902231ad5c9399da32ed63..1f43db69382b8413a1204bf921f5d963951f803b 100755 (executable)
--- a/od-win32/ahidsound.c
+++ b/od-win32/ahidsound.c
@@ -497,6 +497,7 @@ int ahi_open_sound (void)
 
 static void *bswap_buffer = NULL;
 static uae_u32 bswap_buffer_size = 0;
+static double syncdivisor;
 
 uae_u32 ahi_demux (void)
 {
@@ -701,10 +702,11 @@ uae_u32 ahi_demux (void)
            dllname = (char *)get_real_address ((uae_u32)dllname);
            result=(uae_u32) LoadLibrary(dllname);
            write_log("%s windows dll/alib loaded at %d (0 mean failure)\n",dllname,result); 
+           syncdivisor = (3580000.0 * CYCLE_UNIT) / (double)syncbase;
        return result;
        }
 
-       case 101:      //get dll label
+       case 101:       //get dll label
        {
            HMODULE m;
            char *funcname;
@@ -714,10 +716,24 @@ uae_u32 ahi_demux (void)
        return (uae_u32) GetProcAddress(m,funcname);
        }
 
-       case 102:      //execute native code
-           return emulib_ExecuteNativeCode2 ();
+       case 102:       //execute native code
+       {
+           uae_u32 ret;
+           unsigned long rate1;
+           double v;
+           rate1 = read_processor_time(); 
+           ret = emulib_ExecuteNativeCode2 ();
+           rate1 = read_processor_time() - rate1;
+           v = syncdivisor * rate1;
+           if (v > 0) {
+               if (v > 1000000 * CYCLE_UNIT)
+                   v = 1000000 * CYCLE_UNIT;
+               do_extra_cycles((unsigned long)(syncdivisor * rate1)); //compensate the time stay in native func
+           }
+           return ret;
+       }
 
-       case 103:      //close dll
+       case 103:       //close dll
        {
            HMODULE libaddr;
            libaddr = (HMODULE) m68k_dreg (regs, 1);

diff --git a/od-win32/avioutput.c b/od-win32/avioutput.c
index e15116cfbd966803852621a457c759e416816e09..02938efb780ba0d8c383eee0d70bace59082de35 100755 (executable)
--- a/od-win32/avioutput.c
+++ b/od-win32/avioutput.c
@@ -895,7 +895,6 @@ void AVIOutput_Initialize(void)
 void frame_drawn(void)
 {
     double diff, skipmode;
-    static int frame;
 
     if (!avioutput_video || !avioutput_enabled)
        return;

diff --git a/od-win32/win32.h b/od-win32/win32.h
index d3aed57117bb0a988b363cc5edee27309273e4c6..6854b165a951b55c94310383734c35fc995fd2b4 100755 (executable)
--- a/od-win32/win32.h
+++ b/od-win32/win32.h
@@ -22,7 +22,7 @@ extern int manual_palette_refresh_needed;
 extern int mouseactive, focus;
 extern int ignore_messages_all;
 #define WINUAEBETA 1
-#define WINUAEBETASTR " Beta 3"
+#define WINUAEBETASTR " Beta 5"
 
 extern char start_path_exe[MAX_DPATH];
 extern char start_path_data[MAX_DPATH];

diff --git a/od-win32/winuae_msvc/winuae_msvc.vcproj b/od-win32/winuae_msvc/winuae_msvc.vcproj
index 61b0450660c30449d4052269360c1490e0c69897..240ce5336d30bdd91c6e3bd78e0e6c5831c2a874 100755 (executable)
--- a/od-win32/winuae_msvc/winuae_msvc.vcproj
+++ b/od-win32/winuae_msvc/winuae_msvc.vcproj
@@ -5,7 +5,6 @@
        Name="winuae"
        ProjectGUID="{4ADAA943-1AC8-4FB5-82E5-4FB753B6C2DA}"
        RootNamespace="winuae"
-       SignManifests="true"
        >
        <Platforms>
                <Platform
@@ -259,7 +258,7 @@
                        <Tool
                                Name="VCCLCompilerTool"
                                Optimization="3"
-                               InlineFunctionExpansion="1"
+                               InlineFunctionExpansion="2"
                                EnableIntrinsicFunctions="true"
                                FavorSizeOrSpeed="1"
                                OmitFramePointers="true"