X-Git-Url: http://shamusworld.gotdns.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fblitter.cpp;h=13960996e6e9acf7afe55f786ac1a2ef6bfede3a;hb=09b508a759ca512da810722a0d4786dd1695d47c;hp=34c4aaac81d8ad479372d10a1b2ceae6d2307040;hpb=21010310cfb50e9bd58f55ebecbf26b990c785bd;p=virtualjaguar diff --git a/src/blitter.cpp b/src/blitter.cpp index 34c4aaa..1396099 100644 --- a/src/blitter.cpp +++ b/src/blitter.cpp @@ -1,20 +1,38 @@ // // Blitter core // -// by James L. Hammons +// by James Hammons +// (C) 2010 Underground Software +// +// JLH = James Hammons +// +// Who When What +// --- ---------- ------------------------------------------------------------- +// JLH 01/16/2010 Created this log ;-) +// + // // I owe a debt of gratitude to Curt Vendel and to John Mathieson--to Curt // for supplying the Oberon ASIC nets and to John for making them available // to Curt. ;-) Without that excellent documentation which shows *exactly* // what's going on inside the TOM chip, we'd all still be guessing as to how // the wily blitter and other pieces of the Jaguar puzzle actually work. +// Now how about those JERRY ASIC nets gentlemen...? [We have those now!] ;-) // -#include "jaguar.h" #include "blitter.h" +#include +#include +#include +#include "jaguar.h" +#include "log.h" +//#include "memory.h" + // Various conditional compilation goodies... +//#define LOG_BLITS + //#define USE_ORIGINAL_BLITTER //#define USE_MIDSUMMER_BLITTER #define USE_MIDSUMMER_BLITTER_MKII @@ -47,39 +65,39 @@ void BlitterMidsummer2(void); // Blitter registers (offsets from F02200) -#define A1_BASE ((UINT32)0x00) -#define A1_FLAGS ((UINT32)0x04) -#define A1_CLIP ((UINT32)0x08) // Height and width values for clipping -#define A1_PIXEL ((UINT32)0x0C) // Integer part of the pixel (Y.i and X.i) -#define A1_STEP ((UINT32)0x10) // Integer part of the step -#define A1_FSTEP ((UINT32)0x14) // Fractional part of the step -#define A1_FPIXEL ((UINT32)0x18) // Fractional part of the pixel (Y.f and X.f) -#define A1_INC ((UINT32)0x1C) // Integer part of the increment -#define A1_FINC ((UINT32)0x20) // Fractional part of the increment -#define A2_BASE ((UINT32)0x24) -#define A2_FLAGS ((UINT32)0x28) -#define A2_MASK ((UINT32)0x2C) // Modulo values for x and y (M.y and M.x) -#define A2_PIXEL ((UINT32)0x30) // Integer part of the pixel (no fractional part for A2) -#define A2_STEP ((UINT32)0x34) // Integer part of the step (no fractional part for A2) -#define COMMAND ((UINT32)0x38) -#define PIXLINECOUNTER ((UINT32)0x3C) // Inner & outer loop values -#define SRCDATA ((UINT32)0x40) -#define DSTDATA ((UINT32)0x48) -#define DSTZ ((UINT32)0x50) -#define SRCZINT ((UINT32)0x58) -#define SRCZFRAC ((UINT32)0x60) -#define PATTERNDATA ((UINT32)0x68) -#define INTENSITYINC ((UINT32)0x70) -#define ZINC ((UINT32)0x74) -#define COLLISIONCTRL ((UINT32)0x78) -#define PHRASEINT0 ((UINT32)0x7C) -#define PHRASEINT1 ((UINT32)0x80) -#define PHRASEINT2 ((UINT32)0x84) -#define PHRASEINT3 ((UINT32)0x88) -#define PHRASEZ0 ((UINT32)0x8C) -#define PHRASEZ1 ((UINT32)0x90) -#define PHRASEZ2 ((UINT32)0x94) -#define PHRASEZ3 ((UINT32)0x98) +#define A1_BASE ((uint32)0x00) +#define A1_FLAGS ((uint32)0x04) +#define A1_CLIP ((uint32)0x08) // Height and width values for clipping +#define A1_PIXEL ((uint32)0x0C) // Integer part of the pixel (Y.i and X.i) +#define A1_STEP ((uint32)0x10) // Integer part of the step +#define A1_FSTEP ((uint32)0x14) // Fractional part of the step +#define A1_FPIXEL ((uint32)0x18) // Fractional part of the pixel (Y.f and X.f) +#define A1_INC ((uint32)0x1C) // Integer part of the increment +#define A1_FINC ((uint32)0x20) // Fractional part of the increment +#define A2_BASE ((uint32)0x24) +#define A2_FLAGS ((uint32)0x28) +#define A2_MASK ((uint32)0x2C) // Modulo values for x and y (M.y and M.x) +#define A2_PIXEL ((uint32)0x30) // Integer part of the pixel (no fractional part for A2) +#define A2_STEP ((uint32)0x34) // Integer part of the step (no fractional part for A2) +#define COMMAND ((uint32)0x38) +#define PIXLINECOUNTER ((uint32)0x3C) // Inner & outer loop values +#define SRCDATA ((uint32)0x40) +#define DSTDATA ((uint32)0x48) +#define DSTZ ((uint32)0x50) +#define SRCZINT ((uint32)0x58) +#define SRCZFRAC ((uint32)0x60) +#define PATTERNDATA ((uint32)0x68) +#define INTENSITYINC ((uint32)0x70) +#define ZINC ((uint32)0x74) +#define COLLISIONCTRL ((uint32)0x78) +#define PHRASEINT0 ((uint32)0x7C) +#define PHRASEINT1 ((uint32)0x80) +#define PHRASEINT2 ((uint32)0x84) +#define PHRASEINT3 ((uint32)0x88) +#define PHRASEZ0 ((uint32)0x8C) +#define PHRASEZ1 ((uint32)0x90) +#define PHRASEZ2 ((uint32)0x94) +#define PHRASEZ3 ((uint32)0x98) // Blitter command bits @@ -142,34 +160,34 @@ void BlitterMidsummer2(void); //Put 'em back, once we fix the problem!!! [KO] // 1 bpp pixel read #define PIXEL_SHIFT_1(a) (((~a##_x) >> 16) & 7) -#define PIXEL_OFFSET_1(a) (((((UINT32)a##_y >> 16) * a##_width / 8) + (((UINT32)a##_x >> 19) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 19) & 7)) +#define PIXEL_OFFSET_1(a) (((((uint32)a##_y >> 16) * a##_width / 8) + (((uint32)a##_x >> 19) & ~7)) * (1 + a##_pitch) + (((uint32)a##_x >> 19) & 7)) #define READ_PIXEL_1(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a), BLITTER) >> PIXEL_SHIFT_1(a)) & 0x01) //#define READ_PIXEL_1(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_1(a)) >> PIXEL_SHIFT_1(a)) & 0x01) // 2 bpp pixel read #define PIXEL_SHIFT_2(a) (((~a##_x) >> 15) & 6) -#define PIXEL_OFFSET_2(a) (((((UINT32)a##_y >> 16) * a##_width / 4) + (((UINT32)a##_x >> 18) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 18) & 7)) +#define PIXEL_OFFSET_2(a) (((((uint32)a##_y >> 16) * a##_width / 4) + (((uint32)a##_x >> 18) & ~7)) * (1 + a##_pitch) + (((uint32)a##_x >> 18) & 7)) #define READ_PIXEL_2(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a), BLITTER) >> PIXEL_SHIFT_2(a)) & 0x03) //#define READ_PIXEL_2(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_2(a)) >> PIXEL_SHIFT_2(a)) & 0x03) // 4 bpp pixel read #define PIXEL_SHIFT_4(a) (((~a##_x) >> 14) & 4) -#define PIXEL_OFFSET_4(a) (((((UINT32)a##_y >> 16) * (a##_width/2)) + (((UINT32)a##_x >> 17) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 17) & 7)) +#define PIXEL_OFFSET_4(a) (((((uint32)a##_y >> 16) * (a##_width/2)) + (((uint32)a##_x >> 17) & ~7)) * (1 + a##_pitch) + (((uint32)a##_x >> 17) & 7)) #define READ_PIXEL_4(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a), BLITTER) >> PIXEL_SHIFT_4(a)) & 0x0f) //#define READ_PIXEL_4(a) ((JaguarReadByte(a##_addr+PIXEL_OFFSET_4(a)) >> PIXEL_SHIFT_4(a)) & 0x0f) // 8 bpp pixel read -#define PIXEL_OFFSET_8(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~7)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 7)) +#define PIXEL_OFFSET_8(a) (((((uint32)a##_y >> 16) * a##_width) + (((uint32)a##_x >> 16) & ~7)) * (1 + a##_pitch) + (((uint32)a##_x >> 16) & 7)) #define READ_PIXEL_8(a) (JaguarReadByte(a##_addr+PIXEL_OFFSET_8(a), BLITTER)) //#define READ_PIXEL_8(a) (JaguarReadByte(a##_addr+PIXEL_OFFSET_8(a))) // 16 bpp pixel read -#define PIXEL_OFFSET_16(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~3)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 3)) +#define PIXEL_OFFSET_16(a) (((((uint32)a##_y >> 16) * a##_width) + (((uint32)a##_x >> 16) & ~3)) * (1 + a##_pitch) + (((uint32)a##_x >> 16) & 3)) #define READ_PIXEL_16(a) (JaguarReadWord(a##_addr+(PIXEL_OFFSET_16(a)<<1), BLITTER)) //#define READ_PIXEL_16(a) (JaguarReadWord(a##_addr+(PIXEL_OFFSET_16(a)<<1))) // 32 bpp pixel read -#define PIXEL_OFFSET_32(a) (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1)) +#define PIXEL_OFFSET_32(a) (((((uint32)a##_y >> 16) * a##_width) + (((uint32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((uint32)a##_x >> 16) & 1)) #define READ_PIXEL_32(a) (JaguarReadLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), BLITTER)) //#define READ_PIXEL_32(a) (JaguarReadLong(a##_addr+(PIXEL_OFFSET_32(a)<<2))) @@ -195,25 +213,25 @@ void BlitterMidsummer2(void); //#define WRITE_ZDATA_16(a,d) { JaguarWriteWord(a##_addr+(ZDATA_OFFSET_16(a)<<1), d); } // z data write -#define WRITE_ZDATA(a,f,d) WRITE_ZDATA_16(a,d); +#define WRITE_ZDATA(a,f,d) WRITE_ZDATA_16(a,d); // 1 bpp r data read -#define READ_RDATA_1(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 19) & 0x04))) >> (((UINT32)a##_x >> 16) & 0x1F)) & 0x0001 : (REG(r) & 0x0001)) +#define READ_RDATA_1(r,a,p) ((p) ? ((REG(r+(((uint32)a##_x >> 19) & 0x04))) >> (((uint32)a##_x >> 16) & 0x1F)) & 0x0001 : (REG(r) & 0x0001)) // 2 bpp r data read -#define READ_RDATA_2(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 18) & 0x04))) >> (((UINT32)a##_x >> 15) & 0x3E)) & 0x0003 : (REG(r) & 0x0003)) +#define READ_RDATA_2(r,a,p) ((p) ? ((REG(r+(((uint32)a##_x >> 18) & 0x04))) >> (((uint32)a##_x >> 15) & 0x3E)) & 0x0003 : (REG(r) & 0x0003)) // 4 bpp r data read -#define READ_RDATA_4(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 17) & 0x04))) >> (((UINT32)a##_x >> 14) & 0x28)) & 0x000F : (REG(r) & 0x000F)) +#define READ_RDATA_4(r,a,p) ((p) ? ((REG(r+(((uint32)a##_x >> 17) & 0x04))) >> (((uint32)a##_x >> 14) & 0x28)) & 0x000F : (REG(r) & 0x000F)) // 8 bpp r data read -#define READ_RDATA_8(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 16) & 0x04))) >> (((UINT32)a##_x >> 13) & 0x18)) & 0x00FF : (REG(r) & 0x00FF)) +#define READ_RDATA_8(r,a,p) ((p) ? ((REG(r+(((uint32)a##_x >> 16) & 0x04))) >> (((uint32)a##_x >> 13) & 0x18)) & 0x00FF : (REG(r) & 0x00FF)) // 16 bpp r data read -#define READ_RDATA_16(r,a,p) ((p) ? ((REG(r+(((UINT32)a##_x >> 15) & 0x04))) >> (((UINT32)a##_x >> 12) & 0x10)) & 0xFFFF : (REG(r) & 0xFFFF)) +#define READ_RDATA_16(r,a,p) ((p) ? ((REG(r+(((uint32)a##_x >> 15) & 0x04))) >> (((uint32)a##_x >> 12) & 0x10)) & 0xFFFF : (REG(r) & 0xFFFF)) // 32 bpp r data read -#define READ_RDATA_32(r,a,p) ((p) ? REG(r+(((UINT32)a##_x >> 14) & 0x04)) : REG(r)) +#define READ_RDATA_32(r,a,p) ((p) ? REG(r+(((uint32)a##_x >> 14) & 0x04)) : REG(r)) // register data read #define READ_RDATA(r,a,f,p) (\ @@ -246,8 +264,8 @@ void BlitterMidsummer2(void); //#define WRITE_PIXEL_16(a,d) { JaguarWriteWord(a##_addr+(PIXEL_OFFSET_16(a)<<1), d); if (specialLog) WriteLog("Pixel write address: %08X\n", a##_addr+(PIXEL_OFFSET_16(a)<<1)); } // 32 bpp pixel write -#define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d, BLITTER); } -//#define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d); } +#define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d, BLITTER); } +//#define WRITE_PIXEL_32(a,d) { JaguarWriteLong(a##_addr+(PIXEL_OFFSET_32(a)<<2), d); } // pixel write #define WRITE_PIXEL(a,f,d) {\ @@ -267,8 +285,8 @@ void BlitterMidsummer2(void); // as a floating point bit pattern being followed by a number of zeroes. So, e.g., 001101 translates to // 1.01 (the "1." being implied) x (2 ^ 3) or 1010 -> 10 in base 10 (i.e., 1.01 with the decimal place // being shifted to the right 3 places). -/*static uint32 blitter_scanline_width[48] = -{ +/*static uint32 blitter_scanline_width[48] = +{ 0, 0, 0, 0, // Note: This would really translate to 1, 1, 1, 1 2, 0, 0, 0, 4, 0, 6, 0, @@ -351,7 +369,7 @@ void blitter_generic(uint32 cmd) { /* Blit! (0018FA70 <- 008DDC40) count: 2 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -2 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -427,7 +445,7 @@ if (specialLog) srczdata = READ_RDATA(SRCZINT, a2, REG(A2_FLAGS), a2_phrase_mode); } - // load dst data and Z + // load dst data and Z if (DSTEN) { dstdata = READ_PIXEL(a1, REG(A1_FLAGS)); @@ -453,14 +471,14 @@ if (specialLog) inhibit = 1; }//*/ - if (GOURZ) + if (GOURZ) srczdata = z_i[colour_index] >> 16; // apply z comparator if (Z_OP_INF && srczdata < dstzdata) inhibit = 1; if (Z_OP_EQU && srczdata == dstzdata) inhibit = 1; if (Z_OP_SUP && srczdata > dstzdata) inhibit = 1; - + // apply data comparator // Note: DCOMPEN only works in 8/16 bpp modes! !!! FIX !!! // Does BCOMPEN only work in 1 bpp mode??? @@ -490,7 +508,7 @@ if (specialLog) Interesting (Hover Strike--large letter): Blit! (0018FA70 <- 008DDC40) count: 2 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -2 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -498,7 +516,7 @@ Blit! (0018FA70 <- 008DDC40) count: 2 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: A1 x/y: 100/12, A2 x/y: 106/0 Pattern: 000000F300000000 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -8 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -506,7 +524,7 @@ Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: A1 x/y: 102/12, A2 x/y: 107/0 Pattern: 000000F300000000 Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -1 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -514,7 +532,7 @@ Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: A1 x/y: 118/12, A2 x/y: 70/0 Pattern: 000000F300000000 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -8 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -522,7 +540,7 @@ Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: A1 x/y: 119/12, A2 x/y: 71/0 Pattern: 000000F300000000 Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -1 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -530,7 +548,7 @@ Blit! (0018FA70 <- 008DDC40) count: 1 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: A1 x/y: 127/12, A2 x/y: 66/0 Pattern: 000000F300000000 Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: 1401060C] - CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN + CMD -> src: SRCENX dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: BCOMPEN BKGWREN A1 step values: -8 (X), 1 (Y) A2 step values: -1 (X), 1 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 1 phrases, depth: 8bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -545,7 +563,7 @@ Blit! (0018FA70 <- 008DDC40) count: 8 x 13, A1/2_FLAGS: 00014218/00013C18 [cmd: // compare source pixel with pattern pixel /* Blit! (000B8250 <- 0012C3A0) count: 16 x 1, A1/2_FLAGS: 00014420/00012000 [cmd: 05810001] - CMD -> src: SRCEN dst: misc: a1ctl: mode: ity: PATDSEL z-op: op: LFU_REPLACE ctrl: BCOMPEN + CMD -> src: SRCEN dst: misc: a1ctl: mode: ity: PATDSEL z-op: op: LFU_REPLACE ctrl: BCOMPEN A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 384 (22), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 1bpp, z-off: 0, width: 16 (10), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD x/y: 0/20 @@ -592,12 +610,12 @@ Blit! (000B8250 <- 0012C3A0) count: 16 x 1, A1/2_FLAGS: 00014420/00012000 [cmd: // compute the write data and store if (!inhibit) - { + { // Houston, we have a problem... // Look here, at PATDSEL and GOURD. If both are active (as they are on the BIOS intro), then there's // a conflict! E.g.: //Blit! (00100000 <- 000095D0) count: 3 x 1, A1/2_FLAGS: 00014220/00004020 [cmd: 00011008] -// CMD -> src: dst: DSTEN misc: a1ctl: mode: GOURD ity: PATDSEL z-op: op: LFU_CLEAR ctrl: +// CMD -> src: dst: DSTEN misc: a1ctl: mode: GOURD ity: PATDSEL z-op: op: LFU_CLEAR ctrl: // A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD // A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 256 (20), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD // A1 x/y: 90/171, A2 x/y: 808/0 Pattern: 776D770077007700 @@ -630,7 +648,7 @@ Blit! (000B8250 <- 0012C3A0) count: 16 x 1, A1/2_FLAGS: 00014420/00012000 [cmd: Hover Strike ADDDSEL blit: Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cmd: 00020208] - CMD -> src: dst: DSTEN misc: a1ctl: UPDA1 mode: ity: ADDDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: DSTEN misc: a1ctl: UPDA1 mode: ity: ADDDSEL z-op: op: LFU_CLEAR ctrl: A1 step values: -320 (X), 1 (Y) A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 256 (20), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -675,10 +693,10 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm //According to JTRM, this is part of the four things the blitter does with the write data (the other //three being PATDSEL, ADDDSEL, and LFU (default). I'm not sure which gets precedence, this or PATDSEL //(see above blit example)... - if (GOURD) + if (GOURD) writedata = ((gd_c[colour_index]) << 8) | (gd_i[colour_index] >> 16); - if (SRCSHADE) + if (SRCSHADE) { int intensity = srcdata & 0xFF; int ia = gd_ia >> 16; @@ -705,7 +723,7 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm /*if (((REG(A1_FLAGS) >> 3) & 0x07) == 5) { uint32 offset = a1_addr+(PIXEL_OFFSET_32(a1)<<2); -// (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1)) +// (((((uint32)a##_y >> 16) * a##_width) + (((uint32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((uint32)a##_x >> 16) & 1)) if ((offset >= 0x1FF020 && offset <= 0x1FF03F) || (offset >= 0x1FF820 && offset <= 0x1FF83F)) WriteLog("32bpp pixel write: A1 Phrase mode --> "); }//*/ @@ -733,7 +751,7 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm srczdata = READ_RDATA(SRCZINT, a1, REG(A1_FLAGS), a1_phrase_mode); } - // load dst data and Z + // load dst data and Z if (DSTEN) { dstdata = READ_PIXEL(a2, REG(A2_FLAGS)); @@ -749,14 +767,14 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm dstzdata = READ_RDATA(DSTZ, a2, REG(A2_FLAGS), a2_phrase_mode); } - if (GOURZ) + if (GOURZ) srczdata = z_i[colour_index] >> 16; // apply z comparator if (Z_OP_INF && srczdata < dstzdata) inhibit = 1; if (Z_OP_EQU && srczdata == dstzdata) inhibit = 1; if (Z_OP_SUP && srczdata > dstzdata) inhibit = 1; - + // apply data comparator //NOTE: The bit comparator (BCOMPEN) is NOT the same at the data comparator! if (DCOMPEN | BCOMPEN) @@ -795,7 +813,7 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm // if (a1_phrase_mode || a2_phrase_mode) // inhibit = !inhibit; } - + if (CLIPA1) { inhibit |= (((a1_x >> 16) < a1_clip_x && (a1_x >> 16) >= 0 @@ -804,7 +822,7 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm // compute the write data and store if (!inhibit) - { + { if (PATDSEL) { // use pattern data for write data @@ -833,10 +851,10 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm writedata |= srcdata & dstdata; } - if (GOURD) + if (GOURD) writedata = ((gd_c[colour_index]) << 8) | (gd_i[colour_index] >> 16); - if (SRCSHADE) + if (SRCSHADE) { int intensity = srcdata & 0xFF; int ia = gd_ia >> 16; @@ -861,7 +879,7 @@ Blit! (00098D90 <- 0081DDC0) count: 320 x 287, A1/2_FLAGS: 00004220/00004020 [cm /*if (logGo) { uint32 offset = a2_addr+(PIXEL_OFFSET_16(a2)<<1); -// (((((UINT32)a##_y >> 16) * a##_width) + (((UINT32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((UINT32)a##_x >> 16) & 1)) +// (((((uint32)a##_y >> 16) * a##_width) + (((uint32)a##_x >> 16) & ~1)) * (1 + a##_pitch) + (((uint32)a##_x >> 16) & 1)) WriteLog("[%08X:%04X] ", offset, writedata); }//*/ // write to the destination @@ -940,7 +958,7 @@ Below fixes it, but then borks: ; O Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -15 (X), 1 (Y) A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1048,8 +1066,8 @@ Lesse, with pre-add we'd have: a2_y += a2_step_y;//*/ #endif } - - // write values back to registers + + // write values back to registers WREG(A1_PIXEL, (a1_y & 0xFFFF0000) | ((a1_x >> 16) & 0xFFFF)); WREG(A1_FPIXEL, (a1_y << 16) | (a1_x & 0xFFFF)); WREG(A2_PIXEL, (a2_y & 0xFFFF0000) | ((a2_x >> 16) & 0xFFFF)); @@ -1085,7 +1103,7 @@ void blitter_blit(uint32 cmd) a1_zoffs = (REG(A1_FLAGS) >> 6) & 7; a2_zoffs = (REG(A2_FLAGS) >> 6) & 7; - + xadd_a1_control = (REG(A1_FLAGS) >> 16) & 0x03; xadd_a2_control = (REG(A2_FLAGS) >> 16) & 0x03; @@ -1107,7 +1125,7 @@ void blitter_blit(uint32 cmd) // a1_width = blitter_scanline_width[((REG(A1_FLAGS) & 0x00007E00) >> 9)]; // According to JTRM, this must give a *whole number* of phrases in the current // pixel size (this means the lookup above is WRONG)... !!! FIX !!! - UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; + uint32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; a1_width = ((0x04 | m) << e) >> 2;//*/ a2_x = (REG(A2_PIXEL) & 0x0000FFFF) << 16; @@ -1154,7 +1172,7 @@ void blitter_blit(uint32 cmd) // add pixelsize (1) to X a1_xadd = 1 << 16; break; - case XADD0: + case XADD0: // add zero (for those nice vertical lines) a1_xadd = 0; break; @@ -1170,7 +1188,7 @@ void blitter_blit(uint32 cmd) // A1 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 128 (1C), addctl: XADDINC YADD1 XSIGNADD YSIGNADD // A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 320 (21), addctl: XADD0 YADD1 XSIGNADD YSIGNADD //if (YADD1_A1 && YADD1_A2 && xadd_a2_control == XADD0 && xadd_a1_control == XADDINC)// && -// UINT32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS); +// uint32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS); //Ok, so this ISN'T it... Prolly the XADDPHR code above that's doing it... //if (REG(A1_FLAGS) == 0x00073820 && REG(A2_FLAGS) == 0x00064220 && cmd == 0x41802801) // A1 x/y: 14368/7, A2 x/y: 150/36 @@ -1198,7 +1216,7 @@ void blitter_blit(uint32 cmd) // add pixelsize (1) to X a2_xadd = 1 << 16; break; - case XADD0: + case XADD0: // add zero (for those nice vertical lines) a2_xadd = 0; break; @@ -1276,7 +1294,7 @@ WriteLog("BLIT: Asked to use invalid bit combo (XADDINC) for A2...\n"); | ((uint32)blitter_ram[SRCDATA + 0] << 8) | blitter_ram[SRCDATA + 1]; gouraud_add = REG(INTENSITYINC); - + gd_ia = gouraud_add & 0x00FFFFFF; if (gd_ia & 0x00800000) gd_ia = 0xFF000000 | gd_ia; @@ -1346,7 +1364,7 @@ WriteLog("BLIT: Asked to use invalid bit combo (XADDINC) for A2...\n"); WriteLog(" GOURZ = %i\n",GOURZ); WriteLog(" GOURD = %i\n",GOURD); WriteLog(" SRCSHADE= %i\n",SRCSHADE); - } + } #endif //NOTE: Pitch is ignored! @@ -1356,24 +1374,24 @@ WriteLog("BLIT: Asked to use invalid bit combo (XADDINC) for A2...\n"); //Black is short by 3, pink is short by 1... /* Blit! (00110000 <- 000BF010) count: 9 x 31, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD A1 x/y: 262/124, A2 x/y: 128/0 Blit! (00110000 <- 000BF010) count: 5 x 38, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD A1 x/y: 264/117, A2 x/y: 407/0 Blit! (00110000 <- 000BF010) count: 9 x 23, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 step values: -10 (X), 1 (Y) A1 -> pitch: 4(2) phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1(0) phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD A1 x/y: 262/132, A2 x/y: 129/0 Blit! (00110000 <- 000BF010) count: 5 x 27, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 step values: -8 (X), 1 (Y) A1 -> pitch: 4(2) phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1(0) phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -1392,7 +1410,7 @@ Fixed! Now for more: ; This looks like the ship icon in the upper left corner... Blit! (00110000 <- 0010B2A8) count: 11 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -12 (X), 1 (Y) A2 step values: 0 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1409,7 +1427,7 @@ Actually, if you look at the A1 step values, there IS a discrepancy! ; D Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -14 (X), 1 (Y) A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1420,7 +1438,7 @@ Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: ; E Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -13 (X), 1 (Y) A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1430,7 +1448,7 @@ Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: ; M Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -12 (X), 1 (Y) A2 step values: 0 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1440,7 +1458,7 @@ Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: ; O Blit! (00110000 <- 0010B2A8) count: 12 x 12, A1/2_FLAGS: 000042E2/00000020 [cmd: 09800609] - CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN + CMD -> src: SRCEN dst: DSTEN misc: a1ctl: UPDA1 UPDA2 mode: ity: z-op: op: LFU_REPLACE ctrl: DCOMPEN A1 step values: -15 (X), 1 (Y) A2 step values: -4 (X), 0 (Y) [mask (unused): 00000000 - FFFFFFFF/FFFFFFFF] A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD @@ -1458,14 +1476,14 @@ if (blit_start_log) uint32 /*src = cmd & 0x07, dst = (cmd >> 3) & 0x07, misc = (cmd >> 6) & 0x03, a1ctl = (cmd >> 8) & 0x07,*/ mode = (cmd >> 11) & 0x07/*, ity = (cmd >> 14) & 0x0F, zop = (cmd >> 18) & 0x07, op = (cmd >> 21) & 0x0F, ctrl = (cmd >> 25) & 0x3F*/; - UINT32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS); + uint32 a1f = REG(A1_FLAGS), a2f = REG(A2_FLAGS); uint32 p1 = a1f & 0x07, p2 = a2f & 0x07, d1 = (a1f >> 3) & 0x07, d2 = (a2f >> 3) & 0x07, zo1 = (a1f >> 6) & 0x07, zo2 = (a2f >> 6) & 0x07, w1 = (a1f >> 9) & 0x3F, w2 = (a2f >> 9) & 0x3F, ac1 = (a1f >> 16) & 0x1F, ac2 = (a2f >> 16) & 0x1F; - UINT32 iw1 = ((0x04 | (w1 & 0x03)) << ((w1 & 0x3C) >> 2)) >> 2; - UINT32 iw2 = ((0x04 | (w2 & 0x03)) << ((w2 & 0x3C) >> 2)) >> 2; + uint32 iw1 = ((0x04 | (w1 & 0x03)) << ((w1 & 0x3C) >> 2)) >> 2; + uint32 iw2 = ((0x04 | (w2 & 0x03)) << ((w2 & 0x3C) >> 2)) >> 2; WriteLog("Blit! (%08X %s %08X) count: %d x %d, A1/2_FLAGS: %08X/%08X [cmd: %08X]\n", a1_addr, (mode&0x01 ? "->" : "<-"), a2_addr, n_pixels, n_lines, a1f, a2f, cmd); // WriteLog(" CMD -> src: %d, dst: %d, misc: %d, a1ctl: %d, mode: %d, ity: %1X, z-op: %d, op: %1X, ctrl: %02X\n", src, dst, misc, a1ctl, mode, ity, zop, op, ctrl); @@ -1519,17 +1537,17 @@ if (blit_start_log) ********************** STUFF CUT ABOVE THIS LINE! ****************************** *******************************************************************************/ -void blitter_init(void) +void BlitterInit(void) { - blitter_reset(); + BlitterReset(); } -void blitter_reset(void) +void BlitterReset(void) { memset(blitter_ram, 0x00, 0xA0); } -void blitter_done(void) +void BlitterDone(void) { WriteLog("BLIT: Done.\n"); } @@ -1541,10 +1559,17 @@ uint8 BlitterReadByte(uint32 offset, uint32 who/*=UNKNOWN*/) // status register //This isn't cycle accurate--how to fix? !!! FIX !!! //Probably have to do some multi-threaded implementation or at least a reentrant safe implementation... +//Real hardware returns $00000805, just like the JTRM says. + if (offset == (0x38 + 0)) + return 0x00; + if (offset == (0x38 + 1)) + return 0x00; + if (offset == (0x38 + 2)) + return 0x08; if (offset == (0x38 + 3)) - return 0x01; // always idle + return 0x05; // always idle/never stopped (collision detection ignored!) -// CHECK HERE ONCE THIS FIX HAS BEEN TESTED: [ ] +// CHECK HERE ONCE THIS FIX HAS BEEN TESTED: [X] //Fix for AvP: if (offset >= 0x04 && offset <= 0x07) //This is it. I wonder if it just ignores the lower three bits? @@ -1596,12 +1621,12 @@ void BlitterWriteByte(uint32 offset, uint8 data, uint32 who/*=UNKNOWN*/) case 0x81: blitter_ram[PATTERNDATA + 5] = data; break; case 0x82: blitter_ram[SRCDATA + 4] = data; break; case 0x83: blitter_ram[SRCDATA + 5] = data; break; - + case 0x84: break; case 0x85: blitter_ram[PATTERNDATA + 3] = data; break; case 0x86: blitter_ram[SRCDATA + 2] = data; break; case 0x87: blitter_ram[SRCDATA + 3] = data; break; - + case 0x88: break; case 0x89: blitter_ram[PATTERNDATA + 1] = data; break; case 0x8A: blitter_ram[SRCDATA + 0] = data; break; @@ -1618,12 +1643,12 @@ void BlitterWriteByte(uint32 offset, uint8 data, uint32 who/*=UNKNOWN*/) case 0x91: blitter_ram[SRCZINT + 5] = data; break; case 0x92: blitter_ram[SRCZFRAC + 4] = data; break; case 0x93: blitter_ram[SRCZFRAC + 5] = data; break; - + case 0x94: blitter_ram[SRCZINT + 2] = data; break; case 0x95: blitter_ram[SRCZINT + 3] = data; break; case 0x96: blitter_ram[SRCZFRAC + 2] = data; break; case 0x97: blitter_ram[SRCZFRAC + 3] = data; break; - + case 0x98: blitter_ram[SRCZINT + 0] = data; break; case 0x99: blitter_ram[SRCZINT + 1] = data; break; case 0x9A: blitter_ram[SRCZFRAC + 0] = data; break; @@ -1724,13 +1749,13 @@ doGPUDis = true; void LogBlit(void) { - char * opStr[16] = { "LFU_CLEAR", "LFU_NSAND", "LFU_NSAD", "LFU_NOTS", "LFU_SAND", "LFU_NOTD", "LFU_N_SXORD", "LFU_NSORND", + const char * opStr[16] = { "LFU_CLEAR", "LFU_NSAND", "LFU_NSAD", "LFU_NOTS", "LFU_SAND", "LFU_NOTD", "LFU_N_SXORD", "LFU_NSORND", "LFU_SAD", "LFU_XOR", "LFU_D", "LFU_NSORD", "LFU_REPLACE", "LFU_SORND", "LFU_SORD", "LFU_ONE" }; uint32 cmd = GET32(blitter_ram, 0x38); - UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; - UINT32 a1_width = ((0x04 | m) << e) >> 2; + uint32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; + uint32 a1_width = ((0x04 | m) << e) >> 2; m = (REG(A2_FLAGS) >> 9) & 0x03, e = (REG(A2_FLAGS) >> 11) & 0x0F; - UINT32 a2_width = ((0x04 | m) << e) >> 2; + uint32 a2_width = ((0x04 | m) << e) >> 2; WriteLog("Blit!\n"); WriteLog(" COMMAND = %08X\n", cmd); @@ -1809,9 +1834,9 @@ void LogBlit(void) WriteLog(" UPDA2 = %s\n", (UPDA2 ? "1" : "0")); WriteLog(" DSTA2 = %s\n", (DSTA2 ? "1" : "0")); WriteLog(" ZOP = %s %s %s\n", (Z_OP_INF ? "<" : ""), (Z_OP_EQU ? "=" : ""), (Z_OP_SUP ? ">" : "")); - WriteLog("--LFUFUNC = %s\n", opStr[(cmd >> 21) & 0x0F]); + WriteLog("+-LFUFUNC = %s\n", opStr[(cmd >> 21) & 0x0F]); WriteLog("| PATDSEL = %s (PD=%08X%08X)\n", (PATDSEL ? "1" : "0"), REG(PATTERNDATA), REG(PATTERNDATA + 4)); - WriteLog("--ADDDSEL = %s\n", (ADDDSEL ? "1" : "0")); + WriteLog("+-ADDDSEL = %s\n", (ADDDSEL ? "1" : "0")); WriteLog(" CMPDST = %s\n", (CMPDST ? "1" : "0")); WriteLog(" BCOMPEN = %s\n", (BCOMPEN ? "1" : "0")); WriteLog(" DCOMPEN = %s\n", (DCOMPEN ? "1" : "0")); @@ -1838,6 +1863,9 @@ void LogBlit(void) void BlitterMidsummer(uint32 cmd) { +#ifdef LOG_BLITS + LogBlit(); +#endif uint32 outer_loop, inner_loop, a1_addr, a2_addr; int32 a1_x, a1_y, a2_x, a2_y, a1_width, a2_width; uint8 a1_phrase_mode, a2_phrase_mode; @@ -1846,7 +1874,7 @@ uint8 a1_phrase_mode, a2_phrase_mode; a2_addr = REG(A2_BASE) & 0xFFFFFFF8; a1_x = (REG(A1_PIXEL) << 16) | (REG(A1_FPIXEL) & 0xFFFF); a1_y = (REG(A1_PIXEL) & 0xFFFF0000) | (REG(A1_FPIXEL) >> 16); - UINT32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; + uint32 m = (REG(A1_FLAGS) >> 9) & 0x03, e = (REG(A1_FLAGS) >> 11) & 0x0F; a1_width = ((0x04 | m) << e) >> 2;//*/ a2_x = (REG(A2_PIXEL) & 0x0000FFFF) << 16; a2_y = (REG(A2_PIXEL) & 0xFFFF0000); @@ -2009,7 +2037,7 @@ sread: // Source data read //pointing at. Likewise, the pixel (if in BPP 1, 2 & 4, chopped) otherwise. It probably still //transfers an entire phrase even in pixel mode. //Odd thought: Does it expand, e.g., 1 BPP pixels into 32 BPP internally? Hmm... -//No. +//No. /* a1_addr = REG(A1_BASE) & 0xFFFFFFF8; a2_addr = REG(A2_BASE) & 0xFFFFFFF8; @@ -2030,7 +2058,7 @@ sread: // Source data read a1_width = ((0x04 | m) << e) >> 2; a2_width = ((0x04 | m) << e) >> 2; - // write values back to registers + // write values back to registers WREG(A1_PIXEL, (a1_y & 0xFFFF0000) | ((a1_x >> 16) & 0xFFFF)); WREG(A1_FPIXEL, (a1_y << 16) | (a1_x & 0xFFFF)); WREG(A2_PIXEL, (a2_y & 0xFFFF0000) | ((a2_x >> 16) & 0xFFFF)); @@ -2245,7 +2273,7 @@ Blit! } // Figure out what gets written... - + if (PATDSEL) { writeData = GET64(blitter_ram, PATTERNDATA); @@ -2268,7 +2296,7 @@ Blit! else // LFUFUNC is the default... { writeData = 0; - + if (LFU_NAN) writeData |= ~srcData & ~dstData; if (LFU_NA) @@ -2370,7 +2398,7 @@ inhibitWrite://Should this go here? or on the other side of the X/Y incrementing a2_x += (blitter_ram[A2_FLAGS + 1] & 0x08 ? -1 << 16 : 1 << 16); /* else if ((blitter_ram[A2_FLAGS + 1] & 0x03) == 2) a2_x += 0 << 16; */ - + if (blitter_ram[A2_FLAGS + 1] & 0x04) a2_y += (blitter_ram[A2_FLAGS + 1] & 0x10 ? -1 << 16 : 1 << 16); @@ -2475,10 +2503,10 @@ a1fupdate A1 step fraction is added to A1 pointer fraction goto a1update */ /* -#define A1_PIXEL ((UINT32)0x0C) // Integer part of the pixel (Y.i and X.i) -#define A1_STEP ((UINT32)0x10) // Integer part of the step -#define A1_FSTEP ((UINT32)0x14) // Fractional part of the step -#define A1_FPIXEL ((UINT32)0x18) // Fractional part of the pixel (Y.f and X.f) +#define A1_PIXEL ((uint32)0x0C) // Integer part of the pixel (Y.i and X.i) +#define A1_STEP ((uint32)0x10) // Integer part of the step +#define A1_FSTEP ((uint32)0x14) // Fractional part of the step +#define A1_FPIXEL ((uint32)0x18) // Fractional part of the pixel (Y.f and X.f) */ // This is all kinda murky. All we have are the Midsummer docs to give us any guidance, @@ -2646,7 +2674,8 @@ void DATA(uint64 &wdata, uint8 &dcomp, uint8 &zcomp, bool &nowrite, bool big_pix, bool cmpdst, uint8 daddasel, uint8 daddbsel, uint8 daddmode, bool daddq_sel, uint8 data_sel, uint8 dbinh, uint8 dend, uint8 dstart, uint64 dstd, uint32 iinc, uint8 lfu_func, uint64 &patd, bool patdadd, bool phrase_mode, uint64 srcd, bool srcdread, bool srczread, bool srcz2add, uint8 zmode, - bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize); + bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize, + uint64 &srcz, uint64 dstz, uint32 zinc); void COMP_CTRL(uint8 &dbinh, bool &nowrite, bool bcompen, bool big_pix, bool bkgwren, uint8 dcomp, bool dcompen, uint8 icount, uint8 pixsize, bool phrase_mode, uint8 srcd, uint8 zcomp); @@ -2655,6 +2684,9 @@ bool logBlit = false; void BlitterMidsummer2(void) { +#ifdef LOG_BLITS + LogBlit(); +#endif // Here's what the specs say the state machine does. Note that this can probably be // greatly simplified (also, it's different from what John has in his Oberon docs): //Will remove stuff that isn't in Jaguar I once fully described (stuff like texture won't @@ -2662,13 +2694,10 @@ void BlitterMidsummer2(void) uint32 cmd = GET32(blitter_ram, COMMAND); -// $01800005 has SRCENX, may have to investigate further... -// $00011008 has GOURD & DSTEN. -// $41802F41 has SRCSHADE, CLIPA1 -/*logBlit = false; +logBlit = false; if ( cmd != 0x00010200 && // PATDSEL - cmd != 0x01800001 + cmd != 0x01800001 // SRCEN LFUFUNC=C && cmd != 0x01800005 //Boot ROM ATARI letters: && cmd != 0x00011008 // DSTEN GOURD PATDSEL @@ -2683,7 +2712,7 @@ if ( //Static pic on title screen: && cmd != 0x01800601 // SRCEN UPDA1 UPDA2 LFUFUNC=C //Turning letters on Cybermorph intro screen: - && cmd != 0x09800F41 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 LFUFUNC=C DCOMPEN +// && cmd != 0x09800F41 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 LFUFUNC=C DCOMPEN && cmd != 0x00113078 // DSTEN DSTENZ DSTWRZ CLIP_A1 GOURD GOURZ PATDSEL ZMODE=4 && cmd != 0x09900F39 // SRCEN DSTEN DSTENZ DSTWRZ UPDA1 UPDA1F UPDA2 DSTA2 ZMODE=4 LFUFUNC=C DCOMPEN && cmd != 0x09800209 // SRCEN DSTEN UPDA1 LFUFUNC=C DCOMPEN @@ -2692,14 +2721,25 @@ if ( && cmd != 0x00010000 // PATDSEL //Hover Strike text: && cmd != 0x1401060C // SRCENX DSTEN UPDA1 UPDA2 PATDSEL BCOMPEN BKGWREN +//Hover Strike 3D stuff + && cmd != 0x01902839 // SRCEN DSTEN DSTENZ DSTWRZ DSTA2 GOURZ ZMODE=4 LFUFUNC=C +//Hover Strike darkening on intro to play (briefing) screen + && cmd != 0x00020208 // DSTEN UPDA1 ADDDSEL //Trevor McFur stuff: && cmd != 0x05810601 // SRCEN UPDA1 UPDA2 PATDSEL BCOMPEN && cmd != 0x01800201 // SRCEN UPDA1 LFUFUNC=C //T2K: && cmd != 0x00011000 // GOURD PATDSEL && cmd != 0x00011040 // CLIP_A1 GOURD PATDSEL +//Checkered flag: + && cmd != 0x01800000 // LFUFUNC=C + && cmd != 0x01800401 // + && cmd != 0x01800040 // + && cmd != 0x00020008 // +// && cmd != 0x09800F41 // SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 DSTA2 LFUFUNC=C DCOMPEN ) logBlit = true;//*/ +//logBlit = true; if (blit_start_log == 0) // Wait for the signal... logBlit = false;//*/ /* @@ -2709,6 +2749,17 @@ logBlit = F, cmd = 00011000 logBlit = F, cmd = 00011040 logBlit = F, cmd = 01800005 * logBlit = F, cmd = 09800741 * + +Hover Strike mission selection screen: +Blit! (CMD = 01902839) // SRCEN DSTEN DSTENZ DSTWRZ DSTA2 GOURZ ZMODE=4 LFUFUNC=C + +Checkered Flag blits in the screw up zone: +Blit! (CMD = 01800001) // SRCEN LFUFUNC=C +Blit! (CMD = 01800000) // LFUFUNC=C +Blit! (CMD = 00010000) // PATDSEL + +Wolfenstein 3D in the fuckup zone: +Blit! (CMD = 01800000) // LFUFUNC=C */ //printf("logBlit = %s, cmd = %08X\n", (logBlit ? "T" : "F"), cmd); @@ -2781,7 +2832,7 @@ fflush(stdout); #endif // Lines that don't exist in Jaguar I (and will never be asserted) - + bool polygon = false, datinit = false, a1_stepld = false, a2_stepld = false, ext_int = false; bool istepadd = false, istepfadd = false, finneradd = false, inneradd = false; bool zstepfadd = false, zstepadd = false; @@ -2906,7 +2957,7 @@ if ((cmd == 0x00010200) && (GET16(blitter_ram, PIXLINECOUNTER + 2) == 9)) ; Pink altimeter bar Blit! (00110000 <- 000BF010) count: 9 x 23, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 step values: -10 (X), 1 (Y) A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -2917,7 +2968,7 @@ Blit! (00110000 <- 000BF010) count: 9 x 23, A1/2_FLAGS: 000042E2/00010020 [cmd: ; Black altimeter bar Blit! (00110000 <- 000BF010) count: 5 x 29, A1/2_FLAGS: 000042E2/00010020 [cmd: 00010200] - CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: + CMD -> src: dst: misc: a1ctl: UPDA1 mode: ity: PATDSEL z-op: op: LFU_CLEAR ctrl: A1 step values: -8 (X), 1 (Y) A1 -> pitch: 4 phrases, depth: 16bpp, z-off: 3, width: 320 (21), addctl: XADDPHR YADD0 XSIGNADD YSIGNADD A2 -> pitch: 1 phrases, depth: 16bpp, z-off: 0, width: 1 (00), addctl: XADDPIX YADD0 XSIGNADD YSIGNADD @@ -2961,7 +3012,7 @@ Flags: UPDA1 PATDSEL */ // Bugs in Jaguar I - + a2addy = a1addy; // A2 channel Y add bit is tied to A1's //if (logBlit && (ocount > 20)) logBlit = false; @@ -2987,7 +3038,7 @@ printf(" srcz1=%08X%08X srcz2=%08X%08X dstz=%08X%08X zinc=%08X, coll=%X\n", (uint32)(srcz2 >> 32), (uint32)(srcz2 & 0xFFFFFFFF), (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF), zinc, collision); } -#endif +#endif // Various state lines set up by user @@ -3008,7 +3059,7 @@ fflush(stdout); while (true) { // IDLE - + if ((idle && !go) || (inner && outer0 && indone)) { #ifdef VERBOSE_BLITTER_LOGGING @@ -3020,11 +3071,13 @@ fflush(stdout); #endif idlei = true; - return; +//Instead of a return, let's try breaking out of the loop... +break; +// return; } else idlei = false; - + // INNER LOOP ACTIVE /* Entering DWRITE state... (icount=0000, inc=4) @@ -3035,7 +3088,7 @@ fflush(stdout); Now: [in=F a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] */ - + if ((idle && go && !datinit) || (inner && !indone) || (inner && indone && !outer0 && !upda1f && !upda1 && notgzandp && !upda2 && !datinit) @@ -3049,9 +3102,9 @@ Now: } else inneri = false; - + // A1 FRACTION UPDATE - + if (inner && indone && !outer0 && upda1f) { a1fupdatei = true; @@ -3098,9 +3151,9 @@ Now: } else a2updatei = false; - + // INITIALIZE INTENSITY FRACTION - + if ((zupdate && !upda2 && datinit) || (a1update && !upda2 && datinit && notgzandp) || (inner && indone && !outer0 && !upda1f && !upda1 && notgzandp && !upda2 && datinit) @@ -3111,34 +3164,34 @@ Now: } else init_ifi = false; - + // INITIALIZE INTENSITY INTEGER - + if (init_if) { init_iii = true; } else init_iii = false; - + // INITIALIZE Z FRACTION - + if (init_ii && gourz) { init_zfi = true; } else init_zfi = false; - + // INITIALIZE Z INTEGER - + if (init_zf) { init_zii = true; } else init_zii = false; - + // Here we move the fooi into their foo counterparts in order to simulate the moving // of data into the various FDSYNCs... Each time we loop we simulate one clock cycle... @@ -3376,7 +3429,7 @@ sshftld = idle_inner; bool dsta_addi = (dwritei && !dstwrz) || dzwritei; bool gensrc = sreadxi || szreadxi || sreadi || szreadi; - bool gendst = dreadi || szreadi || dwritei || dzwritei; + bool gendst = dreadi || dzreadi || dwritei || dzwritei; bool gena2i = (gensrc && !dsta2) || (gendst && dsta2); bool zaddr = szreadx || szread || dzread || dzwrite; @@ -3576,14 +3629,14 @@ daddbsel |= (istepadd && istepfadd && zstepadd && zstepfadd ? 0x08 : 0x00); /* Data adder mode control 000 16-bit normal add 001 16-bit saturating add with carry -010 8-bit saturating add with carry, carry into top byte is +010 8-bit saturating add with carry, carry into top byte is inhibited (YCrCb) -011 8-bit saturating add with carry, carry into top byte and +011 8-bit saturating add with carry, carry into top byte and between top nybbles is inhibited (CRY) 100 16-bit normal add with carry 101 16-bit saturating add 110 8-bit saturating add, carry into top byte is inhibited -111 8-bit saturating add, carry into top byte and between top +111 8-bit saturating add, carry into top byte and between top nybbles is inhibited The first five are used for Gouraud calculations, the latter three @@ -3602,7 +3655,7 @@ Bit 0 = dzwrite . gourz . atick[1] + init_ii . /topnen . /topben . /ext_int + init_ii . topnen . topben . /ext_int + init_zi - + Bit 1 = dwrite . gourd . atick[1] . /topben . /ext_int + istepadd . /topben . /ext_int + /gourd . /gourz . /topben @@ -3627,8 +3680,8 @@ daddmode |= ((dwrite && gourd && !topben && !ext_int) || (istepadd && !topben && || (init_ii && !topben && !ext_int) ? 0x02 : 0x00); daddmode |= ((!gourd && !gourz) || shadeadd || (dwrite && gourd && ext_int) || (istepadd && ext_int) || (init_ii && ext_int) ? 0x04 : 0x00); -/* Data add load controls -Pattern fraction (dest data) is loaded on +/* Data add load controls +Pattern fraction (dest data) is loaded on dwrite . gourd . atick[0] + istepfadd . /datinit + init_if @@ -3636,11 +3689,11 @@ Pattern data is loaded on dwrite . gourd . atick[1] + istepadd . /datinit . /datinit + init_ii -Source z1 is loaded on +Source z1 is loaded on dzwrite . gourz . atick[1] + zstepadd . /datinit . /datinit + init_zi -Source z2 is loaded on +Source z2 is loaded on dzwrite . gourz . atick[0] + zstepfadd + init_zf @@ -3681,17 +3734,17 @@ if (!justify) /* Generate source alignment shift ------------------------------- -The source alignment shift for data move is the difference between -the source and destination X pointers, multiplied by the pixel -size. Only the low six bits of the pointers are of interest, as -pixel sizes are always a power of 2 and window rows are always -phrase aligned. +The source alignment shift for data move is the difference between +the source and destination X pointers, multiplied by the pixel +size. Only the low six bits of the pointers are of interest, as +pixel sizes are always a power of 2 and window rows are always +phrase aligned. When not in phrase mode, the top 3 bits of the shift value are set to zero (2/26). Source shifting is also used to extract bits for bit-to-byte -expansion in phrase mode. This involves only the bottom three +expansion in phrase mode. This involves only the bottom three bits of the shift value, and is based on the offset within the phrase of the destination X pointer, in pixels. @@ -3732,18 +3785,26 @@ fflush(stdout); //ADDRGEN(srcAddr, pixAddr, gena2i, zaddr, // a1_x, a1_y, a1_base, a1_pitch, a1_pixsize, a1_width, a1_zoffset, // a2_x, a2_y, a2_base, a2_pitch, a2_pixsize, a2_width, a2_zoffset); -srcd2 = srcd1; -srcd1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER); + srcd2 = srcd1; + srcd1 = ((uint64)JaguarReadLong(address + 0, BLITTER) << 32) + | (uint64)JaguarReadLong(address + 4, BLITTER); //Kludge to take pixel size into account... +//Hmm. If we're not in phrase mode, this is most likely NOT going to be used... +//Actually, it would be--because of BCOMPEN expansion, for example... if (!phrase_mode) { - if (pixsize == 5) - srcd1 >>= 32; - else if (pixsize == 4) - srcd1 >>= 48; - else + if (bcompen) srcd1 >>= 56; -} + else + { + if (pixsize == 5) + srcd1 >>= 32; + else if (pixsize == 4) + srcd1 >>= 48; + else + srcd1 >>= 56; + } +}//*/ #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { @@ -3759,9 +3820,19 @@ fflush(stdout); #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Entering SZREADX state...\n"); +printf(" Entering SZREADX state..."); fflush(stdout); } +#endif + srcz2 = srcz1; + srcz1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER); +#ifdef VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf(" Src Z extra read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, + (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF)); + fflush(stdout); +} #endif } @@ -3799,7 +3870,7 @@ if (!phrase_mode) #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Source read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, +printf(" Source read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, (uint32)(srcd1 >> 32), (uint32)(srcd1 & 0xFFFFFFFF)); fflush(stdout); } @@ -3811,9 +3882,23 @@ fflush(stdout); #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Entering SZREAD state...\n"); +printf(" Entering SZREAD state..."); fflush(stdout); } +#endif + srcz2 = srcz1; + srcz1 = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER); +//Kludge to take pixel size into account... I believe that it only has to take 16BPP mode into account. Not sure tho. +if (!phrase_mode && pixsize == 4) + srcz1 >>= 48; + +#ifdef VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf(" Src Z read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, + (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF)); + fflush(stdout); +} #endif } @@ -3822,7 +3907,7 @@ fflush(stdout); #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Entering DREAD state...\n"); +printf(" Entering DREAD state..."); fflush(stdout); } #endif @@ -3844,7 +3929,7 @@ if (!phrase_mode) #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Dest read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, +printf(" Dest read address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, (uint32)(dstd >> 32), (uint32)(dstd & 0xFFFFFFFF)); fflush(stdout); } @@ -3862,6 +3947,10 @@ if (logBlit) } #endif dstz = ((uint64)JaguarReadLong(address, BLITTER) << 32) | (uint64)JaguarReadLong(address + 4, BLITTER); +//Kludge to take pixel size into account... I believe that it only has to take 16BPP mode into account. Not sure tho. +if (!phrase_mode && pixsize == 4) + dstz >>= 48; + #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { @@ -3872,13 +3961,19 @@ if (logBlit) #endif } +// These vars should probably go further up in the code... !!! FIX !!! +// We can't preassign these unless they're static... +//uint64 srcz = 0; // These are assigned to shut up stupid compiler warnings--dwrite is ALWAYS asserted +//bool winhibit = false; +uint64 srcz; +bool winhibit; //NOTE: SRCSHADE requires GOURZ to be set to work properly--another Jaguar I bug if (dwrite) { #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Entering DWRITE state...\n"); +printf(" Entering DWRITE state..."); fflush(stdout); } #endif @@ -3995,11 +4090,8 @@ uint8 pwidth = (((dend | dstart) & 0x07) == 0 ? 0x08 : (dend - dstart) & 0x07); #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Dest write address/pix address: %08X/%1X", address, pixAddr); -printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]", dstart, dend, pwidth, srcshift); -fflush(stdout); -printf("[daas=%X dabs=%X dam=%X ds=%X daq=%s]", daddasel, daddbsel, daddmode, data_sel, (daddq_sel ? "T" : "F")); -fflush(stdout); + printf(" Dest write address/pix address: %08X/%1X", address, pixAddr); + fflush(stdout); } #endif @@ -4021,6 +4113,61 @@ uint64 srcd = (srcd2 << (64 - srcshift)) | (srcd1 >> srcshift); if (srcshift == 0) srcd = srcd1; +//NOTE: This only works with pixel sizes less than 8BPP... +//DOUBLE NOTE: Still need to do regression testing to ensure that this doesn't break other stuff... !!! CHECK !!! +if (!phrase_mode && srcshift != 0) + srcd = ((srcd2 & 0xFF) << (8 - srcshift)) | ((srcd1 & 0xFF) >> srcshift); + +//Z DATA() stuff done here... And it has to be done before any Z shifting... +//Note that we need to have phrase mode start/end support here... (Not since we moved it from dzwrite...!) +/* +Here are a couple of Cybermorph blits with Z: +$00113078 // DSTEN DSTENZ DSTWRZ CLIP_A1 GOURD GOURZ PATDSEL ZMODE=4 +$09900F39 // SRCEN DSTEN DSTENZ DSTWRZ UPDA1 UPDA1F UPDA2 DSTA2 ZMODE=4 LFUFUNC=C DCOMPEN + +We're having the same phrase mode overwrite problem we had with the pixels... !!! FIX !!! +Odd. It's equating 0 with 0... Even though ZMODE is $04 (less than)! +*/ +if (gourz) +{ +/* +void ADDARRAY(uint16 * addq, uint8 daddasel, uint8 daddbsel, uint8 daddmode, + uint64 dstd, uint32 iinc, uint8 initcin[], uint64 initinc, uint16 initpix, + uint32 istep, uint64 patd, uint64 srcd, uint64 srcz1, uint64 srcz2, + uint32 zinc, uint32 zstep) +*/ + uint16 addq[4]; + uint8 initcin[4] = { 0, 0, 0, 0 }; + ADDARRAY(addq, 7/*daddasel*/, 6/*daddbsel*/, 0/*daddmode*/, 0, 0, initcin, 0, 0, 0, 0, 0, srcz1, srcz2, zinc, 0); + srcz2 = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0]; + ADDARRAY(addq, 6/*daddasel*/, 7/*daddbsel*/, 1/*daddmode*/, 0, 0, initcin, 0, 0, 0, 0, 0, srcz1, srcz2, zinc, 0); + srcz1 = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0]; + +#if 0//def VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf("\n[srcz1=%08X%08X, srcz2=%08X%08X, zinc=%08X", + (uint32)(srcz1 >> 32), (uint32)(srcz1 & 0xFFFFFFFF), + (uint32)(srcz2 >> 32), (uint32)(srcz2 & 0xFFFFFFFF), zinc); + fflush(stdout); +} +#endif +} + +uint8 zSrcShift = srcshift & 0x30; +srcz = (srcz2 << (64 - zSrcShift)) | (srcz1 >> zSrcShift); +//bleh, ugly ugly ugly +if (zSrcShift == 0) + srcz = srcz1; + +#if 0//def VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf(" srcz=%08X%08X]\n", (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF)); + fflush(stdout); +} +#endif + //When in SRCSHADE mode, it adds the IINC to the read source (from LFU???) //According to following line, it gets LFU mode. But does it feed the source into the LFU //after the add? @@ -4042,7 +4189,7 @@ if (srcshade) //Temporary kludge, to see if the fractional pattern does anything... //This works, BTW //But it seems to mess up in Cybermorph... the shading should be smooth but it isn't... -//Seems the carry out is lost again... !!! FIX !!! +//Seems the carry out is lost again... !!! FIX !!! [DONE--see below] if (patfadd) { uint16 addq[4]; @@ -4053,15 +4200,17 @@ if (patfadd) //Note that we still don't take atick[0] & [1] into account here, so this will skip half of the data needed... !!! FIX !!! //Not yet enumerated: dbinh, srcdread, srczread +//Also, should do srcshift on the z value in phrase mode... !!! FIX !!! [DONE] +//As well as add a srcz variable we can set external to this state... !!! FIX !!! [DONE] -bool winhibit;// = false; uint64 wdata; uint8 dcomp, zcomp; DATA(wdata, dcomp, zcomp, winhibit, true, cmpdst, daddasel, daddbsel, daddmode, daddq_sel, data_sel, 0/*dbinh*/, dend, dstart, dstd, iinc, lfufunc, patd, patdadd, phrase_mode, srcd, false/*srcdread*/, false/*srczread*/, srcz2add, zmode, - bcompen, bkgwren, dcompen, icount & 0x07, pixsize); + bcompen, bkgwren, dcompen, icount & 0x07, pixsize, + srcz, dstz, zinc); /* Seems that the phrase mode writes with DCOMPEN and DSTEN are corrupting inside of DATA: !!! FIX !!! It's fairly random as well. 7CFE -> 7DFE, 7FCA -> 78CA, 7FA4 -> 78A4, 7F88 -> 8F88 @@ -4143,6 +4292,7 @@ A1_outside := OR6 (a1_outside, a1_x{15}, a1xgr, a1xeq, a1_y{15}, a1ygr, a1yeq); */ //NOTE: There seems to be an off-by-one bug here in the clip_a1 section... !!! FIX !!! // Actually, seems to be related to phrase mode writes... +// Or is it? Could be related to non-15-bit compares as above? if (clip_a1 && ((a1_x & 0x8000) || (a1_y & 0x8000) || (a1_x >= a1_win_x) || (a1_y >= a1_win_y))) winhibit = true; @@ -4162,36 +4312,129 @@ if (!winhibit) else JaguarWriteByte(address, wdata & 0x000000FF, BLITTER); } +} #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { - printf(" [%08X%08X]", (uint32)(wdata >> 32), (uint32)(wdata & 0xFFFFFFFF)); + printf(" [%08X%08X]%s", (uint32)(wdata >> 32), (uint32)(wdata & 0xFFFFFFFF), (winhibit ? "[X]" : "")); + printf(" (icount=%04X, inc=%u)\n", icount, (uint16)inc); + printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]", dstart, dend, pwidth, srcshift); + printf("[daas=%X dabs=%X dam=%X ds=%X daq=%s]\n", daddasel, daddbsel, daddmode, data_sel, (daddq_sel ? "T" : "F")); fflush(stdout); } #endif -} + } + if (dzwrite) + { +// OK, here's the big insight: When NOT in GOURZ mode, srcz1 & 2 function EXACTLY the same way that +// srcd1 & 2 work--there's an implicit shift from srcz1 to srcz2 whenever srcz1 is read. +// OTHERWISE, srcz1 is the integer for the computed Z and srcz2 is the fractional part. +// Writes to srcz1 & 2 follow the same pattern as the other 64-bit registers--low 32 at the low address, +// high 32 at the high address (little endian!). +// NOTE: GOURZ is still not properly supported. Check patd/patf handling... +// Phrase mode start/end masks are not properly supported either... #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" (icount=%04X, inc=%u)\n", icount, (uint16)inc); -fflush(stdout); + printf(" Entering DZWRITE state..."); + printf(" Dest Z write address/pix address: %08X/%1X [%08X%08X]\n", address, pixAddr, + (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF)); + fflush(stdout); } #endif - } - - if (dzwrite) - { +//This is not correct... !!! FIX !!! +//Should be OK now... We'll see... +//Nope. Having the same starstep write problems in phrase mode as we had with pixels... !!! FIX !!! +//This is not causing the problem in Hover Strike... :-/ +//The problem was with the SREADX not shifting. Still problems with Z comparisons & other text in pregame screen... +if (!winhibit) +{ + if (phrase_mode) + { + JaguarWriteLong(address + 0, srcz >> 32, BLITTER); + JaguarWriteLong(address + 4, srcz & 0xFFFFFFFF, BLITTER); + } + else + { + if (pixsize == 4) + JaguarWriteWord(address, srcz & 0x0000FFFF, BLITTER); + } +}//*/ #ifdef VERBOSE_BLITTER_LOGGING if (logBlit) { -printf(" Entering DZWRITE state...\n"); -fflush(stdout); +// printf(" [%08X%08X]\n", (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF)); +// fflush(stdout); +//printf(" [dstart=%X dend=%X pwidth=%X srcshift=%X]", dstart, dend, pwidth, srcshift); + printf(" [dstart=? dend=? pwidth=? srcshift=%X]", srcshift); + printf("[daas=%X dabs=%X dam=%X ds=%X daq=%s]\n", daddasel, daddbsel, daddmode, data_sel, (daddq_sel ? "T" : "F")); + fflush(stdout); } #endif } +/* +This is because the address generator was using only 15 bits of the X when it should have +used 16! + +There's a slight problem here: The X pointer isn't wrapping like it should when it hits +the edge of the window... Notice how the X isn't reset at the edge of the window: + +Blit! (CMD = 00010000) +Flags: PATDSEL + count = 160 x 261 + a1_base = 000E8008, a2_base = 0001FA68 + a1_x = 0000, a1_y = 0000, a1_frac_x = 0000, a1_frac_y = 0000, a2_x = 0000, a2_y = 0000 + a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000 + a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000 + a1_win_x = 0000, a1_win_y = 0000, a2_mask_x = 0000, a2_mask_y = 0000 + a2_mask=F a1add=+phr/+0 a2add=+phr/+0 + a1_pixsize = 5, a2_pixsize = 5 + srcd=7717771777177717 dstd=0000000000000000 patd=7730773077307730 iinc=00000000 + srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0 + Phrase mode is ON + [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] + Entering INNER state... + Entering DWRITE state... Dest write address/pix address: 000E8008/0 [7730773077307730] (icount=009E, inc=2) + srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00] +[srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF] + [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] + Entering A1_ADD state [a1_x=0000, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + Entering DWRITE state... Dest write address/pix address: 000E8018/0 [7730773077307730] (icount=009C, inc=2) + srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00] +[srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF] + [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] + Entering A1_ADD state [a1_x=0002, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + +... + + Entering A1_ADD state [a1_x=009C, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + Entering DWRITE state... Dest write address/pix address: 000E84F8/0 [7730773077307730] (icount=0000, inc=2) + srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00] +[srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF] + [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] + Entering A1_ADD state [a1_x=009E, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + Entering IDLE_INNER state... + + Leaving INNER state... (ocount=0104) + [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] + + Entering INNER state... + Entering DWRITE state... Dest write address/pix address: 000E8508/0 [7730773077307730] (icount=009E, inc=2) + srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00] +[srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF] + [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] + Entering A1_ADD state [a1_x=00A0, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + Entering DWRITE state... Dest write address/pix address: 000E8518/0 [7730773077307730] (icount=009C, inc=2) + srcz=0000000000000000][dcomp=AA zcomp=00 dbinh=00] +[srcz=0000000000000000 dstz=0000000000000000 zwdata=0000000000000000 mask=7FFF] + [dstart=0 dend=40 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=0 daq=F] + Entering A1_ADD state [a1_x=00A2, a1_y=0000, addasel=0, addbsel=0, modx=1, addareg=F, adda_xconst=1, adda_yconst=0]... + +*/ + if (a1_add) { #ifdef VERBOSE_BLITTER_LOGGING @@ -4351,7 +4594,7 @@ cause the inner state to go active */ //Since we don't get here until the inner loop is finished (indone = true) we can get //away with doing it here...! ocount--; - + if (ocount == 0) outer0 = true; #ifdef VERBOSE_BLITTER_LOGGING @@ -4399,7 +4642,7 @@ fflush(stdout); } #endif } - + if (a2update) { #ifdef VERBOSE_BLITTER_LOGGING @@ -4421,122 +4664,86 @@ fflush(stdout); } } +// We never get here! !!! FIX !!! + +#ifdef VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf("Done!\na1_x=%04X a1_y=%04X a1_frac_x=%04X a1_frac_y=%04X a2_x=%04X a2_y%04X\n", + GET16(blitter_ram, A1_PIXEL + 2), + GET16(blitter_ram, A1_PIXEL + 0), + GET16(blitter_ram, A1_FPIXEL + 2), + GET16(blitter_ram, A1_FPIXEL + 0), + GET16(blitter_ram, A2_PIXEL + 2), + GET16(blitter_ram, A2_PIXEL + 0)); + fflush(stdout); +} +#endif + // Write values back to registers (in real blitter, these are continuously updated) - SET16(blitter_ram, A1_PIXEL + 0, a1_y); SET16(blitter_ram, A1_PIXEL + 2, a1_x); - SET16(blitter_ram, A1_FPIXEL + 0, a1_frac_y); + SET16(blitter_ram, A1_PIXEL + 0, a1_y); SET16(blitter_ram, A1_FPIXEL + 2, a1_frac_x); - SET16(blitter_ram, A2_PIXEL + 0, a2_y); + SET16(blitter_ram, A1_FPIXEL + 0, a1_frac_y); SET16(blitter_ram, A2_PIXEL + 2, a2_x); -} + SET16(blitter_ram, A2_PIXEL + 0, a2_y); +#ifdef VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf("Writeback!\na1_x=%04X a1_y=%04X a1_frac_x=%04X a1_frac_y=%04X a2_x=%04X a2_y%04X\n", + GET16(blitter_ram, A1_PIXEL + 2), + GET16(blitter_ram, A1_PIXEL + 0), + GET16(blitter_ram, A1_FPIXEL + 2), + GET16(blitter_ram, A1_FPIXEL + 0), + GET16(blitter_ram, A2_PIXEL + 2), + GET16(blitter_ram, A2_PIXEL + 0)); + fflush(stdout); +} +#endif +} /* -The latest that doesn't work properly: - -Blit! (CMD = 09800741) -Flags: SRCEN CLIP_A1 UPDA1 UPDA1F UPDA2 LFUFUNC=C DCOMPEN - count = 15 x 18 - a1_base = 00050000, a2_base = 0083F400 - a1_x = 003D, a1_y = 00AD, a1_frac_x = 8000, a1_frac_y = 0000, a2_x = 0027, a2_y = 00A4 - a1_step_x = FFF1, a1_step_y = 0001, a1_stepf_x = 0000, a1_stepf_y = 0000, a2_step_x = FFF1, a2_step_y = 0001 - a1_inc_x = 0001, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000 - a1_win_x = 0180, a1_win_y = 0118, a2_mask_x = 0000, a2_mask_y = 0000 - a2_mask=F a1add=+inc/+0 a2add=+1/+0 - a1_pixsize = 4, a2_pixsize = 4 - srcd=0000000000000000 dstd=0000000000000000 patd=0000000000000000 iinc=00FFF000 - srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, col=2 - Phrase mode is off - [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] - Entering INNER state... - Entering SREAD state... Source read address/pix address: 00858E4E/0 [0000000000000000] - Entering A2_ADD state [a2_x=0027, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007077A/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=000E, inc=1) - Entering A1_ADD state [a1_x=003D, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E50/0 [0000000000000000] - Entering A2_ADD state [a2_x=0028, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007077C/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=000D, inc=1) - Entering A1_ADD state [a1_x=003E, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E52/0 [0000000000000000] - Entering A2_ADD state [a2_x=0029, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007077E/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=000C, inc=1) - Entering A1_ADD state [a1_x=003F, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E54/0 [000000000000014A] - Entering A2_ADD state [a2_x=002A, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070780/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=000B, inc=1) - Entering A1_ADD state [a1_x=0040, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E56/0 [000000000000014A] - Entering A2_ADD state [a2_x=002B, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070782/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=000A, inc=1) - Entering A1_ADD state [a1_x=0041, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E58/0 [000000000000014A] - Entering A2_ADD state [a2_x=002C, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070784/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0009, inc=1) - Entering A1_ADD state [a1_x=0042, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E5A/0 [000000000000014A] - Entering A2_ADD state [a2_x=002D, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070786/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0008, inc=1) - Entering A1_ADD state [a1_x=0043, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E5C/0 [000000000000014A] - Entering A2_ADD state [a2_x=002E, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070788/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0007, inc=1) - Entering A1_ADD state [a1_x=0044, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E5E/0 [000000000000014A] - Entering A2_ADD state [a2_x=002F, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007078A/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0006, inc=1) - Entering A1_ADD state [a1_x=0045, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E60/0 [000000000000014A] - Entering A2_ADD state [a2_x=0030, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007078C/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0005, inc=1) - Entering A1_ADD state [a1_x=0046, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E62/0 [000000000000014A] - Entering A2_ADD state [a2_x=0031, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 0007078E/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0004, inc=1) - Entering A1_ADD state [a1_x=0047, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E64/0 [000000000000014A] - Entering A2_ADD state [a2_x=0032, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070790/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [000000000000014A] (icount=0003, inc=1) - Entering A1_ADD state [a1_x=0048, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E66/0 [0000000000000000] - Entering A2_ADD state [a2_x=0033, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070792/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=0002, inc=1) - Entering A1_ADD state [a1_x=0049, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E68/0 [0000000000000000] - Entering A2_ADD state [a2_x=0034, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070794/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=0001, inc=1) - Entering A1_ADD state [a1_x=004A, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering SREAD state... Source read address/pix address: 00858E6A/0 [0000000000000000] - Entering A2_ADD state [a2_x=0035, a2_y=00A4, addasel=0, addbsel=1, modx=0, addareg=F, adda_xconst=0, adda_yconst=0]... - Entering DWRITE state... - Dest write address/pix address: 00070796/0 [dstart=0 dend=10 pwidth=8 srcshift=0][daas=0 dabs=0 dam=7 ds=1 daq=F] [0000000000000000] (icount=0000, inc=1) - Entering A1_ADD state [a1_x=004B, a1_y=00AD, addasel=3, addbsel=2, modx=0, addareg=T, adda_xconst=7, adda_yconst=0]... - Entering IDLE_INNER state... - Leaving INNER state... (ocount=0011) - [in=F a1f=T a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] - Entering A1FUPDATE state... - [in=F a1f=F a1=T zf=F z=F a2=F iif=F iii=F izf=F izi=F] - Entering A1UPDATE state... (76/173 -> 61/174) - [in=F a1f=F a1=F zf=F z=F a2=T iif=F iii=F izf=F izi=F] - Entering A2UPDATE state... (54/164 -> 39/165) - [in=T a1f=F a1=F zf=F z=F a2=F iif=F iii=F izf=F izi=F] - Entering INNER state... + int16 a1_x = (int16)GET16(blitter_ram, A1_PIXEL + 2); + int16 a1_y = (int16)GET16(blitter_ram, A1_PIXEL + 0); + uint16 a1_frac_x = GET16(blitter_ram, A1_FPIXEL + 2); + uint16 a1_frac_y = GET16(blitter_ram, A1_FPIXEL + 0); + int16 a2_x = (int16)GET16(blitter_ram, A2_PIXEL + 2); + int16 a2_y = (int16)GET16(blitter_ram, A2_PIXEL + 0); + +Seems that the ending a1_x should be written between blits, but it doesn't seem to be... + +Blit! (CMD = 01800000) +Flags: LFUFUNC=C + count = 28672 x 1 + a1_base = 00050000, a2_base = 00070000 + a1_x = 0000, a1_y = 0000, a1_frac_x = 49CD, a1_frac_y = 0000, a2_x = 0033, a2_y = 0001 + a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 939A, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000 + a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000 + a1_win_x = 0100, a1_win_y = 0020, a2_mask_x = 0000, a2_mask_y = 0000 + a2_mask=F a1add=+phr/+0 a2add=+phr/+0 + a1_pixsize = 4, a2_pixsize = 3 + srcd=DEDEDEDEDEDEDEDE dstd=0000000000000000 patd=0000000000000000 iinc=00000000 + srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0 + Phrase mode is ON + +Blit! (CMD = 01800000) +Flags: LFUFUNC=C + count = 28672 x 1 + a1_base = 00050000, a2_base = 00070000 + a1_x = 0000, a1_y = 0000, a1_frac_x = 49CD, a1_frac_y = 0000, a2_x = 0033, a2_y = 0001 + a1_step_x = 0000, a1_step_y = 0000, a1_stepf_x = 939A, a1_stepf_y = 0000, a2_step_x = 0000, a2_step_y = 0000 + a1_inc_x = 0000, a1_inc_y = 0000, a1_incf_x = 0000, a1_incf_y = 0000 + a1_win_x = 0100, a1_win_y = 0020, a2_mask_x = 0000, a2_mask_y = 0000 + a2_mask=F a1add=+phr/+0 a2add=+phr/+0 + a1_pixsize = 4, a2_pixsize = 3 + srcd=D6D6D6D6D6D6D6D6 dstd=0000000000000000 patd=0000000000000000 iinc=00000000 + srcz1=0000000000000000 srcz2=0000000000000000 dstz=0000000000000000 zinc=00000000, coll=0 + Phrase mode is ON */ + // Various pieces of the blitter puzzle are teased out here... @@ -4571,7 +4778,8 @@ void ADDRGEN(uint32 &address, uint32 &pixa, bool gena2, bool zaddr, uint16 a1_x, uint16 a1_y, uint32 a1_base, uint8 a1_pitch, uint8 a1_pixsize, uint8 a1_width, uint8 a1_zoffset, uint16 a2_x, uint16 a2_y, uint32 a2_base, uint8 a2_pitch, uint8 a2_pixsize, uint8 a2_width, uint8 a2_zoffset) { - uint16 x = (gena2 ? a2_x : a1_x) & 0x7FFF; +// uint16 x = (gena2 ? a2_x : a1_x) & 0x7FFF; + uint16 x = (gena2 ? a2_x : a1_x) & 0xFFFF; // Actually uses all 16 bits to generate address...! uint16 y = (gena2 ? a2_y : a1_y) & 0x0FFF; uint8 width = (gena2 ? a2_width : a1_width); uint8 pixsize = (gena2 ? a2_pixsize : a1_pixsize); @@ -4936,7 +5144,7 @@ addasel[0..2] select the register to add adda_xconst[0..2] generate a power of 2 in the range 1-64 or all zeroes when they are all 1. -addareg selects register value to be added as opposed to constant +addareg selects register value to be added as opposed to constant value. suba_x, suba_y complement the X and Y values @@ -4971,7 +5179,7 @@ void ADDAMUX(int16 &adda_x, int16 &adda_y, uint8 addasel, int16 a1_step_x, int16 bool adda_yconst, bool addareg, bool suba_x, bool suba_y) { -/*INT16/ addac_x, addac_y, addar_x, addar_y, addart_x, addart_y, +/*INT16/ addac_x, addac_y, addar_x, addar_y, addart_x, addart_y, INT16/ addas_x, addas_y, suba_x16, suba_y16 :LOCAL; BEGIN @@ -4994,7 +5202,7 @@ Addar_y := MX2 (addar_y, addart_y, a2_step_y, addaselb[2]);*/ int16 addar_y = (addasel & 0x04 ? a2_step_y : yterm[addasel & 0x03]); ////////////////////////////////////////////////////////////////////////////////////// -/* Generate a constant value - this is a power of 2 in the range +/* Generate a constant value - this is a power of 2 in the range 0-64, or zero. The control bits are adda_xconst[0..2], when they are all 1 the result is 0. Constants for Y can only be 0 or 1 */ @@ -5003,7 +5211,7 @@ Constants for Y can only be 0 or 1 */ Unused[0] := DUMMY (unused[0]); Addac_x := JOIN (addac_x, addac_x[0..6], zero, zero, zero, zero, zero, zero, zero, zero, zero); -Addac_y := JOIN (addac_y, adda_yconst, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, +Addac_y := JOIN (addac_y, adda_yconst, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero);*/ ////////////////////////////////////// C++ CODE ////////////////////////////////////// int16 addac_x = (adda_xconst == 0x07 ? 0 : 1 << adda_xconst); @@ -5021,9 +5229,9 @@ Addas_y := MX2 (addas_y, addac_y, addar_y, addareg);*/ /* Complement these values (complement flag gives adder carry in)*/ -/*Suba_x16 := JOIN (suba_x16, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, +/*Suba_x16 := JOIN (suba_x16, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x, suba_x); -Suba_y16 := JOIN (suba_y16, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, +Suba_y16 := JOIN (suba_y16, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y, suba_y); Adda_x := EO (adda_x, suba_x16, addas_x); Adda_y := EO (adda_y, suba_y16, addas_y);*/ @@ -5037,9 +5245,9 @@ Adda_y := EO (adda_y, suba_y16, addas_y);*/ /** ADDBMUX - Address adder input B selection ******************* -This module selects the register to be updated by the address -adder. This can be one of three registers, the A1 and A2 -pointers, or the A1 fractional part. It can also be zero, so that the step +This module selects the register to be updated by the address +adder. This can be one of three registers, the A1 and A2 +pointers, or the A1 fractional part. It can also be zero, so that the step registers load directly into the pointers. */ @@ -5062,7 +5270,7 @@ void ADDBMUX(int16 &addb_x, int16 &addb_y, uint8 addbsel, int16 a1_x, int16 a1_y { /*Zero := TIE0 (zero); -Zero16 := JOIN (zero16, zero, zero, zero, zero, zero, zero, zero, +Zero16 := JOIN (zero16, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero, zero); Addbselb[0-1] := BUF8 (addbselb[0-1], addbsel[0-1]); Addb_x := MX4 (addb_x, a1_x, a2_x, a1_frac_x, zero16, addbselb[0..1]); @@ -5118,10 +5326,10 @@ addradd Blitter Address Adder --------------------- -The blitter address adder is a pair of sixteen bit adders, one -each for X and Y. The multiplexing of the input terms is -performed elsewhere, but this adder can also perform modulo -arithmetic to align X-addresses onto phrase boundaries. +The blitter address adder is a pair of sixteen bit adders, one +each for X and Y. The multiplexing of the input terms is +performed elsewhere, but this adder can also perform modulo +arithmetic to align X-addresses onto phrase boundaries. modx[0..2] take values 000 no mask @@ -5147,7 +5355,7 @@ INT16/ addb_y modx[0..2] suba_x suba_y - :IN); + :IN); BEGIN @@ -5211,7 +5419,7 @@ DEF DATA ( dcomp[0..7] // data byte equal flags srcd[0..7] // bits to use for bit to byte expansion zcomp[0..3] // output from Z comparators - :OUT; + :OUT; a1_x[0..1] // low two bits of A1 X pointer big_pix // pixel organisation is big-endian blitter_active // blitter is active @@ -5274,14 +5482,15 @@ void DATA(uint64 &wdata, uint8 &dcomp, uint8 &zcomp, bool &nowrite, bool big_pix, bool cmpdst, uint8 daddasel, uint8 daddbsel, uint8 daddmode, bool daddq_sel, uint8 data_sel, uint8 dbinh, uint8 dend, uint8 dstart, uint64 dstd, uint32 iinc, uint8 lfu_func, uint64 &patd, bool patdadd, bool phrase_mode, uint64 srcd, bool srcdread, bool srczread, bool srcz2add, uint8 zmode, - bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize) + bool bcompen, bool bkgwren, bool dcompen, uint8 icount, uint8 pixsize, + uint64 &srcz, uint64 dstz, uint32 zinc) { /* Stuff we absolutely *need* to have passed in/out: IN: patdadd, dstd, srcd, patd, daddasel, daddbsel, daddmode, iinc, srcz1, srcz2, big_pix, phrase_mode, cmpdst OUT: - changed patd (wdata I guess...) + changed patd (wdata I guess...) (Nope. We pass it back directly now...) */ // Source data registers @@ -5293,13 +5502,13 @@ Srcd[0-7] := JOIN (srcd[0-7], srcdlo{0-7}); Srcd[8-31] := JOIN (srcd[8-31], srcdlo{8-31}); Srcd[32-63] := JOIN (srcd[32-63], srcdhi{0-31});*/ -// Destination data registers +// Destination data registers /*Data_dst := DATA_DST (dstd[0..63], dstz[0..1], clk, dstdld[0..1], dstzld[0..1], load_data[0..1]); Dstdlo := JOIN (dstdlo, dstd[0..31]); Dstdhi := JOIN (dstdhi, dstd[32..63]);*/ -// Pattern and Color data registers +// Pattern and Color data registers // Looks like this is simply another register file for the pattern data registers. No adding or anything funky // going on. Note that patd & patdv will output the same info. @@ -5313,7 +5522,7 @@ Patdhi := JOIN (patdhi, patd[32..63]);*/ // Multiplying data Mixer (NOT IN JAGUAR I) -/*Datamix := DATAMIX (patdo[0..1], clk, colord[0..15], dpipe[1], dstd[0..63], int0dp[8..10], int1dp[8..10], +/*Datamix := DATAMIX (patdo[0..1], clk, colord[0..15], dpipe[1], dstd[0..63], int0dp[8..10], int1dp[8..10], int2dp[8..10], int3dp[8..10], mixsel[0..2], patd[0..63], pdsel[0..1], srcd[0..63], textrgb, txtd[0..63]);*/ // Logic function unit @@ -5329,7 +5538,7 @@ Patdhi := JOIN (patdhi, patd[32..63]);*/ ////////////////////////////////////////////////////////////////////////////////////// // Increment and Step Registers - + // Does it do anything without the step add lines? Check it! // No. This is pretty much just a register file without the Jaguar II lines... /*Inc_step := INC_STEP (iinc, istep[0..31], zinc, zstep[0..31], clk, ext_int, gpu_din, iincld, iincldx, istepadd, @@ -5360,29 +5569,72 @@ Zstep := JOIN (zstep, zstep[0..31]);*/ dcomp |= 0x40; if ((cmpd & 0xFF00000000000000LL) == 0) dcomp |= 0x80; +////////////////////////////////////////////////////////////////////////////////////// + +// Zed comparator for Z-buffer operations + +/*Zedcomp := ZEDCOMP (zcomp[0..3], srczp[0..1], dstz[0..1], zmode[0..2]);*/ +////////////////////////////////////// C++ CODE ////////////////////////////////////// +//srczp is srcz pipelined, also it goes through a source shift as well... +/*The shift is basically like so (each piece is 16 bits long): + + 0 1 2 3 4 5 6 + srcz1lolo srcz1lohi srcz1hilo srcz1hihi srcrz2lolo srcz2lohi srcz2hilo + +with srcshift bits 4 & 5 selecting the start position +*/ +//So... basically what we have here is: + zcomp = 0; + + if ((((srcz & 0x000000000000FFFFLL) < (dstz & 0x000000000000FFFFLL)) && (zmode & 0x01)) + || (((srcz & 0x000000000000FFFFLL) == (dstz & 0x000000000000FFFFLL)) && (zmode & 0x02)) + || (((srcz & 0x000000000000FFFFLL) > (dstz & 0x000000000000FFFFLL)) && (zmode & 0x04))) + zcomp |= 0x01; + + if ((((srcz & 0x00000000FFFF0000LL) < (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x01)) + || (((srcz & 0x00000000FFFF0000LL) == (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x02)) + || (((srcz & 0x00000000FFFF0000LL) > (dstz & 0x00000000FFFF0000LL)) && (zmode & 0x04))) + zcomp |= 0x02; + + if ((((srcz & 0x0000FFFF00000000LL) < (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x01)) + || (((srcz & 0x0000FFFF00000000LL) == (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x02)) + || (((srcz & 0x0000FFFF00000000LL) > (dstz & 0x0000FFFF00000000LL)) && (zmode & 0x04))) + zcomp |= 0x04; + + if ((((srcz & 0xFFFF000000000000LL) < (dstz & 0xFFFF000000000000LL)) && (zmode & 0x01)) + || (((srcz & 0xFFFF000000000000LL) == (dstz & 0xFFFF000000000000LL)) && (zmode & 0x02)) + || (((srcz & 0xFFFF000000000000LL) > (dstz & 0xFFFF000000000000LL)) && (zmode & 0x04))) + zcomp |= 0x08; + +//TEMP, TO TEST IF ZCOMP IS THE CULPRIT... +//Nope, this is NOT the problem... +//zcomp=0; // We'll do the comparison/bit/byte inhibits here, since that's they way it happens // in the real thing (dcomp goes out to COMP_CTRL and back into DATA through dbinh)... #if 1 uint8 dbinht; // bool nowrite; COMP_CTRL(dbinht, nowrite, - bcompen, true/*big_pix*/, bkgwren, dcomp, dcompen, icount, pixsize, phrase_mode, srcd & 0xFF, 0);//zcomp); + bcompen, true/*big_pix*/, bkgwren, dcomp, dcompen, icount, pixsize, phrase_mode, srcd & 0xFF, zcomp); dbinh = dbinht; // dbinh = 0x00; #endif -////////////////////////////////////////////////////////////////////////////////////// - -// Zed comparator for Z-buffer operations - -/*Zedcomp := ZEDCOMP (zcomp[0..3], srczp[0..1], dstz[0..1], zmode[0..2]);*/ -////////////////////////////////////// C++ CODE ////////////////////////////////////// +#if 1 +#ifdef VERBOSE_BLITTER_LOGGING +if (logBlit) +{ + printf("\n[dcomp=%02X zcomp=%02X dbinh=%02X]\n", dcomp, zcomp, dbinh); + fflush(stdout); +}//*/ +#endif +#endif ////////////////////////////////////////////////////////////////////////////////////// // 22 Mar 94 // The data initializer - allows all four initial values to be computed from one (NOT IN JAGUAR I) -/*Datinit := DATINIT (initcin[0..3], initinc[0..63], initpix[0..15], a1_x[0..1], big_pix, clk, iinc, init_if, init_ii, +/*Datinit := DATINIT (initcin[0..3], initinc[0..63], initpix[0..15], a1_x[0..1], big_pix, clk, iinc, init_if, init_ii, init_zf, istep[0..31], zinc, zstep[0..31]);*/ // Adder array for Z and intensity increments @@ -5401,7 +5653,7 @@ Zstep := JOIN (zstep, zstep[0..31]);*/ //This is normally done asynchronously above (thru local_data) when in patdadd mode... //And now it's passed back to the caller to be persistent between calls...! -//But it's causing some serious fuck-ups in T2K now... !!! FIX !!! +//But it's causing some serious fuck-ups in T2K now... !!! FIX !!! [DONE--???] //Weird! It doesn't anymore...! if (patdadd) patd = ((uint64)addq[3] << 48) | ((uint64)addq[2] << 32) | ((uint64)addq[1] << 16) | (uint64)addq[0]; @@ -5511,7 +5763,7 @@ Maskt[9-14] := OAN1P (maskt[9-14], maskt[8-13], s_coarse[2-7], e_coarse\[2-7]);* ////////////////////////////////////////////////////////////////////////////////////// /* The bit terms are mirrored for big-endian pixels outside phrase -mode. The byte terms are mirrored for big-endian pixels in phrase +mode. The byte terms are mirrored for big-endian pixels in phrase mode. */ /*Mirror_bit := AN2M (mir_bit, phrase_mode\, big_pix); @@ -5627,6 +5879,25 @@ Dat[56-63] := MX4 (dat[56-63], dstdhi{24-31}, ddathi{24-31}, dstzhi{24-31}, srcz (uint32)(wdata >> 32), (uint32)(wdata & 0xFFFFFFFF), mask); fflush(stdout); }//*/ +//This is a crappy way of handling this, but it should work for now... + uint64 zwdata; + zwdata = ((srcz & mask) | (dstz & ~mask)) & 0x00000000000000FFLL; + zwdata |= (mask & 0x0100 ? srcz : dstz) & 0x000000000000FF00LL; + zwdata |= (mask & 0x0200 ? srcz : dstz) & 0x0000000000FF0000LL; + zwdata |= (mask & 0x0400 ? srcz : dstz) & 0x00000000FF000000LL; + zwdata |= (mask & 0x0800 ? srcz : dstz) & 0x000000FF00000000LL; + zwdata |= (mask & 0x1000 ? srcz : dstz) & 0x0000FF0000000000LL; + zwdata |= (mask & 0x2000 ? srcz : dstz) & 0x00FF000000000000LL; + zwdata |= (mask & 0x4000 ? srcz : dstz) & 0xFF00000000000000LL; +if (logBlit) +{ + printf("\n[srcz=%08X%08X dstz=%08X%08X zwdata=%08X%08X mask=%04X]\n", + (uint32)(srcz >> 32), (uint32)(srcz & 0xFFFFFFFF), + (uint32)(dstz >> 32), (uint32)(dstz & 0xFFFFFFFF), + (uint32)(zwdata >> 32), (uint32)(zwdata & 0xFFFFFFFF), mask); + fflush(stdout); +}//*/ + srcz = zwdata; ////////////////////////////////////////////////////////////////////////////////////// /*Data_enab[0-1] := BUF8 (data_enab[0-1], data_ena); @@ -5641,7 +5912,7 @@ END;*/ /** COMP_CTRL - Comparator output control logic ***************** This block is responsible for taking the comparator outputs and -using them as appropriate to inhibit writes. Two methods are +using them as appropriate to inhibit writes. Two methods are supported for inhibiting write data: - suppression of the inner loop controlled write operation @@ -5654,13 +5925,13 @@ and sixteen bit pixel modes. Writes can be suppressed by data being equal, by the Z comparator conditions being met, or by the bit to pixel expansion scheme. -Pipe-lining issues: the data derived comparator outputs are stable +Pipe-lining issues: the data derived comparator outputs are stable until the next data read, well after the affected write from this operation. However, the inner counter bits can count immediately -before the ack for the last write. Therefore, it is necessary to +before the ack for the last write. Therefore, it is necessary to delay bcompbit select terms by one inner loop pipe-line stage, when generating the select for the data control - the output is -delayed one further tick to give it write data timing (2/34). +delayed one further tick to give it write data timing (2/34). There is also a problem with computed data - the new values are calculated before the write associated with the old value has been @@ -5709,7 +5980,7 @@ In phrase mode, the eight bits are used directly, and this mode is only applicable to 8-bit pixel mode (2/34) */ /*Bcompselt[0-2] := EO (bcompselt[0-2], icount[0-2], big_pix); -Bcompbit := MX8 (bcompbit, srcd[7], srcd[6], srcd[5], +Bcompbit := MX8 (bcompbit, srcd[7], srcd[6], srcd[5], srcd[4], srcd[3], srcd[2], srcd[1], srcd[0], bcompselt[0..2]); Bcompbit\ := INV1 (bcompbit\, bcompbit);*/ ////////////////////////////////////// C++ CODE ////////////////////////////////////// @@ -5728,14 +5999,14 @@ if (logBlit) /* pipe-line the count */ /*Bcompsel[0-2] := FDSYNC (bcompsel[0-2], bcompselt[0-2], step_inner, clk); -Bcompbt := MX8 (bcompbitpt, srcd[7], srcd[6], srcd[5], +Bcompbt := MX8 (bcompbitpt, srcd[7], srcd[6], srcd[5], srcd[4], srcd[3], srcd[2], srcd[1], srcd[0], bcompsel[0..2]); Bcompbitp := FD1Q (bcompbitp, bcompbitpt, clk); Bcompbitp\ := INV1 (bcompbitp\, bcompbitp);*/ /* For pixel mode, generate the write inhibit signal for all modes on bit inhibit, for 8 and 16 bit modes on comparator inhibit, and -for 16 bit mode on Z inhibit +for 16 bit mode on Z inhibit Nowrite = bcompen . /bcompbit . /phrase_mode + dcompen . dcomp[0] . /phrase_mode . pixsize = 011 @@ -5777,7 +6048,7 @@ if (logBlit) ////////////////////////////////////////////////////////////////////////////////////// /* For phrase mode, generate the byte inhibit signals for eight bit -mode 011, or sixteen bit mode 100 +mode 011, or sixteen bit mode 100 dbinh\[0] = pixsize[2] . zcomp[0] + pixsize[2] . dcomp[0] . dcomp[1] . dcompen + /pixsize[2] . dcomp[0] . dcompen @@ -5969,4 +6240,9 @@ if (logBlit) ////////////////////////////////////// C++ CODE ////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////// +// !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! +// !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! +// !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! TESTING !!! + #endif +