core, testpico fixes

diff --git a/pico/memory.c b/pico/memory.c
index 33e2aa5..3cc2fee 100644 (file)
--- a/pico/memory.c
+++ b/pico/memory.c
@@ -1039,10 +1039,24 @@ static void m68k_mem_setup(void)
 static int get_scanline(int is_from_z80)\r
 {\r
   if (is_from_z80) {\r
-    int mclk_z80 = (z80_cyclesLeft<0 ? Pico.t.z80c_aim : z80_cyclesDone()) * 15;\r
-    int mclk_line = Pico.t.z80_scanline * 488 * 7;\r
-    while (mclk_z80 - mclk_line >= 488 * 7)\r
-      Pico.t.z80_scanline++, mclk_line += 488 * 7;\r
+    // ugh... compute by dividing cycles since frame start by cycles per line\r
+    // need some fractional resolution here, else there may be an extra line\r
+    int cycles_line = cycles_68k_to_z80(488 << 8); // cycles per line, as Q8\r
+    int cycles_z80 = (z80_cyclesLeft<0 ? Pico.t.z80c_aim:z80_cyclesDone())<<8;\r
+    int cycles = cycles_line * Pico.t.z80_scanline;\r
+    // approximation by multiplying with inverse\r
+    if (cycles_z80 - cycles >= 2*cycles_line) {\r
+      // compute 1/cycles_line, storing the result to avoid future dividing\r
+      static int cycles_line_o, cycles_line_i;\r
+      if (cycles_line_o != cycles_line)\r
+        { cycles_line_o = cycles_line, cycles_line_i = (1<<22) / cycles_line; }\r
+      // compute lines = diff/cycles_line = diff*(1/cycles_line)\r
+      int lines = ((cycles_z80 - cycles) * cycles_line_i) >> 22;\r
+      Pico.t.z80_scanline += lines, cycles += cycles_line * lines;\r
+    }\r
+    // handle any rounding leftover\r
+    while (cycles_z80 - cycles >= cycles_line)\r
+      Pico.t.z80_scanline ++, cycles += cycles_line;\r
     return Pico.t.z80_scanline;\r
   }\r
 \r
@@ -1342,7 +1356,8 @@ void PicoWrite16_32x(u32 a, u32 d) {}
 static unsigned char z80_md_vdp_read(unsigned short a)\r
 {\r
   if ((a & 0xff00) == 0x7f00) {\r
-    z80_subCLeft(3);\r
+    static int f; f = (f&0xff) + 0x8c; // 0.6\r
+    z80_subCLeft(2+(f>>8)); // 3.3 per kabuto, but notaz' test implies 2.6 ?!?\r
     Pico.t.z80_buscycles += 7;\r
 \r
     switch (a & 0x0d)\r
@@ -1368,9 +1383,10 @@ static unsigned char z80_md_bank_read(unsigned short a)
   unsigned char ret;\r
 \r
   // account for 68K bus access on both CPUs.\r
+  static int f; f = (f&0xff) + 0x4c; // 0.3\r
+  z80_subCLeft(3+(f>>8)); // 3.3 per kabuto\r
   // don't use SekCyclesBurn(7) here since the Z80 doesn't run in cycle lock to\r
   // the 68K. Count the stolen cycles to be accounted later in the 68k CPU runs\r
-  z80_subCLeft(3);\r
   Pico.t.z80_buscycles += 7;\r
 \r
   addr68k = Pico.m.z80_bank68k << 15;\r
@@ -1413,9 +1429,10 @@ static void z80_md_bank_write(unsigned int a, unsigned char data)
   unsigned int addr68k;\r
 \r
   // account for 68K bus access on both CPUs.\r
+  static int f; f = (f&0xff) + 0x4c; // 0.3\r
+  z80_subCLeft(3+(f>>8)); // 3.3 per kabuto\r
   // don't use SekCyclesBurn(7) here since the Z80 doesn't run in cycle lock to\r
   // the 68K. Count the stolen cycles to be accounted later in the 68K CPU runs\r
-  z80_subCLeft(3);\r
   Pico.t.z80_buscycles += 7;\r
 \r
   addr68k = Pico.m.z80_bank68k << 15;\r

diff --git a/pico/pico_int.h b/pico/pico_int.h
index 3ed7c0f..f6618b9 100644 (file)
--- a/pico/pico_int.h
+++ b/pico/pico_int.h
@@ -205,7 +205,7 @@ extern struct DrZ80 drZ80;
   (Pico.t.z80c_aim - z80_cyclesLeft)\r
 \r
 // one line has 488 68K cycles and 228 Z80 cycles, 228/488*8192=3827\r
-#define cycles_68k_to_z80(x) ((x) * 3857 >> 13)\r
+#define cycles_68k_to_z80(x) ((x) * 3847 >> 13)\r
 \r
 // ----------------------- SH2 CPU -----------------------\r
 \r

diff --git a/pico/videoport.c b/pico/videoport.c
index c39a560..6fb47a7 100644 (file)
--- a/pico/videoport.c
+++ b/pico/videoport.c
@@ -17,7 +17,8 @@ enum { clkdiv = 2 };    // CPU clock granularity: one of 1,2,4,8
 \r
 // VDP Slot timing, taken from http://gendev.spritesmind.net/\r
 //     forum/viewtopic.php?f=22&t=851&sid=d5701a71396ee7f700c74fb7cd85cb09\r
-// Thank you very much for the great work, Nemesis!\r
+//     http://plutiedev.com/mirror/kabuto-hardware-notes\r
+// Thank you very much for the great work, Nemesis, Kabuto!\r
 \r
 // Slot clock is sysclock/20 for h32 and sysclock/16 for h40.\r
 // One scanline is 63.7us/63.5us (h32/h40) long which is 488.6/487.4 68k cycles.\r
@@ -30,13 +31,15 @@ enum { slcpu = 488 };
 // is generated also depends on the video mode.\r
 enum { hint32 = 0x85, gapstart32 = 0x94, gapend32 = 0xe9};\r
 enum { hint40 = 0xa5, gapstart40 = 0xb7, gapend40 = 0xe5};\r
+// XXX Kabuto says gapend40 is 0xe4, but then a line would've 211 slots, while\r
+// it's 210 in all other sources I looked at?\r
 \r
 // The horizontal sync period (HBLANK) is 30/37 slots (h32/h40):\r
 // h32: 4 slots front porch (1.49us), 13 HSYNC (4.84us), 13 back porch (4.84us)\r
 // h40: 5 slots front porch (1.49us), 16 HSYNC (4.77us), 16 back porch (4.77us)\r
-// HBLANK starts in slot 0x93/0xb4, according to Nemesis' measurements.\r
+// HBLANK starts in slot 0x93/0xb3 and ends after slot 0x05 (from Kabuto's doc)\r
 enum { hboff32 = 0x93-hint32, hblen32 = 0xf8-(gapend32-gapstart32)-hint32};//30\r
-enum { hboff40 = 0xb4-hint40, hblen40 = 0xf8-(gapend40-gapstart40)-hint40};//37\r
+enum { hboff40 = 0xb3-hint40, hblen40 = 0xf8-(gapend40-gapstart40)-hint40};//37\r
 \r
 // number of slots in a scanline\r
 #define slots32        (0x100-(gapend32-gapstart32)) // 171\r
@@ -1030,7 +1033,8 @@ static u32 VideoSr(const struct PicoVideo *pv)
 {\r
   unsigned int hp = pv->reg[12]&1 ? hboff40*488/slots40 : hboff32*488/slots32;\r
   unsigned int hl = pv->reg[12]&1 ? hblen40*488/slots40 : hblen32*488/slots32;\r
-  unsigned int c = SekCyclesDone() - Pico.t.m68c_line_start;\r
+  // XXX -2 is to please notaz' testpico, but why is this?\r
+  unsigned int c = SekCyclesDone()-2 - Pico.t.m68c_line_start;\r
   u32 d;\r
 \r
   PicoVideoFIFOSync(c);\r