X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=plugins%2Fgpu_unai%2Fgpu_inner.h;h=8cb4bd534b2364934372b786cc12fa1cdcff0b8c;hb=4cfc568dce10c4ff9a476e229c56b17f8f1ed678;hp=6c18c63991f7911afb5816aebebe03d692e10e19;hpb=86aad47b0418b1715a4d223adf8f59aa92619d15;p=pcsx_rearmed.git diff --git a/plugins/gpu_unai/gpu_inner.h b/plugins/gpu_unai/gpu_inner.h index 6c18c639..8cb4bd53 100644 --- a/plugins/gpu_unai/gpu_inner.h +++ b/plugins/gpu_unai/gpu_inner.h @@ -1,6 +1,7 @@ /*************************************************************************** * Copyright (C) 2010 PCSX4ALL Team * * Copyright (C) 2010 Unai * +* Copyright (C) 2016 Senquack (dansilsby gmail com) * * * * This program is free software; you can redistribute it and/or modify * * it under the terms of the GNU General Public License as published by * @@ -18,410 +19,716 @@ * 51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA. * ***************************************************************************/ -/////////////////////////////////////////////////////////////////////////////// -// Inner loop driver instanciation file +#ifndef __GPU_UNAI_GPU_INNER_H__ +#define __GPU_UNAI_GPU_INNER_H__ /////////////////////////////////////////////////////////////////////////////// -// Option Masks -#define L ((CF>>0)&1) -#define B ((CF>>1)&1) -#define M ((CF>>2)&1) -#define BM ((CF>>3)&3) -#define TM ((CF>>5)&3) -#define G ((CF>>7)&1) +// Inner loop driver instantiation file -#define AH ((CF>>7)&1) +/////////////////////////////////////////////////////////////////////////////// +// Option Masks (CF template paramter) +#define CF_LIGHT ((CF>> 0)&1) // Lighting +#define CF_BLEND ((CF>> 1)&1) // Blending +#define CF_MASKCHECK ((CF>> 2)&1) // Mask bit check +#define CF_BLENDMODE ((CF>> 3)&3) // Blend mode 0..3 +#define CF_TEXTMODE ((CF>> 5)&3) // Texture mode 1..3 (0: texturing disabled) +#define CF_GOURAUD ((CF>> 7)&1) // Gouraud shading +#define CF_MASKSET ((CF>> 8)&1) // Mask bit set +#define CF_DITHER ((CF>> 9)&1) // Dithering +#define CF_BLITMASK ((CF>>10)&1) // blit_mask check (skip rendering pixels + // that wouldn't end up displayed on + // low-res screen using simple downscaler) -#define MB ((CF>>8)&1) +//#ifdef __arm__ +//#ifndef ENABLE_GPU_ARMV7 +/* ARMv5 */ +//#include "gpu_inner_blend_arm5.h" +//#else +/* ARMv7 optimized */ +//#include "gpu_inner_blend_arm7.h" +//#endif +//#else +//#include "gpu_inner_blend.h" +//#endif #include "gpu_inner_blend.h" +#include "gpu_inner_quantization.h" #include "gpu_inner_light.h" +#ifdef __arm__ +#include "gpu_inner_blend_arm.h" +#include "gpu_inner_light_arm.h" +#define gpuBlending gpuBlendingARM +#define gpuLightingRGB gpuLightingRGBARM +#define gpuLightingTXT gpuLightingTXTARM +#define gpuLightingTXTGouraud gpuLightingTXTGouraudARM +// Non-dithering lighting and blending functions preserve uSrc +// MSB. This saves a few operations and useless load/stores. +#define MSB_PRESERVED (!CF_DITHER) +#else +#define gpuBlending gpuBlendingGeneric +#define gpuLightingRGB gpuLightingRGBGeneric +#define gpuLightingTXT gpuLightingTXTGeneric +#define gpuLightingTXTGouraud gpuLightingTXTGouraudGeneric +#define MSB_PRESERVED 0 +#endif + + +// If defined, Gouraud colors are fixed-point 5.11, otherwise they are 8.16 +// This is only for debugging/verification of low-precision colors in C. +// Low-precision Gouraud is intended for use by SIMD-optimized inner drivers +// which get/use Gouraud colors in SIMD registers. +//#define GPU_GOURAUD_LOW_PRECISION + +// How many bits of fixed-point precision GouraudColor uses +#ifdef GPU_GOURAUD_LOW_PRECISION +#define GPU_GOURAUD_FIXED_BITS 11 +#else +#define GPU_GOURAUD_FIXED_BITS 16 +#endif + +// Used to pass Gouraud colors to gpuPixelSpanFn() (lines) +struct GouraudColor { +#ifdef GPU_GOURAUD_LOW_PRECISION + u16 r, g, b; + s16 r_incr, g_incr, b_incr; +#else + u32 r, g, b; + s32 r_incr, g_incr, b_incr; +#endif +}; + +static inline u16 gpuGouraudColor15bpp(u32 r, u32 g, u32 b) +{ + r >>= GPU_GOURAUD_FIXED_BITS; + g >>= GPU_GOURAUD_FIXED_BITS; + b >>= GPU_GOURAUD_FIXED_BITS; + +#ifndef GPU_GOURAUD_LOW_PRECISION + // High-precision Gouraud colors are 8-bit + fractional + r >>= 3; g >>= 3; b >>= 3; +#endif + + return r | (g << 5) | (b << 10); +} + /////////////////////////////////////////////////////////////////////////////// -// GPU Pixel opperations generator -template -INLINE void gpuPixelFn(u16 *pixel,const u16 data) +// GPU Pixel span operations generator gpuPixelSpanFn<> +// Oct 2016: Created/adapted from old gpuPixelFn by senquack: +// Original gpuPixelFn was used to draw lines one pixel at a time. I wrote +// new line algorithms that draw lines using horizontal/vertical/diagonal +// spans of pixels, necessitating new pixel-drawing function that could +// not only render spans of pixels, but gouraud-shade them as well. +// This speeds up line rendering and would allow tile-rendering (untextured +// rectangles) to use the same set of functions. Since tiles are always +// monochrome, they simply wouldn't use the extra set of 32 gouraud-shaded +// gpuPixelSpanFn functions (TODO?). +// +// NOTE: While the PS1 framebuffer is 16 bit, we use 8-bit pointers here, +// so that pDst can be incremented directly by 'incr' parameter +// without having to shift it before use. +template +static u8* gpuPixelSpanFn(u8* pDst, uintptr_t data, ptrdiff_t incr, size_t len) { - if ((!M)&&(!B)) - { - if(MB) { *pixel = data | 0x8000; } - else { *pixel = data; } + // Blend func can save an operation if it knows uSrc MSB is + // unset. For untextured prims, this is always true. + const bool skip_uSrc_mask = true; + + u16 col; + struct GouraudColor * gcPtr; + u32 r, g, b; + s32 r_incr, g_incr, b_incr; + + if (CF_GOURAUD) { + gcPtr = (GouraudColor*)data; + r = gcPtr->r; r_incr = gcPtr->r_incr; + g = gcPtr->g; g_incr = gcPtr->g_incr; + b = gcPtr->b; b_incr = gcPtr->b_incr; + } else { + col = (u16)data; } - else if ((M)&&(!B)) - { - if (!(*pixel&0x8000)) - { - if(MB) { *pixel = data | 0x8000; } - else { *pixel = data; } + + do { + if (!CF_GOURAUD) + { // NO GOURAUD + if (!CF_MASKCHECK && !CF_BLEND) { + if (CF_MASKSET) { *(u16*)pDst = col | 0x8000; } + else { *(u16*)pDst = col; } + } else if (CF_MASKCHECK && !CF_BLEND) { + if (!(*(u16*)pDst & 0x8000)) { + if (CF_MASKSET) { *(u16*)pDst = col | 0x8000; } + else { *(u16*)pDst = col; } + } + } else { + uint_fast16_t uDst = *(u16*)pDst; + if (CF_MASKCHECK) { if (uDst & 0x8000) goto endpixel; } + + uint_fast16_t uSrc = col; + + if (CF_BLEND) + uSrc = gpuBlending(uSrc, uDst); + + if (CF_MASKSET) { *(u16*)pDst = uSrc | 0x8000; } + else { *(u16*)pDst = uSrc; } + } + + } else + { // GOURAUD + + if (!CF_MASKCHECK && !CF_BLEND) { + col = gpuGouraudColor15bpp(r, g, b); + if (CF_MASKSET) { *(u16*)pDst = col | 0x8000; } + else { *(u16*)pDst = col; } + } else if (CF_MASKCHECK && !CF_BLEND) { + col = gpuGouraudColor15bpp(r, g, b); + if (!(*(u16*)pDst & 0x8000)) { + if (CF_MASKSET) { *(u16*)pDst = col | 0x8000; } + else { *(u16*)pDst = col; } + } + } else { + uint_fast16_t uDst = *(u16*)pDst; + if (CF_MASKCHECK) { if (uDst & 0x8000) goto endpixel; } + col = gpuGouraudColor15bpp(r, g, b); + + uint_fast16_t uSrc = col; + + // Blend func can save an operation if it knows uSrc MSB is + // unset. For untextured prims, this is always true. + const bool skip_uSrc_mask = true; + + if (CF_BLEND) + uSrc = gpuBlending(uSrc, uDst); + + if (CF_MASKSET) { *(u16*)pDst = uSrc | 0x8000; } + else { *(u16*)pDst = uSrc; } + } } + +endpixel: + if (CF_GOURAUD) { + r += r_incr; + g += g_incr; + b += b_incr; + } + pDst += incr; + } while (len-- > 1); + + // Note from senquack: Normally, I'd prefer to write a 'do {} while (--len)' + // loop, or even a for() loop, however, on MIPS platforms anything but the + // 'do {} while (len-- > 1)' tends to generate very unoptimal asm, with + // many unneeded MULs/ADDs/branches at the ends of these functions. + // If you change the loop structure above, be sure to compare the quality + // of the generated code!! + + if (CF_GOURAUD) { + gcPtr->r = r; + gcPtr->g = g; + gcPtr->b = b; } - else - { - u16 uDst = *pixel; - if(M) { if (uDst&0x8000) return; } - u16 uSrc = data; - u32 uMsk; if (BM==0) uMsk=0x7BDE; - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); - if(MB) { *pixel = uSrc | 0x8000; } - else { *pixel = uSrc; } - } + return pDst; +} + +static u8* PixelSpanNULL(u8* pDst, uintptr_t data, ptrdiff_t incr, size_t len) +{ + #ifdef ENABLE_GPU_LOG_SUPPORT + fprintf(stdout,"PixelSpanNULL()\n"); + #endif + return pDst; } -/////////////////////////////////////////////////////////////////////////////// /////////////////////////////////////////////////////////////////////////////// -// Pixel drawing drivers, for lines (only blending) -typedef void (*PD)(u16 *pixel,const u16 data); -const PD gpuPixelDrivers[32] = // We only generate pixel op for MASKING/BLEND_ENABLE/BLEND_MODE +// PixelSpan (lines) innerloops driver +typedef u8* (*PSD)(u8* dst, uintptr_t data, ptrdiff_t incr, size_t len); + +const PSD gpuPixelSpanDrivers[64] = { - gpuPixelFn<0x00<<1>,gpuPixelFn<0x01<<1>,gpuPixelFn<0x02<<1>,gpuPixelFn<0x03<<1>, - NULL,gpuPixelFn<0x05<<1>,NULL,gpuPixelFn<0x07<<1>, - NULL,gpuPixelFn<0x09<<1>,NULL,gpuPixelFn<0x0B<<1>, - NULL,gpuPixelFn<0x0D<<1>,NULL,gpuPixelFn<0x0F<<1>, - - gpuPixelFn<(0x00<<1)|256>,gpuPixelFn<(0x01<<1)|256>,gpuPixelFn<(0x02<<1)|256>,gpuPixelFn<(0x03<<1)|256>, - NULL,gpuPixelFn<(0x05<<1)|256>,NULL,gpuPixelFn<(0x07<<1)|256>, - NULL,gpuPixelFn<(0x09<<1)|256>,NULL,gpuPixelFn<(0x0B<<1)|256>, - NULL,gpuPixelFn<(0x0D<<1)|256>,NULL,gpuPixelFn<(0x0F<<1)|256> + // Array index | 'CF' template field | Field value + // ------------+---------------------+---------------- + // Bit 0 | CF_BLEND | off (0), on (1) + // Bit 1 | CF_MASKCHECK | off (0), on (1) + // Bit 3:2 | CF_BLENDMODE | 0..3 + // Bit 4 | CF_MASKSET | off (0), on (1) + // Bit 5 | CF_GOURAUD | off (0), on (1) + // + // NULL entries are ones for which blending is disabled and blend-mode + // field is non-zero, which is obviously invalid. + + // Flat-shaded + gpuPixelSpanFn<0x00<<1>, gpuPixelSpanFn<0x01<<1>, gpuPixelSpanFn<0x02<<1>, gpuPixelSpanFn<0x03<<1>, + PixelSpanNULL, gpuPixelSpanFn<0x05<<1>, PixelSpanNULL, gpuPixelSpanFn<0x07<<1>, + PixelSpanNULL, gpuPixelSpanFn<0x09<<1>, PixelSpanNULL, gpuPixelSpanFn<0x0B<<1>, + PixelSpanNULL, gpuPixelSpanFn<0x0D<<1>, PixelSpanNULL, gpuPixelSpanFn<0x0F<<1>, + + // Flat-shaded + PixelMSB (CF_MASKSET) + gpuPixelSpanFn<(0x00<<1)|0x100>, gpuPixelSpanFn<(0x01<<1)|0x100>, gpuPixelSpanFn<(0x02<<1)|0x100>, gpuPixelSpanFn<(0x03<<1)|0x100>, + PixelSpanNULL, gpuPixelSpanFn<(0x05<<1)|0x100>, PixelSpanNULL, gpuPixelSpanFn<(0x07<<1)|0x100>, + PixelSpanNULL, gpuPixelSpanFn<(0x09<<1)|0x100>, PixelSpanNULL, gpuPixelSpanFn<(0x0B<<1)|0x100>, + PixelSpanNULL, gpuPixelSpanFn<(0x0D<<1)|0x100>, PixelSpanNULL, gpuPixelSpanFn<(0x0F<<1)|0x100>, + + // Gouraud-shaded (CF_GOURAUD) + gpuPixelSpanFn<(0x00<<1)|0x80>, gpuPixelSpanFn<(0x01<<1)|0x80>, gpuPixelSpanFn<(0x02<<1)|0x80>, gpuPixelSpanFn<(0x03<<1)|0x80>, + PixelSpanNULL, gpuPixelSpanFn<(0x05<<1)|0x80>, PixelSpanNULL, gpuPixelSpanFn<(0x07<<1)|0x80>, + PixelSpanNULL, gpuPixelSpanFn<(0x09<<1)|0x80>, PixelSpanNULL, gpuPixelSpanFn<(0x0B<<1)|0x80>, + PixelSpanNULL, gpuPixelSpanFn<(0x0D<<1)|0x80>, PixelSpanNULL, gpuPixelSpanFn<(0x0F<<1)|0x80>, + + // Gouraud-shaded (CF_GOURAUD) + PixelMSB (CF_MASKSET) + gpuPixelSpanFn<(0x00<<1)|0x180>, gpuPixelSpanFn<(0x01<<1)|0x180>, gpuPixelSpanFn<(0x02<<1)|0x180>, gpuPixelSpanFn<(0x03<<1)|0x180>, + PixelSpanNULL, gpuPixelSpanFn<(0x05<<1)|0x180>, PixelSpanNULL, gpuPixelSpanFn<(0x07<<1)|0x180>, + PixelSpanNULL, gpuPixelSpanFn<(0x09<<1)|0x180>, PixelSpanNULL, gpuPixelSpanFn<(0x0B<<1)|0x180>, + PixelSpanNULL, gpuPixelSpanFn<(0x0D<<1)|0x180>, PixelSpanNULL, gpuPixelSpanFn<(0x0F<<1)|0x180> }; /////////////////////////////////////////////////////////////////////////////// // GPU Tiles innerloops generator -template -INLINE void gpuTileSpanFn(u16 *pDst, u32 count, u16 data) +template +static void gpuTileSpanFn(u16 *pDst, u32 count, u16 data) { - if ((!M)&&(!B)) - { - if (MB) { data = data | 0x8000; } + if (!CF_MASKCHECK && !CF_BLEND) { + if (CF_MASKSET) { data = data | 0x8000; } do { *pDst++ = data; } while (--count); - } - else if ((M)&&(!B)) - { - if (MB) { data = data | 0x8000; } + } else if (CF_MASKCHECK && !CF_BLEND) { + if (CF_MASKSET) { data = data | 0x8000; } do { if (!(*pDst&0x8000)) { *pDst = data; } pDst++; } while (--count); - } - else + } else { - u16 uSrc; - u16 uDst; - u32 uMsk; if (BM==0) uMsk=0x7BDE; + // Blend func can save an operation if it knows uSrc MSB is + // unset. For untextured prims, this is always true. + const bool skip_uSrc_mask = true; + + uint_fast16_t uSrc, uDst; do { - // MASKING - uDst = *pDst; - if(M) { if (uDst&0x8000) goto endtile; } + if (CF_MASKCHECK || CF_BLEND) { uDst = *pDst; } + if (CF_MASKCHECK) { if (uDst&0x8000) goto endtile; } + uSrc = data; - // BLEND - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); + if (CF_BLEND) + uSrc = gpuBlending(uSrc, uDst); - if (MB) { *pDst = uSrc | 0x8000; } - else { *pDst = uSrc; } - endtile: pDst++; + if (CF_MASKSET) { *pDst = uSrc | 0x8000; } + else { *pDst = uSrc; } + + //senquack - Did not apply "Silent Hill" mask-bit fix to here. + // It is hard to tell from scarce documentation available and + // lack of comments in code, but I believe the tile-span + // functions here should not bother to preserve any source MSB, + // as they are not drawing from a texture. +endtile: + pDst++; } while (--count); } } +static void TileNULL(u16 *pDst, u32 count, u16 data) +{ + #ifdef ENABLE_GPU_LOG_SUPPORT + fprintf(stdout,"TileNULL()\n"); + #endif +} + /////////////////////////////////////////////////////////////////////////////// // Tiles innerloops driver typedef void (*PT)(u16 *pDst, u32 count, u16 data); -const PT gpuTileSpanDrivers[64] = -{ - gpuTileSpanFn<0x00>,NULL,gpuTileSpanFn<0x02>,NULL, gpuTileSpanFn<0x04>,NULL,gpuTileSpanFn<0x06>,NULL, NULL,NULL,gpuTileSpanFn<0x0A>,NULL, NULL,NULL,gpuTileSpanFn<0x0E>,NULL, - NULL,NULL,gpuTileSpanFn<0x12>,NULL, NULL,NULL,gpuTileSpanFn<0x16>,NULL, NULL,NULL,gpuTileSpanFn<0x1A>,NULL, NULL,NULL,gpuTileSpanFn<0x1E>,NULL, - gpuTileSpanFn<0x100>,NULL,gpuTileSpanFn<0x102>,NULL, gpuTileSpanFn<0x104>,NULL,gpuTileSpanFn<0x106>,NULL, NULL,NULL,gpuTileSpanFn<0x10A>,NULL, NULL,NULL,gpuTileSpanFn<0x10E>,NULL, - NULL,NULL,gpuTileSpanFn<0x112>,NULL, NULL,NULL,gpuTileSpanFn<0x116>,NULL, NULL,NULL,gpuTileSpanFn<0x11A>,NULL, NULL,NULL,gpuTileSpanFn<0x11E>,NULL, +// Template instantiation helper macros +#define TI(cf) gpuTileSpanFn<(cf)> +#define TN TileNULL +#define TIBLOCK(ub) \ + TI((ub)|0x00), TI((ub)|0x02), TI((ub)|0x04), TI((ub)|0x06), \ + TN, TI((ub)|0x0a), TN, TI((ub)|0x0e), \ + TN, TI((ub)|0x12), TN, TI((ub)|0x16), \ + TN, TI((ub)|0x1a), TN, TI((ub)|0x1e) + +const PT gpuTileSpanDrivers[32] = { + TIBLOCK(0<<8), TIBLOCK(1<<8) }; +#undef TI +#undef TN +#undef TIBLOCK + + /////////////////////////////////////////////////////////////////////////////// // GPU Sprites innerloops generator -template -INLINE void gpuSpriteSpanFn(u16 *pDst, u32 count, u32 u0, const u32 mask) +template +static void gpuSpriteSpanFn(u16 *pDst, u32 count, u8* pTxt, u32 u0) { - u16 uSrc; - u16 uDst; - const u16* pTxt = TBA+(u0&~0x1ff); u0=u0&0x1ff; - const u16 *_CBA; if(TM!=3) _CBA=CBA; - u32 lCol; if(L) { lCol = ((u32)(b4<< 2)&(0x03ff)) | ((u32)(g4<<13)&(0x07ff<<10)) | ((u32)(r4<<24)&(0x07ff<<21)); } - u8 rgb; if (TM==1) rgb = ((u8*)pTxt)[u0>>1]; - u32 uMsk; if ((B)&&(BM==0)) uMsk=0x7BDE; + // Blend func can save an operation if it knows uSrc MSB is unset. + // Untextured prims can always skip (source color always comes with MSB=0). + // For textured prims, the generic lighting funcs always return it unset. (bonus!) + const bool skip_uSrc_mask = MSB_PRESERVED ? (!CF_TEXTMODE) : (!CF_TEXTMODE) || CF_LIGHT; + + uint_fast16_t uSrc, uDst, srcMSB; + bool should_blend; + u32 u0_mask = gpu_senquack.TextureWindow[2]; + + u8 r5, g5, b5; + if (CF_LIGHT) { + r5 = gpu_senquack.r5; + g5 = gpu_senquack.g5; + b5 = gpu_senquack.b5; + } + + if (CF_TEXTMODE==3) { + // Texture is accessed byte-wise, so adjust mask if 16bpp + u0_mask <<= 1; + } + + const u16 *CBA_; if (CF_TEXTMODE!=3) CBA_ = gpu_senquack.CBA; do { - // MASKING - if(M) { uDst = *pDst; if (uDst&0x8000) { u0=(u0+1)&mask; goto endsprite; } } - - // TEXTURE MAPPING - if (TM==1) { if (!(u0&1)) rgb = ((u8*)pTxt)[u0>>1]; uSrc = _CBA[(rgb>>((u0&1)<<2))&0xf]; u0=(u0+1)&mask; } - if (TM==2) { uSrc = _CBA[((u8*)pTxt)[u0]]; u0=(u0+1)&mask; } - if (TM==3) { uSrc = pTxt[u0]; u0=(u0+1)&mask; } - if(!AH) { if (!uSrc) goto endsprite; } + if (CF_MASKCHECK || CF_BLEND) { uDst = *pDst; } + if (CF_MASKCHECK) if (uDst&0x8000) { goto endsprite; } - // BLEND - if(B) - { - if(uSrc&0x8000) - { - // LIGHTING CALCULATIONS - if(L) { gpuLightingTXT(uSrc, lCol); } - - if(!M) { uDst = *pDst; } - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); - } - else - { - // LIGHTING CALCULATIONS - if(L) { gpuLightingTXT(uSrc, lCol); } - } + if (CF_TEXTMODE==1) { // 4bpp (CLUT) + u8 rgb = pTxt[(u0 & u0_mask)>>1]; + uSrc = CBA_[(rgb>>((u0&1)<<2))&0xf]; } - else - { - // LIGHTING CALCULATIONS - if(L) { gpuLightingTXT(uSrc, lCol); } else - { if(!MB) uSrc&= 0x7fff; } + if (CF_TEXTMODE==2) { // 8bpp (CLUT) + uSrc = CBA_[pTxt[u0 & u0_mask]]; + } + if (CF_TEXTMODE==3) { // 16bpp + uSrc = *(u16*)(&pTxt[u0 & u0_mask]); } - if (MB) { *pDst = uSrc | 0x8000; } - else { *pDst = uSrc; } + if (!uSrc) goto endsprite; + + //senquack - save source MSB, as blending or lighting macros will not + // (Silent Hill gray rectangles mask bit bug) + if (CF_BLEND || CF_LIGHT) srcMSB = uSrc & 0x8000; - endsprite: pDst++; + if (CF_LIGHT) + uSrc = gpuLightingTXT(uSrc, r5, g5, b5); + + should_blend = MSB_PRESERVED ? uSrc & 0x8000 : srcMSB; + + if (CF_BLEND && should_blend) + uSrc = gpuBlending(uSrc, uDst); + + if (CF_MASKSET) { *pDst = uSrc | 0x8000; } + else if (!MSB_PRESERVED && (CF_BLEND || CF_LIGHT)) { *pDst = uSrc | srcMSB; } + else { *pDst = uSrc; } + +endsprite: + u0 += (CF_TEXTMODE==3) ? 2 : 1; + pDst++; } while (--count); } + +static void SpriteNULL(u16 *pDst, u32 count, u8* pTxt, u32 u0) +{ + #ifdef ENABLE_GPU_LOG_SUPPORT + fprintf(stdout,"SpriteNULL()\n"); + #endif +} + /////////////////////////////////////////////////////////////////////////////// /////////////////////////////////////////////////////////////////////////////// // Sprite innerloops driver -typedef void (*PS)(u16 *pDst, u32 count, u32 u0, const u32 mask); -const PS gpuSpriteSpanDrivers[512] = -{ - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - gpuSpriteSpanFn<0x20>,gpuSpriteSpanFn<0x21>,gpuSpriteSpanFn<0x22>,gpuSpriteSpanFn<0x23>, gpuSpriteSpanFn<0x24>,gpuSpriteSpanFn<0x25>,gpuSpriteSpanFn<0x26>,gpuSpriteSpanFn<0x27>, NULL,NULL,gpuSpriteSpanFn<0x2A>,gpuSpriteSpanFn<0x2B>, NULL,NULL,gpuSpriteSpanFn<0x2E>,gpuSpriteSpanFn<0x2F>, - NULL,NULL,gpuSpriteSpanFn<0x32>,gpuSpriteSpanFn<0x33>, NULL,NULL,gpuSpriteSpanFn<0x36>,gpuSpriteSpanFn<0x37>, NULL,NULL,gpuSpriteSpanFn<0x3A>,gpuSpriteSpanFn<0x3B>, NULL,NULL,gpuSpriteSpanFn<0x3E>,gpuSpriteSpanFn<0x3F>, - gpuSpriteSpanFn<0x40>,gpuSpriteSpanFn<0x41>,gpuSpriteSpanFn<0x42>,gpuSpriteSpanFn<0x43>, gpuSpriteSpanFn<0x44>,gpuSpriteSpanFn<0x45>,gpuSpriteSpanFn<0x46>,gpuSpriteSpanFn<0x47>, NULL,NULL,gpuSpriteSpanFn<0x4A>,gpuSpriteSpanFn<0x4B>, NULL,NULL,gpuSpriteSpanFn<0x4E>,gpuSpriteSpanFn<0x4F>, - NULL,NULL,gpuSpriteSpanFn<0x52>,gpuSpriteSpanFn<0x53>, NULL,NULL,gpuSpriteSpanFn<0x56>,gpuSpriteSpanFn<0x57>, NULL,NULL,gpuSpriteSpanFn<0x5A>,gpuSpriteSpanFn<0x5B>, NULL,NULL,gpuSpriteSpanFn<0x5E>,gpuSpriteSpanFn<0x5F>, - gpuSpriteSpanFn<0x60>,gpuSpriteSpanFn<0x61>,gpuSpriteSpanFn<0x62>,gpuSpriteSpanFn<0x63>, gpuSpriteSpanFn<0x64>,gpuSpriteSpanFn<0x65>,gpuSpriteSpanFn<0x66>,gpuSpriteSpanFn<0x67>, NULL,NULL,gpuSpriteSpanFn<0x6A>,gpuSpriteSpanFn<0x6B>, NULL,NULL,gpuSpriteSpanFn<0x6E>,gpuSpriteSpanFn<0x6F>, - NULL,NULL,gpuSpriteSpanFn<0x72>,gpuSpriteSpanFn<0x73>, NULL,NULL,gpuSpriteSpanFn<0x76>,gpuSpriteSpanFn<0x77>, NULL,NULL,gpuSpriteSpanFn<0x7A>,gpuSpriteSpanFn<0x7B>, NULL,NULL,gpuSpriteSpanFn<0x7E>,gpuSpriteSpanFn<0x7F>, - - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - gpuSpriteSpanFn<0xa0>,gpuSpriteSpanFn<0xa1>,gpuSpriteSpanFn<0xa2>,gpuSpriteSpanFn<0xa3>, gpuSpriteSpanFn<0xa4>,gpuSpriteSpanFn<0xa5>,gpuSpriteSpanFn<0xa6>,gpuSpriteSpanFn<0xa7>, NULL,NULL,gpuSpriteSpanFn<0xaA>,gpuSpriteSpanFn<0xaB>, NULL,NULL,gpuSpriteSpanFn<0xaE>,gpuSpriteSpanFn<0xaF>, - NULL,NULL,gpuSpriteSpanFn<0xb2>,gpuSpriteSpanFn<0xb3>, NULL,NULL,gpuSpriteSpanFn<0xb6>,gpuSpriteSpanFn<0xb7>, NULL,NULL,gpuSpriteSpanFn<0xbA>,gpuSpriteSpanFn<0xbB>, NULL,NULL,gpuSpriteSpanFn<0xbE>,gpuSpriteSpanFn<0xbF>, - gpuSpriteSpanFn<0xc0>,gpuSpriteSpanFn<0xc1>,gpuSpriteSpanFn<0xc2>,gpuSpriteSpanFn<0xc3>, gpuSpriteSpanFn<0xc4>,gpuSpriteSpanFn<0xc5>,gpuSpriteSpanFn<0xc6>,gpuSpriteSpanFn<0xc7>, NULL,NULL,gpuSpriteSpanFn<0xcA>,gpuSpriteSpanFn<0xcB>, NULL,NULL,gpuSpriteSpanFn<0xcE>,gpuSpriteSpanFn<0xcF>, - NULL,NULL,gpuSpriteSpanFn<0xd2>,gpuSpriteSpanFn<0xd3>, NULL,NULL,gpuSpriteSpanFn<0xd6>,gpuSpriteSpanFn<0xd7>, NULL,NULL,gpuSpriteSpanFn<0xdA>,gpuSpriteSpanFn<0xdB>, NULL,NULL,gpuSpriteSpanFn<0xdE>,gpuSpriteSpanFn<0xdF>, - gpuSpriteSpanFn<0xe0>,gpuSpriteSpanFn<0xe1>,gpuSpriteSpanFn<0xe2>,gpuSpriteSpanFn<0xe3>, gpuSpriteSpanFn<0xe4>,gpuSpriteSpanFn<0xe5>,gpuSpriteSpanFn<0xe6>,gpuSpriteSpanFn<0xe7>, NULL,NULL,gpuSpriteSpanFn<0xeA>,gpuSpriteSpanFn<0xeB>, NULL,NULL,gpuSpriteSpanFn<0xeE>,gpuSpriteSpanFn<0xeF>, - NULL,NULL,gpuSpriteSpanFn<0xf2>,gpuSpriteSpanFn<0xf3>, NULL,NULL,gpuSpriteSpanFn<0xf6>,gpuSpriteSpanFn<0xf7>, NULL,NULL,gpuSpriteSpanFn<0xfA>,gpuSpriteSpanFn<0xfB>, NULL,NULL,gpuSpriteSpanFn<0xfE>,gpuSpriteSpanFn<0xfF>, - - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - gpuSpriteSpanFn<0x120>,gpuSpriteSpanFn<0x121>,gpuSpriteSpanFn<0x122>,gpuSpriteSpanFn<0x123>, gpuSpriteSpanFn<0x124>,gpuSpriteSpanFn<0x125>,gpuSpriteSpanFn<0x126>,gpuSpriteSpanFn<0x127>, NULL,NULL,gpuSpriteSpanFn<0x12A>,gpuSpriteSpanFn<0x12B>, NULL,NULL,gpuSpriteSpanFn<0x12E>,gpuSpriteSpanFn<0x12F>, - NULL,NULL,gpuSpriteSpanFn<0x132>,gpuSpriteSpanFn<0x133>, NULL,NULL,gpuSpriteSpanFn<0x136>,gpuSpriteSpanFn<0x137>, NULL,NULL,gpuSpriteSpanFn<0x13A>,gpuSpriteSpanFn<0x13B>, NULL,NULL,gpuSpriteSpanFn<0x13E>,gpuSpriteSpanFn<0x13F>, - gpuSpriteSpanFn<0x140>,gpuSpriteSpanFn<0x141>,gpuSpriteSpanFn<0x142>,gpuSpriteSpanFn<0x143>, gpuSpriteSpanFn<0x144>,gpuSpriteSpanFn<0x145>,gpuSpriteSpanFn<0x146>,gpuSpriteSpanFn<0x147>, NULL,NULL,gpuSpriteSpanFn<0x14A>,gpuSpriteSpanFn<0x14B>, NULL,NULL,gpuSpriteSpanFn<0x14E>,gpuSpriteSpanFn<0x14F>, - NULL,NULL,gpuSpriteSpanFn<0x152>,gpuSpriteSpanFn<0x153>, NULL,NULL,gpuSpriteSpanFn<0x156>,gpuSpriteSpanFn<0x157>, NULL,NULL,gpuSpriteSpanFn<0x15A>,gpuSpriteSpanFn<0x15B>, NULL,NULL,gpuSpriteSpanFn<0x15E>,gpuSpriteSpanFn<0x15F>, - gpuSpriteSpanFn<0x160>,gpuSpriteSpanFn<0x161>,gpuSpriteSpanFn<0x162>,gpuSpriteSpanFn<0x163>, gpuSpriteSpanFn<0x164>,gpuSpriteSpanFn<0x165>,gpuSpriteSpanFn<0x166>,gpuSpriteSpanFn<0x167>, NULL,NULL,gpuSpriteSpanFn<0x16A>,gpuSpriteSpanFn<0x16B>, NULL,NULL,gpuSpriteSpanFn<0x16E>,gpuSpriteSpanFn<0x16F>, - NULL,NULL,gpuSpriteSpanFn<0x172>,gpuSpriteSpanFn<0x173>, NULL,NULL,gpuSpriteSpanFn<0x176>,gpuSpriteSpanFn<0x177>, NULL,NULL,gpuSpriteSpanFn<0x17A>,gpuSpriteSpanFn<0x17B>, NULL,NULL,gpuSpriteSpanFn<0x17E>,gpuSpriteSpanFn<0x17F>, - - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL, - gpuSpriteSpanFn<0x1a0>,gpuSpriteSpanFn<0x1a1>,gpuSpriteSpanFn<0x1a2>,gpuSpriteSpanFn<0x1a3>, gpuSpriteSpanFn<0x1a4>,gpuSpriteSpanFn<0x1a5>,gpuSpriteSpanFn<0x1a6>,gpuSpriteSpanFn<0x1a7>, NULL,NULL,gpuSpriteSpanFn<0x1aA>,gpuSpriteSpanFn<0x1aB>, NULL,NULL,gpuSpriteSpanFn<0x1aE>,gpuSpriteSpanFn<0x1aF>, - NULL,NULL,gpuSpriteSpanFn<0x1b2>,gpuSpriteSpanFn<0x1b3>, NULL,NULL,gpuSpriteSpanFn<0x1b6>,gpuSpriteSpanFn<0x1b7>, NULL,NULL,gpuSpriteSpanFn<0x1bA>,gpuSpriteSpanFn<0x1bB>, NULL,NULL,gpuSpriteSpanFn<0x1bE>,gpuSpriteSpanFn<0x1bF>, - gpuSpriteSpanFn<0x1c0>,gpuSpriteSpanFn<0x1c1>,gpuSpriteSpanFn<0x1c2>,gpuSpriteSpanFn<0x1c3>, gpuSpriteSpanFn<0x1c4>,gpuSpriteSpanFn<0x1c5>,gpuSpriteSpanFn<0x1c6>,gpuSpriteSpanFn<0x1c7>, NULL,NULL,gpuSpriteSpanFn<0x1cA>,gpuSpriteSpanFn<0x1cB>, NULL,NULL,gpuSpriteSpanFn<0x1cE>,gpuSpriteSpanFn<0x1cF>, - NULL,NULL,gpuSpriteSpanFn<0x1d2>,gpuSpriteSpanFn<0x1d3>, NULL,NULL,gpuSpriteSpanFn<0x1d6>,gpuSpriteSpanFn<0x1d7>, NULL,NULL,gpuSpriteSpanFn<0x1dA>,gpuSpriteSpanFn<0x1dB>, NULL,NULL,gpuSpriteSpanFn<0x1dE>,gpuSpriteSpanFn<0x1dF>, - gpuSpriteSpanFn<0x1e0>,gpuSpriteSpanFn<0x1e1>,gpuSpriteSpanFn<0x1e2>,gpuSpriteSpanFn<0x1e3>, gpuSpriteSpanFn<0x1e4>,gpuSpriteSpanFn<0x1e5>,gpuSpriteSpanFn<0x1e6>,gpuSpriteSpanFn<0x1e7>, NULL,NULL,gpuSpriteSpanFn<0x1eA>,gpuSpriteSpanFn<0x1eB>, NULL,NULL,gpuSpriteSpanFn<0x1eE>,gpuSpriteSpanFn<0x1eF>, - NULL,NULL,gpuSpriteSpanFn<0x1f2>,gpuSpriteSpanFn<0x1f3>, NULL,NULL,gpuSpriteSpanFn<0x1f6>,gpuSpriteSpanFn<0x1f7>, NULL,NULL,gpuSpriteSpanFn<0x1fA>,gpuSpriteSpanFn<0x1fB>, NULL,NULL,gpuSpriteSpanFn<0x1fE>,gpuSpriteSpanFn<0x1fF> +typedef void (*PS)(u16 *pDst, u32 count, u8* pTxt, u32 u0); + +// Template instantiation helper macros +#define TI(cf) gpuSpriteSpanFn<(cf)> +#define TN SpriteNULL +#define TIBLOCK(ub) \ + TN, TN, TN, TN, TN, TN, TN, TN, \ + TN, TN, TN, TN, TN, TN, TN, TN, \ + TN, TN, TN, TN, TN, TN, TN, TN, \ + TN, TN, TN, TN, TN, TN, TN, TN, \ + TI((ub)|0x20), TI((ub)|0x21), TI((ub)|0x22), TI((ub)|0x23), TI((ub)|0x24), TI((ub)|0x25), TI((ub)|0x26), TI((ub)|0x27), \ + TN, TN, TI((ub)|0x2a), TI((ub)|0x2b), TN, TN, TI((ub)|0x2e), TI((ub)|0x2f), \ + TN, TN, TI((ub)|0x32), TI((ub)|0x33), TN, TN, TI((ub)|0x36), TI((ub)|0x37), \ + TN, TN, TI((ub)|0x3a), TI((ub)|0x3b), TN, TN, TI((ub)|0x3e), TI((ub)|0x3f), \ + TI((ub)|0x40), TI((ub)|0x41), TI((ub)|0x42), TI((ub)|0x43), TI((ub)|0x44), TI((ub)|0x45), TI((ub)|0x46), TI((ub)|0x47), \ + TN, TN, TI((ub)|0x4a), TI((ub)|0x4b), TN, TN, TI((ub)|0x4e), TI((ub)|0x4f), \ + TN, TN, TI((ub)|0x52), TI((ub)|0x53), TN, TN, TI((ub)|0x56), TI((ub)|0x57), \ + TN, TN, TI((ub)|0x5a), TI((ub)|0x5b), TN, TN, TI((ub)|0x5e), TI((ub)|0x5f), \ + TI((ub)|0x60), TI((ub)|0x61), TI((ub)|0x62), TI((ub)|0x63), TI((ub)|0x64), TI((ub)|0x65), TI((ub)|0x66), TI((ub)|0x67), \ + TN, TN, TI((ub)|0x6a), TI((ub)|0x6b), TN, TN, TI((ub)|0x6e), TI((ub)|0x6f), \ + TN, TN, TI((ub)|0x72), TI((ub)|0x73), TN, TN, TI((ub)|0x76), TI((ub)|0x77), \ + TN, TN, TI((ub)|0x7a), TI((ub)|0x7b), TN, TN, TI((ub)|0x7e), TI((ub)|0x7f) + +const PS gpuSpriteSpanDrivers[256] = { + TIBLOCK(0<<8), TIBLOCK(1<<8) }; +#undef TI +#undef TN +#undef TIBLOCK + /////////////////////////////////////////////////////////////////////////////// // GPU Polygon innerloops generator -template -INLINE void gpuPolySpanFn(u16 *pDst, u32 count) + +//senquack - Newer version with following changes: +// * Adapted to work with new poly routings in gpu_raster_polygon.h +// adapted from DrHell GPU. They are less glitchy and use 22.10 +// fixed-point instead of original UNAI's 16.16. +// * Texture coordinates are no longer packed together into one +// unsigned int. This seems to lose too much accuracy (they each +// end up being only 8.7 fixed-point that way) and pixel-droupouts +// were noticeable both with original code and current DrHell +// adaptations. An example would be the sky in NFS3. Now, they are +// stored in separate ints, using separate masks. +// * Function is no longer INLINE, as it was always called +// through a function pointer. +// * Function now ensures the mask bit of source texture is preserved +// across calls to blending functions (Silent Hill rectangles fix) +// * November 2016: Large refactoring of blending/lighting when +// JohnnyF added dithering. See gpu_inner_quantization.h and +// relevant blend/light headers. +// (see README_senquack.txt) +template +static void gpuPolySpanFn(const gpu_senquack_t &gpu_senquack, u16 *pDst, u32 count) { - if (!TM) - { - // NO TEXTURE - if (!G) + // Blend func can save an operation if it knows uSrc MSB is unset. + // Untextured prims can always skip this (src color MSB is always 0). + // For textured prims, the generic lighting funcs always return it unset. (bonus!) + const bool skip_uSrc_mask = MSB_PRESERVED ? (!CF_TEXTMODE) : (!CF_TEXTMODE) || CF_LIGHT; + bool should_blend; + + u32 bMsk; if (CF_BLITMASK) bMsk = gpu_senquack.blit_mask; + + if (!CF_TEXTMODE) + { + if (!CF_GOURAUD) { - // NO GOURAUD - u16 data; - if (L) { u32 lCol=((u32)(b4<< 2)&(0x03ff)) | ((u32)(g4<<13)&(0x07ff<<10)) | ((u32)(r4<<24)&(0x07ff<<21)); gpuLightingRGB(data,lCol); } - else data=PixelData; - if ((!M)&&(!B)) - { - if (MB) { data = data | 0x8000; } - do { *pDst++ = data; } while (--count); - } - else if ((M)&&(!B)) - { - if (MB) { data = data | 0x8000; } - do { if (!(*pDst&0x8000)) { *pDst = data; } pDst++; } while (--count); - } - else - { - u16 uSrc; - u16 uDst; - u32 uMsk; if (BM==0) uMsk=0x7BDE; - do - { - // masking - uDst = *pDst; - if(M) { if (uDst&0x8000) goto endtile; } - uSrc = data; - // blend - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); - if (MB) { *pDst = uSrc | 0x8000; } - else { *pDst = uSrc; } - endtile: pDst++; - } - while (--count); - } + // UNTEXTURED, NO GOURAUD + const u16 pix15 = gpu_senquack.PixelData; + do { + uint_fast16_t uSrc, uDst; + + // NOTE: Don't enable CF_BLITMASK pixel skipping (speed hack) + // on untextured polys. It seems to do more harm than good: see + // gravestone text at end of Medieval intro sequence. -senquack + //if (CF_BLITMASK) { if ((bMsk>>((((uintptr_t)pDst)>>1)&7))&1) { goto endpolynotextnogou; } } + + if (CF_BLEND || CF_MASKCHECK) uDst = *pDst; + if (CF_MASKCHECK) { if (uDst&0x8000) { goto endpolynotextnogou; } } + + uSrc = pix15; + + if (CF_BLEND) + uSrc = gpuBlending(uSrc, uDst); + + if (CF_MASKSET) { *pDst = uSrc | 0x8000; } + else { *pDst = uSrc; } + +endpolynotextnogou: + pDst++; + } while(--count); } else { - // GOURAUD - u16 uDst; - u16 uSrc; - u32 linc=lInc; - u32 lCol=((u32)(b4>>14)&(0x03ff)) | ((u32)(g4>>3)&(0x07ff<<10)) | ((u32)(r4<<8)&(0x07ff<<21)); - u32 uMsk; if ((B)&&(BM==0)) uMsk=0x7BDE; - do - { - // masking - if(M) { uDst = *pDst; if (uDst&0x8000) goto endgou; } - // blend - if(B) - { - // light - gpuLightingRGB(uSrc,lCol); - if(!M) { uDst = *pDst; } - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); - } - else - { - // light - gpuLightingRGB(uSrc,lCol); + // UNTEXTURED, GOURAUD + u32 l_gCol = gpu_senquack.gCol; + u32 l_gInc = gpu_senquack.gInc; + + do { + uint_fast16_t uDst, uSrc; + + // See note in above loop regarding CF_BLITMASK + //if (CF_BLITMASK) { if ((bMsk>>((((uintptr_t)pDst)>>1)&7))&1) goto endpolynotextgou; } + + if (CF_BLEND || CF_MASKCHECK) uDst = *pDst; + if (CF_MASKCHECK) { if (uDst&0x8000) goto endpolynotextgou; } + + if (CF_DITHER) { + // GOURAUD, DITHER + + u32 uSrc24 = gpuLightingRGB24(l_gCol); + if (CF_BLEND) + uSrc24 = gpuBlending24(uSrc24, uDst); + uSrc = gpuColorQuantization24(uSrc24, pDst); + } else { + // GOURAUD, NO DITHER + + uSrc = gpuLightingRGB(l_gCol); + + if (CF_BLEND) + uSrc = gpuBlending(uSrc, uDst); } - if (MB) { *pDst = uSrc | 0x8000; } - else { *pDst = uSrc; } - endgou: pDst++; lCol=(lCol+linc); + + if (CF_MASKSET) { *pDst = uSrc | 0x8000; } + else { *pDst = uSrc; } + +endpolynotextgou: + pDst++; + l_gCol += l_gInc; } while (--count); } } else { - // TEXTURE - u16 uDst; - u16 uSrc; - u32 linc; if (L&&G) linc=lInc; - u32 tinc=tInc; - u32 tmsk=tMsk; - u32 tCor = ((u32)( u4<<7)&0x7fff0000) | ((u32)( v4>>9)&0x00007fff); tCor&= tmsk; - const u16* _TBA=TBA; - const u16* _CBA; if (TM!=3) _CBA=CBA; - u32 lCol; - if(L && !G) { lCol = ((u32)(b4<< 2)&(0x03ff)) | ((u32)(g4<<13)&(0x07ff<<10)) | ((u32)(r4<<24)&(0x07ff<<21)); } - else if(L && G) { lCol = ((u32)(b4>>14)&(0x03ff)) | ((u32)(g4>>3)&(0x07ff<<10)) | ((u32)(r4<<8)&(0x07ff<<21)); } - u32 uMsk; if ((B)&&(BM==0)) uMsk=0x7BDE; + // TEXTURED + + uint_fast16_t uDst, uSrc, srcMSB; + + //senquack - note: original UNAI code had gpu_senquack.{u4/v4} packed into + // one 32-bit unsigned int, but this proved to lose too much accuracy + // (pixel drouputs noticeable in NFS3 sky), so now are separate vars. + u32 l_u_msk = gpu_senquack.u_msk; u32 l_v_msk = gpu_senquack.v_msk; + u32 l_u = gpu_senquack.u & l_u_msk; u32 l_v = gpu_senquack.v & l_v_msk; + s32 l_u_inc = gpu_senquack.u_inc; s32 l_v_inc = gpu_senquack.v_inc; + + const u16* TBA_ = gpu_senquack.TBA; + const u16* CBA_; if (CF_TEXTMODE!=3) CBA_ = gpu_senquack.CBA; + + u8 r5, g5, b5; + u8 r8, g8, b8; + + u32 l_gInc, l_gCol; + + if (CF_LIGHT) { + if (CF_GOURAUD) { + l_gInc = gpu_senquack.gInc; + l_gCol = gpu_senquack.gCol; + } else { + if (CF_DITHER) { + r8 = gpu_senquack.r8; + g8 = gpu_senquack.g8; + b8 = gpu_senquack.b8; + } else { + r5 = gpu_senquack.r5; + g5 = gpu_senquack.g5; + b5 = gpu_senquack.b5; + } + } + } + do { - // masking - if(M) { uDst = *pDst; if (uDst&0x8000) goto endpoly; } - // texture - if (TM==1) { u32 tu=(tCor>>23); u32 tv=(tCor<<4)&(0xff<<11); u8 rgb=((u8*)_TBA)[tv+(tu>>1)]; uSrc=_CBA[(rgb>>((tu&1)<<2))&0xf]; if(!uSrc) goto endpoly; } - if (TM==2) { uSrc = _CBA[(((u8*)_TBA)[(tCor>>23)+((tCor<<4)&(0xff<<11))])]; if(!uSrc) goto endpoly; } - if (TM==3) { uSrc = _TBA[(tCor>>23)+((tCor<<3)&(0xff<<10))]; if(!uSrc) goto endpoly; } - // blend - if(B) - { - if (uSrc&0x8000) - { - // light - if(L) gpuLightingTXT(uSrc, lCol); - if(!M) { uDst = *pDst; } - if (BM==0) gpuBlending00(uSrc, uDst); - if (BM==1) gpuBlending01(uSrc, uDst); - if (BM==2) gpuBlending02(uSrc, uDst); - if (BM==3) gpuBlending03(uSrc, uDst); - } - else - { - // light - if(L) gpuLightingTXT(uSrc, lCol); - } + if (CF_BLITMASK) { if ((bMsk>>((((uintptr_t)pDst)>>1)&7))&1) goto endpolytext; } + if (CF_MASKCHECK || CF_BLEND) { uDst = *pDst; } + if (CF_MASKCHECK) if (uDst&0x8000) { goto endpolytext; } + + //senquack - adapted to work with new 22.10 fixed point routines: + // (UNAI originally used 16.16) + if (CF_TEXTMODE==1) { // 4bpp (CLUT) + u32 tu=(l_u>>10); + u32 tv=(l_v<<1)&(0xff<<11); + u8 rgb=((u8*)TBA_)[tv+(tu>>1)]; + uSrc=CBA_[(rgb>>((tu&1)<<2))&0xf]; + if (!uSrc) goto endpolytext; } - else + if (CF_TEXTMODE==2) { // 8bpp (CLUT) + uSrc = CBA_[(((u8*)TBA_)[(l_u>>10)+((l_v<<1)&(0xff<<11))])]; + if (!uSrc) goto endpolytext; + } + if (CF_TEXTMODE==3) { // 16bpp + uSrc = TBA_[(l_u>>10)+((l_v)&(0xff<<10))]; + if (!uSrc) goto endpolytext; + } + + // Save source MSB, as blending or lighting will not (Silent Hill) + if (CF_BLEND || CF_LIGHT) srcMSB = uSrc & 0x8000; + + // When textured, only dither when LIGHT (texture blend) is enabled + // LIGHT && BLEND => dither + // LIGHT && !BLEND => dither + //!LIGHT && BLEND => no dither + //!LIGHT && !BLEND => no dither + + if (CF_DITHER && CF_LIGHT) { + u32 uSrc24; + if ( CF_GOURAUD) + uSrc24 = gpuLightingTXT24Gouraud(uSrc, l_gCol); + if (!CF_GOURAUD) + uSrc24 = gpuLightingTXT24(uSrc, r8, g8, b8); + + if (CF_BLEND && srcMSB) + uSrc24 = gpuBlending24(uSrc24, uDst); + + uSrc = gpuColorQuantization24(uSrc24, pDst); + } else { - // light - if(L) { gpuLightingTXT(uSrc, lCol); } else if(!MB) { uSrc&= 0x7fff; } + if (CF_LIGHT) { + if ( CF_GOURAUD) + uSrc = gpuLightingTXTGouraud(uSrc, l_gCol); + if (!CF_GOURAUD) + uSrc = gpuLightingTXT(uSrc, r5, g5, b5); + } + + should_blend = MSB_PRESERVED ? uSrc & 0x8000 : srcMSB; + if (CF_BLEND && should_blend) + uSrc = gpuBlending(uSrc, uDst); } - if (MB) { *pDst = uSrc | 0x8000; } - else { *pDst = uSrc; } - endpoly: pDst++; - tCor=(tCor+tinc)&tmsk; - if (L&&G) lCol=(lCol+linc); + + if (CF_MASKSET) { *pDst = uSrc | 0x8000; } + else if (!MSB_PRESERVED && (CF_BLEND || CF_LIGHT)) { *pDst = uSrc | srcMSB; } + else { *pDst = uSrc; } +endpolytext: + pDst++; + l_u = (l_u + l_u_inc) & l_u_msk; + l_v = (l_v + l_v_inc) & l_v_msk; + if (CF_LIGHT && CF_GOURAUD) l_gCol += l_gInc; } while (--count); } } -/////////////////////////////////////////////////////////////////////////////// + +static void PolyNULL(const gpu_senquack_t &gpu_senquack, u16 *pDst, u32 count) +{ + #ifdef ENABLE_GPU_LOG_SUPPORT + fprintf(stdout,"PolyNULL()\n"); + #endif +} /////////////////////////////////////////////////////////////////////////////// // Polygon innerloops driver -typedef void (*PP)(u16 *pDst, u32 count); -const PP gpuPolySpanDrivers[512] = -{ - gpuPolySpanFn<0x00>,gpuPolySpanFn<0x01>,gpuPolySpanFn<0x02>,gpuPolySpanFn<0x03>, gpuPolySpanFn<0x04>,gpuPolySpanFn<0x05>,gpuPolySpanFn<0x06>,gpuPolySpanFn<0x07>, NULL,NULL,gpuPolySpanFn<0x0A>,gpuPolySpanFn<0x0B>, NULL,NULL,gpuPolySpanFn<0x0E>,gpuPolySpanFn<0x0F>, - NULL,NULL,gpuPolySpanFn<0x12>,gpuPolySpanFn<0x13>, NULL,NULL,gpuPolySpanFn<0x16>,gpuPolySpanFn<0x17>, NULL,NULL,gpuPolySpanFn<0x1A>,gpuPolySpanFn<0x1B>, NULL,NULL,gpuPolySpanFn<0x1E>,gpuPolySpanFn<0x1F>, - gpuPolySpanFn<0x20>,gpuPolySpanFn<0x21>,gpuPolySpanFn<0x22>,gpuPolySpanFn<0x23>, gpuPolySpanFn<0x24>,gpuPolySpanFn<0x25>,gpuPolySpanFn<0x26>,gpuPolySpanFn<0x27>, NULL,NULL,gpuPolySpanFn<0x2A>,gpuPolySpanFn<0x2B>, NULL,NULL,gpuPolySpanFn<0x2E>,gpuPolySpanFn<0x2F>, - NULL,NULL,gpuPolySpanFn<0x32>,gpuPolySpanFn<0x33>, NULL,NULL,gpuPolySpanFn<0x36>,gpuPolySpanFn<0x37>, NULL,NULL,gpuPolySpanFn<0x3A>,gpuPolySpanFn<0x3B>, NULL,NULL,gpuPolySpanFn<0x3E>,gpuPolySpanFn<0x3F>, - gpuPolySpanFn<0x40>,gpuPolySpanFn<0x41>,gpuPolySpanFn<0x42>,gpuPolySpanFn<0x43>, gpuPolySpanFn<0x44>,gpuPolySpanFn<0x45>,gpuPolySpanFn<0x46>,gpuPolySpanFn<0x47>, NULL,NULL,gpuPolySpanFn<0x4A>,gpuPolySpanFn<0x4B>, NULL,NULL,gpuPolySpanFn<0x4E>,gpuPolySpanFn<0x4F>, - NULL,NULL,gpuPolySpanFn<0x52>,gpuPolySpanFn<0x53>, NULL,NULL,gpuPolySpanFn<0x56>,gpuPolySpanFn<0x57>, NULL,NULL,gpuPolySpanFn<0x5A>,gpuPolySpanFn<0x5B>, NULL,NULL,gpuPolySpanFn<0x5E>,gpuPolySpanFn<0x5F>, - gpuPolySpanFn<0x60>,gpuPolySpanFn<0x61>,gpuPolySpanFn<0x62>,gpuPolySpanFn<0x63>, gpuPolySpanFn<0x64>,gpuPolySpanFn<0x65>,gpuPolySpanFn<0x66>,gpuPolySpanFn<0x67>, NULL,NULL,gpuPolySpanFn<0x6A>,gpuPolySpanFn<0x6B>, NULL,NULL,gpuPolySpanFn<0x6E>,gpuPolySpanFn<0x6F>, - NULL,NULL,gpuPolySpanFn<0x72>,gpuPolySpanFn<0x73>, NULL,NULL,gpuPolySpanFn<0x76>,gpuPolySpanFn<0x77>, NULL,NULL,gpuPolySpanFn<0x7A>,gpuPolySpanFn<0x7B>, NULL,NULL,gpuPolySpanFn<0x7E>,gpuPolySpanFn<0x7F>, - - NULL,gpuPolySpanFn<0x81>,NULL,gpuPolySpanFn<0x83>, NULL,gpuPolySpanFn<0x85>,NULL,gpuPolySpanFn<0x87>, NULL,NULL,NULL,gpuPolySpanFn<0x8B>, NULL,NULL,NULL,gpuPolySpanFn<0x8F>, - NULL,NULL,NULL,gpuPolySpanFn<0x93>, NULL,NULL,NULL,gpuPolySpanFn<0x97>, NULL,NULL,NULL,gpuPolySpanFn<0x9B>, NULL,NULL,NULL,gpuPolySpanFn<0x9F>, - NULL,gpuPolySpanFn<0xa1>,NULL,gpuPolySpanFn<0xa3>, NULL,gpuPolySpanFn<0xa5>,NULL,gpuPolySpanFn<0xa7>, NULL,NULL,NULL,gpuPolySpanFn<0xaB>, NULL,NULL,NULL,gpuPolySpanFn<0xaF>, - NULL,NULL,NULL,gpuPolySpanFn<0xb3>, NULL,NULL,NULL,gpuPolySpanFn<0xb7>, NULL,NULL,NULL,gpuPolySpanFn<0xbB>, NULL,NULL,NULL,gpuPolySpanFn<0xbF>, - NULL,gpuPolySpanFn<0xc1>,NULL,gpuPolySpanFn<0xc3>, NULL,gpuPolySpanFn<0xc5>,NULL,gpuPolySpanFn<0xc7>, NULL,NULL,NULL,gpuPolySpanFn<0xcB>, NULL,NULL,NULL,gpuPolySpanFn<0xcF>, - NULL,NULL,NULL,gpuPolySpanFn<0xd3>, NULL,NULL,NULL,gpuPolySpanFn<0xd7>, NULL,NULL,NULL,gpuPolySpanFn<0xdB>, NULL,NULL,NULL,gpuPolySpanFn<0xdF>, - NULL,gpuPolySpanFn<0xe1>,NULL,gpuPolySpanFn<0xe3>, NULL,gpuPolySpanFn<0xe5>,NULL,gpuPolySpanFn<0xe7>, NULL,NULL,NULL,gpuPolySpanFn<0xeB>, NULL,NULL,NULL,gpuPolySpanFn<0xeF>, - NULL,NULL,NULL,gpuPolySpanFn<0xf3>, NULL,NULL,NULL,gpuPolySpanFn<0xf7>, NULL,NULL,NULL,gpuPolySpanFn<0xfB>, NULL,NULL,NULL,gpuPolySpanFn<0xfF>, - - gpuPolySpanFn<0x100>,gpuPolySpanFn<0x101>,gpuPolySpanFn<0x102>,gpuPolySpanFn<0x103>, gpuPolySpanFn<0x104>,gpuPolySpanFn<0x105>,gpuPolySpanFn<0x106>,gpuPolySpanFn<0x107>, NULL,NULL,gpuPolySpanFn<0x10A>,gpuPolySpanFn<0x10B>, NULL,NULL,gpuPolySpanFn<0x10E>,gpuPolySpanFn<0x10F>, - NULL,NULL,gpuPolySpanFn<0x112>,gpuPolySpanFn<0x113>, NULL,NULL,gpuPolySpanFn<0x116>,gpuPolySpanFn<0x117>, NULL,NULL,gpuPolySpanFn<0x11A>,gpuPolySpanFn<0x11B>, NULL,NULL,gpuPolySpanFn<0x11E>,gpuPolySpanFn<0x11F>, - gpuPolySpanFn<0x120>,gpuPolySpanFn<0x121>,gpuPolySpanFn<0x122>,gpuPolySpanFn<0x123>, gpuPolySpanFn<0x124>,gpuPolySpanFn<0x125>,gpuPolySpanFn<0x126>,gpuPolySpanFn<0x127>, NULL,NULL,gpuPolySpanFn<0x12A>,gpuPolySpanFn<0x12B>, NULL,NULL,gpuPolySpanFn<0x12E>,gpuPolySpanFn<0x12F>, - NULL,NULL,gpuPolySpanFn<0x132>,gpuPolySpanFn<0x133>, NULL,NULL,gpuPolySpanFn<0x136>,gpuPolySpanFn<0x137>, NULL,NULL,gpuPolySpanFn<0x13A>,gpuPolySpanFn<0x13B>, NULL,NULL,gpuPolySpanFn<0x13E>,gpuPolySpanFn<0x13F>, - gpuPolySpanFn<0x140>,gpuPolySpanFn<0x141>,gpuPolySpanFn<0x142>,gpuPolySpanFn<0x143>, gpuPolySpanFn<0x144>,gpuPolySpanFn<0x145>,gpuPolySpanFn<0x146>,gpuPolySpanFn<0x147>, NULL,NULL,gpuPolySpanFn<0x14A>,gpuPolySpanFn<0x14B>, NULL,NULL,gpuPolySpanFn<0x14E>,gpuPolySpanFn<0x14F>, - NULL,NULL,gpuPolySpanFn<0x152>,gpuPolySpanFn<0x153>, NULL,NULL,gpuPolySpanFn<0x156>,gpuPolySpanFn<0x157>, NULL,NULL,gpuPolySpanFn<0x15A>,gpuPolySpanFn<0x15B>, NULL,NULL,gpuPolySpanFn<0x15E>,gpuPolySpanFn<0x15F>, - gpuPolySpanFn<0x160>,gpuPolySpanFn<0x161>,gpuPolySpanFn<0x162>,gpuPolySpanFn<0x163>, gpuPolySpanFn<0x164>,gpuPolySpanFn<0x165>,gpuPolySpanFn<0x166>,gpuPolySpanFn<0x167>, NULL,NULL,gpuPolySpanFn<0x16A>,gpuPolySpanFn<0x16B>, NULL,NULL,gpuPolySpanFn<0x16E>,gpuPolySpanFn<0x16F>, - NULL,NULL,gpuPolySpanFn<0x172>,gpuPolySpanFn<0x173>, NULL,NULL,gpuPolySpanFn<0x176>,gpuPolySpanFn<0x177>, NULL,NULL,gpuPolySpanFn<0x17A>,gpuPolySpanFn<0x17B>, NULL,NULL,gpuPolySpanFn<0x17E>,gpuPolySpanFn<0x17F>, - - NULL,gpuPolySpanFn<0x181>,NULL,gpuPolySpanFn<0x183>, NULL,gpuPolySpanFn<0x185>,NULL,gpuPolySpanFn<0x187>, NULL,NULL,NULL,gpuPolySpanFn<0x18B>, NULL,NULL,NULL,gpuPolySpanFn<0x18F>, - NULL,NULL,NULL,gpuPolySpanFn<0x193>, NULL,NULL,NULL,gpuPolySpanFn<0x197>, NULL,NULL,NULL,gpuPolySpanFn<0x19B>, NULL,NULL,NULL,gpuPolySpanFn<0x19F>, - NULL,gpuPolySpanFn<0x1a1>,NULL,gpuPolySpanFn<0x1a3>, NULL,gpuPolySpanFn<0x1a5>,NULL,gpuPolySpanFn<0x1a7>, NULL,NULL,NULL,gpuPolySpanFn<0x1aB>, NULL,NULL,NULL,gpuPolySpanFn<0x1aF>, - NULL,NULL,NULL,gpuPolySpanFn<0x1b3>, NULL,NULL,NULL,gpuPolySpanFn<0x1b7>, NULL,NULL,NULL,gpuPolySpanFn<0x1bB>, NULL,NULL,NULL,gpuPolySpanFn<0x1bF>, - NULL,gpuPolySpanFn<0x1c1>,NULL,gpuPolySpanFn<0x1c3>, NULL,gpuPolySpanFn<0x1c5>,NULL,gpuPolySpanFn<0x1c7>, NULL,NULL,NULL,gpuPolySpanFn<0x1cB>, NULL,NULL,NULL,gpuPolySpanFn<0x1cF>, - NULL,NULL,NULL,gpuPolySpanFn<0x1d3>, NULL,NULL,NULL,gpuPolySpanFn<0x1d7>, NULL,NULL,NULL,gpuPolySpanFn<0x1dB>, NULL,NULL,NULL,gpuPolySpanFn<0x1dF>, - NULL,gpuPolySpanFn<0x1e1>,NULL,gpuPolySpanFn<0x1e3>, NULL,gpuPolySpanFn<0x1e5>,NULL,gpuPolySpanFn<0x1e7>, NULL,NULL,NULL,gpuPolySpanFn<0x1eB>, NULL,NULL,NULL,gpuPolySpanFn<0x1eF>, - NULL,NULL,NULL,gpuPolySpanFn<0x1f3>, NULL,NULL,NULL,gpuPolySpanFn<0x1f7>, NULL,NULL,NULL,gpuPolySpanFn<0x1fB>, NULL,NULL,NULL,gpuPolySpanFn<0x1fF> +typedef void (*PP)(const gpu_senquack_t &gpu_senquack, u16 *pDst, u32 count); + +// Template instantiation helper macros +#define TI(cf) gpuPolySpanFn<(cf)> +#define TN PolyNULL +#define TIBLOCK(ub) \ + TI((ub)|0x00), TI((ub)|0x01), TI((ub)|0x02), TI((ub)|0x03), TI((ub)|0x04), TI((ub)|0x05), TI((ub)|0x06), TI((ub)|0x07), \ + TN, TN, TI((ub)|0x0a), TI((ub)|0x0b), TN, TN, TI((ub)|0x0e), TI((ub)|0x0f), \ + TN, TN, TI((ub)|0x12), TI((ub)|0x13), TN, TN, TI((ub)|0x16), TI((ub)|0x17), \ + TN, TN, TI((ub)|0x1a), TI((ub)|0x1b), TN, TN, TI((ub)|0x1e), TI((ub)|0x1f), \ + TI((ub)|0x20), TI((ub)|0x21), TI((ub)|0x22), TI((ub)|0x23), TI((ub)|0x24), TI((ub)|0x25), TI((ub)|0x26), TI((ub)|0x27), \ + TN, TN, TI((ub)|0x2a), TI((ub)|0x2b), TN, TN, TI((ub)|0x2e), TI((ub)|0x2f), \ + TN, TN, TI((ub)|0x32), TI((ub)|0x33), TN, TN, TI((ub)|0x36), TI((ub)|0x37), \ + TN, TN, TI((ub)|0x3a), TI((ub)|0x3b), TN, TN, TI((ub)|0x3e), TI((ub)|0x3f), \ + TI((ub)|0x40), TI((ub)|0x41), TI((ub)|0x42), TI((ub)|0x43), TI((ub)|0x44), TI((ub)|0x45), TI((ub)|0x46), TI((ub)|0x47), \ + TN, TN, TI((ub)|0x4a), TI((ub)|0x4b), TN, TN, TI((ub)|0x4e), TI((ub)|0x4f), \ + TN, TN, TI((ub)|0x52), TI((ub)|0x53), TN, TN, TI((ub)|0x56), TI((ub)|0x57), \ + TN, TN, TI((ub)|0x5a), TI((ub)|0x5b), TN, TN, TI((ub)|0x5e), TI((ub)|0x5f), \ + TI((ub)|0x60), TI((ub)|0x61), TI((ub)|0x62), TI((ub)|0x63), TI((ub)|0x64), TI((ub)|0x65), TI((ub)|0x66), TI((ub)|0x67), \ + TN, TN, TI((ub)|0x6a), TI((ub)|0x6b), TN, TN, TI((ub)|0x6e), TI((ub)|0x6f), \ + TN, TN, TI((ub)|0x72), TI((ub)|0x73), TN, TN, TI((ub)|0x76), TI((ub)|0x77), \ + TN, TN, TI((ub)|0x7a), TI((ub)|0x7b), TN, TN, TI((ub)|0x7e), TI((ub)|0x7f), \ + TN, TI((ub)|0x81), TN, TI((ub)|0x83), TN, TI((ub)|0x85), TN, TI((ub)|0x87), \ + TN, TN, TN, TI((ub)|0x8b), TN, TN, TN, TI((ub)|0x8f), \ + TN, TN, TN, TI((ub)|0x93), TN, TN, TN, TI((ub)|0x97), \ + TN, TN, TN, TI((ub)|0x9b), TN, TN, TN, TI((ub)|0x9f), \ + TN, TI((ub)|0xa1), TN, TI((ub)|0xa3), TN, TI((ub)|0xa5), TN, TI((ub)|0xa7), \ + TN, TN, TN, TI((ub)|0xab), TN, TN, TN, TI((ub)|0xaf), \ + TN, TN, TN, TI((ub)|0xb3), TN, TN, TN, TI((ub)|0xb7), \ + TN, TN, TN, TI((ub)|0xbb), TN, TN, TN, TI((ub)|0xbf), \ + TN, TI((ub)|0xc1), TN, TI((ub)|0xc3), TN, TI((ub)|0xc5), TN, TI((ub)|0xc7), \ + TN, TN, TN, TI((ub)|0xcb), TN, TN, TN, TI((ub)|0xcf), \ + TN, TN, TN, TI((ub)|0xd3), TN, TN, TN, TI((ub)|0xd7), \ + TN, TN, TN, TI((ub)|0xdb), TN, TN, TN, TI((ub)|0xdf), \ + TN, TI((ub)|0xe1), TN, TI((ub)|0xe3), TN, TI((ub)|0xe5), TN, TI((ub)|0xe7), \ + TN, TN, TN, TI((ub)|0xeb), TN, TN, TN, TI((ub)|0xef), \ + TN, TN, TN, TI((ub)|0xf3), TN, TN, TN, TI((ub)|0xf7), \ + TN, TN, TN, TI((ub)|0xfb), TN, TN, TN, TI((ub)|0xff) + +const PP gpuPolySpanDrivers[2048] = { + TIBLOCK(0<<8), TIBLOCK(1<<8), TIBLOCK(2<<8), TIBLOCK(3<<8), + TIBLOCK(4<<8), TIBLOCK(5<<8), TIBLOCK(6<<8), TIBLOCK(7<<8) }; + +#undef TI +#undef TN +#undef TIBLOCK + +#endif /* __GPU_UNAI_GPU_INNER_H__ */