X-Git-Url: http://git.xonotic.org/?a=blobdiff_plain;f=dpsoftrast.c;h=179a4d0c5fe93cae3b4668e1f4d9221c6a79082c;hb=1b6062f6c60a2525501e1133052d87307ee02e36;hp=84785eec18f06ec32e67c6b201ab103a435b1e7d;hpb=6e32ef9b2e2b8ed0cb9862869d18e80d60c7b958;p=xonotic%2Fdarkplaces.git diff --git a/dpsoftrast.c b/dpsoftrast.c index 84785eec..179a4d0c 100644 --- a/dpsoftrast.c +++ b/dpsoftrast.c @@ -3,12 +3,11 @@ #define _USE_MATH_DEFINES #include #include "quakedef.h" +#include "thread.h" #include "dpsoftrast.h" -//#define USETHREADS -#ifdef USETHREADS -#include -#include +#ifdef _MSC_VER +#pragma warning(disable : 4324) #endif #ifndef __cplusplus @@ -18,36 +17,86 @@ typedef qboolean bool; #define ALIGN_SIZE 16 #define ATOMIC_SIZE 32 -#if defined(__GNUC__) -#define ALIGN(var) var __attribute__((__aligned__(16))) -#define ATOMIC(var) var __attribute__((__aligned__(32))) -#define MEMORY_BARRIER (_mm_sfence()) -//(__sync_synchronize()) -#elif defined(_MSC_VER) -#define ALIGN(var) __declspec(align(16)) var -#define ATOMIC(var) __declspec(align(32)) var -#define MEMORY_BARRIER (_mm_sfence()) -//(MemoryBarrier()) -#else +#ifdef SSE_POSSIBLE + #if defined(__APPLE__) + #include + #define ALIGN(var) var __attribute__((__aligned__(16))) + #define ATOMIC(var) var __attribute__((__aligned__(32))) + #define MEMORY_BARRIER (_mm_sfence()) + #define ATOMIC_COUNTER volatile int32_t + #define ATOMIC_INCREMENT(counter) (OSAtomicIncrement32Barrier(&(counter))) + #define ATOMIC_DECREMENT(counter) (OSAtomicDecrement32Barrier(&(counter))) + #define ATOMIC_ADD(counter, val) ((void)OSAtomicAdd32Barrier((val), &(counter))) + #elif defined(__GNUC__) && defined(WIN32) + #define ALIGN(var) var __attribute__((__aligned__(16))) + #define ATOMIC(var) var __attribute__((__aligned__(32))) + #define MEMORY_BARRIER (_mm_sfence()) + //(__sync_synchronize()) + #define ATOMIC_COUNTER volatile LONG + // this LONG * cast serves to fix an issue with broken mingw + // packages on Ubuntu; these only declare the function to take + // a LONG *, causing a compile error here. This seems to be + // error- and warn-free on platforms that DO declare + // InterlockedIncrement correctly, like mingw on Windows. + #define ATOMIC_INCREMENT(counter) (InterlockedIncrement((LONG *) &(counter))) + #define ATOMIC_DECREMENT(counter) (InterlockedDecrement((LONG *) &(counter))) + #define ATOMIC_ADD(counter, val) ((void)InterlockedExchangeAdd((LONG *) &(counter), (val))) + #elif defined(__GNUC__) + #define ALIGN(var) var __attribute__((__aligned__(16))) + #define ATOMIC(var) var __attribute__((__aligned__(32))) + #define MEMORY_BARRIER (_mm_sfence()) + //(__sync_synchronize()) + #define ATOMIC_COUNTER volatile int + #define ATOMIC_INCREMENT(counter) (__sync_add_and_fetch(&(counter), 1)) + #define ATOMIC_DECREMENT(counter) (__sync_add_and_fetch(&(counter), -1)) + #define ATOMIC_ADD(counter, val) ((void)__sync_fetch_and_add(&(counter), (val))) + #elif defined(_MSC_VER) + #define ALIGN(var) __declspec(align(16)) var + #define ATOMIC(var) __declspec(align(32)) var + #define MEMORY_BARRIER (_mm_sfence()) + //(MemoryBarrier()) + #define ATOMIC_COUNTER volatile LONG + #define ATOMIC_INCREMENT(counter) (InterlockedIncrement(&(counter))) + #define ATOMIC_DECREMENT(counter) (InterlockedDecrement(&(counter))) + #define ATOMIC_ADD(counter, val) ((void)InterlockedExchangeAdd(&(counter), (val))) + #endif +#endif + +#ifndef ALIGN #define ALIGN(var) var +#endif +#ifndef ATOMIC #define ATOMIC(var) var -#define MEMORY_BARRIER ((void)0) #endif - -#ifndef USETHREADS -#undef MEMORY_BARRIER +#ifndef MEMORY_BARRIER #define MEMORY_BARRIER ((void)0) #endif +#ifndef ATOMIC_COUNTER +#define ATOMIC_COUNTER int +#endif +#ifndef ATOMIC_INCREMENT +#define ATOMIC_INCREMENT(counter) (++(counter)) +#endif +#ifndef ATOMIC_DECREMENT +#define ATOMIC_DECREMENT(counter) (--(counter)) +#endif +#ifndef ATOMIC_ADD +#define ATOMIC_ADD(counter, val) ((void)((counter) += (val))) +#endif -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE #include +#if defined(__GNUC__) && (__GNUC < 4 || __GNUC_MINOR__ < 6) && !defined(__clang__) + #define _mm_cvtss_f32(val) (__builtin_ia32_vec_ext_v4sf ((__v4sf)(val), 0)) +#endif + #define MM_MALLOC(size) _mm_malloc(size, ATOMIC_SIZE) static void *MM_CALLOC(size_t nmemb, size_t size) { void *ptr = _mm_malloc(nmemb*size, ATOMIC_SIZE); - if(ptr != NULL) memset(ptr, 0, nmemb*size); + if (ptr != NULL) memset(ptr, 0, nmemb*size); return ptr; } @@ -84,6 +133,7 @@ typedef struct DPSOFTRAST_Texture_s DPSOFTRAST_TEXTURE_FILTER filter; int mipmaps; int size; + ATOMIC_COUNTER binds; unsigned char *bytes; int mipmap[DPSOFTRAST_MAXMIPMAPS][5]; } @@ -94,7 +144,8 @@ DPSOFTRAST_Texture; typedef COMMAND_ALIGN(struct DPSOFTRAST_Command_s { - int opcode; + unsigned char opcode; + unsigned short commandsize; } DPSOFTRAST_Command); @@ -104,11 +155,13 @@ enum { DPSOFTRAST_OPCODE_Reset = 0 }; enum { DPSOFTRAST_OPCODE_##name = opcodeval }; \ typedef COMMAND_ALIGN(struct DPSOFTRAST_Command_##name##_s \ { \ - int opcode; \ + unsigned char opcode; \ + unsigned short commandsize; \ fields \ } DPSOFTRAST_Command_##name ); #define DPSOFTRAST_DRAW_MAXCOMMANDPOOL 2097152 +#define DPSOFTRAST_DRAW_MAXCOMMANDSIZE 16384 typedef ATOMIC(struct DPSOFTRAST_State_Command_Pool_s { @@ -120,14 +173,8 @@ DPSOFTRAST_State_Command_Pool); typedef ATOMIC(struct DPSOFTRAST_State_Triangle_s { - int commandoffset; unsigned char mip[DPSOFTRAST_MAXTEXTUREUNITS]; // texcoord to screen space density values (for picking mipmap of textures) - int starty; - int endy; - int numpoints; float w[3]; - ALIGN(float coords[4][4]); - ALIGN(int ycoords[4]); ALIGN(float attribs[DPSOFTRAST_ARRAY_TOTAL][3][4]); } DPSOFTRAST_State_Triangle); @@ -156,25 +203,17 @@ typedef ALIGN(struct DPSOFTRAST_State_Span_s int triangle; // triangle this span was generated by int x; // framebuffer x coord int y; // framebuffer y coord - int length; // pixel count int startx; // usable range (according to pixelmask) int endx; // usable range (according to pixelmask) unsigned char *pixelmask; // true for pixels that passed depth test, false for others + int depthbase; // depthbuffer value at x (add depthslope*startx to get first pixel's depthbuffer value) + int depthslope; // depthbuffer value pixel delta } DPSOFTRAST_State_Span); #define DPSOFTRAST_DRAW_MAXSPANS 1024 - -#define DPSOFTRAST_DRAW_MAXTRIANGLEPOOL 4096 -#define DPSOFTRAST_DRAW_FLUSHPROCESSTRIANGLES 64 - -typedef ATOMIC(struct DPSOFTRAST_State_Triangle_Pool_s -{ - int freetriangle; - int usedtriangles; - ATOMIC(DPSOFTRAST_State_Triangle triangles[DPSOFTRAST_DRAW_MAXTRIANGLEPOOL]); -} -DPSOFTRAST_State_Triangle_Pool); +#define DPSOFTRAST_DRAW_MAXTRIANGLES 128 +#define DPSOFTRAST_DRAW_MAXSPANLENGTH 256 #define DPSOFTRAST_VALIDATE_FB 1 #define DPSOFTRAST_VALIDATE_DEPTHFUNC 2 @@ -192,17 +231,17 @@ typedef enum DPSOFTRAST_BLENDMODE_e DPSOFTRAST_BLENDMODE_MUL2, DPSOFTRAST_BLENDMODE_SUBALPHA, DPSOFTRAST_BLENDMODE_PSEUDOALPHA, + DPSOFTRAST_BLENDMODE_INVADD, DPSOFTRAST_BLENDMODE_TOTAL } DPSOFTRAST_BLENDMODE; typedef ATOMIC(struct DPSOFTRAST_State_Thread_s { -#ifdef USETHREADS - SDL_Thread *thread; -#endif + void *thread; int index; + int cullface; int colormask[4]; int blendfunc[2]; int blendsubtract; @@ -210,16 +249,16 @@ typedef ATOMIC(struct DPSOFTRAST_State_Thread_s int depthtest; int depthfunc; int scissortest; - int alphatest; - int alphafunc; - float alphavalue; - int scissor[4]; int viewport[4]; + int scissor[4]; float depthrange[2]; float polygonoffset[2]; + float clipplane[4]; + ALIGN(float fb_clipplane[4]); int shader_mode; int shader_permutation; + int shader_exactspecularmath; DPSOFTRAST_Texture *texbound[DPSOFTRAST_MAXTEXTUREUNITS]; @@ -231,7 +270,9 @@ typedef ATOMIC(struct DPSOFTRAST_State_Thread_s // derived values (DPSOFTRAST_VALIDATE_FB) int fb_colormask; - int fb_clearscissor[4]; + int fb_scissor[4]; + ALIGN(float fb_viewportcenter[4]); + ALIGN(float fb_viewportscale[4]); // derived values (DPSOFTRAST_VALIDATE_DEPTHFUNC) int fb_depthfunc; @@ -239,16 +280,25 @@ typedef ATOMIC(struct DPSOFTRAST_State_Thread_s // derived values (DPSOFTRAST_VALIDATE_BLENDFUNC) int fb_blendmode; - ATOMIC(int commandoffset); - int triangleoffset; + // band boundaries + int miny1; + int maxy1; + int miny2; + int maxy2; - bool waiting; -#ifdef USETHREADS - SDL_cond *waitcond; -#endif + ATOMIC(volatile int commandoffset); + + volatile bool waiting; + volatile bool starving; + void *waitcond; + void *drawcond; + void *drawmutex; int numspans; + int numtriangles; DPSOFTRAST_State_Span spans[DPSOFTRAST_DRAW_MAXSPANS]; + DPSOFTRAST_State_Triangle triangles[DPSOFTRAST_DRAW_MAXTRIANGLES]; + unsigned char pixelmaskarray[DPSOFTRAST_DRAW_MAXSPANLENGTH+4]; // LordHavoc: padded to allow some termination bytes } DPSOFTRAST_State_Thread); @@ -267,8 +317,6 @@ typedef ATOMIC(struct DPSOFTRAST_State_s ALIGN(float uniform4f[DPSOFTRAST_UNIFORM_TOTAL*4]); int uniform1i[DPSOFTRAST_UNIFORM_TOTAL]; - int cullface; - const float *pointer_vertex3f; const float *pointer_color4f; const unsigned char *pointer_color4ub; @@ -279,14 +327,17 @@ typedef ATOMIC(struct DPSOFTRAST_State_s int components_texcoord[DPSOFTRAST_MAXTEXCOORDARRAYS]; DPSOFTRAST_Texture *texbound[DPSOFTRAST_MAXTEXTUREUNITS]; + int firstvertex; int numvertices; - int maxvertices; - float *in_array4f[DPSOFTRAST_ARRAY_TOTAL]; float *post_array4f[DPSOFTRAST_ARRAY_TOTAL]; float *screencoord4f; - + int drawstarty; + int drawendy; + int drawclipped; + int shader_mode; int shader_permutation; + int shader_exactspecularmath; int texture_max; int texture_end; @@ -298,31 +349,65 @@ typedef ATOMIC(struct DPSOFTRAST_State_s // error reporting const char *errorstring; + bool usethreads; + int interlace; int numthreads; DPSOFTRAST_State_Thread *threads; -#ifdef USETHREADS - SDL_mutex *trianglemutex; - SDL_cond *trianglecond; -#endif - ATOMIC(int drawtriangle); + ATOMIC(volatile int drawcommand); DPSOFTRAST_State_Command_Pool commandpool; - DPSOFTRAST_State_Triangle_Pool trianglepool; } DPSOFTRAST_State); DPSOFTRAST_State dpsoftrast; -extern int dpsoftrast_test; - #define DPSOFTRAST_DEPTHSCALE (1024.0f*1048576.0f) #define DPSOFTRAST_DEPTHOFFSET (128.0f) #define DPSOFTRAST_BGRA8_FROM_RGBA32F(r,g,b,a) (((int)(r * 255.0f + 0.5f) << 16) | ((int)(g * 255.0f + 0.5f) << 8) | (int)(b * 255.0f + 0.5f) | ((int)(a * 255.0f + 0.5f) << 24)) #define DPSOFTRAST_DEPTH32_FROM_DEPTH32F(d) ((int)(DPSOFTRAST_DEPTHSCALE * (1-d))) -#define DPSOFTRAST_DRAW_MAXSPANLENGTH 256 -void DPSOFTRAST_RecalcFB(DPSOFTRAST_State_Thread *thread) +static void DPSOFTRAST_Draw_DepthTest(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_State_Span *span); +static void DPSOFTRAST_Draw_DepthWrite(const DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Span *span); + +static void DPSOFTRAST_RecalcViewport(const int *viewport, float *fb_viewportcenter, float *fb_viewportscale) +{ + fb_viewportcenter[1] = viewport[0] + 0.5f * viewport[2] - 0.5f; + fb_viewportcenter[2] = dpsoftrast.fb_height - viewport[1] - 0.5f * viewport[3] - 0.5f; + fb_viewportcenter[3] = 0.5f; + fb_viewportcenter[0] = 0.0f; + fb_viewportscale[1] = 0.5f * viewport[2]; + fb_viewportscale[2] = -0.5f * viewport[3]; + fb_viewportscale[3] = 0.5f; + fb_viewportscale[0] = 1.0f; +} + +static void DPSOFTRAST_RecalcThread(DPSOFTRAST_State_Thread *thread) +{ + if (dpsoftrast.interlace) + { + thread->miny1 = (thread->index*dpsoftrast.fb_height)/(2*dpsoftrast.numthreads); + thread->maxy1 = ((thread->index+1)*dpsoftrast.fb_height)/(2*dpsoftrast.numthreads); + thread->miny2 = ((dpsoftrast.numthreads+thread->index)*dpsoftrast.fb_height)/(2*dpsoftrast.numthreads); + thread->maxy2 = ((dpsoftrast.numthreads+thread->index+1)*dpsoftrast.fb_height)/(2*dpsoftrast.numthreads); + } + else + { + thread->miny1 = thread->miny2 = (thread->index*dpsoftrast.fb_height)/dpsoftrast.numthreads; + thread->maxy1 = thread->maxy2 = ((thread->index+1)*dpsoftrast.fb_height)/dpsoftrast.numthreads; + } +} + +static void DPSOFTRAST_RecalcClipPlane(DPSOFTRAST_State_Thread *thread) +{ + thread->fb_clipplane[0] = thread->clipplane[0] / thread->fb_viewportscale[1]; + thread->fb_clipplane[1] = thread->clipplane[1] / thread->fb_viewportscale[2]; + thread->fb_clipplane[2] = thread->clipplane[2] / thread->fb_viewportscale[3]; + thread->fb_clipplane[3] = thread->clipplane[3] / thread->fb_viewportscale[0]; + thread->fb_clipplane[3] -= thread->fb_viewportcenter[1]*thread->fb_clipplane[0] + thread->fb_viewportcenter[2]*thread->fb_clipplane[1] + thread->fb_viewportcenter[3]*thread->fb_clipplane[2] + thread->fb_viewportcenter[0]*thread->fb_clipplane[3]; +} + +static void DPSOFTRAST_RecalcFB(DPSOFTRAST_State_Thread *thread) { // calculate framebuffer scissor, viewport, viewport clipped by scissor, // and viewport projection values @@ -337,18 +422,22 @@ void DPSOFTRAST_RecalcFB(DPSOFTRAST_State_Thread *thread) if (x2 > dpsoftrast.fb_width) x2 = dpsoftrast.fb_width; if (y1 < 0) y1 = 0; if (y2 > dpsoftrast.fb_height) y2 = dpsoftrast.fb_height; - thread->fb_clearscissor[0] = x1; - thread->fb_clearscissor[1] = y1; - thread->fb_clearscissor[2] = x2 - x1; - thread->fb_clearscissor[3] = y2 - y1; + thread->fb_scissor[0] = x1; + thread->fb_scissor[1] = y1; + thread->fb_scissor[2] = x2 - x1; + thread->fb_scissor[3] = y2 - y1; + + DPSOFTRAST_RecalcViewport(thread->viewport, thread->fb_viewportcenter, thread->fb_viewportscale); + DPSOFTRAST_RecalcClipPlane(thread); + DPSOFTRAST_RecalcThread(thread); } -void DPSOFTRAST_RecalcDepthFunc(DPSOFTRAST_State_Thread *thread) +static void DPSOFTRAST_RecalcDepthFunc(DPSOFTRAST_State_Thread *thread) { thread->fb_depthfunc = thread->depthtest ? thread->depthfunc : GL_ALWAYS; } -void DPSOFTRAST_RecalcBlendFunc(DPSOFTRAST_State_Thread *thread) +static void DPSOFTRAST_RecalcBlendFunc(DPSOFTRAST_State_Thread *thread) { if (thread->blendsubtract) { @@ -356,14 +445,14 @@ void DPSOFTRAST_RecalcBlendFunc(DPSOFTRAST_State_Thread *thread) { #define BLENDFUNC(sfactor, dfactor, blendmode) \ case (sfactor<<16)|dfactor: thread->fb_blendmode = blendmode; break; - BLENDFUNC(GL_SRC_COLOR, GL_ONE, DPSOFTRAST_BLENDMODE_SUBALPHA) + BLENDFUNC(GL_SRC_ALPHA, GL_ONE, DPSOFTRAST_BLENDMODE_SUBALPHA) default: thread->fb_blendmode = DPSOFTRAST_BLENDMODE_OPAQUE; break; } } else { - switch ((thread->blendfunc[0]<<16)|thread->blendfunc[1]) - { + switch ((thread->blendfunc[0]<<16)|thread->blendfunc[1]) + { BLENDFUNC(GL_ONE, GL_ZERO, DPSOFTRAST_BLENDMODE_OPAQUE) BLENDFUNC(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, DPSOFTRAST_BLENDMODE_ALPHA) BLENDFUNC(GL_SRC_ALPHA, GL_ONE, DPSOFTRAST_BLENDMODE_ADDALPHA) @@ -373,15 +462,15 @@ void DPSOFTRAST_RecalcBlendFunc(DPSOFTRAST_State_Thread *thread) BLENDFUNC(GL_DST_COLOR, GL_ZERO, DPSOFTRAST_BLENDMODE_MUL) BLENDFUNC(GL_DST_COLOR, GL_SRC_COLOR, DPSOFTRAST_BLENDMODE_MUL2) BLENDFUNC(GL_ONE, GL_ONE_MINUS_SRC_ALPHA, DPSOFTRAST_BLENDMODE_PSEUDOALPHA) - BLENDFUNC(GL_SRC_COLOR, GL_ONE, DPSOFTRAST_BLENDMODE_SUBALPHA) + BLENDFUNC(GL_ONE_MINUS_DST_COLOR, GL_ONE, DPSOFTRAST_BLENDMODE_INVADD) default: thread->fb_blendmode = DPSOFTRAST_BLENDMODE_OPAQUE; break; - } + } } } #define DPSOFTRAST_ValidateQuick(thread, f) ((thread->validate & (f)) ? (DPSOFTRAST_Validate(thread, f), 0) : 0) -void DPSOFTRAST_Validate(DPSOFTRAST_State_Thread *thread, int mask) +static void DPSOFTRAST_Validate(DPSOFTRAST_State_Thread *thread, int mask) { mask &= thread->validate; if (!mask) @@ -410,6 +499,31 @@ DPSOFTRAST_Texture *DPSOFTRAST_Texture_GetByIndex(int index) return NULL; } +static void DPSOFTRAST_Texture_Grow(void) +{ + DPSOFTRAST_Texture *oldtexture = dpsoftrast.texture; + DPSOFTRAST_State_Thread *thread; + int i; + int j; + DPSOFTRAST_Flush(); + // expand texture array as needed + if (dpsoftrast.texture_max < 1024) + dpsoftrast.texture_max = 1024; + else + dpsoftrast.texture_max *= 2; + dpsoftrast.texture = (DPSOFTRAST_Texture *)realloc(dpsoftrast.texture, dpsoftrast.texture_max * sizeof(DPSOFTRAST_Texture)); + for (i = 0; i < DPSOFTRAST_MAXTEXTUREUNITS; i++) + if (dpsoftrast.texbound[i]) + dpsoftrast.texbound[i] = dpsoftrast.texture + (dpsoftrast.texbound[i] - oldtexture); + for (j = 0; j < dpsoftrast.numthreads; j++) + { + thread = &dpsoftrast.threads[j]; + for (i = 0; i < DPSOFTRAST_MAXTEXTUREUNITS; i++) + if (thread->texbound[i]) + thread->texbound[i] = dpsoftrast.texture + (thread->texbound[i] - oldtexture); + } +} + int DPSOFTRAST_Texture_New(int flags, int width, int height, int depth) { int w; @@ -481,21 +595,13 @@ int DPSOFTRAST_Texture_New(int flags, int width, int height, int depth) dpsoftrast.errorstring = "DPSOFTRAST_Texture_New: dimensions are not power of two"; return 0; } - DPSOFTRAST_Flush(); // find first empty slot in texture array for (texnum = dpsoftrast.texture_firstfree;texnum < dpsoftrast.texture_end;texnum++) if (!dpsoftrast.texture[texnum].bytes) break; dpsoftrast.texture_firstfree = texnum + 1; if (dpsoftrast.texture_max <= texnum) - { - // expand texture array as needed - if (dpsoftrast.texture_max < 1024) - dpsoftrast.texture_max = 1024; - else - dpsoftrast.texture_max *= 2; - dpsoftrast.texture = (DPSOFTRAST_Texture *)realloc(dpsoftrast.texture, dpsoftrast.texture_max * sizeof(DPSOFTRAST_Texture)); - } + DPSOFTRAST_Texture_Grow(); if (dpsoftrast.texture_end <= texnum) dpsoftrast.texture_end = texnum + 1; texture = &dpsoftrast.texture[texnum]; @@ -505,6 +611,7 @@ int DPSOFTRAST_Texture_New(int flags, int width, int height, int depth) texture->height = height; texture->depth = depth; texture->sides = sides; + texture->binds = 0; w = width; h = height; d = depth; @@ -541,7 +648,8 @@ void DPSOFTRAST_Texture_Free(int index) { DPSOFTRAST_Texture *texture; texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return; - DPSOFTRAST_Flush(); + if (texture->binds) + DPSOFTRAST_Flush(); if (texture->bytes) MM_FREE(texture->bytes); texture->bytes = NULL; @@ -636,14 +744,18 @@ void DPSOFTRAST_Texture_UpdatePartial(int index, int mip, const unsigned char *p DPSOFTRAST_Texture *texture; unsigned char *dst; texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return; - DPSOFTRAST_Flush(); - dst = texture->bytes + (blocky * texture->mipmap[0][2] + blockx) * 4; - while (blockheight > 0) + if (texture->binds) + DPSOFTRAST_Flush(); + if (pixels) { - memcpy(dst, pixels, blockwidth * 4); - pixels += blockwidth * 4; - dst += texture->mipmap[0][2] * 4; - blockheight--; + dst = texture->bytes + (blocky * texture->mipmap[0][2] + blockx) * 4; + while (blockheight > 0) + { + memcpy(dst, pixels, blockwidth * 4); + pixels += blockwidth * 4; + dst += texture->mipmap[0][2] * 4; + blockheight--; + } } DPSOFTRAST_Texture_CalculateMipmaps(index); } @@ -651,8 +763,10 @@ void DPSOFTRAST_Texture_UpdateFull(int index, const unsigned char *pixels) { DPSOFTRAST_Texture *texture; texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return; - DPSOFTRAST_Flush(); - memcpy(texture->bytes, pixels, texture->mipmap[0][1]); + if (texture->binds) + DPSOFTRAST_Flush(); + if (pixels) + memcpy(texture->bytes, pixels, texture->mipmap[0][1]); DPSOFTRAST_Texture_CalculateMipmaps(index); } int DPSOFTRAST_Texture_GetWidth(int index, int mip) @@ -677,7 +791,8 @@ unsigned char *DPSOFTRAST_Texture_GetPixelPointer(int index, int mip) { DPSOFTRAST_Texture *texture; texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return 0; - DPSOFTRAST_Flush(); + if (texture->binds) + DPSOFTRAST_Flush(); return texture->bytes + texture->mipmap[mip][0]; } void DPSOFTRAST_Texture_Filter(int index, DPSOFTRAST_TEXTURE_FILTER filter) @@ -689,91 +804,20 @@ void DPSOFTRAST_Texture_Filter(int index, DPSOFTRAST_TEXTURE_FILTER filter) dpsoftrast.errorstring = "DPSOFTRAST_Texture_Filter: requested filter mode requires mipmaps"; return; } - DPSOFTRAST_Flush(); - texture->filter = filter; -} - -void DPSOFTRAST_SetRenderTargets(int width, int height, unsigned int *depthpixels, unsigned int *colorpixels0, unsigned int *colorpixels1, unsigned int *colorpixels2, unsigned int *colorpixels3) -{ - if (width != dpsoftrast.fb_width || height != dpsoftrast.fb_height || depthpixels != dpsoftrast.fb_depthpixels || - colorpixels0 != dpsoftrast.fb_colorpixels[0] || colorpixels1 != dpsoftrast.fb_colorpixels[1] || - colorpixels2 != dpsoftrast.fb_colorpixels[2] || colorpixels3 != dpsoftrast.fb_colorpixels[3]) + if (texture->binds) DPSOFTRAST_Flush(); - dpsoftrast.fb_width = width; - dpsoftrast.fb_height = height; - dpsoftrast.fb_depthpixels = depthpixels; - dpsoftrast.fb_colorpixels[0] = colorpixels0; - dpsoftrast.fb_colorpixels[1] = colorpixels1; - dpsoftrast.fb_colorpixels[2] = colorpixels2; - dpsoftrast.fb_colorpixels[3] = colorpixels3; + texture->filter = filter; } -void DPSOFTRAST_Draw_FlushThreads(void); - -void DPSOFTRAST_Draw_FreeTrianglePool(int space) -{ - DPSOFTRAST_State_Thread *thread; - int i; - int freetriangle = dpsoftrast.trianglepool.freetriangle; - int usedtriangles = dpsoftrast.trianglepool.usedtriangles; - if (usedtriangles <= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-space) - return; -#ifdef USETHREADS - SDL_LockMutex(dpsoftrast.trianglemutex); -#endif - for(;;) - { - int waitindex = -1; - int triangleoffset; - usedtriangles = 0; - for (i = 0; i < dpsoftrast.numthreads; i++) - { - thread = &dpsoftrast.threads[i]; - triangleoffset = freetriangle - thread->triangleoffset; - if (triangleoffset < 0) - triangleoffset += DPSOFTRAST_DRAW_MAXTRIANGLEPOOL; - if (triangleoffset > usedtriangles) - { - waitindex = i; - usedtriangles = triangleoffset; - } - } - if (usedtriangles <= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-space || waitindex < 0) - break; -#ifdef USETHREADS - thread = &dpsoftrast.threads[waitindex]; - thread->waiting = true; - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_CondWait(thread->waitcond, dpsoftrast.trianglemutex); - thread->waiting = false; -#endif - } -#ifdef USETHREADS - SDL_UnlockMutex(dpsoftrast.trianglemutex); -#endif - dpsoftrast.trianglepool.usedtriangles = usedtriangles; -} +static void DPSOFTRAST_Draw_FlushThreads(void); -void DPSOFTRAST_Draw_SyncCommands(void) +static void DPSOFTRAST_Draw_SyncCommands(void) { - DPSOFTRAST_State_Triangle *triangle; - if (dpsoftrast.trianglepool.usedtriangles >= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1) -#ifdef USETHREADS - DPSOFTRAST_Draw_FreeTrianglePool(DPSOFTRAST_DRAW_MAXTRIANGLEPOOL/8); -#else - DPSOFTRAST_Draw_FlushThreads(); -#endif - triangle = &dpsoftrast.trianglepool.triangles[dpsoftrast.trianglepool.freetriangle]; - triangle->commandoffset = dpsoftrast.commandpool.freecommand; - triangle->starty = -1; - triangle->endy = -1; - dpsoftrast.trianglepool.freetriangle = dpsoftrast.trianglepool.freetriangle < DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1 ? dpsoftrast.trianglepool.freetriangle + 1 : 0; - dpsoftrast.trianglepool.usedtriangles++; - MEMORY_BARRIER; - dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle; + if(dpsoftrast.usethreads) MEMORY_BARRIER; + dpsoftrast.drawcommand = dpsoftrast.commandpool.freecommand; } -void DPSOFTRAST_Draw_FreeCommandPool(int space) +static void DPSOFTRAST_Draw_FreeCommandPool(int space) { DPSOFTRAST_State_Thread *thread; int i; @@ -782,9 +826,6 @@ void DPSOFTRAST_Draw_FreeCommandPool(int space) if (usedcommands <= DPSOFTRAST_DRAW_MAXCOMMANDPOOL-space) return; DPSOFTRAST_Draw_SyncCommands(); -#ifdef USETHREADS - SDL_LockMutex(dpsoftrast.trianglemutex); -#endif for(;;) { int waitindex = -1; @@ -804,42 +845,43 @@ void DPSOFTRAST_Draw_FreeCommandPool(int space) } if (usedcommands <= DPSOFTRAST_DRAW_MAXCOMMANDPOOL-space || waitindex < 0) break; -#ifdef USETHREADS thread = &dpsoftrast.threads[waitindex]; - thread->waiting = true; - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_CondWait(thread->waitcond, dpsoftrast.trianglemutex); - thread->waiting = false; -#endif + Thread_LockMutex(thread->drawmutex); + if (thread->commandoffset != dpsoftrast.drawcommand) + { + thread->waiting = true; + if (thread->starving) Thread_CondSignal(thread->drawcond); + Thread_CondWait(thread->waitcond, thread->drawmutex); + thread->waiting = false; + } + Thread_UnlockMutex(thread->drawmutex); } -#ifdef USETHREADS - SDL_UnlockMutex(dpsoftrast.trianglemutex); -#endif dpsoftrast.commandpool.usedcommands = usedcommands; } +#define DPSOFTRAST_ALIGNCOMMAND(size) \ + ((size) + ((COMMAND_SIZE - ((size)&(COMMAND_SIZE-1))) & (COMMAND_SIZE-1))) #define DPSOFTRAST_ALLOCATECOMMAND(name) \ - ((DPSOFTRAST_Command_##name *) DPSOFTRAST_AllocateCommand(sizeof( DPSOFTRAST_Command_##name ) + ((COMMAND_SIZE - (sizeof( DPSOFTRAST_Command_##name )&(COMMAND_SIZE-1))) & (COMMAND_SIZE-1)))) + ((DPSOFTRAST_Command_##name *) DPSOFTRAST_AllocateCommand( DPSOFTRAST_OPCODE_##name , DPSOFTRAST_ALIGNCOMMAND(sizeof( DPSOFTRAST_Command_##name )))) -static void *DPSOFTRAST_AllocateCommand(int size) +static void *DPSOFTRAST_AllocateCommand(int opcode, int size) { DPSOFTRAST_Command *command; int freecommand = dpsoftrast.commandpool.freecommand; int usedcommands = dpsoftrast.commandpool.usedcommands; int extra = sizeof(DPSOFTRAST_Command); - if(DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand < size) + if (DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand < size) extra += DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand; - if(usedcommands > DPSOFTRAST_DRAW_MAXCOMMANDPOOL - (size + extra)) + if (usedcommands > DPSOFTRAST_DRAW_MAXCOMMANDPOOL - (size + extra)) { -#ifdef USETHREADS - DPSOFTRAST_Draw_FreeCommandPool(size + extra); -#else - DPSOFTRAST_Draw_FlushThreads(); -#endif + if (dpsoftrast.usethreads) + DPSOFTRAST_Draw_FreeCommandPool(size + extra); + else + DPSOFTRAST_Draw_FlushThreads(); freecommand = dpsoftrast.commandpool.freecommand; usedcommands = dpsoftrast.commandpool.usedcommands; } - if(DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand < size) + if (DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand < size) { command = (DPSOFTRAST_Command *) &dpsoftrast.commandpool.commands[freecommand]; command->opcode = DPSOFTRAST_OPCODE_Reset; @@ -847,15 +889,28 @@ static void *DPSOFTRAST_AllocateCommand(int size) freecommand = 0; } command = (DPSOFTRAST_Command *) &dpsoftrast.commandpool.commands[freecommand]; + command->opcode = opcode; + command->commandsize = size; freecommand += size; if (freecommand >= DPSOFTRAST_DRAW_MAXCOMMANDPOOL) freecommand = 0; - dpsoftrast.commandpool.freecommand = freecommand; dpsoftrast.commandpool.usedcommands = usedcommands + size; return command; } - + +static void DPSOFTRAST_UndoCommand(int size) +{ + int freecommand = dpsoftrast.commandpool.freecommand; + int usedcommands = dpsoftrast.commandpool.usedcommands; + freecommand -= size; + if (freecommand < 0) + freecommand += DPSOFTRAST_DRAW_MAXCOMMANDPOOL; + usedcommands -= size; + dpsoftrast.commandpool.freecommand = freecommand; + dpsoftrast.commandpool.usedcommands = usedcommands; +} + DEFCOMMAND(1, Viewport, int x; int y; int width; int height;) static void DPSOFTRAST_Interpret_Viewport(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_Command_Viewport *command) { @@ -868,7 +923,6 @@ static void DPSOFTRAST_Interpret_Viewport(DPSOFTRAST_State_Thread *thread, const void DPSOFTRAST_Viewport(int x, int y, int width, int height) { DPSOFTRAST_Command_Viewport *command = DPSOFTRAST_ALLOCATECOMMAND(Viewport); - command->opcode = DPSOFTRAST_OPCODE_Viewport; command->x = x; command->y = y; command->width = width; @@ -878,31 +932,28 @@ void DPSOFTRAST_Viewport(int x, int y, int width, int height) dpsoftrast.viewport[1] = y; dpsoftrast.viewport[2] = width; dpsoftrast.viewport[3] = height; - dpsoftrast.fb_viewportcenter[1] = dpsoftrast.viewport[0] + 0.5f * dpsoftrast.viewport[2] - 0.5f; - dpsoftrast.fb_viewportcenter[2] = dpsoftrast.fb_height - dpsoftrast.viewport[1] - 0.5f * dpsoftrast.viewport[3] - 0.5f; - dpsoftrast.fb_viewportcenter[3] = 0.5f; - dpsoftrast.fb_viewportcenter[0] = 0.0f; - dpsoftrast.fb_viewportscale[1] = 0.5f * dpsoftrast.viewport[2]; - dpsoftrast.fb_viewportscale[2] = -0.5f * dpsoftrast.viewport[3]; - dpsoftrast.fb_viewportscale[3] = 0.5f; - dpsoftrast.fb_viewportscale[0] = 1.0f; + DPSOFTRAST_RecalcViewport(dpsoftrast.viewport, dpsoftrast.fb_viewportcenter, dpsoftrast.fb_viewportscale); } DEFCOMMAND(2, ClearColor, float r; float g; float b; float a;) static void DPSOFTRAST_Interpret_ClearColor(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_Command_ClearColor *command) { - int i, x1, y1, x2, y2, w, h, x, y, t1, t2; + int i, x1, y1, x2, y2, w, h, x, y; + int miny1, maxy1, miny2, maxy2; + int bandy; unsigned int *p; unsigned int c; DPSOFTRAST_Validate(thread, DPSOFTRAST_VALIDATE_FB); - x1 = thread->fb_clearscissor[0]; - y1 = thread->fb_clearscissor[1]; - x2 = thread->fb_clearscissor[2]; - y2 = thread->fb_clearscissor[1] + thread->fb_clearscissor[3]; - t1 = (thread->index*dpsoftrast.fb_height)/dpsoftrast.numthreads; - t2 = ((thread->index+1)*dpsoftrast.fb_height)/dpsoftrast.numthreads; - if(y1 < t1) y1 = t1; - if(y2 > t2) y2 = t2; + miny1 = thread->miny1; + maxy1 = thread->maxy1; + miny2 = thread->miny2; + maxy2 = thread->maxy2; + x1 = thread->fb_scissor[0]; + y1 = thread->fb_scissor[1]; + x2 = thread->fb_scissor[0] + thread->fb_scissor[2]; + y2 = thread->fb_scissor[1] + thread->fb_scissor[3]; + if (y1 < miny1) y1 = miny1; + if (y2 > maxy2) y2 = maxy2; w = x2 - x1; h = y2 - y1; if (w < 1 || h < 1) @@ -913,7 +964,8 @@ static void DPSOFTRAST_Interpret_ClearColor(DPSOFTRAST_State_Thread *thread, con { if (!dpsoftrast.fb_colorpixels[i]) continue; - for (y = y1;y < y2;y++) + for (y = y1, bandy = min(y2, maxy1); y < y2; bandy = min(y2, maxy2), y = max(y, miny2)) + for (;y < bandy;y++) { p = dpsoftrast.fb_colorpixels[i] + y * dpsoftrast.fb_width; for (x = x1;x < x2;x++) @@ -924,7 +976,6 @@ static void DPSOFTRAST_Interpret_ClearColor(DPSOFTRAST_State_Thread *thread, con void DPSOFTRAST_ClearColor(float r, float g, float b, float a) { DPSOFTRAST_Command_ClearColor *command = DPSOFTRAST_ALLOCATECOMMAND(ClearColor); - command->opcode = DPSOFTRAST_OPCODE_ClearColor; command->r = r; command->g = g; command->b = b; @@ -934,24 +985,29 @@ void DPSOFTRAST_ClearColor(float r, float g, float b, float a) DEFCOMMAND(3, ClearDepth, float depth;) static void DPSOFTRAST_Interpret_ClearDepth(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_ClearDepth *command) { - int x1, y1, x2, y2, w, h, x, y, t1, t2; + int x1, y1, x2, y2, w, h, x, y; + int miny1, maxy1, miny2, maxy2; + int bandy; unsigned int *p; unsigned int c; DPSOFTRAST_Validate(thread, DPSOFTRAST_VALIDATE_FB); - x1 = thread->fb_clearscissor[0]; - y1 = thread->fb_clearscissor[1]; - x2 = thread->fb_clearscissor[2]; - y2 = thread->fb_clearscissor[1] + thread->fb_clearscissor[3]; - t1 = (thread->index*dpsoftrast.fb_height)/dpsoftrast.numthreads; - t2 = ((thread->index+1)*dpsoftrast.fb_height)/dpsoftrast.numthreads; - if(y1 < t1) y1 = t1; - if(y2 > t2) y2 = t2; + miny1 = thread->miny1; + maxy1 = thread->maxy1; + miny2 = thread->miny2; + maxy2 = thread->maxy2; + x1 = thread->fb_scissor[0]; + y1 = thread->fb_scissor[1]; + x2 = thread->fb_scissor[0] + thread->fb_scissor[2]; + y2 = thread->fb_scissor[1] + thread->fb_scissor[3]; + if (y1 < miny1) y1 = miny1; + if (y2 > maxy2) y2 = maxy2; w = x2 - x1; h = y2 - y1; if (w < 1 || h < 1) return; c = DPSOFTRAST_DEPTH32_FROM_DEPTH32F(command->depth); - for (y = y1;y < y2;y++) + for (y = y1, bandy = min(y2, maxy1); y < y2; bandy = min(y2, maxy2), y = max(y, miny2)) + for (;y < bandy;y++) { p = dpsoftrast.fb_depthpixels + y * dpsoftrast.fb_width; for (x = x1;x < x2;x++) @@ -961,7 +1017,6 @@ static void DPSOFTRAST_Interpret_ClearDepth(DPSOFTRAST_State_Thread *thread, DPS void DPSOFTRAST_ClearDepth(float d) { DPSOFTRAST_Command_ClearDepth *command = DPSOFTRAST_ALLOCATECOMMAND(ClearDepth); - command->opcode = DPSOFTRAST_OPCODE_ClearDepth; command->depth = d; } @@ -977,7 +1032,6 @@ static void DPSOFTRAST_Interpret_ColorMask(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_ColorMask(int r, int g, int b, int a) { DPSOFTRAST_Command_ColorMask *command = DPSOFTRAST_ALLOCATECOMMAND(ColorMask); - command->opcode = DPSOFTRAST_OPCODE_ColorMask; command->r = r; command->g = g; command->b = b; @@ -993,7 +1047,6 @@ static void DPSOFTRAST_Interpret_DepthTest(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_DepthTest(int enable) { DPSOFTRAST_Command_DepthTest *command = DPSOFTRAST_ALLOCATECOMMAND(DepthTest); - command->opcode = DPSOFTRAST_OPCODE_DepthTest; command->enable = enable; } @@ -1006,7 +1059,6 @@ static void DPSOFTRAST_Interpret_ScissorTest(DPSOFTRAST_State_Thread *thread, DP void DPSOFTRAST_ScissorTest(int enable) { DPSOFTRAST_Command_ScissorTest *command = DPSOFTRAST_ALLOCATECOMMAND(ScissorTest); - command->opcode = DPSOFTRAST_OPCODE_ScissorTest; command->enable = enable; } @@ -1022,7 +1074,6 @@ static void DPSOFTRAST_Interpret_Scissor(DPSOFTRAST_State_Thread *thread, DPSOFT void DPSOFTRAST_Scissor(float x, float y, float width, float height) { DPSOFTRAST_Command_Scissor *command = DPSOFTRAST_ALLOCATECOMMAND(Scissor); - command->opcode = DPSOFTRAST_OPCODE_Scissor; command->x = x; command->y = y; command->width = width; @@ -1039,7 +1090,6 @@ static void DPSOFTRAST_Interpret_BlendFunc(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_BlendFunc(int sfactor, int dfactor) { DPSOFTRAST_Command_BlendFunc *command = DPSOFTRAST_ALLOCATECOMMAND(BlendFunc); - command->opcode = DPSOFTRAST_OPCODE_BlendFunc; command->sfactor = sfactor; command->dfactor = dfactor; } @@ -1053,7 +1103,6 @@ static void DPSOFTRAST_Interpret_BlendSubtract(DPSOFTRAST_State_Thread *thread, void DPSOFTRAST_BlendSubtract(int enable) { DPSOFTRAST_Command_BlendSubtract *command = DPSOFTRAST_ALLOCATECOMMAND(BlendSubtract); - command->opcode = DPSOFTRAST_OPCODE_BlendSubtract; command->enable = enable; } @@ -1065,7 +1114,6 @@ static void DPSOFTRAST_Interpret_DepthMask(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_DepthMask(int enable) { DPSOFTRAST_Command_DepthMask *command = DPSOFTRAST_ALLOCATECOMMAND(DepthMask); - command->opcode = DPSOFTRAST_OPCODE_DepthMask; command->enable = enable; } @@ -1077,7 +1125,6 @@ static void DPSOFTRAST_Interpret_DepthFunc(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_DepthFunc(int func) { DPSOFTRAST_Command_DepthFunc *command = DPSOFTRAST_ALLOCATECOMMAND(DepthFunc); - command->opcode = DPSOFTRAST_OPCODE_DepthFunc; command->func = func; } @@ -1090,7 +1137,6 @@ static void DPSOFTRAST_Interpret_DepthRange(DPSOFTRAST_State_Thread *thread, DPS void DPSOFTRAST_DepthRange(float nearval, float farval) { DPSOFTRAST_Command_DepthRange *command = DPSOFTRAST_ALLOCATECOMMAND(DepthRange); - command->opcode = DPSOFTRAST_OPCODE_DepthRange; command->nearval = nearval; command->farval = farval; } @@ -1104,40 +1150,19 @@ static void DPSOFTRAST_Interpret_PolygonOffset(DPSOFTRAST_State_Thread *thread, void DPSOFTRAST_PolygonOffset(float alongnormal, float intoview) { DPSOFTRAST_Command_PolygonOffset *command = DPSOFTRAST_ALLOCATECOMMAND(PolygonOffset); - command->opcode = DPSOFTRAST_OPCODE_PolygonOffset; command->alongnormal = alongnormal; command->intoview = intoview; } -void DPSOFTRAST_CullFace(int mode) -{ - dpsoftrast.cullface = mode; -} - -DEFCOMMAND(15, AlphaTest, int enable;) -static void DPSOFTRAST_Interpret_AlphaTest(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_AlphaTest *command) -{ - thread->alphatest = command->enable; -} -void DPSOFTRAST_AlphaTest(int enable) -{ - DPSOFTRAST_Command_AlphaTest *command = DPSOFTRAST_ALLOCATECOMMAND(AlphaTest); - command->opcode = DPSOFTRAST_OPCODE_AlphaTest; - command->enable = enable; -} - -DEFCOMMAND(16, AlphaFunc, int func; float ref;) -static void DPSOFTRAST_Interpret_AlphaFunc(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_AlphaFunc *command) +DEFCOMMAND(14, CullFace, int mode;) +static void DPSOFTRAST_Interpret_CullFace(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_CullFace *command) { - thread->alphafunc = command->func; - thread->alphavalue = command->ref; + thread->cullface = command->mode; } -void DPSOFTRAST_AlphaFunc(int func, float ref) +void DPSOFTRAST_CullFace(int mode) { - DPSOFTRAST_Command_AlphaFunc *command = DPSOFTRAST_ALLOCATECOMMAND(AlphaFunc); - command->opcode = DPSOFTRAST_OPCODE_AlphaFunc; - command->func = func; - command->ref = ref; + DPSOFTRAST_Command_CullFace *command = DPSOFTRAST_ALLOCATECOMMAND(CullFace); + command->mode = mode; } void DPSOFTRAST_Color4f(float r, float g, float b, float a) @@ -1157,7 +1182,6 @@ void DPSOFTRAST_GetPixelsBGRA(int blockx, int blocky, int blockwidth, int blockh int bx2 = blockx + blockwidth; int by2 = blocky + blockheight; int bw; - int bh; int x; int y; unsigned char *inpixels; @@ -1169,7 +1193,6 @@ void DPSOFTRAST_GetPixelsBGRA(int blockx, int blocky, int blockwidth, int blockh if (bx2 > dpsoftrast.fb_width) bx2 = dpsoftrast.fb_width; if (by2 > dpsoftrast.fb_height) by2 = dpsoftrast.fb_height; bw = bx2 - bx1; - bh = by2 - by1; inpixels = (unsigned char *)dpsoftrast.fb_colorpixels[0]; if (dpsoftrast.bigendian) { @@ -1246,8 +1269,9 @@ void DPSOFTRAST_CopyRectangleToTexture(int index, int mip, int tx, int ty, int s if (th > sh) th = sh; if (tw < 1 || th < 1) return; + sy1 = sheight - 1 - sy1; for (y = 0;y < th;y++) - memcpy(tpixels + ((ty1 + y) * twidth + tx1), spixels + ((sy1 + y) * swidth + sx1), tw*4); + memcpy(tpixels + ((ty1 + y) * twidth + tx1), spixels + ((sy1 - y) * swidth + sx1), tw*4); if (texture->mipmaps > 1) DPSOFTRAST_Texture_CalculateMipmaps(index); } @@ -1255,6 +1279,8 @@ void DPSOFTRAST_CopyRectangleToTexture(int index, int mip, int tx, int ty, int s DEFCOMMAND(17, SetTexture, int unitnum; DPSOFTRAST_Texture *texture;) static void DPSOFTRAST_Interpret_SetTexture(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_SetTexture *command) { + if (thread->texbound[command->unitnum]) + ATOMIC_DECREMENT(thread->texbound[command->unitnum]->binds); thread->texbound[command->unitnum] = command->texture; } void DPSOFTRAST_SetTexture(int unitnum, int index) @@ -1274,11 +1300,11 @@ void DPSOFTRAST_SetTexture(int unitnum, int index) } command = DPSOFTRAST_ALLOCATECOMMAND(SetTexture); - command->opcode = DPSOFTRAST_OPCODE_SetTexture; command->unitnum = unitnum; command->texture = texture; dpsoftrast.texbound[unitnum] = texture; + ATOMIC_ADD(texture->binds, dpsoftrast.numthreads); } void DPSOFTRAST_SetVertexPointer(const float *vertex3f, size_t stride) @@ -1305,21 +1331,23 @@ void DPSOFTRAST_SetTexCoordPointer(int unitnum, int numcomponents, size_t stride dpsoftrast.stride_texcoord[unitnum] = stride; } -DEFCOMMAND(18, SetShader, int mode; int permutation;) +DEFCOMMAND(18, SetShader, int mode; int permutation; int exactspecularmath;) static void DPSOFTRAST_Interpret_SetShader(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_SetShader *command) { thread->shader_mode = command->mode; thread->shader_permutation = command->permutation; + thread->shader_exactspecularmath = command->exactspecularmath; } -void DPSOFTRAST_SetShader(int mode, int permutation) +void DPSOFTRAST_SetShader(int mode, int permutation, int exactspecularmath) { DPSOFTRAST_Command_SetShader *command = DPSOFTRAST_ALLOCATECOMMAND(SetShader); - command->opcode = DPSOFTRAST_OPCODE_SetShader; command->mode = mode; command->permutation = permutation; + command->exactspecularmath = exactspecularmath; dpsoftrast.shader_mode = mode; dpsoftrast.shader_permutation = permutation; + dpsoftrast.shader_exactspecularmath = exactspecularmath; } DEFCOMMAND(19, Uniform4f, DPSOFTRAST_UNIFORM index; float val[4];) @@ -1330,7 +1358,6 @@ static void DPSOFTRAST_Interpret_Uniform4f(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_Uniform4f(DPSOFTRAST_UNIFORM index, float v0, float v1, float v2, float v3) { DPSOFTRAST_Command_Uniform4f *command = DPSOFTRAST_ALLOCATECOMMAND(Uniform4f); - command->opcode = DPSOFTRAST_OPCODE_Uniform4f; command->index = index; command->val[0] = v0; command->val[1] = v1; @@ -1345,7 +1372,6 @@ void DPSOFTRAST_Uniform4f(DPSOFTRAST_UNIFORM index, float v0, float v1, float v2 void DPSOFTRAST_Uniform4fv(DPSOFTRAST_UNIFORM index, const float *v) { DPSOFTRAST_Command_Uniform4f *command = DPSOFTRAST_ALLOCATECOMMAND(Uniform4f); - command->opcode = DPSOFTRAST_OPCODE_Uniform4f; command->index = index; memcpy(command->val, v, sizeof(command->val)); @@ -1359,14 +1385,13 @@ static void DPSOFTRAST_Interpret_UniformMatrix4f(DPSOFTRAST_State_Thread *thread } void DPSOFTRAST_UniformMatrix4fv(DPSOFTRAST_UNIFORM uniform, int arraysize, int transpose, const float *v) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int i, index; for (i = 0, index = (int)uniform;i < arraysize;i++, index += 4, v += 16) { __m128 m0, m1, m2, m3; DPSOFTRAST_Command_UniformMatrix4f *command = DPSOFTRAST_ALLOCATECOMMAND(UniformMatrix4f); - command->opcode = DPSOFTRAST_OPCODE_UniformMatrix4f; - command->index = index; + command->index = (DPSOFTRAST_UNIFORM)index; if (((size_t)v)&(ALIGN_SIZE-1)) { m0 = _mm_loadu_ps(v); @@ -1413,14 +1438,28 @@ static void DPSOFTRAST_Interpret_Uniform1i(DPSOFTRAST_State_Thread *thread, DPSO void DPSOFTRAST_Uniform1i(DPSOFTRAST_UNIFORM index, int i0) { DPSOFTRAST_Command_Uniform1i *command = DPSOFTRAST_ALLOCATECOMMAND(Uniform1i); - command->opcode = DPSOFTRAST_OPCODE_Uniform1i; command->index = index; command->val = i0; dpsoftrast.uniform1i[command->index] = i0; } -#ifdef SSE2_PRESENT +DEFCOMMAND(24, ClipPlane, float clipplane[4];) +static void DPSOFTRAST_Interpret_ClipPlane(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_ClipPlane *command) +{ + memcpy(thread->clipplane, command->clipplane, 4*sizeof(float)); + thread->validate |= DPSOFTRAST_VALIDATE_FB; +} +void DPSOFTRAST_ClipPlane(float x, float y, float z, float w) +{ + DPSOFTRAST_Command_ClipPlane *command = DPSOFTRAST_ALLOCATECOMMAND(ClipPlane); + command->clipplane[0] = x; + command->clipplane[1] = y; + command->clipplane[2] = z; + command->clipplane[3] = w; +} + +#ifdef SSE_POSSIBLE static void DPSOFTRAST_Load4fTo4f(float *dst, const unsigned char *src, int size, int stride) { float *end = dst + size*4; @@ -1573,24 +1612,24 @@ static void DPSOFTRAST_Load4bTo4f(float *dst, const unsigned char *src, int size __m128i v = _mm_loadu_si128((const __m128i *)src), v1 = _mm_unpacklo_epi8(v, _mm_setzero_si128()), v2 = _mm_unpackhi_epi8(v, _mm_setzero_si128()); _mm_store_ps(dst, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v1, _mm_setzero_si128())), scale)); _mm_store_ps(dst + 4, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v1, _mm_setzero_si128())), scale)); - _mm_store_ps(dst + 8, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v2, _mm_setzero_si128())), scale)); - _mm_store_ps(dst + 12, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v2, _mm_setzero_si128())), scale)); + _mm_store_ps(dst + 8, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v2, _mm_setzero_si128())), scale)); + _mm_store_ps(dst + 12, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v2, _mm_setzero_si128())), scale)); dst += 16; src += 4*sizeof(unsigned char[4]); } } else { - while (dst < end4) - { - __m128i v = _mm_load_si128((const __m128i *)src), v1 = _mm_unpacklo_epi8(v, _mm_setzero_si128()), v2 = _mm_unpackhi_epi8(v, _mm_setzero_si128()); - _mm_store_ps(dst, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v1, _mm_setzero_si128())), scale)); - _mm_store_ps(dst + 4, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v1, _mm_setzero_si128())), scale)); - _mm_store_ps(dst + 8, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v2, _mm_setzero_si128())), scale)); - _mm_store_ps(dst + 12, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v2, _mm_setzero_si128())), scale)); - dst += 16; - src += 4*sizeof(unsigned char[4]); - } + while (dst < end4) + { + __m128i v = _mm_load_si128((const __m128i *)src), v1 = _mm_unpacklo_epi8(v, _mm_setzero_si128()), v2 = _mm_unpackhi_epi8(v, _mm_setzero_si128()); + _mm_store_ps(dst, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v1, _mm_setzero_si128())), scale)); + _mm_store_ps(dst + 4, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v1, _mm_setzero_si128())), scale)); + _mm_store_ps(dst + 8, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(v2, _mm_setzero_si128())), scale)); + _mm_store_ps(dst + 12, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpackhi_epi16(v2, _mm_setzero_si128())), scale)); + dst += 16; + src += 4*sizeof(unsigned char[4]); + } } } while (dst < end) @@ -1614,95 +1653,19 @@ static void DPSOFTRAST_Fill4f(float *dst, const float *src, int size) } #endif -void DPSOFTRAST_Draw_LoadVertices(int firstvertex, int numvertices, bool needcolors) -{ -#ifdef SSE2_PRESENT - int i; - int j; - int stride; - const float *v; - float *p; - float *data; - const unsigned char *b; - dpsoftrast.numvertices = numvertices; - if (dpsoftrast.maxvertices < dpsoftrast.numvertices) - { - if (dpsoftrast.maxvertices < 4096) - dpsoftrast.maxvertices = 4096; - while (dpsoftrast.maxvertices < dpsoftrast.numvertices) - dpsoftrast.maxvertices *= 2; - if (dpsoftrast.in_array4f[0]) - MM_FREE(dpsoftrast.in_array4f[0]); - data = (float *)MM_CALLOC(1, dpsoftrast.maxvertices * sizeof(float[4])*(DPSOFTRAST_ARRAY_TOTAL*2 + 1)); - for (i = 0;i < DPSOFTRAST_ARRAY_TOTAL;i++, data += dpsoftrast.maxvertices * 4) - dpsoftrast.in_array4f[i] = data; - for (i = 0;i < DPSOFTRAST_ARRAY_TOTAL;i++, data += dpsoftrast.maxvertices * 4) - dpsoftrast.post_array4f[i] = data; - dpsoftrast.screencoord4f = data; - data += dpsoftrast.maxvertices * 4; - } - stride = dpsoftrast.stride_vertex; - v = (const float *)((unsigned char *)dpsoftrast.pointer_vertex3f + firstvertex * stride); - p = dpsoftrast.in_array4f[0]; - DPSOFTRAST_Load3fTo4f(p, (const unsigned char *)v, numvertices, stride); - if (needcolors) - { - if (dpsoftrast.pointer_color4f) - { - stride = dpsoftrast.stride_color; - v = (const float *)((const unsigned char *)dpsoftrast.pointer_color4f + firstvertex * stride); - p = dpsoftrast.in_array4f[1]; - DPSOFTRAST_Load4fTo4f(p, (const unsigned char *)v, numvertices, stride); - } - else if (dpsoftrast.pointer_color4ub) - { - stride = dpsoftrast.stride_color; - b = (const unsigned char *)((const unsigned char *)dpsoftrast.pointer_color4ub + firstvertex * stride); - p = dpsoftrast.in_array4f[1]; - DPSOFTRAST_Load4bTo4f(p, b, numvertices, stride); - } - else - { - p = dpsoftrast.in_array4f[1]; - DPSOFTRAST_Fill4f(p, dpsoftrast.color, numvertices); - } - } - for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL-2;j++) - { - if (dpsoftrast.pointer_texcoordf[j]) - { - stride = dpsoftrast.stride_texcoord[j]; - v = (const float *)((const unsigned char *)dpsoftrast.pointer_texcoordf[j] + firstvertex * stride); - p = dpsoftrast.in_array4f[j+2]; - switch(dpsoftrast.components_texcoord[j]) - { - case 2: - DPSOFTRAST_Load2fTo4f(p, (const unsigned char *)v, numvertices, stride); - break; - case 3: - DPSOFTRAST_Load3fTo4f(p, (const unsigned char *)v, numvertices, stride); - break; - case 4: - DPSOFTRAST_Load4fTo4f(p, (const unsigned char *)v, numvertices, stride); - break; - } - } - } -#endif -} - -void DPSOFTRAST_Array_Transform(float *out4f, const float *in4f, int numitems, const float *inmatrix16f) +void DPSOFTRAST_Vertex_Transform(float *out4f, const float *in4f, int numitems, const float *inmatrix16f) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE static const float identitymatrix[4][4] = {{1,0,0,0},{0,1,0,0},{0,0,1,0},{0,0,0,1}}; __m128 m0, m1, m2, m3; - float *end = out4f + numitems*4; + float *end; if (!memcmp(identitymatrix, inmatrix16f, sizeof(float[16]))) { // fast case for identity matrix - memcpy(out4f, in4f, numitems * sizeof(float[4])); + if (out4f != in4f) memcpy(out4f, in4f, numitems * sizeof(float[4])); return; } + end = out4f + numitems*4; m0 = _mm_loadu_ps(inmatrix16f); m1 = _mm_loadu_ps(inmatrix16f + 4); m2 = _mm_loadu_ps(inmatrix16f + 8); @@ -1738,53 +1701,168 @@ void DPSOFTRAST_Array_Transform(float *out4f, const float *in4f, int numitems, c #endif } -void DPSOFTRAST_Array_Copy(float *out4f, const float *in4f, int numitems) +void DPSOFTRAST_Vertex_Copy(float *out4f, const float *in4f, int numitems) { memcpy(out4f, in4f, numitems * sizeof(float[4])); } -#ifdef SSE2_PRESENT -static __m128 DPSOFTRAST_Draw_ProjectVertex(__m128 v) +#ifdef SSE_POSSIBLE +#define DPSOFTRAST_PROJECTVERTEX(out, in, viewportcenter, viewportscale) \ +{ \ + __m128 p = (in), w = _mm_shuffle_ps(p, p, _MM_SHUFFLE(3, 3, 3, 3)); \ + p = _mm_move_ss(_mm_shuffle_ps(p, p, _MM_SHUFFLE(2, 1, 0, 3)), _mm_set_ss(1.0f)); \ + p = _mm_add_ps(viewportcenter, _mm_div_ps(_mm_mul_ps(viewportscale, p), w)); \ + out = _mm_shuffle_ps(p, p, _MM_SHUFFLE(0, 3, 2, 1)); \ +} + +#define DPSOFTRAST_PROJECTY(out, in, viewportcenter, viewportscale) \ +{ \ + __m128 p = (in), w = _mm_shuffle_ps(p, p, _MM_SHUFFLE(3, 3, 3, 3)); \ + p = _mm_move_ss(_mm_shuffle_ps(p, p, _MM_SHUFFLE(2, 1, 0, 3)), _mm_set_ss(1.0f)); \ + p = _mm_add_ps(viewportcenter, _mm_div_ps(_mm_mul_ps(viewportscale, p), w)); \ + out = _mm_shuffle_ps(p, p, _MM_SHUFFLE(0, 3, 2, 1)); \ +} + +#define DPSOFTRAST_TRANSFORMVERTEX(out, in, m0, m1, m2, m3) \ +{ \ + __m128 p = (in); \ + out = _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(p, p, _MM_SHUFFLE(0, 0, 0, 0)), m0), \ + _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(p, p, _MM_SHUFFLE(1, 1, 1, 1)), m1), \ + _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(p, p, _MM_SHUFFLE(2, 2, 2, 2)), m2), \ + _mm_mul_ps(_mm_shuffle_ps(p, p, _MM_SHUFFLE(3, 3, 3, 3)), m3)))); \ +} + +static int DPSOFTRAST_Vertex_BoundY(int *starty, int *endy, const float *minposf, const float *maxposf, const float *inmatrix16f) { + int clipmask = 0xFF; __m128 viewportcenter = _mm_load_ps(dpsoftrast.fb_viewportcenter), viewportscale = _mm_load_ps(dpsoftrast.fb_viewportscale); - __m128 w = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)); - v = _mm_move_ss(_mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 1, 0, 3)), _mm_set_ss(1.0f)); - v = _mm_add_ps(viewportcenter, _mm_div_ps(_mm_mul_ps(viewportscale, v), w)); - v = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1)); - return v; + __m128 bb[8], clipdist[8], minproj = _mm_set_ss(2.0f), maxproj = _mm_set_ss(-2.0f); + __m128 m0 = _mm_loadu_ps(inmatrix16f), m1 = _mm_loadu_ps(inmatrix16f + 4), m2 = _mm_loadu_ps(inmatrix16f + 8), m3 = _mm_loadu_ps(inmatrix16f + 12); + __m128 minpos = _mm_load_ps(minposf), maxpos = _mm_load_ps(maxposf); + m0 = _mm_shuffle_ps(m0, m0, _MM_SHUFFLE(3, 2, 0, 1)); + m1 = _mm_shuffle_ps(m1, m1, _MM_SHUFFLE(3, 2, 0, 1)); + m2 = _mm_shuffle_ps(m2, m2, _MM_SHUFFLE(3, 2, 0, 1)); + m3 = _mm_shuffle_ps(m3, m3, _MM_SHUFFLE(3, 2, 0, 1)); + #define BBFRONT(k, pos) \ + { \ + DPSOFTRAST_TRANSFORMVERTEX(bb[k], pos, m0, m1, m2, m3); \ + clipdist[k] = _mm_add_ss(_mm_shuffle_ps(bb[k], bb[k], _MM_SHUFFLE(2, 2, 2, 2)), _mm_shuffle_ps(bb[k], bb[k], _MM_SHUFFLE(3, 3, 3, 3))); \ + if (_mm_ucomige_ss(clipdist[k], _mm_setzero_ps())) \ + { \ + __m128 proj; \ + clipmask &= ~(1<startx; - int endx = span->endx; +#ifdef SSE_POSSIBLE + float *outf = dpsoftrast.post_array4f[outarray]; + const unsigned char *inb; + int firstvertex = dpsoftrast.firstvertex; + int numvertices = dpsoftrast.numvertices; + int stride; + switch(inarray) + { + case DPSOFTRAST_ARRAY_POSITION: + stride = dpsoftrast.stride_vertex; + inb = (unsigned char *)dpsoftrast.pointer_vertex3f + firstvertex * stride; + DPSOFTRAST_Load3fTo4f(outf, inb, numvertices, stride); + break; + case DPSOFTRAST_ARRAY_COLOR: + stride = dpsoftrast.stride_color; + if (dpsoftrast.pointer_color4f) + { + inb = (const unsigned char *)dpsoftrast.pointer_color4f + firstvertex * stride; + DPSOFTRAST_Load4fTo4f(outf, inb, numvertices, stride); + } + else if (dpsoftrast.pointer_color4ub) + { + stride = dpsoftrast.stride_color; + inb = (const unsigned char *)dpsoftrast.pointer_color4ub + firstvertex * stride; + DPSOFTRAST_Load4bTo4f(outf, inb, numvertices, stride); + } + else + { + DPSOFTRAST_Fill4f(outf, dpsoftrast.color, numvertices); + } + break; + default: + stride = dpsoftrast.stride_texcoord[inarray-DPSOFTRAST_ARRAY_TEXCOORD0]; + if (dpsoftrast.pointer_texcoordf[inarray-DPSOFTRAST_ARRAY_TEXCOORD0]) + { + inb = (const unsigned char *)dpsoftrast.pointer_texcoordf[inarray-DPSOFTRAST_ARRAY_TEXCOORD0] + firstvertex * stride; + switch(dpsoftrast.components_texcoord[inarray-DPSOFTRAST_ARRAY_TEXCOORD0]) + { + case 2: + DPSOFTRAST_Load2fTo4f(outf, inb, numvertices, stride); + break; + case 3: + DPSOFTRAST_Load3fTo4f(outf, inb, numvertices, stride); + break; + case 4: + DPSOFTRAST_Load4fTo4f(outf, inb, numvertices, stride); + break; + } + } + break; + } + return outf; +#else + return NULL; +#endif +} + +static float *DPSOFTRAST_Array_Transform(int outarray, int inarray, const float *inmatrix16f) +{ + float *data = inarray >= 0 ? DPSOFTRAST_Array_Load(outarray, inarray) : dpsoftrast.post_array4f[outarray]; + DPSOFTRAST_Vertex_Transform(data, data, dpsoftrast.numvertices, inmatrix16f); + return data; +} + +#if 0 +static float *DPSOFTRAST_Array_Project(int outarray, int inarray) +{ +#ifdef SSE_POSSIBLE + float *data = inarray >= 0 ? DPSOFTRAST_Array_Load(outarray, inarray) : dpsoftrast.post_array4f[outarray]; + dpsoftrast.drawclipped = DPSOFTRAST_Vertex_Project(data, dpsoftrast.screencoord4f, &dpsoftrast.drawstarty, &dpsoftrast.drawendy, data, dpsoftrast.numvertices); + return data; +#else + return NULL; +#endif +} +#endif + +static float *DPSOFTRAST_Array_TransformProject(int outarray, int inarray, const float *inmatrix16f) +{ +#ifdef SSE_POSSIBLE + float *data = inarray >= 0 ? DPSOFTRAST_Array_Load(outarray, inarray) : dpsoftrast.post_array4f[outarray]; + dpsoftrast.drawclipped = DPSOFTRAST_Vertex_TransformProject(data, dpsoftrast.screencoord4f, &dpsoftrast.drawstarty, &dpsoftrast.drawendy, data, dpsoftrast.numvertices, inmatrix16f); + return data; +#else + return NULL; +#endif +} + +void DPSOFTRAST_Draw_Span_Begin(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *zf) +{ + int x; + int startx = span->startx; + int endx = span->endx; float wslope = triangle->w[0]; float w = triangle->w[2] + span->x*wslope + span->y*triangle->w[1]; float endz = 1.0f / (w + wslope * startx); + if (triangle->w[0] == 0) + { + // LordHavoc: fast flat polygons (HUD/menu) + for (x = startx;x < endx;x++) + zf[x] = endz; + return; + } for (x = startx;x < endx;) { int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1; float z = endz, dz; - if(nextsub >= endx) nextsub = endsub = endx-1; + if (nextsub >= endx) nextsub = endsub = endx-1; endz = 1.0f / (w + wslope * nextsub); dz = x < nextsub ? (endz - z) / (nextsub - x) : 0.0f; for (; x <= endsub; x++, z += dz) @@ -1852,305 +2032,318 @@ void DPSOFTRAST_Draw_Span_Begin(DPSOFTRAST_State_Thread *thread, const DPSOFTRAS } } -void DPSOFTRAST_Draw_Span_Finish(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, const float * RESTRICT in4f) +void DPSOFTRAST_Draw_Span_FinishBGRA8(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, const unsigned char* RESTRICT in4ub) { +#ifdef SSE_POSSIBLE int x; int startx = span->startx; int endx = span->endx; - int d[4]; - float a, b; + int maskx; + int subx; + const unsigned int * RESTRICT ini = (const unsigned int *)in4ub; unsigned char * RESTRICT pixelmask = span->pixelmask; unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0]; + unsigned int * RESTRICT pixeli = (unsigned int *)dpsoftrast.fb_colorpixels[0]; if (!pixel) return; pixel += (span->y * dpsoftrast.fb_width + span->x) * 4; + pixeli += span->y * dpsoftrast.fb_width + span->x; // handle alphatest now (this affects depth writes too) - if (thread->alphatest) + if (thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) for (x = startx;x < endx;x++) - if (in4f[x*4+3] < 0.5f) + if (in4ub[x*4+3] < 128) pixelmask[x] = false; - // FIXME: this does not handle bigendian + // LordHavoc: clear pixelmask for some pixels in alphablend cases, this + // helps sprites, text and hud artwork switch(thread->fb_blendmode) { - case DPSOFTRAST_BLENDMODE_OPAQUE: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - d[0] = (int)(in4f[x*4+2]*255.0f);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*255.0f);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*255.0f);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*255.0f);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } - break; case DPSOFTRAST_BLENDMODE_ALPHA: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - a = in4f[x*4+3] * 255.0f; - b = 1.0f - in4f[x*4+3]; - d[0] = (int)(in4f[x*4+2]*a+pixel[x*4+0]*b);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*a+pixel[x*4+1]*b);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*a+pixel[x*4+2]*b);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*a+pixel[x*4+3]*b);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } - break; case DPSOFTRAST_BLENDMODE_ADDALPHA: + case DPSOFTRAST_BLENDMODE_SUBALPHA: + maskx = startx; for (x = startx;x < endx;x++) { - if (!pixelmask[x]) - continue; - a = in4f[x*4+3] * 255.0f; - d[0] = (int)(in4f[x*4+2]*a+pixel[x*4+0]);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*a+pixel[x*4+1]);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*a+pixel[x*4+2]);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*a+pixel[x*4+3]);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; + if (in4ub[x*4+3] >= 1) + { + startx = x; + for (;;) + { + while (++x < endx && in4ub[x*4+3] >= 1) ; + maskx = x; + if (x >= endx) break; + ++x; + while (++x < endx && in4ub[x*4+3] < 1) pixelmask[x] = false; + if (x >= endx) break; + } + break; + } } + endx = maskx; break; + case DPSOFTRAST_BLENDMODE_OPAQUE: case DPSOFTRAST_BLENDMODE_ADD: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - d[0] = (int)(in4f[x*4+2]*255.0f+pixel[x*4+0]);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*255.0f+pixel[x*4+1]);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*255.0f+pixel[x*4+2]);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*255.0f+pixel[x*4+3]);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } - break; case DPSOFTRAST_BLENDMODE_INVMOD: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - d[0] = (int)((1.0f-in4f[x*4+2])*pixel[x*4+0]);if (d[0] > 255) d[0] = 255; - d[1] = (int)((1.0f-in4f[x*4+1])*pixel[x*4+1]);if (d[1] > 255) d[1] = 255; - d[2] = (int)((1.0f-in4f[x*4+0])*pixel[x*4+2]);if (d[2] > 255) d[2] = 255; - d[3] = (int)((1.0f-in4f[x*4+3])*pixel[x*4+3]);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } - break; case DPSOFTRAST_BLENDMODE_MUL: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - d[0] = (int)(in4f[x*4+2]*pixel[x*4+0]);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*pixel[x*4+1]);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*pixel[x*4+2]);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*pixel[x*4+3]);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } - break; case DPSOFTRAST_BLENDMODE_MUL2: - for (x = startx;x < endx;x++) - { - if (!pixelmask[x]) - continue; - d[0] = (int)(in4f[x*4+2]*pixel[x*4+0]*2.0f);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*pixel[x*4+1]*2.0f);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*pixel[x*4+2]*2.0f);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*pixel[x*4+3]*2.0f);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; - } + case DPSOFTRAST_BLENDMODE_PSEUDOALPHA: + case DPSOFTRAST_BLENDMODE_INVADD: break; - case DPSOFTRAST_BLENDMODE_SUBALPHA: - for (x = startx;x < endx;x++) + } + // put some special values at the end of the mask to ensure the loops end + pixelmask[endx] = 1; + pixelmask[endx+1] = 0; + // LordHavoc: use a double loop to identify subspans, this helps the + // optimized copy/blend loops to perform at their best, most triangles + // have only one run of pixels, and do the search using wide reads... + x = startx; + while (x < endx) + { + // if this pixel is masked off, it's probably not alone... + if (!pixelmask[x]) { - if (!pixelmask[x]) - continue; - a = in4f[x*4+3] * -255.0f; - d[0] = (int)(in4f[x*4+2]*a+pixel[x*4+0]);if (d[0] > 255) d[0] = 255;if (d[0] < 0) d[0] = 0; - d[1] = (int)(in4f[x*4+1]*a+pixel[x*4+1]);if (d[1] > 255) d[1] = 255;if (d[1] < 0) d[1] = 0; - d[2] = (int)(in4f[x*4+0]*a+pixel[x*4+2]);if (d[2] > 255) d[2] = 255;if (d[2] < 0) d[2] = 0; - d[3] = (int)(in4f[x*4+3]*a+pixel[x*4+3]);if (d[3] > 255) d[3] = 255;if (d[3] < 0) d[3] = 0; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; + x++; +#if 1 + if (x + 8 < endx) + { + // the 4-item search must be aligned or else it stalls badly + if ((x & 3) && !pixelmask[x]) + { + if(pixelmask[x]) goto endmasked; + x++; + if (x & 3) + { + if(pixelmask[x]) goto endmasked; + x++; + if (x & 3) + { + if(pixelmask[x]) goto endmasked; + x++; + } + } + } + while (*(unsigned int *)&pixelmask[x] == 0x00000000) + x += 4; + } +#endif + for (;!pixelmask[x];x++) + ; + // rather than continue the loop, just check the end variable + if (x >= endx) + break; } - break; - case DPSOFTRAST_BLENDMODE_PSEUDOALPHA: - for (x = startx;x < endx;x++) + endmasked: + // find length of subspan + subx = x + 1; +#if 1 + if (subx + 8 < endx) { - if (!pixelmask[x]) - continue; - a = 255.0f; - b = 1.0f - in4f[x*4+3]; - d[0] = (int)(in4f[x*4+2]*a+pixel[x*4+0]*b);if (d[0] > 255) d[0] = 255; - d[1] = (int)(in4f[x*4+1]*a+pixel[x*4+1]*b);if (d[1] > 255) d[1] = 255; - d[2] = (int)(in4f[x*4+0]*a+pixel[x*4+2]*b);if (d[2] > 255) d[2] = 255; - d[3] = (int)(in4f[x*4+3]*a+pixel[x*4+3]*b);if (d[3] > 255) d[3] = 255; - pixel[x*4+0] = d[0]; - pixel[x*4+1] = d[1]; - pixel[x*4+2] = d[2]; - pixel[x*4+3] = d[3]; + if (subx & 3) + { + if(!pixelmask[subx]) goto endunmasked; + subx++; + if (subx & 3) + { + if(!pixelmask[subx]) goto endunmasked; + subx++; + if (subx & 3) + { + if(!pixelmask[subx]) goto endunmasked; + subx++; + } + } + } + while (*(unsigned int *)&pixelmask[subx] == 0x01010101) + subx += 4; } - break; - } -} - -void DPSOFTRAST_Draw_Span_FinishBGRA8(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, const unsigned char* RESTRICT in4ub) -{ -#ifdef SSE2_PRESENT - int x; - int startx = span->startx; - int endx = span->endx; - const unsigned int * RESTRICT ini = (const unsigned int *)in4ub; - unsigned char * RESTRICT pixelmask = span->pixelmask; - unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0]; - unsigned int * RESTRICT pixeli = (unsigned int *)dpsoftrast.fb_colorpixels[0]; - if (!pixel) - return; - pixel += (span->y * dpsoftrast.fb_width + span->x) * 4; - pixeli += span->y * dpsoftrast.fb_width + span->x; - // handle alphatest now (this affects depth writes too) - if (thread->alphatest) - for (x = startx;x < endx;x++) - if (in4ub[x*4+3] < 0.5f) - pixelmask[x] = false; - // FIXME: this does not handle bigendian - switch(thread->fb_blendmode) - { - case DPSOFTRAST_BLENDMODE_OPAQUE: - for (x = startx;x + 4 <= endx;) +#endif + for (;pixelmask[subx];subx++) + ; + // the checks can overshoot, so make sure to clip it... + if (subx > endx) + subx = endx; + endunmasked: + // now that we know the subspan length... process! + switch(thread->fb_blendmode) { - if (*(const unsigned int *)&pixelmask[x] == 0x01010101) + case DPSOFTRAST_BLENDMODE_OPAQUE: +#if 0 + if (subx - x >= 16) { - _mm_storeu_si128((__m128i *)&pixeli[x], _mm_loadu_si128((const __m128i *)&ini[x])); - x += 4; + memcpy(pixeli + x, ini + x, (subx - x) * sizeof(pixeli[x])); + x = subx; } else +#elif 1 + while (x + 16 <= subx) + { + _mm_storeu_si128((__m128i *)&pixeli[x], _mm_loadu_si128((const __m128i *)&ini[x])); + _mm_storeu_si128((__m128i *)&pixeli[x+4], _mm_loadu_si128((const __m128i *)&ini[x+4])); + _mm_storeu_si128((__m128i *)&pixeli[x+8], _mm_loadu_si128((const __m128i *)&ini[x+8])); + _mm_storeu_si128((__m128i *)&pixeli[x+12], _mm_loadu_si128((const __m128i *)&ini[x+12])); + x += 16; + } +#endif { - if (pixelmask[x]) + while (x + 4 <= subx) + { + _mm_storeu_si128((__m128i *)&pixeli[x], _mm_loadu_si128((const __m128i *)&ini[x])); + x += 4; + } + if (x + 2 <= subx) + { + pixeli[x] = ini[x]; + pixeli[x+1] = ini[x+1]; + x += 2; + } + if (x < subx) + { pixeli[x] = ini[x]; - x++; + x++; + } } - } - for (;x < endx;x++) - if (pixelmask[x]) - pixeli[x] = ini[x]; - break; - case DPSOFTRAST_BLENDMODE_ALPHA: - #define FINISHBLEND(blend2, blend1) \ - for (x = startx;x + 2 <= endx;x += 2) \ - { \ - __m128i src, dst; \ - switch (*(const unsigned short*)&pixelmask[x]) \ + break; + case DPSOFTRAST_BLENDMODE_ALPHA: + #define FINISHBLEND(blend2, blend1) \ + for (;x + 1 < subx;x += 2) \ { \ - case 0x0101: \ + __m128i src, dst; \ src = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&ini[x]), _mm_setzero_si128()); \ dst = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixeli[x]), _mm_setzero_si128()); \ blend2; \ _mm_storel_epi64((__m128i *)&pixeli[x], _mm_packus_epi16(dst, dst)); \ - continue; \ - case 0x0100: \ - src = _mm_unpacklo_epi8(_mm_cvtsi32_si128(ini[x+1]), _mm_setzero_si128()); \ - dst = _mm_unpacklo_epi8(_mm_cvtsi32_si128(pixeli[x+1]), _mm_setzero_si128()); \ - blend1; \ - pixeli[x+1] = _mm_cvtsi128_si32(_mm_packus_epi16(dst, dst)); \ - continue; \ - case 0x0001: \ + } \ + if (x < subx) \ + { \ + __m128i src, dst; \ src = _mm_unpacklo_epi8(_mm_cvtsi32_si128(ini[x]), _mm_setzero_si128()); \ dst = _mm_unpacklo_epi8(_mm_cvtsi32_si128(pixeli[x]), _mm_setzero_si128()); \ blend1; \ pixeli[x] = _mm_cvtsi128_si32(_mm_packus_epi16(dst, dst)); \ - continue; \ - } \ - break; \ - } \ - for(;x < endx; x++) \ - { \ - __m128i src, dst; \ - if (!pixelmask[x]) \ - continue; \ - src = _mm_unpacklo_epi8(_mm_cvtsi32_si128(ini[x]), _mm_setzero_si128()); \ - dst = _mm_unpacklo_epi8(_mm_cvtsi32_si128(pixeli[x]), _mm_setzero_si128()); \ - blend1; \ - pixeli[x] = _mm_cvtsi128_si32(_mm_packus_epi16(dst, dst)); \ + x++; \ + } + FINISHBLEND({ + __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(src, dst), 4), _mm_slli_epi16(blend, 4))); + }, { + __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(src, dst), 4), _mm_slli_epi16(blend, 4))); + }); + break; + case DPSOFTRAST_BLENDMODE_ADDALPHA: + FINISHBLEND({ + __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); + }, { + __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); + }); + break; + case DPSOFTRAST_BLENDMODE_ADD: + FINISHBLEND({ dst = _mm_add_epi16(src, dst); }, { dst = _mm_add_epi16(src, dst); }); + break; + case DPSOFTRAST_BLENDMODE_INVMOD: + FINISHBLEND({ + dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, src), 8)); + }, { + dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, src), 8)); + }); + break; + case DPSOFTRAST_BLENDMODE_MUL: + FINISHBLEND({ dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 8); }, { dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 8); }); + break; + case DPSOFTRAST_BLENDMODE_MUL2: + FINISHBLEND({ dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 7); }, { dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 7); }); + break; + case DPSOFTRAST_BLENDMODE_SUBALPHA: + FINISHBLEND({ + __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); + }, { + __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); + }); + break; + case DPSOFTRAST_BLENDMODE_PSEUDOALPHA: + FINISHBLEND({ + __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(src, _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, blend), 8))); + }, { + __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); + dst = _mm_add_epi16(src, _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, blend), 8))); + }); + break; + case DPSOFTRAST_BLENDMODE_INVADD: + FINISHBLEND({ + dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(_mm_set1_epi16(255), dst), 4), _mm_slli_epi16(src, 4))); + }, { + dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(_mm_set1_epi16(255), dst), 4), _mm_slli_epi16(src, 4))); + }); + break; } - - FINISHBLEND({ - __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(src, dst), 4), _mm_slli_epi16(blend, 4))); - }, { - __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(dst, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(src, dst), 4), _mm_slli_epi16(blend, 4))); - }); - break; - case DPSOFTRAST_BLENDMODE_ADDALPHA: - FINISHBLEND({ - __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); - }, { - __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); - }); - break; - case DPSOFTRAST_BLENDMODE_ADD: - FINISHBLEND({ dst = _mm_add_epi16(src, dst); }, { dst = _mm_add_epi16(src, dst); }); - break; - case DPSOFTRAST_BLENDMODE_INVMOD: - FINISHBLEND({ - dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, src), 8)); - }, { - dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, src), 8)); - }); - break; - case DPSOFTRAST_BLENDMODE_MUL: - FINISHBLEND({ dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 8); }, { dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 8); }); - break; - case DPSOFTRAST_BLENDMODE_MUL2: - FINISHBLEND({ dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 7); }, { dst = _mm_srli_epi16(_mm_mullo_epi16(src, dst), 7); }); - break; - case DPSOFTRAST_BLENDMODE_SUBALPHA: - FINISHBLEND({ - __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); - }, { - __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(src, blend), 8)); - }); - break; - case DPSOFTRAST_BLENDMODE_PSEUDOALPHA: - FINISHBLEND({ - __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(src, _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, blend), 8))); - }, { - __m128i blend = _mm_shufflelo_epi16(src, _MM_SHUFFLE(3, 3, 3, 3)); - dst = _mm_add_epi16(src, _mm_sub_epi16(dst, _mm_srli_epi16(_mm_mullo_epi16(dst, blend), 8))); - }); - break; } #endif } +static void DPSOFTRAST_Texture2DBGRA8(DPSOFTRAST_Texture *texture, int mip, float x, float y, unsigned char c[4]) + // warning: this is SLOW, only use if the optimized per-span functions won't do +{ + const unsigned char * RESTRICT pixelbase; + const unsigned char * RESTRICT pixel[4]; + int width = texture->mipmap[mip][2], height = texture->mipmap[mip][3]; + int wrapmask[2] = { width-1, height-1 }; + pixelbase = (unsigned char *)texture->bytes + texture->mipmap[mip][0]; + if(texture->filter & DPSOFTRAST_TEXTURE_FILTER_LINEAR) + { + unsigned int tc[2] = { x * (width<<12) - 2048, y * (height<<12) - 2048}; + unsigned int frac[2] = { tc[0]&0xFFF, tc[1]&0xFFF }; + unsigned int ifrac[2] = { 0x1000 - frac[0], 0x1000 - frac[1] }; + unsigned int lerp[4] = { ifrac[0]*ifrac[1], frac[0]*ifrac[1], ifrac[0]*frac[1], frac[0]*frac[1] }; + int tci[2] = { tc[0]>>12, tc[1]>>12 }; + int tci1[2] = { tci[0] + 1, tci[1] + 1 }; + if (texture->flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE) + { + tci[0] = tci[0] >= 0 ? (tci[0] <= wrapmask[0] ? tci[0] : wrapmask[0]) : 0; + tci[1] = tci[1] >= 0 ? (tci[1] <= wrapmask[1] ? tci[1] : wrapmask[1]) : 0; + tci1[0] = tci1[0] >= 0 ? (tci1[0] <= wrapmask[0] ? tci1[0] : wrapmask[0]) : 0; + tci1[1] = tci1[1] >= 0 ? (tci1[1] <= wrapmask[1] ? tci1[1] : wrapmask[1]) : 0; + } + else + { + tci[0] &= wrapmask[0]; + tci[1] &= wrapmask[1]; + tci1[0] &= wrapmask[0]; + tci1[1] &= wrapmask[1]; + } + pixel[0] = pixelbase + 4 * (tci[1]*width+tci[0]); + pixel[1] = pixelbase + 4 * (tci[1]*width+tci1[0]); + pixel[2] = pixelbase + 4 * (tci1[1]*width+tci[0]); + pixel[3] = pixelbase + 4 * (tci1[1]*width+tci1[0]); + c[0] = (pixel[0][0]*lerp[0]+pixel[1][0]*lerp[1]+pixel[2][0]*lerp[2]+pixel[3][0]*lerp[3])>>24; + c[1] = (pixel[0][1]*lerp[0]+pixel[1][1]*lerp[1]+pixel[2][1]*lerp[2]+pixel[3][1]*lerp[3])>>24; + c[2] = (pixel[0][2]*lerp[0]+pixel[1][2]*lerp[1]+pixel[2][2]*lerp[2]+pixel[3][2]*lerp[3])>>24; + c[3] = (pixel[0][3]*lerp[0]+pixel[1][3]*lerp[1]+pixel[2][3]*lerp[2]+pixel[3][3]*lerp[3])>>24; + } + else + { + int tci[2] = { x * width, y * height }; + if (texture->flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE) + { + tci[0] = tci[0] >= 0 ? (tci[0] <= wrapmask[0] ? tci[0] : wrapmask[0]) : 0; + tci[1] = tci[1] >= 0 ? (tci[1] <= wrapmask[1] ? tci[1] : wrapmask[1]) : 0; + } + else + { + tci[0] &= wrapmask[0]; + tci[1] &= wrapmask[1]; + } + pixel[0] = pixelbase + 4 * (tci[1]*width+tci[0]); + c[0] = pixel[0][0]; + c[1] = pixel[0][1]; + c[2] = pixel[0][2]; + c[3] = pixel[0][3]; + } +} + void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float * RESTRICT out4f, int texunitindex, int arrayindex, const float * RESTRICT zf) { int x; @@ -2215,28 +2408,38 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons tcimax[1] = texture->mipmap[mip][3]-1; tciwrapmask[0] = texture->mipmap[mip][2]-1; tciwrapmask[1] = texture->mipmap[mip][3]-1; - endtc[0] = (data[0] + slope[0]*startx) * zf[startx] * tcscale[0] - 0.5f; - endtc[1] = (data[1] + slope[1]*startx) * zf[startx] * tcscale[1] - 0.5f; + endtc[0] = (data[0] + slope[0]*startx) * zf[startx] * tcscale[0]; + endtc[1] = (data[1] + slope[1]*startx) * zf[startx] * tcscale[1]; + if (filter) + { + endtc[0] -= 0.5f; + endtc[1] -= 0.5f; + } for (x = startx;x < endx;) { unsigned int subtc[2]; unsigned int substep[2]; - float subscale = 65536.0f/DPSOFTRAST_DRAW_MAXSUBSPAN; + float subscale = 4096.0f/DPSOFTRAST_DRAW_MAXSUBSPAN; int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1; - if(nextsub >= endx) + if (nextsub >= endx) { nextsub = endsub = endx-1; - if(x < nextsub) subscale = 65536.0f / (nextsub - x); + if (x < nextsub) subscale = 4096.0f / (nextsub - x); } tc[0] = endtc[0]; tc[1] = endtc[1]; - endtc[0] = (data[0] + slope[0]*nextsub) * zf[nextsub] * tcscale[0] - 0.5f; - endtc[1] = (data[1] + slope[1]*nextsub) * zf[nextsub] * tcscale[1] - 0.5f; + endtc[0] = (data[0] + slope[0]*nextsub) * zf[nextsub] * tcscale[0]; + endtc[1] = (data[1] + slope[1]*nextsub) * zf[nextsub] * tcscale[1]; + if (filter) + { + endtc[0] -= 0.5f; + endtc[1] -= 0.5f; + } substep[0] = (endtc[0] - tc[0]) * subscale; substep[1] = (endtc[1] - tc[1]) * subscale; - subtc[0] = tc[0] * (1<<16); - subtc[1] = tc[1] * (1<<16); - if(filter) + subtc[0] = tc[0] * (1<<12); + subtc[1] = tc[1] * (1<<12); + if (filter) { if (flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE) { @@ -2245,8 +2448,8 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons unsigned int frac[2] = { subtc[0]&0xFFF, subtc[1]&0xFFF }; unsigned int ifrac[2] = { 0x1000 - frac[0], 0x1000 - frac[1] }; unsigned int lerp[4] = { ifrac[0]*ifrac[1], frac[0]*ifrac[1], ifrac[0]*frac[1], frac[0]*frac[1] }; - tci[0] = subtc[0]>>16; - tci[1] = subtc[1]>>16; + tci[0] = subtc[0]>>12; + tci[1] = subtc[1]>>12; tci1[0] = tci[0] + 1; tci1[1] = tci[1] + 1; tci[0] = tci[0] >= tcimin[0] ? (tci[0] <= tcimax[0] ? tci[0] : tcimax[0]) : tcimin[0]; @@ -2274,8 +2477,8 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons unsigned int frac[2] = { subtc[0]&0xFFF, subtc[1]&0xFFF }; unsigned int ifrac[2] = { 0x1000 - frac[0], 0x1000 - frac[1] }; unsigned int lerp[4] = { ifrac[0]*ifrac[1], frac[0]*ifrac[1], ifrac[0]*frac[1], frac[0]*frac[1] }; - tci[0] = subtc[0]>>16; - tci[1] = subtc[1]>>16; + tci[0] = subtc[0]>>12; + tci[1] = subtc[1]>>12; tci1[0] = tci[0] + 1; tci1[1] = tci[1] + 1; tci[0] &= tciwrapmask[0]; @@ -2301,8 +2504,8 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons { for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1]) { - tci[0] = subtc[0]>>16; - tci[1] = subtc[1]>>16; + tci[0] = subtc[0]>>12; + tci[1] = subtc[1]>>12; tci[0] = tci[0] >= tcimin[0] ? (tci[0] <= tcimax[0] ? tci[0] : tcimax[0]) : tcimin[0]; tci[1] = tci[1] >= tcimin[1] ? (tci[1] <= tcimax[1] ? tci[1] : tcimax[1]) : tcimin[1]; pixel[0] = pixelbase + 4 * (tci[1]*tciwidth+tci[0]); @@ -2320,8 +2523,8 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons { for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1]) { - tci[0] = subtc[0]>>16; - tci[1] = subtc[1]>>16; + tci[0] = subtc[0]>>12; + tci[1] = subtc[1]>>12; tci[0] &= tciwrapmask[0]; tci[1] &= tciwrapmask[1]; pixel[0] = pixelbase + 4 * (tci[1]*tciwidth+tci[0]); @@ -2340,7 +2543,7 @@ void DPSOFTRAST_Draw_Span_Texture2DVarying(DPSOFTRAST_State_Thread *thread, cons void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char * RESTRICT out4ub, int texunitindex, int arrayindex, const float * RESTRICT zf) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x; int startx = span->startx; int endx = span->endx; @@ -2351,13 +2554,14 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, __m128i subtc, substep, endsubtc; int filter; int mip; - unsigned int *outi = (unsigned int *)out4ub; + int affine; // LordHavoc: optimized affine texturing case + unsigned int * RESTRICT outi = (unsigned int *)out4ub; const unsigned char * RESTRICT pixelbase; DPSOFTRAST_Texture *texture = thread->texbound[texunitindex]; // if no texture is bound, just fill it with white if (!texture) { - memset(out4ub + startx*4, 255, span->length*4); + memset(out4ub + startx*4, 255, (span->endx - span->startx)*4); return; } mip = triangle->mip[texunitindex]; @@ -2370,6 +2574,7 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, outi[x] = k; return; } + affine = zf[startx] == zf[endx-1]; filter = texture->filter & DPSOFTRAST_TEXTURE_FILTER_LINEAR; DPSOFTRAST_CALCATTRIB(triangle, span, data, slope, arrayindex); flags = texture->flags; @@ -2378,22 +2583,26 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, tcscale = _mm_cvtepi32_ps(tcsize); data = _mm_mul_ps(_mm_movelh_ps(data, data), tcscale); slope = _mm_mul_ps(_mm_movelh_ps(slope, slope), tcscale); - endtc = _mm_sub_ps(_mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))), _mm_load1_ps(&zf[startx])), _mm_set1_ps(0.5f)); + endtc = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))), _mm_load1_ps(&zf[startx])); + if (filter) + endtc = _mm_sub_ps(endtc, _mm_set1_ps(0.5f)); endsubtc = _mm_cvtps_epi32(_mm_mul_ps(endtc, _mm_set1_ps(65536.0f))); tcoffset = _mm_add_epi32(_mm_slli_epi32(_mm_shuffle_epi32(tcsize, _MM_SHUFFLE(0, 0, 0, 0)), 18), _mm_set1_epi32(4)); - tcmax = filter ? _mm_packs_epi32(tcmask, tcmask) : _mm_slli_epi32(tcmask, 16); + tcmax = _mm_packs_epi32(tcmask, tcmask); for (x = startx;x < endx;) { int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1; __m128 subscale = _mm_set1_ps(65536.0f/DPSOFTRAST_DRAW_MAXSUBSPAN); - if(nextsub >= endx) + if (nextsub >= endx || affine) { nextsub = endsub = endx-1; - if(x < nextsub) subscale = _mm_set1_ps(65536.0f / (nextsub - x)); + if (x < nextsub) subscale = _mm_set1_ps(65536.0f / (nextsub - x)); } tc = endtc; subtc = endsubtc; - endtc = _mm_sub_ps(_mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(nextsub))), _mm_load1_ps(&zf[nextsub])), _mm_set1_ps(0.5f)); + endtc = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(nextsub))), _mm_load1_ps(&zf[nextsub])); + if (filter) + endtc = _mm_sub_ps(endtc, _mm_set1_ps(0.5f)); substep = _mm_cvtps_epi32(_mm_mul_ps(_mm_sub_ps(endtc, tc), subscale)); endsubtc = _mm_cvtps_epi32(_mm_mul_ps(endtc, _mm_set1_ps(65536.0f))); subtc = _mm_unpacklo_epi64(subtc, _mm_add_epi32(subtc, substep)); @@ -2403,14 +2612,18 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, __m128i tcrange = _mm_srai_epi32(_mm_unpacklo_epi64(subtc, _mm_add_epi32(endsubtc, substep)), 16); if (_mm_movemask_epi8(_mm_andnot_si128(_mm_cmplt_epi32(tcrange, _mm_setzero_si128()), _mm_cmplt_epi32(tcrange, tcmask))) == 0xFFFF) { + int stride = _mm_cvtsi128_si32(tcoffset)>>16; for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep)) { + const unsigned char * RESTRICT ptr1, * RESTRICT ptr2; __m128i tci = _mm_shufflehi_epi16(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 1, 3, 1)), pix1, pix2, pix3, pix4, fracm; - tci = _mm_madd_epi16(_mm_add_epi16(tci, _mm_setr_epi32(0, 0x10000, 0, 0x10000)), tcoffset); - pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(tci)]), _mm_setzero_si128()); - pix2 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))]), _mm_setzero_si128()); - pix3 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]), _mm_setzero_si128()); - pix4 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))]), _mm_setzero_si128()); + tci = _mm_madd_epi16(tci, tcoffset); + ptr1 = pixelbase + _mm_cvtsi128_si32(tci); + ptr2 = pixelbase + _mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2))); + pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)ptr1), _mm_setzero_si128()); + pix2 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(ptr1 + stride)), _mm_setzero_si128()); + pix3 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)ptr2), _mm_setzero_si128()); + pix4 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(ptr2 + stride)), _mm_setzero_si128()); fracm = _mm_srli_epi16(subtc, 1); pix1 = _mm_add_epi16(pix1, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1), @@ -2427,10 +2640,12 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, } if (x <= endsub) { + const unsigned char * RESTRICT ptr1; __m128i tci = _mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), pix1, pix2, fracm; - tci = _mm_madd_epi16(_mm_add_epi16(tci, _mm_setr_epi32(0, 0x10000, 0, 0)), tcoffset); - pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(tci)]), _mm_setzero_si128()); - pix2 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))]), _mm_setzero_si128()); + tci = _mm_madd_epi16(tci, tcoffset); + ptr1 = pixelbase + _mm_cvtsi128_si32(tci); + pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)ptr1), _mm_setzero_si128()); + pix2 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)(ptr1 + stride)), _mm_setzero_si128()); fracm = _mm_srli_epi16(subtc, 1); pix1 = _mm_add_epi16(pix1, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1), @@ -2568,16 +2783,16 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, { for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep)) { - __m128i tci = _mm_min_epi16(_mm_max_epi16(subtc, _mm_setzero_si128()), tcmax); - tci = _mm_shufflehi_epi16(_mm_shufflelo_epi16(tci, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 1, 3, 1)); + __m128i tci = _mm_shufflehi_epi16(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 1, 3, 1)); + tci = _mm_min_epi16(_mm_max_epi16(tci, _mm_setzero_si128()), tcmax); tci = _mm_madd_epi16(tci, tcoffset); outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]; - outi[x+1] = *(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))]; + outi[x+1] = *(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]; } if (x <= endsub) { - __m128i tci = _mm_min_epi16(_mm_max_epi16(subtc, _mm_setzero_si128()), tcmax); - tci = _mm_shufflelo_epi16(tci, _MM_SHUFFLE(3, 1, 3, 1)); + __m128i tci = _mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)); + tci =_mm_min_epi16(_mm_max_epi16(tci, _mm_setzero_si128()), tcmax); tci = _mm_madd_epi16(tci, tcoffset); outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]; x++; @@ -2587,16 +2802,16 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, { for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep)) { - __m128i tci = _mm_and_si128(subtc, tcmax); - tci = _mm_shufflehi_epi16(_mm_shufflelo_epi16(tci, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 1, 3, 1)); + __m128i tci = _mm_shufflehi_epi16(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 1, 3, 1)); + tci = _mm_and_si128(tci, tcmax); tci = _mm_madd_epi16(tci, tcoffset); outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]; - outi[x+1] = *(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))]; + outi[x+1] = *(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]; } if (x <= endsub) { - __m128i tci = _mm_and_si128(subtc, tcmax); - tci = _mm_shufflelo_epi16(tci, _MM_SHUFFLE(3, 1, 3, 1)); + __m128i tci = _mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)); + tci = _mm_and_si128(tci, tcmax); tci = _mm_madd_epi16(tci, tcoffset); outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]; x++; @@ -2610,7 +2825,7 @@ void DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(DPSOFTRAST_State_Thread *thread, void DPSOFTRAST_Draw_Span_TextureCubeVaryingBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char * RESTRICT out4ub, int texunitindex, int arrayindex, const float * RESTRICT zf) { // TODO: IMPLEMENT - memset(out4ub, 255, span->length*4); + memset(out4ub + span->startx*4, 255, (span->startx - span->endx)*4); } float DPSOFTRAST_SampleShadowmap(const float *vector) @@ -2750,86 +2965,114 @@ void DPSOFTRAST_Draw_Span_MixUniformColor(const DPSOFTRAST_State_Triangle * REST void DPSOFTRAST_Draw_Span_MultiplyVaryingBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *in4ub, int arrayindex, const float *zf) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x; int startx = span->startx; int endx = span->endx; __m128 data, slope; + __m128 mod, endmod; + __m128i submod, substep, endsubmod; DPSOFTRAST_CALCATTRIB(triangle, span, data, slope, arrayindex); data = _mm_shuffle_ps(data, data, _MM_SHUFFLE(3, 0, 1, 2)); slope = _mm_shuffle_ps(slope, slope, _MM_SHUFFLE(3, 0, 1, 2)); - data = _mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))); - data = _mm_mul_ps(data, _mm_set1_ps(256.0f)); - slope = _mm_mul_ps(slope, _mm_set1_ps(256.0f)); - for (x = startx;x+2 <= endx;x += 2, data = _mm_add_ps(data, slope)) - { - __m128i pix = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_loadl_epi64((const __m128i *)&in4ub[x*4])); - __m128i mod = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x]))), mod2; - data = _mm_add_ps(data, slope); - mod2 = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x+1]))); - mod = _mm_unpacklo_epi64(_mm_packs_epi32(mod, mod), _mm_packs_epi32(mod2, mod2)); - pix = _mm_mulhi_epu16(pix, mod); - _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix, pix)); - } - for (;x < endx;x++, data = _mm_add_ps(data, slope)) + endmod = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))), _mm_load1_ps(&zf[startx])); + endsubmod = _mm_cvtps_epi32(_mm_mul_ps(endmod, _mm_set1_ps(256.0f))); + for (x = startx; x < endx;) { - __m128i pix = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&in4ub[x*4])); - __m128i mod = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x]))); - mod = _mm_packs_epi32(mod, mod); - pix = _mm_mulhi_epu16(pix, mod); - *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); - } -#endif -} - + int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1; + __m128 subscale = _mm_set1_ps(256.0f/DPSOFTRAST_DRAW_MAXSUBSPAN); + if (nextsub >= endx) + { + nextsub = endsub = endx-1; + if (x < nextsub) subscale = _mm_set1_ps(256.0f / (nextsub - x)); + } + mod = endmod; + submod = endsubmod; + endmod = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(nextsub))), _mm_load1_ps(&zf[nextsub])); + substep = _mm_cvtps_epi32(_mm_mul_ps(_mm_sub_ps(endmod, mod), subscale)); + endsubmod = _mm_cvtps_epi32(_mm_mul_ps(endmod, _mm_set1_ps(256.0f))); + submod = _mm_packs_epi32(submod, _mm_add_epi32(submod, substep)); + substep = _mm_packs_epi32(substep, substep); + for (; x + 1 <= endsub; x += 2, submod = _mm_add_epi16(submod, substep)) + { + __m128i pix = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_loadl_epi64((const __m128i *)&in4ub[x*4])); + pix = _mm_mulhi_epu16(pix, submod); + _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix, pix)); + } + if (x <= endsub) + { + __m128i pix = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&in4ub[x*4])); + pix = _mm_mulhi_epu16(pix, submod); + *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); + x++; + } + } +#endif +} + void DPSOFTRAST_Draw_Span_VaryingBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, int arrayindex, const float *zf) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x; int startx = span->startx; int endx = span->endx; __m128 data, slope; + __m128 mod, endmod; + __m128i submod, substep, endsubmod; DPSOFTRAST_CALCATTRIB(triangle, span, data, slope, arrayindex); data = _mm_shuffle_ps(data, data, _MM_SHUFFLE(3, 0, 1, 2)); slope = _mm_shuffle_ps(slope, slope, _MM_SHUFFLE(3, 0, 1, 2)); - data = _mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))); - data = _mm_mul_ps(data, _mm_set1_ps(255.0f)); - slope = _mm_mul_ps(slope, _mm_set1_ps(255.0f)); - for (x = startx;x+2 <= endx;x += 2, data = _mm_add_ps(data, slope)) - { - __m128i pix = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x]))), pix2; - data = _mm_add_ps(data, slope); - pix2 = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x+1]))); - pix = _mm_unpacklo_epi64(_mm_packs_epi32(pix, pix), _mm_packs_epi32(pix2, pix2)); - _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix, pix)); - } - for (;x < endx;x++, data = _mm_add_ps(data, slope)) + endmod = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(startx))), _mm_load1_ps(&zf[startx])); + endsubmod = _mm_cvtps_epi32(_mm_mul_ps(endmod, _mm_set1_ps(4095.0f))); + for (x = startx; x < endx;) { - __m128i pix = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&zf[x]))); - pix = _mm_packs_epi32(pix, pix); - *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); + int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1; + __m128 subscale = _mm_set1_ps(4095.0f/DPSOFTRAST_DRAW_MAXSUBSPAN); + if (nextsub >= endx) + { + nextsub = endsub = endx-1; + if (x < nextsub) subscale = _mm_set1_ps(4095.0f / (nextsub - x)); + } + mod = endmod; + submod = endsubmod; + endmod = _mm_mul_ps(_mm_add_ps(data, _mm_mul_ps(slope, _mm_set1_ps(nextsub))), _mm_load1_ps(&zf[nextsub])); + substep = _mm_cvtps_epi32(_mm_mul_ps(_mm_sub_ps(endmod, mod), subscale)); + endsubmod = _mm_cvtps_epi32(_mm_mul_ps(endmod, _mm_set1_ps(4095.0f))); + submod = _mm_packs_epi32(submod, _mm_add_epi32(submod, substep)); + substep = _mm_packs_epi32(substep, substep); + for (; x + 1 <= endsub; x += 2, submod = _mm_add_epi16(submod, substep)) + { + __m128i pix = _mm_srai_epi16(submod, 4); + _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix, pix)); + } + if (x <= endsub) + { + __m128i pix = _mm_srai_epi16(submod, 4); + *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); + x++; + } } #endif } void DPSOFTRAST_Draw_Span_AddBloomBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *ina4ub, const unsigned char *inb4ub, const float *subcolor) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; __m128i localcolor = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_loadu_ps(subcolor), _mm_set1_ps(255.0f))), _MM_SHUFFLE(3, 0, 1, 2)); - localcolor = _mm_shuffle_epi32(_mm_packs_epi32(localcolor, localcolor), _MM_SHUFFLE(1, 0, 1, 0)); + localcolor = _mm_packs_epi32(localcolor, localcolor); for (x = startx;x+2 <= endx;x+=2) { __m128i pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&inb4ub[x*4]), _mm_setzero_si128()); - pix1 = _mm_add_epi16(pix1, _mm_sub_epi16(pix2, localcolor)); + pix1 = _mm_add_epi16(pix1, _mm_subs_epu16(pix2, localcolor)); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1)); } - if(x < endx) + if (x < endx) { __m128i pix1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&inb4ub[x*4]), _mm_setzero_si128()); - pix1 = _mm_add_epi16(pix1, _mm_sub_epi16(pix2, localcolor)); + pix1 = _mm_add_epi16(pix1, _mm_subs_epu16(pix2, localcolor)); *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1)); } #endif @@ -2837,7 +3080,7 @@ void DPSOFTRAST_Draw_Span_AddBloomBGRA8(const DPSOFTRAST_State_Triangle * RESTRI void DPSOFTRAST_Draw_Span_MultiplyBuffersBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *ina4ub, const unsigned char *inb4ub) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; for (x = startx;x+2 <= endx;x+=2) { @@ -2846,7 +3089,7 @@ void DPSOFTRAST_Draw_Span_MultiplyBuffersBGRA8(const DPSOFTRAST_State_Triangle * pix1 = _mm_mulhi_epu16(pix1, pix2); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1)); } - if(x < endx) + if (x < endx) { __m128i pix1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&inb4ub[x*4])); @@ -2858,7 +3101,7 @@ void DPSOFTRAST_Draw_Span_MultiplyBuffersBGRA8(const DPSOFTRAST_State_Triangle * void DPSOFTRAST_Draw_Span_AddBuffersBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *ina4ub, const unsigned char *inb4ub) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; for (x = startx;x+2 <= endx;x+=2) { @@ -2867,7 +3110,7 @@ void DPSOFTRAST_Draw_Span_AddBuffersBGRA8(const DPSOFTRAST_State_Triangle * REST pix1 = _mm_add_epi16(pix1, pix2); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1)); } - if(x < endx) + if (x < endx) { __m128i pix1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&inb4ub[x*4]), _mm_setzero_si128()); @@ -2879,10 +3122,10 @@ void DPSOFTRAST_Draw_Span_AddBuffersBGRA8(const DPSOFTRAST_State_Triangle * REST void DPSOFTRAST_Draw_Span_TintedAddBuffersBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *ina4ub, const unsigned char *inb4ub, const float *inbtintbgra) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; __m128i tint = _mm_cvtps_epi32(_mm_mul_ps(_mm_loadu_ps(inbtintbgra), _mm_set1_ps(256.0f))); - tint = _mm_shuffle_epi32(_mm_packs_epi32(tint, tint), _MM_SHUFFLE(1, 0, 1, 0)); + tint = _mm_packs_epi32(tint, tint); for (x = startx;x+2 <= endx;x+=2) { __m128i pix1 = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&ina4ub[x*4]), _mm_setzero_si128()); @@ -2890,7 +3133,7 @@ void DPSOFTRAST_Draw_Span_TintedAddBuffersBGRA8(const DPSOFTRAST_State_Triangle pix1 = _mm_add_epi16(pix1, _mm_mulhi_epu16(tint, pix2)); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1)); } - if(x < endx) + if (x < endx) { __m128i pix1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&inb4ub[x*4])); @@ -2902,7 +3145,7 @@ void DPSOFTRAST_Draw_Span_TintedAddBuffersBGRA8(const DPSOFTRAST_State_Triangle void DPSOFTRAST_Draw_Span_MixBuffersBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *ina4ub, const unsigned char *inb4ub) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; for (x = startx;x+2 <= endx;x+=2) { @@ -2912,7 +3155,7 @@ void DPSOFTRAST_Draw_Span_MixBuffersBGRA8(const DPSOFTRAST_State_Triangle * REST pix1 = _mm_add_epi16(pix1, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 4), _mm_slli_epi16(blend, 4))); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1)); } - if(x < endx) + if (x < endx) { __m128i pix1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&ina4ub[x*4]), _mm_setzero_si128()); __m128i pix2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&inb4ub[x*4]), _mm_setzero_si128()); @@ -2925,10 +3168,10 @@ void DPSOFTRAST_Draw_Span_MixBuffersBGRA8(const DPSOFTRAST_State_Triangle * REST void DPSOFTRAST_Draw_Span_MixUniformColorBGRA8(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, unsigned char *out4ub, const unsigned char *in4ub, const float *color) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE int x, startx = span->startx, endx = span->endx; __m128i localcolor = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_loadu_ps(color), _mm_set1_ps(255.0f))), _MM_SHUFFLE(3, 0, 1, 2)), blend; - localcolor = _mm_shuffle_epi32(_mm_packs_epi32(localcolor, localcolor), _MM_SHUFFLE(1, 0, 1, 0)); + localcolor = _mm_packs_epi32(localcolor, localcolor); blend = _mm_slli_epi16(_mm_shufflehi_epi16(_mm_shufflelo_epi16(localcolor, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3)), 4); for (x = startx;x+2 <= endx;x+=2) { @@ -2936,7 +3179,7 @@ void DPSOFTRAST_Draw_Span_MixUniformColorBGRA8(const DPSOFTRAST_State_Triangle * pix = _mm_add_epi16(pix, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(localcolor, pix), 4), blend)); _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix, pix)); } - if(x < endx) + if (x < endx) { __m128i pix = _mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const int *)&in4ub[x*4]), _mm_setzero_si128()); pix = _mm_add_epi16(pix, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(localcolor, pix), 4), blend)); @@ -2949,11 +3192,11 @@ void DPSOFTRAST_Draw_Span_MixUniformColorBGRA8(const DPSOFTRAST_State_Triangle * void DPSOFTRAST_VertexShader_Generic(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.numvertices); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_COLOR, DPSOFTRAST_ARRAY_COLOR); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0); if (dpsoftrast.shader_permutation & SHADERPERMUTATION_SPECULAR) - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.numvertices); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD1); } void DPSOFTRAST_PixelShader_Generic(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -2996,9 +3239,9 @@ void DPSOFTRAST_PixelShader_Generic(DPSOFTRAST_State_Thread *thread, const DPSOF void DPSOFTRAST_VertexShader_PostProcess(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.numvertices); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD4); } void DPSOFTRAST_PixelShader_PostProcess(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -3030,7 +3273,7 @@ void DPSOFTRAST_PixelShader_PostProcess(DPSOFTRAST_State_Thread *thread, const D void DPSOFTRAST_VertexShader_Depth_Or_Shadow(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } void DPSOFTRAST_PixelShader_Depth_Or_Shadow(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -3039,7 +3282,7 @@ void DPSOFTRAST_PixelShader_Depth_Or_Shadow(DPSOFTRAST_State_Thread *thread, con float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + memset(buffer_FragColorbgra8 + span->startx*4, 0, (span->endx - span->startx)*4); DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -3047,45 +3290,64 @@ void DPSOFTRAST_PixelShader_Depth_Or_Shadow(DPSOFTRAST_State_Thread *thread, con void DPSOFTRAST_VertexShader_FlatColor(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); } void DPSOFTRAST_PixelShader_FlatColor(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { +#ifdef SSE_POSSIBLE + unsigned char * RESTRICT pixelmask = span->pixelmask; + unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0] + (span->y * dpsoftrast.fb_width + span->x) * 4; int x, startx = span->startx, endx = span->endx; - int Color_Ambienti[4]; + __m128i Color_Ambientm; float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; - Color_Ambienti[2] = (int)(thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0]*256.0f); - Color_Ambienti[1] = (int)(thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1]*256.0f); - Color_Ambienti[0] = (int)(thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2]*256.0f); - Color_Ambienti[3] = (int)(thread->uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0] *256.0f); DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, 2, buffer_z); + if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE) + pixel = buffer_FragColorbgra8; + Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2)); + Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0)); + Color_Ambientm = _mm_or_si128(Color_Ambientm, _mm_setr_epi32(0, 0, 0, (int)(thread->uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0]*255.0f))); + Color_Ambientm = _mm_packs_epi32(Color_Ambientm, Color_Ambientm); for (x = startx;x < endx;x++) { - buffer_FragColorbgra8[x*4+0] = (buffer_texture_colorbgra8[x*4+0] * Color_Ambienti[0])>>8; - buffer_FragColorbgra8[x*4+1] = (buffer_texture_colorbgra8[x*4+1] * Color_Ambienti[1])>>8; - buffer_FragColorbgra8[x*4+2] = (buffer_texture_colorbgra8[x*4+2] * Color_Ambienti[2])>>8; - buffer_FragColorbgra8[x*4+3] = (buffer_texture_colorbgra8[x*4+3] * Color_Ambienti[3])>>8; + __m128i color, pix; + if (x + 4 <= endx && *(const unsigned int *)&pixelmask[x] == 0x01010101) + { + __m128i pix2; + color = _mm_loadu_si128((const __m128i *)&buffer_texture_colorbgra8[x*4]); + pix = _mm_mulhi_epu16(Color_Ambientm, _mm_unpacklo_epi8(_mm_setzero_si128(), color)); + pix2 = _mm_mulhi_epu16(Color_Ambientm, _mm_unpackhi_epi8(_mm_setzero_si128(), color)); + _mm_storeu_si128((__m128i *)&pixel[x*4], _mm_packus_epi16(pix, pix2)); + x += 3; + continue; + } + if (!pixelmask[x]) + continue; + color = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_colorbgra8[x*4])); + pix = _mm_mulhi_epu16(Color_Ambientm, color); + *(int *)&pixel[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); } - DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); + if (pixel == buffer_FragColorbgra8) + DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); +#endif } void DPSOFTRAST_VertexShader_VertexColor(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.numvertices); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_COLOR, DPSOFTRAST_ARRAY_COLOR); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); } void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE unsigned char * RESTRICT pixelmask = span->pixelmask; unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0] + (span->y * dpsoftrast.fb_width + span->x) * 4; int x, startx = span->startx, endx = span->endx; @@ -3097,7 +3359,7 @@ void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const D int arrayindex = DPSOFTRAST_ARRAY_COLOR; DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, 2, buffer_z); - if (thread->alphatest || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE) + if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE) pixel = buffer_FragColorbgra8; Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2)); Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0)); @@ -3135,7 +3397,7 @@ void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const D x += 3; continue; } - if(!pixelmask[x]) + if (!pixelmask[x]) continue; color = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_colorbgra8[x*4])); mod = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&buffer_z[x]))); @@ -3143,7 +3405,7 @@ void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const D pix = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(mod, Color_Diffusem), Color_Ambientm), color); *(int *)&pixel[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); } - if(pixel == buffer_FragColorbgra8) + if (pixel == buffer_FragColorbgra8) DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); #endif } @@ -3152,14 +3414,14 @@ void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const D void DPSOFTRAST_VertexShader_Lightmap(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.numvertices); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD4); } void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { -#ifdef SSE2_PRESENT +#ifdef SSE_POSSIBLE unsigned char * RESTRICT pixelmask = span->pixelmask; unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0] + (span->y * dpsoftrast.fb_width + span->x) * 4; int x, startx = span->startx, endx = span->endx; @@ -3172,7 +3434,7 @@ void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSO DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); - if (thread->alphatest || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE) + if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE) pixel = buffer_FragColorbgra8; Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2)); Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0)); @@ -3207,7 +3469,7 @@ void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSO x += 3; continue; } - if(!pixelmask[x]) + if (!pixelmask[x]) continue; color = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_colorbgra8[x*4])); lightmap = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_lightmapbgra8[x*4])); @@ -3235,7 +3497,7 @@ void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSO x += 3; continue; } - if(!pixelmask[x]) + if (!pixelmask[x]) continue; color = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_colorbgra8[x*4])); lightmap = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_lightmapbgra8[x*4])); @@ -3243,52 +3505,49 @@ void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSO *(int *)&pixel[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix)); } } - if(pixel == buffer_FragColorbgra8) + if (pixel == buffer_FragColorbgra8) DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); #endif } +void DPSOFTRAST_VertexShader_LightDirection(void); +void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span); void DPSOFTRAST_VertexShader_FakeLight(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_VertexShader_LightDirection(); } void DPSOFTRAST_PixelShader_FakeLight(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { - // TODO: IMPLEMENT - float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; - unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; - DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); - DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); + DPSOFTRAST_PixelShader_LightDirection(thread, triangle, span); } void DPSOFTRAST_VertexShader_LightDirectionMap_ModelSpace(void) { - DPSOFTRAST_VertexShader_Lightmap(); + DPSOFTRAST_VertexShader_LightDirection(); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD4); } void DPSOFTRAST_PixelShader_LightDirectionMap_ModelSpace(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { - DPSOFTRAST_PixelShader_Lightmap(thread, triangle, span); - // TODO: IMPLEMENT + DPSOFTRAST_PixelShader_LightDirection(thread, triangle, span); } void DPSOFTRAST_VertexShader_LightDirectionMap_TangentSpace(void) { - DPSOFTRAST_VertexShader_Lightmap(); + DPSOFTRAST_VertexShader_LightDirection(); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD4); } void DPSOFTRAST_PixelShader_LightDirectionMap_TangentSpace(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { - DPSOFTRAST_PixelShader_Lightmap(thread, triangle, span); - // TODO: IMPLEMENT + DPSOFTRAST_PixelShader_LightDirection(thread, triangle, span); } @@ -3314,40 +3573,44 @@ void DPSOFTRAST_VertexShader_LightDirection(void) EyePosition[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+1]; EyePosition[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+2]; EyePosition[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+3]; - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD3); for (i = 0;i < numvertices;i++) { - position[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0]; - position[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1]; - position[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+2]; - svector[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0]; - svector[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1]; - svector[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2]; - tvector[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0]; - tvector[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1]; - tvector[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2]; - normal[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+0]; - normal[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+1]; - normal[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+2]; + position[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0]; + position[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1]; + position[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+2]; + svector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0]; + svector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1]; + svector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2]; + tvector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0]; + tvector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1]; + tvector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2]; + normal[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+0]; + normal[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+1]; + normal[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+2]; LightVector[0] = svector[0] * LightDir[0] + svector[1] * LightDir[1] + svector[2] * LightDir[2]; LightVector[1] = tvector[0] * LightDir[0] + tvector[1] * LightDir[1] + tvector[2] * LightDir[2]; LightVector[2] = normal[0] * LightDir[0] + normal[1] * LightDir[1] + normal[2] * LightDir[2]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0] = LightVector[0]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1] = LightVector[1]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2] = LightVector[2]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+3] = 0.0f; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD5][i*4+0] = LightVector[0]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD5][i*4+1] = LightVector[1]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD5][i*4+2] = LightVector[2]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD5][i*4+3] = 0.0f; EyeVectorModelSpace[0] = EyePosition[0] - position[0]; EyeVectorModelSpace[1] = EyePosition[1] - position[1]; EyeVectorModelSpace[2] = EyePosition[2] - position[2]; EyeVector[0] = svector[0] * EyeVectorModelSpace[0] + svector[1] * EyeVectorModelSpace[1] + svector[2] * EyeVectorModelSpace[2]; EyeVector[1] = tvector[0] * EyeVectorModelSpace[0] + tvector[1] * EyeVectorModelSpace[1] + tvector[2] * EyeVectorModelSpace[2]; EyeVector[2] = normal[0] * EyeVectorModelSpace[0] + normal[1] * EyeVectorModelSpace[1] + normal[2] * EyeVectorModelSpace[2]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0] = EyeVector[0]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1] = EyeVector[1]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2] = EyeVector[2]; - dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+3] = 0.0f; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+0] = EyeVector[0]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+1] = EyeVector[1]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+2] = EyeVector[2]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+3] = 0.0f; } + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, -1, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } #define DPSOFTRAST_Min(a,b) ((a) < (b) ? (a) : (b)) @@ -3378,6 +3641,8 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons unsigned char buffer_texture_glowbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_texture_pantsbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_texture_shirtbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; + unsigned char buffer_texture_deluxemapbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; + unsigned char buffer_texture_lightmapbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; int x, startx = span->startx, endx = span->endx; float Color_Ambient[4], Color_Diffuse[4], Color_Specular[4], Color_Glow[4], Color_Pants[4], Color_Shirt[4], LightColor[4]; @@ -3385,11 +3650,18 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons float LightVectorslope[4]; float EyeVectordata[4]; float EyeVectorslope[4]; + float VectorSdata[4]; + float VectorSslope[4]; + float VectorTdata[4]; + float VectorTslope[4]; + float VectorRdata[4]; + float VectorRslope[4]; float z; float diffusetex[4]; float glosstex[4]; float surfacenormal[4]; float lightnormal[4]; + float lightnormal_modelspace[4]; float eyenormal[4]; float specularnormal[4]; float diffuse; @@ -3433,15 +3705,37 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons LightColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+1]; LightColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+2]; LightColor[3] = 0.0f; - DPSOFTRAST_CALCATTRIB4F(triangle, span, LightVectordata, LightVectorslope, DPSOFTRAST_ARRAY_TEXCOORD1); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); Color_Specular[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Specular*4+0]; Color_Specular[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Specular*4+1]; Color_Specular[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Specular*4+2]; Color_Specular[3] = 0.0f; SpecularPower = thread->uniform4f[DPSOFTRAST_UNIFORM_SpecularPower*4+0] * (1.0f / 255.0f); - DPSOFTRAST_CALCATTRIB4F(triangle, span, EyeVectordata, EyeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_CALCATTRIB4F(triangle, span, EyeVectordata, EyeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD6); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_glossbgra8, GL20TU_GLOSS, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); + + if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE) + { + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorSdata, VectorSslope, DPSOFTRAST_ARRAY_TEXCOORD1); + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorTdata, VectorTslope, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorRdata, VectorRslope, DPSOFTRAST_ARRAY_TEXCOORD3); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_deluxemapbgra8, GL20TU_DELUXEMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + } + else if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE) + { + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_deluxemapbgra8, GL20TU_DELUXEMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + } + else if(thread->shader_mode == SHADERMODE_FAKELIGHT) + { + // nothing of this needed + } + else + { + DPSOFTRAST_CALCATTRIB4F(triangle, span, LightVectordata, LightVectorslope, DPSOFTRAST_ARRAY_TEXCOORD5); + } + for (x = startx;x < endx;x++) { z = buffer_z[x]; @@ -3465,23 +3759,105 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons surfacenormal[2] = buffer_texture_normalbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; DPSOFTRAST_Vector3Normalize(surfacenormal); - lightnormal[0] = (LightVectordata[0] + LightVectorslope[0]*x) * z; - lightnormal[1] = (LightVectordata[1] + LightVectorslope[1]*x) * z; - lightnormal[2] = (LightVectordata[2] + LightVectorslope[2]*x) * z; - DPSOFTRAST_Vector3Normalize(lightnormal); + if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE) + { + // myhalf3 lightnormal_modelspace = myhalf3(dp_texture2D(Texture_Deluxemap, TexCoordSurfaceLightmap.zw)) * 2.0 + myhalf3(-1.0, -1.0, -1.0);\n"; + lightnormal_modelspace[0] = buffer_texture_deluxemapbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + lightnormal_modelspace[1] = buffer_texture_deluxemapbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + lightnormal_modelspace[2] = buffer_texture_deluxemapbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + + // lightnormal.x = dot(lightnormal_modelspace, myhalf3(VectorS));\n" + lightnormal[0] = lightnormal_modelspace[0] * (VectorSdata[0] + VectorSslope[0] * x) + + lightnormal_modelspace[1] * (VectorSdata[1] + VectorSslope[1] * x) + + lightnormal_modelspace[2] * (VectorSdata[2] + VectorSslope[2] * x); + + // lightnormal.y = dot(lightnormal_modelspace, myhalf3(VectorT));\n" + lightnormal[1] = lightnormal_modelspace[0] * (VectorTdata[0] + VectorTslope[0] * x) + + lightnormal_modelspace[1] * (VectorTdata[1] + VectorTslope[1] * x) + + lightnormal_modelspace[2] * (VectorTdata[2] + VectorTslope[2] * x); + + // lightnormal.z = dot(lightnormal_modelspace, myhalf3(VectorR));\n" + lightnormal[2] = lightnormal_modelspace[0] * (VectorRdata[0] + VectorRslope[0] * x) + + lightnormal_modelspace[1] * (VectorRdata[1] + VectorRslope[1] * x) + + lightnormal_modelspace[2] * (VectorRdata[2] + VectorRslope[2] * x); + + // lightnormal = normalize(lightnormal); // VectorS/T/R are not always perfectly normalized, and EXACTSPECULARMATH is very picky about this\n" + DPSOFTRAST_Vector3Normalize(lightnormal); + + // myhalf3 lightcolor = myhalf3(dp_texture2D(Texture_Lightmap, TexCoordSurfaceLightmap.zw));\n"; + { + float f = 1.0f / (256.0f * max(0.25f, lightnormal[2])); + LightColor[0] = buffer_texture_lightmapbgra8[x*4+0] * f; + LightColor[1] = buffer_texture_lightmapbgra8[x*4+1] * f; + LightColor[2] = buffer_texture_lightmapbgra8[x*4+2] * f; + } + } + else if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE) + { + lightnormal[0] = buffer_texture_deluxemapbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + lightnormal[1] = buffer_texture_deluxemapbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + lightnormal[2] = buffer_texture_deluxemapbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + { + float f = 1.0f / 256.0f; + LightColor[0] = buffer_texture_lightmapbgra8[x*4+0] * f; + LightColor[1] = buffer_texture_lightmapbgra8[x*4+1] * f; + LightColor[2] = buffer_texture_lightmapbgra8[x*4+2] * f; + } + } + else if(thread->shader_mode == SHADERMODE_FAKELIGHT) + { + lightnormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + lightnormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + lightnormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(lightnormal); + + LightColor[0] = 1.0; + LightColor[1] = 1.0; + LightColor[2] = 1.0; + } + else + { + lightnormal[0] = (LightVectordata[0] + LightVectorslope[0]*x) * z; + lightnormal[1] = (LightVectordata[1] + LightVectorslope[1]*x) * z; + lightnormal[2] = (LightVectordata[2] + LightVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(lightnormal); + } + + diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f; - eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; - eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; - eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; - DPSOFTRAST_Vector3Normalize(eyenormal); + if(thread->shader_exactspecularmath) + { + // reflect lightnormal at surfacenormal, take the negative of that + // i.e. we want (2*dot(N, i) * N - I) for N=surfacenormal, I=lightnormal + float f; + f = DPSOFTRAST_Vector3Dot(lightnormal, surfacenormal); + specularnormal[0] = 2*f*surfacenormal[0] - lightnormal[0]; + specularnormal[1] = 2*f*surfacenormal[1] - lightnormal[1]; + specularnormal[2] = 2*f*surfacenormal[2] - lightnormal[2]; + + // dot of this and normalize(EyeVectorFogDepth.xyz) + eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(eyenormal); + + specular = DPSOFTRAST_Vector3Dot(eyenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; + } + else + { + eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(eyenormal); - specularnormal[0] = lightnormal[0] + eyenormal[0]; - specularnormal[1] = lightnormal[1] + eyenormal[1]; - specularnormal[2] = lightnormal[2] + eyenormal[2]; - DPSOFTRAST_Vector3Normalize(specularnormal); + specularnormal[0] = lightnormal[0] + eyenormal[0]; + specularnormal[1] = lightnormal[1] + eyenormal[1]; + specularnormal[2] = lightnormal[2] + eyenormal[2]; + DPSOFTRAST_Vector3Normalize(specularnormal); + + specular = DPSOFTRAST_Vector3Dot(surfacenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; + } - diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f; - specular = DPSOFTRAST_Vector3Dot(surfacenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; specular = pow(specular, SpecularPower * glosstex[3]); if (thread->shader_permutation & SHADERPERMUTATION_GLOW) { @@ -3497,6 +3873,7 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons d[2] = (int)( diffusetex[2] * Color_Ambient[2] + (diffusetex[2] * Color_Diffuse[2] * diffuse + glosstex[2] * Color_Specular[2] * specular) * LightColor[2]);if (d[2] > 255) d[2] = 255; d[3] = (int)( diffusetex[3] * Color_Ambient[3]);if (d[3] > 255) d[3] = 255; } + buffer_FragColorbgra8[x*4+0] = d[0]; buffer_FragColorbgra8[x*4+1] = d[1]; buffer_FragColorbgra8[x*4+2] = d[2]; @@ -3513,8 +3890,30 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons LightColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+1]; LightColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+2]; LightColor[3] = 0.0f; - DPSOFTRAST_CALCATTRIB4F(triangle, span, LightVectordata, LightVectorslope, DPSOFTRAST_ARRAY_TEXCOORD1); DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); + + if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE) + { + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorSdata, VectorSslope, DPSOFTRAST_ARRAY_TEXCOORD1); + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorTdata, VectorTslope, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_CALCATTRIB4F(triangle, span, VectorRdata, VectorRslope, DPSOFTRAST_ARRAY_TEXCOORD3); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_deluxemapbgra8, GL20TU_DELUXEMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + } + else if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE) + { + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_deluxemapbgra8, GL20TU_DELUXEMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z); + } + else if(thread->shader_mode == SHADERMODE_FAKELIGHT) + { + DPSOFTRAST_CALCATTRIB4F(triangle, span, EyeVectordata, EyeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD6); + } + else + { + DPSOFTRAST_CALCATTRIB4F(triangle, span, LightVectordata, LightVectorslope, DPSOFTRAST_ARRAY_TEXCOORD5); + } + for (x = startx;x < endx;x++) { z = buffer_z[x]; @@ -3527,10 +3926,69 @@ void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, cons surfacenormal[2] = buffer_texture_normalbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; DPSOFTRAST_Vector3Normalize(surfacenormal); - lightnormal[0] = (LightVectordata[0] + LightVectorslope[0]*x) * z; - lightnormal[1] = (LightVectordata[1] + LightVectorslope[1]*x) * z; - lightnormal[2] = (LightVectordata[2] + LightVectorslope[2]*x) * z; - DPSOFTRAST_Vector3Normalize(lightnormal); + if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE) + { + // myhalf3 lightnormal_modelspace = myhalf3(dp_texture2D(Texture_Deluxemap, TexCoordSurfaceLightmap.zw)) * 2.0 + myhalf3(-1.0, -1.0, -1.0);\n"; + lightnormal_modelspace[0] = buffer_texture_deluxemapbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + lightnormal_modelspace[1] = buffer_texture_deluxemapbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + lightnormal_modelspace[2] = buffer_texture_deluxemapbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + + // lightnormal.x = dot(lightnormal_modelspace, myhalf3(VectorS));\n" + lightnormal[0] = lightnormal_modelspace[0] * (VectorSdata[0] + VectorSslope[0] * x) + + lightnormal_modelspace[1] * (VectorSdata[1] + VectorSslope[1] * x) + + lightnormal_modelspace[2] * (VectorSdata[2] + VectorSslope[2] * x); + + // lightnormal.y = dot(lightnormal_modelspace, myhalf3(VectorT));\n" + lightnormal[1] = lightnormal_modelspace[0] * (VectorTdata[0] + VectorTslope[0] * x) + + lightnormal_modelspace[1] * (VectorTdata[1] + VectorTslope[1] * x) + + lightnormal_modelspace[2] * (VectorTdata[2] + VectorTslope[2] * x); + + // lightnormal.z = dot(lightnormal_modelspace, myhalf3(VectorR));\n" + lightnormal[2] = lightnormal_modelspace[0] * (VectorRdata[0] + VectorRslope[0] * x) + + lightnormal_modelspace[1] * (VectorRdata[1] + VectorRslope[1] * x) + + lightnormal_modelspace[2] * (VectorRdata[2] + VectorRslope[2] * x); + + // lightnormal = normalize(lightnormal); // VectorS/T/R are not always perfectly normalized, and EXACTSPECULARMATH is very picky about this\n" + DPSOFTRAST_Vector3Normalize(lightnormal); + + // myhalf3 lightcolor = myhalf3(dp_texture2D(Texture_Lightmap, TexCoordSurfaceLightmap.zw));\n"; + { + float f = 1.0f / (256.0f * max(0.25f, lightnormal[2])); + LightColor[0] = buffer_texture_lightmapbgra8[x*4+0] * f; + LightColor[1] = buffer_texture_lightmapbgra8[x*4+1] * f; + LightColor[2] = buffer_texture_lightmapbgra8[x*4+2] * f; + } + } + else if(thread->shader_mode == SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE) + { + lightnormal[0] = buffer_texture_deluxemapbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + lightnormal[1] = buffer_texture_deluxemapbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + lightnormal[2] = buffer_texture_deluxemapbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + { + float f = 1.0f / 256.0f; + LightColor[0] = buffer_texture_lightmapbgra8[x*4+0] * f; + LightColor[1] = buffer_texture_lightmapbgra8[x*4+1] * f; + LightColor[2] = buffer_texture_lightmapbgra8[x*4+2] * f; + } + } + else if(thread->shader_mode == SHADERMODE_FAKELIGHT) + { + lightnormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + lightnormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + lightnormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(lightnormal); + + LightColor[0] = 1.0; + LightColor[1] = 1.0; + LightColor[2] = 1.0; + } + else + { + lightnormal[0] = (LightVectordata[0] + LightVectorslope[0]*x) * z; + lightnormal[1] = (LightVectordata[1] + LightVectorslope[1]*x) * z; + lightnormal[2] = (LightVectordata[2] + LightVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(lightnormal); + } diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f; if (thread->shader_permutation & SHADERPERMUTATION_GLOW) @@ -3610,24 +4068,26 @@ void DPSOFTRAST_VertexShader_LightSource(void) EyePosition[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+1]; EyePosition[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+2]; EyePosition[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+3]; - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); - DPSOFTRAST_Array_Transform(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelToLightM1); - DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.numvertices); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD3); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD4); for (i = 0;i < numvertices;i++) { - position[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0]; - position[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1]; - position[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+2]; - svector[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0]; - svector[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1]; - svector[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2]; - tvector[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0]; - tvector[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1]; - tvector[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2]; - normal[0] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+0]; - normal[1] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+1]; - normal[2] = dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+2]; + position[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0]; + position[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1]; + position[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+2]; + svector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0]; + svector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1]; + svector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2]; + tvector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0]; + tvector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1]; + tvector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2]; + normal[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+0]; + normal[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+1]; + normal[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+2]; LightVectorModelSpace[0] = LightPosition[0] - position[0]; LightVectorModelSpace[1] = LightPosition[1] - position[1]; LightVectorModelSpace[2] = LightPosition[2] - position[2]; @@ -3649,10 +4109,13 @@ void DPSOFTRAST_VertexShader_LightSource(void) dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2] = EyeVector[2]; dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+3] = 0.0f; } + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, -1, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelToLightM1); } void DPSOFTRAST_PixelShader_LightSource(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { +#ifdef SSE_POSSIBLE float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_texture_normalbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; @@ -3769,19 +4232,42 @@ void DPSOFTRAST_PixelShader_LightSource(DPSOFTRAST_State_Thread *thread, const D lightnormal[2] = (LightVectordata[2] + LightVectorslope[2]*x) * z; DPSOFTRAST_Vector3Normalize(lightnormal); - eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; - eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; - eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; - DPSOFTRAST_Vector3Normalize(eyenormal); + diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f; + + if(thread->shader_exactspecularmath) + { + // reflect lightnormal at surfacenormal, take the negative of that + // i.e. we want (2*dot(N, i) * N - I) for N=surfacenormal, I=lightnormal + float f; + f = DPSOFTRAST_Vector3Dot(lightnormal, surfacenormal); + specularnormal[0] = 2*f*surfacenormal[0] - lightnormal[0]; + specularnormal[1] = 2*f*surfacenormal[1] - lightnormal[1]; + specularnormal[2] = 2*f*surfacenormal[2] - lightnormal[2]; + + // dot of this and normalize(EyeVectorFogDepth.xyz) + eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(eyenormal); + + specular = DPSOFTRAST_Vector3Dot(eyenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; + } + else + { + eyenormal[0] = (EyeVectordata[0] + EyeVectorslope[0]*x) * z; + eyenormal[1] = (EyeVectordata[1] + EyeVectorslope[1]*x) * z; + eyenormal[2] = (EyeVectordata[2] + EyeVectorslope[2]*x) * z; + DPSOFTRAST_Vector3Normalize(eyenormal); - specularnormal[0] = lightnormal[0] + eyenormal[0]; - specularnormal[1] = lightnormal[1] + eyenormal[1]; - specularnormal[2] = lightnormal[2] + eyenormal[2]; - DPSOFTRAST_Vector3Normalize(specularnormal); + specularnormal[0] = lightnormal[0] + eyenormal[0]; + specularnormal[1] = lightnormal[1] + eyenormal[1]; + specularnormal[2] = lightnormal[2] + eyenormal[2]; + DPSOFTRAST_Vector3Normalize(specularnormal); - diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f; - specular = DPSOFTRAST_Vector3Dot(surfacenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; + specular = DPSOFTRAST_Vector3Dot(surfacenormal, specularnormal);if (specular < 0.0f) specular = 0.0f; + } specular = pow(specular, SpecularPower * glosstex[3]); + if (thread->shader_permutation & SHADERPERMUTATION_CUBEFILTER) { // scale down the attenuation to account for the cubefilter multiplying everything by 255 @@ -3919,22 +4405,95 @@ void DPSOFTRAST_PixelShader_LightSource(DPSOFTRAST_State_Thread *thread, const D } } DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); +#endif } void DPSOFTRAST_VertexShader_Refraction(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } void DPSOFTRAST_PixelShader_Refraction(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { - // TODO: IMPLEMENT float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; + float z; + int x, startx = span->startx, endx = span->endx; + + // texture reads + unsigned char buffer_texture_normalbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; + + // varyings + float ModelViewProjectionPositiondata[4]; + float ModelViewProjectionPositionslope[4]; + + // uniforms + float ScreenScaleRefractReflect[2]; + float ScreenCenterRefractReflect[2]; + float DistortScaleRefractReflect[2]; + float RefractColor[4]; + + DPSOFTRAST_Texture *texture = thread->texbound[GL20TU_REFRACTION]; + if(!texture) return; + + // read textures DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); + + // read varyings + DPSOFTRAST_CALCATTRIB4F(triangle, span, ModelViewProjectionPositiondata, ModelViewProjectionPositionslope, DPSOFTRAST_ARRAY_TEXCOORD4); + + // read uniforms + ScreenScaleRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+0]; + ScreenScaleRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+1]; + ScreenCenterRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+0]; + ScreenCenterRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+1]; + DistortScaleRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+0]; + DistortScaleRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+1]; + RefractColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+2]; + RefractColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+1]; + RefractColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+0]; + RefractColor[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+3]; + + // do stuff + for (x = startx;x < endx;x++) + { + float SafeScreenTexCoord[2]; + float ScreenTexCoord[2]; + float v[3]; + float iw; + unsigned char c[4]; + + z = buffer_z[x]; + + // " vec2 ScreenScaleRefractReflectIW = ScreenScaleRefractReflect.xy * (1.0 / ModelViewProjectionPosition.w);\n" + iw = 1.0f / (ModelViewProjectionPositiondata[3] + ModelViewProjectionPositionslope[3]*x); // / z + + // " vec2 SafeScreenTexCoord = ModelViewProjectionPosition.xy * ScreenScaleRefractReflectIW + ScreenCenterRefractReflect.xy;\n" + SafeScreenTexCoord[0] = (ModelViewProjectionPositiondata[0] + ModelViewProjectionPositionslope[0]*x) * iw * ScreenScaleRefractReflect[0] + ScreenCenterRefractReflect[0]; // * z (disappears) + SafeScreenTexCoord[1] = (ModelViewProjectionPositiondata[1] + ModelViewProjectionPositionslope[1]*x) * iw * ScreenScaleRefractReflect[1] + ScreenCenterRefractReflect[1]; // * z (disappears) + + // " vec2 ScreenTexCoord = SafeScreenTexCoord + vec3(normalize(myhalf3(dp_texture2D(Texture_Normal, TexCoord)) - myhalf3(0.5))).xy * DistortScaleRefractReflect.zw;\n" + v[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + v[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + v[2] = buffer_texture_normalbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + DPSOFTRAST_Vector3Normalize(v); + ScreenTexCoord[0] = SafeScreenTexCoord[0] + v[0] * DistortScaleRefractReflect[0]; + ScreenTexCoord[1] = SafeScreenTexCoord[1] + v[1] * DistortScaleRefractReflect[1]; + + // " dp_FragColor = vec4(dp_texture2D(Texture_Refraction, ScreenTexCoord).rgb, 1.0) * RefractColor;\n" + DPSOFTRAST_Texture2DBGRA8(texture, 0, ScreenTexCoord[0], ScreenTexCoord[1], c); + + buffer_FragColorbgra8[x*4+0] = c[0] * RefractColor[0]; + buffer_FragColorbgra8[x*4+1] = c[1] * RefractColor[1]; + buffer_FragColorbgra8[x*4+2] = c[2] * RefractColor[2]; + buffer_FragColorbgra8[x*4+3] = min(RefractColor[3] * 256, 255); + } + DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -3942,17 +4501,167 @@ void DPSOFTRAST_PixelShader_Refraction(DPSOFTRAST_State_Thread *thread, const DP void DPSOFTRAST_VertexShader_Water(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + int i; + int numvertices = dpsoftrast.numvertices; + float EyePosition[4]; + float EyeVectorModelSpace[4]; + float EyeVector[4]; + float position[4]; + float svector[4]; + float tvector[4]; + float normal[4]; + EyePosition[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+0]; + EyePosition[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+1]; + EyePosition[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+2]; + EyePosition[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_EyePosition*4+3]; + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD1); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD2); + DPSOFTRAST_Array_Load(DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD3); + for (i = 0;i < numvertices;i++) + { + position[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0]; + position[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1]; + position[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+2]; + svector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+0]; + svector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+1]; + svector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1][i*4+2]; + tvector[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+0]; + tvector[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+1]; + tvector[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD2][i*4+2]; + normal[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+0]; + normal[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+1]; + normal[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD3][i*4+2]; + EyeVectorModelSpace[0] = EyePosition[0] - position[0]; + EyeVectorModelSpace[1] = EyePosition[1] - position[1]; + EyeVectorModelSpace[2] = EyePosition[2] - position[2]; + EyeVector[0] = svector[0] * EyeVectorModelSpace[0] + svector[1] * EyeVectorModelSpace[1] + svector[2] * EyeVectorModelSpace[2]; + EyeVector[1] = tvector[0] * EyeVectorModelSpace[0] + tvector[1] * EyeVectorModelSpace[1] + tvector[2] * EyeVectorModelSpace[2]; + EyeVector[2] = normal[0] * EyeVectorModelSpace[0] + normal[1] * EyeVectorModelSpace[1] + normal[2] * EyeVectorModelSpace[2]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+0] = EyeVector[0]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+1] = EyeVector[1]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+2] = EyeVector[2]; + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD6][i*4+3] = 0.0f; + } + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, -1, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_Transform(DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD0, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_TexMatrixM1); } void DPSOFTRAST_PixelShader_Water(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) { - // TODO: IMPLEMENT float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; + float z; + int x, startx = span->startx, endx = span->endx; + + // texture reads + unsigned char buffer_texture_normalbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; + + // varyings + float ModelViewProjectionPositiondata[4]; + float ModelViewProjectionPositionslope[4]; + float EyeVectordata[4]; + float EyeVectorslope[4]; + + // uniforms + float ScreenScaleRefractReflect[4]; + float ScreenCenterRefractReflect[4]; + float DistortScaleRefractReflect[4]; + float RefractColor[4]; + float ReflectColor[4]; + float ReflectFactor; + float ReflectOffset; + + DPSOFTRAST_Texture *texture_refraction = thread->texbound[GL20TU_REFRACTION]; + DPSOFTRAST_Texture *texture_reflection = thread->texbound[GL20TU_REFLECTION]; + if(!texture_refraction || !texture_reflection) return; + + // read textures DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z); + + // read varyings + DPSOFTRAST_CALCATTRIB4F(triangle, span, ModelViewProjectionPositiondata, ModelViewProjectionPositionslope, DPSOFTRAST_ARRAY_TEXCOORD4); + DPSOFTRAST_CALCATTRIB4F(triangle, span, EyeVectordata, EyeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD6); + + // read uniforms + ScreenScaleRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+0]; + ScreenScaleRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+1]; + ScreenScaleRefractReflect[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+2]; + ScreenScaleRefractReflect[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenScaleRefractReflect*4+3]; + ScreenCenterRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+0]; + ScreenCenterRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+1]; + ScreenCenterRefractReflect[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+2]; + ScreenCenterRefractReflect[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_ScreenCenterRefractReflect*4+3]; + DistortScaleRefractReflect[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+0]; + DistortScaleRefractReflect[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+1]; + DistortScaleRefractReflect[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+2]; + DistortScaleRefractReflect[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_DistortScaleRefractReflect*4+3]; + RefractColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+2]; + RefractColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+1]; + RefractColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+0]; + RefractColor[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_RefractColor*4+3]; + ReflectColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectColor*4+2]; + ReflectColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectColor*4+1]; + ReflectColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectColor*4+0]; + ReflectColor[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectColor*4+3]; + ReflectFactor = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectFactor*4+0]; + ReflectOffset = thread->uniform4f[DPSOFTRAST_UNIFORM_ReflectOffset*4+0]; + + // do stuff + for (x = startx;x < endx;x++) + { + float SafeScreenTexCoord[4]; + float ScreenTexCoord[4]; + float v[3]; + float iw; + unsigned char c1[4]; + unsigned char c2[4]; + float Fresnel; + + z = buffer_z[x]; + + // " vec4 ScreenScaleRefractReflectIW = ScreenScaleRefractReflect * (1.0 / ModelViewProjectionPosition.w);\n" + iw = 1.0f / (ModelViewProjectionPositiondata[3] + ModelViewProjectionPositionslope[3]*x); // / z + + // " vec4 SafeScreenTexCoord = ModelViewProjectionPosition.xyxy * ScreenScaleRefractReflectIW + ScreenCenterRefractReflect;\n" + SafeScreenTexCoord[0] = (ModelViewProjectionPositiondata[0] + ModelViewProjectionPositionslope[0]*x) * iw * ScreenScaleRefractReflect[0] + ScreenCenterRefractReflect[0]; // * z (disappears) + SafeScreenTexCoord[1] = (ModelViewProjectionPositiondata[1] + ModelViewProjectionPositionslope[1]*x) * iw * ScreenScaleRefractReflect[1] + ScreenCenterRefractReflect[1]; // * z (disappears) + SafeScreenTexCoord[2] = (ModelViewProjectionPositiondata[0] + ModelViewProjectionPositionslope[0]*x) * iw * ScreenScaleRefractReflect[2] + ScreenCenterRefractReflect[2]; // * z (disappears) + SafeScreenTexCoord[3] = (ModelViewProjectionPositiondata[1] + ModelViewProjectionPositionslope[1]*x) * iw * ScreenScaleRefractReflect[3] + ScreenCenterRefractReflect[3]; // * z (disappears) + + // " vec4 ScreenTexCoord = SafeScreenTexCoord + vec2(normalize(vec3(dp_texture2D(Texture_Normal, TexCoord)) - vec3(0.5))).xyxy * DistortScaleRefractReflect;\n" + v[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f; + v[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f; + v[2] = buffer_texture_normalbgra8[x*4+0] * (1.0f / 128.0f) - 1.0f; + DPSOFTRAST_Vector3Normalize(v); + ScreenTexCoord[0] = SafeScreenTexCoord[0] + v[0] * DistortScaleRefractReflect[0]; + ScreenTexCoord[1] = SafeScreenTexCoord[1] + v[1] * DistortScaleRefractReflect[1]; + ScreenTexCoord[2] = SafeScreenTexCoord[2] + v[0] * DistortScaleRefractReflect[2]; + ScreenTexCoord[3] = SafeScreenTexCoord[3] + v[1] * DistortScaleRefractReflect[3]; + + // " float Fresnel = pow(min(1.0, 1.0 - float(normalize(EyeVector).z)), 2.0) * ReflectFactor + ReflectOffset;\n" + v[0] = (EyeVectordata[0] + EyeVectorslope[0] * x); // * z (disappears) + v[1] = (EyeVectordata[1] + EyeVectorslope[1] * x); // * z (disappears) + v[2] = (EyeVectordata[2] + EyeVectorslope[2] * x); // * z (disappears) + DPSOFTRAST_Vector3Normalize(v); + Fresnel = 1.0f - v[2]; + Fresnel = min(1.0f, Fresnel); + Fresnel = Fresnel * Fresnel * ReflectFactor + ReflectOffset; + + // " dp_FragColor = vec4(dp_texture2D(Texture_Refraction, ScreenTexCoord).rgb, 1.0) * RefractColor;\n" + // " dp_FragColor = mix(vec4(dp_texture2D(Texture_Refraction, ScreenTexCoord.xy).rgb, 1) * RefractColor, vec4(dp_texture2D(Texture_Reflection, ScreenTexCoord.zw).rgb, 1) * ReflectColor, Fresnel);\n" + DPSOFTRAST_Texture2DBGRA8(texture_refraction, 0, ScreenTexCoord[0], ScreenTexCoord[1], c1); + DPSOFTRAST_Texture2DBGRA8(texture_reflection, 0, ScreenTexCoord[2], ScreenTexCoord[3], c2); + + buffer_FragColorbgra8[x*4+0] = (c1[0] * RefractColor[0]) * (1.0f - Fresnel) + (c2[0] * ReflectColor[0]) * Fresnel; + buffer_FragColorbgra8[x*4+1] = (c1[1] * RefractColor[1]) * (1.0f - Fresnel) + (c2[1] * ReflectColor[1]) * Fresnel; + buffer_FragColorbgra8[x*4+2] = (c1[2] * RefractColor[2]) * (1.0f - Fresnel) + (c2[2] * ReflectColor[2]) * Fresnel; + buffer_FragColorbgra8[x*4+3] = min(( RefractColor[3] * (1.0f - Fresnel) + ReflectColor[3] * Fresnel) * 256, 255); + } + DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -3960,7 +4669,7 @@ void DPSOFTRAST_PixelShader_Water(DPSOFTRAST_State_Thread *thread, const DPSOFTR void DPSOFTRAST_VertexShader_ShowDepth(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } void DPSOFTRAST_PixelShader_ShowDepth(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -3969,7 +4678,7 @@ void DPSOFTRAST_PixelShader_ShowDepth(DPSOFTRAST_State_Thread *thread, const DPS float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + memset(buffer_FragColorbgra8 + span->startx*4, 0, (span->endx - span->startx)*4); DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -3977,7 +4686,7 @@ void DPSOFTRAST_PixelShader_ShowDepth(DPSOFTRAST_State_Thread *thread, const DPS void DPSOFTRAST_VertexShader_DeferredGeometry(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } void DPSOFTRAST_PixelShader_DeferredGeometry(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -3986,7 +4695,7 @@ void DPSOFTRAST_PixelShader_DeferredGeometry(DPSOFTRAST_State_Thread *thread, co float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + memset(buffer_FragColorbgra8 + span->startx*4, 0, (span->endx - span->startx)*4); DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -3994,7 +4703,7 @@ void DPSOFTRAST_PixelShader_DeferredGeometry(DPSOFTRAST_State_Thread *thread, co void DPSOFTRAST_VertexShader_DeferredLightSource(void) { - DPSOFTRAST_Array_TransformProject(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.screencoord4f, dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); + DPSOFTRAST_Array_TransformProject(DPSOFTRAST_ARRAY_POSITION, DPSOFTRAST_ARRAY_POSITION, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1); } void DPSOFTRAST_PixelShader_DeferredLightSource(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span) @@ -4003,7 +4712,7 @@ void DPSOFTRAST_PixelShader_DeferredLightSource(DPSOFTRAST_State_Thread *thread, float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH]; unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4]; DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z); - memset(buffer_FragColorbgra8, 0, span->length*4); + memset(buffer_FragColorbgra8 + span->startx*4, 0, (span->endx - span->startx)*4); DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8); } @@ -4027,387 +4736,181 @@ static const DPSOFTRAST_ShaderModeInfo DPSOFTRAST_ShaderModeTable[SHADERMODE_COU {2, DPSOFTRAST_VertexShader_FlatColor, DPSOFTRAST_PixelShader_FlatColor, {DPSOFTRAST_ARRAY_TEXCOORD0, ~0}, {GL20TU_COLOR, ~0}}, {2, DPSOFTRAST_VertexShader_VertexColor, DPSOFTRAST_PixelShader_VertexColor, {DPSOFTRAST_ARRAY_COLOR, DPSOFTRAST_ARRAY_TEXCOORD0, ~0}, {GL20TU_COLOR, ~0}}, {2, DPSOFTRAST_VertexShader_Lightmap, DPSOFTRAST_PixelShader_Lightmap, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD4, ~0}, {GL20TU_COLOR, GL20TU_LIGHTMAP, GL20TU_GLOW, ~0}}, - {2, DPSOFTRAST_VertexShader_FakeLight, DPSOFTRAST_PixelShader_FakeLight, {~0}, {~0}}, - {2, DPSOFTRAST_VertexShader_LightDirectionMap_ModelSpace, DPSOFTRAST_PixelShader_LightDirectionMap_ModelSpace, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD4, ~0}, {GL20TU_COLOR, GL20TU_LIGHTMAP, GL20TU_GLOW, ~0}}, - {2, DPSOFTRAST_VertexShader_LightDirectionMap_TangentSpace, DPSOFTRAST_PixelShader_LightDirectionMap_TangentSpace, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD4, ~0}, {GL20TU_COLOR, GL20TU_LIGHTMAP, GL20TU_GLOW, ~0}}, - {2, DPSOFTRAST_VertexShader_LightDirection, DPSOFTRAST_PixelShader_LightDirection, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, ~0}}, + {2, DPSOFTRAST_VertexShader_FakeLight, DPSOFTRAST_PixelShader_FakeLight, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD5, DPSOFTRAST_ARRAY_TEXCOORD6, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, ~0}}, + {2, DPSOFTRAST_VertexShader_LightDirectionMap_ModelSpace, DPSOFTRAST_PixelShader_LightDirectionMap_ModelSpace, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD5, DPSOFTRAST_ARRAY_TEXCOORD6, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, GL20TU_LIGHTMAP, GL20TU_DELUXEMAP, ~0}}, + {2, DPSOFTRAST_VertexShader_LightDirectionMap_TangentSpace, DPSOFTRAST_PixelShader_LightDirectionMap_TangentSpace, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD5, DPSOFTRAST_ARRAY_TEXCOORD6, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, GL20TU_LIGHTMAP, GL20TU_DELUXEMAP, ~0}}, + {2, DPSOFTRAST_VertexShader_LightDirection, DPSOFTRAST_PixelShader_LightDirection, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD5, DPSOFTRAST_ARRAY_TEXCOORD6, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, ~0}}, {2, DPSOFTRAST_VertexShader_LightSource, DPSOFTRAST_PixelShader_LightSource, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD4, ~0}, {GL20TU_COLOR, GL20TU_PANTS, GL20TU_SHIRT, GL20TU_GLOW, GL20TU_NORMAL, GL20TU_GLOSS, GL20TU_CUBE, ~0}}, - {2, DPSOFTRAST_VertexShader_Refraction, DPSOFTRAST_PixelShader_Refraction, {~0}}, - {2, DPSOFTRAST_VertexShader_Water, DPSOFTRAST_PixelShader_Water, {~0}}, + {2, DPSOFTRAST_VertexShader_Refraction, DPSOFTRAST_PixelShader_Refraction, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD4, ~0}, {GL20TU_NORMAL, GL20TU_REFRACTION, ~0}}, + {2, DPSOFTRAST_VertexShader_Water, DPSOFTRAST_PixelShader_Water, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, DPSOFTRAST_ARRAY_TEXCOORD2, DPSOFTRAST_ARRAY_TEXCOORD3, DPSOFTRAST_ARRAY_TEXCOORD4, DPSOFTRAST_ARRAY_TEXCOORD6, ~0}, {GL20TU_NORMAL, GL20TU_REFLECTION, GL20TU_REFRACTION, ~0}}, {2, DPSOFTRAST_VertexShader_ShowDepth, DPSOFTRAST_PixelShader_ShowDepth, {~0}}, {2, DPSOFTRAST_VertexShader_DeferredGeometry, DPSOFTRAST_PixelShader_DeferredGeometry, {~0}}, - {2, DPSOFTRAST_VertexShader_DeferredLightSource, DPSOFTRAST_PixelShader_DeferredLightSource, {~0}} + {2, DPSOFTRAST_VertexShader_DeferredLightSource, DPSOFTRAST_PixelShader_DeferredLightSource, {~0}}, }; - -int DPSOFTRAST_Draw_InterpretCommands(DPSOFTRAST_State_Thread *thread, int commandoffset, int endoffset) -{ - while (commandoffset != endoffset) - { - DPSOFTRAST_Command *command = (DPSOFTRAST_Command *)&dpsoftrast.commandpool.commands[commandoffset]; - switch (command->opcode) - { -#define INTERPCOMMAND(name) \ - case DPSOFTRAST_OPCODE_##name : \ - DPSOFTRAST_Interpret_##name (thread, (DPSOFTRAST_Command_##name *)command); \ - commandoffset += sizeof( DPSOFTRAST_Command_##name ) + ((COMMAND_SIZE - (sizeof( DPSOFTRAST_Command_##name )&(COMMAND_SIZE-1))) & (COMMAND_SIZE-1)); \ - if (commandoffset >= DPSOFTRAST_DRAW_MAXCOMMANDPOOL) \ - commandoffset = 0; \ - break; - INTERPCOMMAND(Viewport) - INTERPCOMMAND(ClearColor) - INTERPCOMMAND(ClearDepth) - INTERPCOMMAND(ColorMask) - INTERPCOMMAND(DepthTest) - INTERPCOMMAND(ScissorTest) - INTERPCOMMAND(Scissor) - INTERPCOMMAND(BlendFunc) - INTERPCOMMAND(BlendSubtract) - INTERPCOMMAND(DepthMask) - INTERPCOMMAND(DepthFunc) - INTERPCOMMAND(DepthRange) - INTERPCOMMAND(PolygonOffset) - INTERPCOMMAND(AlphaTest) - INTERPCOMMAND(AlphaFunc) - INTERPCOMMAND(SetTexture) - INTERPCOMMAND(SetShader) - INTERPCOMMAND(Uniform4f) - INTERPCOMMAND(UniformMatrix4f) - INTERPCOMMAND(Uniform1i) - - case DPSOFTRAST_OPCODE_Reset: - commandoffset = 0; - break; - } - } - return commandoffset; -} - -int DPSOFTRAST_Draw_ProcessSpans(DPSOFTRAST_State_Thread *thread, int commandoffset) +static void DPSOFTRAST_Draw_DepthTest(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_State_Span *span) { - int i; int x; int startx; int endx; -// unsigned int c; -// unsigned int *colorpixel; unsigned int *depthpixel; - float w; - float wslope; int depth; int depthslope; unsigned int d; + unsigned char *pixelmask; DPSOFTRAST_State_Triangle *triangle; - DPSOFTRAST_State_Span *span; - unsigned char pixelmask[DPSOFTRAST_DRAW_MAXSPANLENGTH]; - for (i = 0; i < thread->numspans; i++) + triangle = &thread->triangles[span->triangle]; + depthpixel = dpsoftrast.fb_depthpixels + span->y * dpsoftrast.fb_width + span->x; + startx = span->startx; + endx = span->endx; + depth = span->depthbase; + depthslope = span->depthslope; + pixelmask = thread->pixelmaskarray; + if (thread->depthtest && dpsoftrast.fb_depthpixels) { - span = &thread->spans[i]; - triangle = &dpsoftrast.trianglepool.triangles[span->triangle]; - if (commandoffset != triangle->commandoffset) - { - commandoffset = DPSOFTRAST_Draw_InterpretCommands(thread, commandoffset, triangle->commandoffset); - DPSOFTRAST_ValidateQuick(thread, DPSOFTRAST_VALIDATE_DRAW); - } - if (thread->depthtest && dpsoftrast.fb_depthpixels) + switch(thread->fb_depthfunc) { - wslope = triangle->w[0]; - w = triangle->w[2] + span->x*wslope + span->y*triangle->w[1]; - depthslope = (int)(wslope*DPSOFTRAST_DEPTHSCALE); - depth = (int)(w*DPSOFTRAST_DEPTHSCALE - DPSOFTRAST_DEPTHOFFSET*(thread->polygonoffset[1] + fabs(wslope)*thread->polygonoffset[0])); - depthpixel = dpsoftrast.fb_depthpixels + span->y * dpsoftrast.fb_width + span->x; - switch(thread->fb_depthfunc) - { - default: - case GL_ALWAYS: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = true; break; - case GL_LESS: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = depthpixel[x] < d; break; - case GL_LEQUAL: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = depthpixel[x] <= d; break; - case GL_EQUAL: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = depthpixel[x] == d; break; - case GL_GEQUAL: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = depthpixel[x] >= d; break; - case GL_GREATER: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = depthpixel[x] > d; break; - case GL_NEVER: for (x = 0, d = depth;x < span->length;x++, d += depthslope) pixelmask[x] = false; break; - } - //colorpixel = dpsoftrast.fb_colorpixels[0] + (span->y * dpsoftrast.fb_width + span->x) * 4;; - //for (x = 0;x < span->length;x++) - // colorpixel[x] = (depthpixel[x] & 0xFF000000) ? (0x00FF0000) : (depthpixel[x] & 0x00FF0000); - // if there is no color buffer, skip pixel shader - startx = 0; - endx = span->length; - while (startx < endx && !pixelmask[startx]) - startx++; - while (endx > startx && !pixelmask[endx-1]) - endx--; - if (startx >= endx) - continue; // no pixels to fill - span->pixelmask = pixelmask; - span->startx = startx; - span->endx = endx; - // run pixel shader if appropriate - // do this before running depthmask code, to allow the pixelshader - // to clear pixelmask values for alpha testing - if (dpsoftrast.fb_colorpixels[0] && thread->fb_colormask) - DPSOFTRAST_ShaderModeTable[thread->shader_mode].Span(thread, triangle, span); - if (thread->depthmask) - for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) - if (pixelmask[x]) - depthpixel[x] = d; - } - else - { - // no depth testing means we're just dealing with color... - // if there is no color buffer, skip pixel shader - if (dpsoftrast.fb_colorpixels[0] && thread->fb_colormask) - { - memset(pixelmask, 1, span->length); - span->pixelmask = pixelmask; - span->startx = 0; - span->endx = span->length; - DPSOFTRAST_ShaderModeTable[thread->shader_mode].Span(thread, triangle, span); - } + default: + case GL_ALWAYS: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = true; break; + case GL_LESS: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = depthpixel[x] < d; break; + case GL_LEQUAL: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = depthpixel[x] <= d; break; + case GL_EQUAL: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = depthpixel[x] == d; break; + case GL_GEQUAL: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = depthpixel[x] >= d; break; + case GL_GREATER: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = depthpixel[x] > d; break; + case GL_NEVER: for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) pixelmask[x] = false; break; } + while (startx < endx && !pixelmask[startx]) + startx++; + while (endx > startx && !pixelmask[endx-1]) + endx--; } - thread->numspans = 0; - return commandoffset; + else + { + // no depth testing means we're just dealing with color... + memset(pixelmask + startx, 1, endx - startx); + } + span->pixelmask = pixelmask; + span->startx = startx; + span->endx = endx; } -void DPSOFTRAST_Draw_GenerateSpans(DPSOFTRAST_State_Thread *thread, int freetriangle) +static void DPSOFTRAST_Draw_DepthWrite(const DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Span *span) { - int miny = (thread->index*dpsoftrast.fb_height)/dpsoftrast.numthreads; - int maxy = ((thread->index+1)*dpsoftrast.fb_height)/dpsoftrast.numthreads; - int commandoffset = thread->commandoffset; - int triangleoffset = thread->triangleoffset; - DPSOFTRAST_State_Triangle *triangle = NULL; - int starty; - int endy; - int y; - int numpoints; - __m128 coords[4]; - __m128i ycoords; - while (triangleoffset != freetriangle) - { - triangle = &dpsoftrast.trianglepool.triangles[triangleoffset]; - if (++triangleoffset >= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL) - triangleoffset = 0; - starty = triangle->starty + 1; - endy = triangle->endy; - if (starty >= maxy || endy <= miny) - continue; - numpoints = triangle->numpoints; - coords[0] = _mm_load_ps(triangle->coords[0]); - coords[1] = _mm_load_ps(triangle->coords[1]); - coords[2] = _mm_load_ps(triangle->coords[2]); - coords[3] = _mm_load_ps(triangle->coords[3]); - ycoords = _mm_load_si128((const __m128i *)triangle->ycoords); - if (starty < miny) - starty = miny; - if (endy > maxy) - endy = maxy; - for (y = starty; y < endy;) - { - __m128 xcoords, xslope; - __m128i ycc = _mm_cmpgt_epi32(_mm_set1_epi32(y), ycoords); - int yccmask = _mm_movemask_epi8(ycc); - int edge0p, edge0n, edge1p, edge1n; - int nexty; - if (numpoints == 4) - { - switch(yccmask) - { - default: - case 0xFFFF: /*0000*/ y = endy; continue; - case 0xFFF0: /*1000*/ edge0p = 3;edge0n = 0;edge1p = 1;edge1n = 0;break; - case 0xFF0F: /*0100*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 1;break; - case 0xFF00: /*1100*/ edge0p = 3;edge0n = 0;edge1p = 2;edge1n = 1;break; - case 0xF0FF: /*0010*/ edge0p = 1;edge0n = 2;edge1p = 3;edge1n = 2;break; - case 0xF0F0: /*1010*/ edge0p = 1;edge0n = 2;edge1p = 3;edge1n = 2;break; // concave - nonsense - case 0xF00F: /*0110*/ edge0p = 0;edge0n = 1;edge1p = 3;edge1n = 2;break; - case 0xF000: /*1110*/ edge0p = 3;edge0n = 0;edge1p = 3;edge1n = 2;break; - case 0x0FFF: /*0001*/ edge0p = 2;edge0n = 3;edge1p = 0;edge1n = 3;break; - case 0x0FF0: /*1001*/ edge0p = 2;edge0n = 3;edge1p = 1;edge1n = 0;break; - case 0x0F0F: /*0101*/ edge0p = 2;edge0n = 3;edge1p = 2;edge1n = 1;break; // concave - nonsense - case 0x0F00: /*1101*/ edge0p = 2;edge0n = 3;edge1p = 2;edge1n = 1;break; - case 0x00FF: /*0011*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 3;break; - case 0x00F0: /*1011*/ edge0p = 1;edge0n = 2;edge1p = 1;edge1n = 0;break; - case 0x000F: /*0111*/ edge0p = 0;edge0n = 1;edge1p = 0;edge1n = 3;break; - case 0x0000: /*1111*/ y++; continue; - } - } - else - { - switch(yccmask) - { - default: - case 0xFFFF: /*000*/ y = endy; continue; - case 0xFFF0: /*100*/ edge0p = 2;edge0n = 0;edge1p = 1;edge1n = 0;break; - case 0xFF0F: /*010*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 1;break; - case 0xFF00: /*110*/ edge0p = 2;edge0n = 0;edge1p = 2;edge1n = 1;break; - case 0x00FF: /*001*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 2;break; - case 0x00F0: /*101*/ edge0p = 1;edge0n = 2;edge1p = 1;edge1n = 0;break; - case 0x000F: /*011*/ edge0p = 0;edge0n = 1;edge1p = 0;edge1n = 2;break; - case 0x0000: /*111*/ y++; continue; - } - } - ycc = _mm_max_epi16(_mm_srli_epi16(ycc, 1), ycoords); - ycc = _mm_min_epi16(ycc, _mm_shuffle_epi32(ycc, _MM_SHUFFLE(1, 0, 3, 2))); - ycc = _mm_min_epi16(ycc, _mm_shuffle_epi32(ycc, _MM_SHUFFLE(2, 3, 0, 1))); - nexty = _mm_extract_epi16(ycc, 0); - if(nexty >= endy) nexty = endy-1; - if (_mm_ucomigt_ss(_mm_max_ss(coords[edge0n], coords[edge0p]), _mm_min_ss(coords[edge1n], coords[edge1p]))) - { - int tmp = edge0n; - edge0n = edge1n; - edge1n = tmp; - tmp = edge0p; - edge0p = edge1p; - edge1p = tmp; - } - xslope = _mm_sub_ps(_mm_movelh_ps(coords[edge0n], coords[edge1n]), _mm_movelh_ps(coords[edge0p], coords[edge1p])); - xslope = _mm_div_ps(xslope, _mm_shuffle_ps(xslope, xslope, _MM_SHUFFLE(3, 3, 1, 1))); - xcoords = _mm_add_ps(_mm_movelh_ps(coords[edge0p], coords[edge1p]), - _mm_mul_ps(xslope, _mm_sub_ps(_mm_set1_ps(y), _mm_shuffle_ps(coords[edge0p], coords[edge1p], _MM_SHUFFLE(1, 1, 1, 1))))); - xcoords = _mm_add_ps(xcoords, _mm_set1_ps(0.5f)); - for(; y <= nexty; y++, xcoords = _mm_add_ps(xcoords, xslope)) - { - int startx, endx, offset; - startx = _mm_cvtss_si32(xcoords); - endx = _mm_cvtss_si32(_mm_movehl_ps(xcoords, xcoords)); - if (startx < 0) startx = 0; - if (endx > dpsoftrast.fb_width) endx = dpsoftrast.fb_width; - if (startx >= endx) continue; - for (offset = startx; offset < endx;) - { - DPSOFTRAST_State_Span *span = &thread->spans[thread->numspans]; - span->triangle = (int)(triangle - dpsoftrast.trianglepool.triangles); - span->x = offset; - span->y = y; - span->length = endx - offset; - if (span -> length > DPSOFTRAST_DRAW_MAXSPANLENGTH) - span -> length = DPSOFTRAST_DRAW_MAXSPANLENGTH; - offset += span->length; - if (++thread->numspans >= DPSOFTRAST_DRAW_MAXSPANS) - commandoffset = DPSOFTRAST_Draw_ProcessSpans(thread, commandoffset); - } - } - } - } - - if (thread->numspans > 0) - commandoffset = DPSOFTRAST_Draw_ProcessSpans(thread, commandoffset); - if (commandoffset != triangle->commandoffset) + int x, d, depth, depthslope, startx, endx; + const unsigned char *pixelmask; + unsigned int *depthpixel; + if (thread->depthmask && thread->depthtest && dpsoftrast.fb_depthpixels) { - commandoffset = DPSOFTRAST_Draw_InterpretCommands(thread, commandoffset, triangle->commandoffset); - DPSOFTRAST_ValidateQuick(thread, DPSOFTRAST_VALIDATE_DRAW); + depth = span->depthbase; + depthslope = span->depthslope; + pixelmask = span->pixelmask; + startx = span->startx; + endx = span->endx; + depthpixel = dpsoftrast.fb_depthpixels + span->y * dpsoftrast.fb_width + span->x; + for (x = startx, d = depth + depthslope*startx;x < endx;x++, d += depthslope) + if (pixelmask[x]) + depthpixel[x] = d; } - - MEMORY_BARRIER; - - thread->commandoffset = commandoffset; - thread->triangleoffset = triangleoffset; } -void DPSOFTRAST_Draw_FlushThreads(void) +void DPSOFTRAST_Draw_ProcessSpans(DPSOFTRAST_State_Thread *thread) { - DPSOFTRAST_State_Thread *thread; int i; - if(dpsoftrast.drawtriangle != dpsoftrast.trianglepool.freetriangle) - { - MEMORY_BARRIER; - dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle; - } -#ifdef USETHREADS - SDL_LockMutex(dpsoftrast.trianglemutex); -#endif - for (i = 0; i < dpsoftrast.numthreads; i++) + DPSOFTRAST_State_Triangle *triangle; + DPSOFTRAST_State_Span *span; + for (i = 0; i < thread->numspans; i++) { - thread = &dpsoftrast.threads[i]; -#ifdef USETHREADS - while (thread->triangleoffset != dpsoftrast.drawtriangle) - { - thread->waiting = true; - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_CondWait(thread->waitcond, dpsoftrast.trianglemutex); - thread->waiting = false; - } -#else - if (thread->triangleoffset != dpsoftrast.drawtriangle) - DPSOFTRAST_Draw_GenerateSpans(thread, dpsoftrast.drawtriangle); -#endif + span = &thread->spans[i]; + triangle = &thread->triangles[span->triangle]; + DPSOFTRAST_Draw_DepthTest(thread, span); + if (span->startx >= span->endx) + continue; + // run pixel shader if appropriate + // do this before running depthmask code, to allow the pixelshader + // to clear pixelmask values for alpha testing + if (dpsoftrast.fb_colorpixels[0] && thread->fb_colormask) + DPSOFTRAST_ShaderModeTable[thread->shader_mode].Span(thread, triangle, span); + DPSOFTRAST_Draw_DepthWrite(thread, span); } -#ifdef USETHREADS - SDL_UnlockMutex(dpsoftrast.trianglemutex); -#endif - dpsoftrast.trianglepool.usedtriangles = 0; - dpsoftrast.commandpool.usedcommands = 0; + thread->numspans = 0; } -#ifdef USETHREADS -static int DPSOFTRAST_Draw_Thread(void *data) -{ - DPSOFTRAST_State_Thread *thread = (DPSOFTRAST_State_Thread *)data; - while(thread->index >= 0) - { - if (thread->triangleoffset != dpsoftrast.drawtriangle) - { - DPSOFTRAST_Draw_GenerateSpans(thread, dpsoftrast.drawtriangle); - } - else - { - SDL_LockMutex(dpsoftrast.trianglemutex); - if (thread->triangleoffset != dpsoftrast.drawtriangle) - { - SDL_UnlockMutex(dpsoftrast.trianglemutex); - continue; - } - if (thread->waiting) SDL_CondSignal(thread->waitcond); - SDL_CondWait(dpsoftrast.trianglecond, dpsoftrast.trianglemutex); - SDL_UnlockMutex(dpsoftrast.trianglemutex); - } - } - return 0; -} -#endif +DEFCOMMAND(22, Draw, int datasize; int starty; int endy; ATOMIC_COUNTER refcount; int clipped; int firstvertex; int numvertices; int numtriangles; float *arrays; int *element3i; unsigned short *element3s;); -void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const int *element3i, const unsigned short *element3s, unsigned char *arraymask, int numarrays) +static void DPSOFTRAST_Interpret_Draw(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_Draw *command) { -#ifdef SSE2_PRESENT - int cullface = dpsoftrast.cullface; - int width = dpsoftrast.fb_width; - int height = dpsoftrast.fb_height; - __m128i fbmax = _mm_sub_epi16(_mm_setr_epi16(width, height, width, height, width, height, width, height), _mm_set1_epi16(1)); - DPSOFTRAST_State_Triangle *triangle; - int numqueued = 0; +#ifdef SSE_POSSIBLE + int cullface = thread->cullface; + int minx, maxx, miny, maxy; + int miny1, maxy1, miny2, maxy2; + __m128i fbmin, fbmax; + __m128 viewportcenter, viewportscale; + int firstvertex = command->firstvertex; + int numvertices = command->numvertices; + int numtriangles = command->numtriangles; + const int *element3i = command->element3i; + const unsigned short *element3s = command->element3s; + int clipped = command->clipped; int i; int j; int k; int y; int e[3]; __m128i screeny; - int starty, endy; + int starty, endy, bandy; int numpoints; int clipcase; float clipdist[4]; + float clip0origin, clip0slope; + int clip0dir; __m128 triangleedge1, triangleedge2, trianglenormal; __m128 clipfrac[3]; __m128 screen[4]; + DPSOFTRAST_State_Triangle *triangle; DPSOFTRAST_Texture *texture; + DPSOFTRAST_ValidateQuick(thread, DPSOFTRAST_VALIDATE_DRAW); + miny = thread->fb_scissor[1]; + maxy = thread->fb_scissor[1] + thread->fb_scissor[3]; + miny1 = bound(miny, thread->miny1, maxy); + maxy1 = bound(miny, thread->maxy1, maxy); + miny2 = bound(miny, thread->miny2, maxy); + maxy2 = bound(miny, thread->maxy2, maxy); + if ((command->starty >= maxy1 || command->endy <= miny1) && (command->starty >= maxy2 || command->endy <= miny2)) + { + if (!ATOMIC_DECREMENT(command->refcount)) + { + if (command->commandsize <= DPSOFTRAST_ALIGNCOMMAND(sizeof(DPSOFTRAST_Command_Draw))) + MM_FREE(command->arrays); + } + return; + } + minx = thread->fb_scissor[0]; + maxx = thread->fb_scissor[0] + thread->fb_scissor[2]; + fbmin = _mm_setr_epi16(minx, miny1, minx, miny1, minx, miny1, minx, miny1); + fbmax = _mm_sub_epi16(_mm_setr_epi16(maxx, maxy2, maxx, maxy2, maxx, maxy2, maxx, maxy2), _mm_set1_epi16(1)); + viewportcenter = _mm_load_ps(thread->fb_viewportcenter); + viewportscale = _mm_load_ps(thread->fb_viewportscale); screen[3] = _mm_setzero_ps(); - clipfrac[0] = clipfrac[1] = clipfrac[2] = _mm_setzero_ps(); + clipfrac[0] = clipfrac[1] = clipfrac[2] = _mm_setzero_ps(); for (i = 0;i < numtriangles;i++) { + const float *screencoord4f = command->arrays; + const float *arrays = screencoord4f + numvertices*4; + // generate the 3 edges of this triangle // generate spans for the triangle - switch based on left split or right split classification of triangle - if (element3i) - { - e[0] = element3i[i*3+0] - firstvertex; - e[1] = element3i[i*3+1] - firstvertex; - e[2] = element3i[i*3+2] - firstvertex; - } - else if (element3s) + if (element3s) { e[0] = element3s[i*3+0] - firstvertex; e[1] = element3s[i*3+1] - firstvertex; e[2] = element3s[i*3+2] - firstvertex; } + else if (element3i) + { + e[0] = element3i[i*3+0] - firstvertex; + e[1] = element3i[i*3+1] - firstvertex; + e[2] = element3i[i*3+2] - firstvertex; + } else { e[0] = i*3+0; @@ -4432,53 +4935,50 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i continue; \ break; \ } - //trianglenormal = _mm_sub_ps(_mm_mul_ps(triangleedge[0], _mm_shuffle_ps(triangleedge[1], triangleedge[1], _MM_SHUFFLE(3, 0, 2, 1))), - // _mm_mul_ps(_mm_shuffle_ps(triangleedge[0], triangleedge[0], _MM_SHUFFLE(3, 0, 2, 1)), triangleedge[1])); - //trianglenormal[2] = triangleedge[0][0] * triangleedge[1][1] - triangleedge[0][1] * triangleedge[1][0]; - //trianglenormal[0] = triangleedge[0][1] * triangleedge[1][2] - triangleedge[0][2] * triangleedge[1][1]; - //trianglenormal[1] = triangleedge[0][2] * triangleedge[1][0] - triangleedge[0][0] * triangleedge[1][2]; - - // macros for clipping vertices #define CLIPPEDVERTEXLERP(k,p1, p2) \ clipfrac[p1] = _mm_set1_ps(clipdist[p1] / (clipdist[p1] - clipdist[p2])); \ { \ - __m128 v1 = _mm_load_ps(&dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[p1]*4]), v2 = _mm_load_ps(&dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[p2]*4]); \ - screen[k] = DPSOFTRAST_Draw_ProjectVertex(_mm_add_ps(v1, _mm_mul_ps(_mm_sub_ps(v2, v1), clipfrac[p1]))); \ + __m128 v1 = _mm_load_ps(&arrays[e[p1]*4]), v2 = _mm_load_ps(&arrays[e[p2]*4]); \ + DPSOFTRAST_PROJECTVERTEX(screen[k], _mm_add_ps(v1, _mm_mul_ps(_mm_sub_ps(v2, v1), clipfrac[p1])), viewportcenter, viewportscale); \ } #define CLIPPEDVERTEXCOPY(k,p1) \ - screen[k] = _mm_load_ps(&dpsoftrast.screencoord4f[e[p1]*4]); + screen[k] = _mm_load_ps(&screencoord4f[e[p1]*4]); -#define GENATTRIBCOPY(j, attrib, p1) \ - attrib = _mm_load_ps(&dpsoftrast.post_array4f[j][e[p1]*4]); -#define GENATTRIBLERP(j, attrib, p1, p2) \ +#define GENATTRIBCOPY(attrib, p1) \ + attrib = _mm_load_ps(&arrays[e[p1]*4]); +#define GENATTRIBLERP(attrib, p1, p2) \ { \ - __m128 v1 = _mm_load_ps(&dpsoftrast.post_array4f[j][e[p1]*4]), v2 = _mm_load_ps(&dpsoftrast.post_array4f[j][e[p2]*4]); \ + __m128 v1 = _mm_load_ps(&arrays[e[p1]*4]), v2 = _mm_load_ps(&arrays[e[p2]*4]); \ attrib = _mm_add_ps(v1, _mm_mul_ps(_mm_sub_ps(v2, v1), clipfrac[p1])); \ } -#define GENATTRIBS(j, attrib0, attrib1, attrib2) \ +#define GENATTRIBS(attrib0, attrib1, attrib2) \ switch(clipcase) \ { \ default: \ - case 0: GENATTRIBCOPY(j, attrib0, 0); GENATTRIBCOPY(j, attrib1, 1); GENATTRIBCOPY(j, attrib2, 2); break; \ - case 1: GENATTRIBCOPY(j, attrib0, 0); GENATTRIBCOPY(j, attrib1, 1); GENATTRIBLERP(j, attrib2, 1, 2); break; \ - case 2: GENATTRIBCOPY(j, attrib0, 0); GENATTRIBLERP(j, attrib1, 0, 1); GENATTRIBLERP(j, attrib2, 1, 2); break; \ - case 3: GENATTRIBCOPY(j, attrib0, 0); GENATTRIBLERP(j, attrib1, 0, 1); GENATTRIBLERP(j, attrib2, 2, 0); break; \ - case 4: GENATTRIBLERP(j, attrib0, 0, 1); GENATTRIBCOPY(j, attrib1, 1); GENATTRIBCOPY(j, attrib2, 2); break; \ - case 5: GENATTRIBLERP(j, attrib0, 0, 1); GENATTRIBCOPY(j, attrib1, 1); GENATTRIBLERP(j, attrib2, 1, 2); break; \ - case 6: GENATTRIBLERP(j, attrib0, 1, 2); GENATTRIBCOPY(j, attrib1, 2); GENATTRIBLERP(j, attrib2, 2, 0); break; \ + case 0: GENATTRIBCOPY(attrib0, 0); GENATTRIBCOPY(attrib1, 1); GENATTRIBCOPY(attrib2, 2); break; \ + case 1: GENATTRIBCOPY(attrib0, 0); GENATTRIBCOPY(attrib1, 1); GENATTRIBLERP(attrib2, 1, 2); break; \ + case 2: GENATTRIBCOPY(attrib0, 0); GENATTRIBLERP(attrib1, 0, 1); GENATTRIBLERP(attrib2, 1, 2); break; \ + case 3: GENATTRIBCOPY(attrib0, 0); GENATTRIBLERP(attrib1, 0, 1); GENATTRIBLERP(attrib2, 2, 0); break; \ + case 4: GENATTRIBLERP(attrib0, 0, 1); GENATTRIBCOPY(attrib1, 1); GENATTRIBCOPY(attrib2, 2); break; \ + case 5: GENATTRIBLERP(attrib0, 0, 1); GENATTRIBCOPY(attrib1, 1); GENATTRIBLERP(attrib2, 1, 2); break; \ + case 6: GENATTRIBLERP(attrib0, 1, 2); GENATTRIBCOPY(attrib1, 2); GENATTRIBLERP(attrib2, 2, 0); break; \ } + if (! clipped) + goto notclipped; + // calculate distance from nearplane - clipdist[0] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+2] + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+3]; - clipdist[1] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+2] + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+3]; - clipdist[2] = dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+2] + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+3]; + clipdist[0] = arrays[e[0]*4+2] + arrays[e[0]*4+3]; + clipdist[1] = arrays[e[1]*4+2] + arrays[e[1]*4+3]; + clipdist[2] = arrays[e[2]*4+2] + arrays[e[2]*4+3]; if (clipdist[0] >= 0.0f) { if (clipdist[1] >= 0.0f) { if (clipdist[2] >= 0.0f) { + notclipped: // triangle is entirely in front of nearplane CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXCOPY(1,1); CLIPPEDVERTEXCOPY(2,2); SKIPBACKFACE; @@ -4493,11 +4993,11 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i clipcase = 1; } } - else + else { if (clipdist[2] >= 0.0f) { - CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXLERP(1,0,1); CLIPPEDVERTEXLERP(2,1,2); CLIPPEDVERTEXCOPY(3,2); + CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXLERP(1,0,1); CLIPPEDVERTEXLERP(2,1,2); CLIPPEDVERTEXCOPY(3,2); SKIPBACKFACE; numpoints = 4; clipcase = 2; @@ -4510,7 +5010,7 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i clipcase = 3; } } - } + } else if (clipdist[1] >= 0.0f) { if (clipdist[2] >= 0.0f) @@ -4540,38 +5040,24 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i { // calculate integer y coords for triangle points __m128i screeni = _mm_packs_epi32(_mm_cvttps_epi32(_mm_movelh_ps(screen[0], screen[1])), _mm_cvttps_epi32(_mm_movelh_ps(screen[2], numpoints > 3 ? screen[3] : screen[2]))), - screenir = _mm_shuffle_epi32(screeni, _MM_SHUFFLE(1, 0, 3, 2)), - screenmin = _mm_min_epi16(screeni, screenir), + screenir = _mm_shuffle_epi32(screeni, _MM_SHUFFLE(1, 0, 3, 2)), + screenmin = _mm_min_epi16(screeni, screenir), screenmax = _mm_max_epi16(screeni, screenir); screenmin = _mm_min_epi16(screenmin, _mm_shufflelo_epi16(screenmin, _MM_SHUFFLE(1, 0, 3, 2))); screenmax = _mm_max_epi16(screenmax, _mm_shufflelo_epi16(screenmax, _MM_SHUFFLE(1, 0, 3, 2))); - screenmin = _mm_max_epi16(screenmin, _mm_setzero_si128()); + screenmin = _mm_max_epi16(screenmin, fbmin); screenmax = _mm_min_epi16(screenmax, fbmax); // skip offscreen triangles if (_mm_cvtsi128_si32(_mm_cmplt_epi16(screenmax, screenmin))) continue; starty = _mm_extract_epi16(screenmin, 1); endy = _mm_extract_epi16(screenmax, 1)+1; + if (starty >= maxy1 && endy <= miny2) + continue; screeny = _mm_srai_epi32(screeni, 16); } - if (dpsoftrast.trianglepool.usedtriangles >= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1) -#ifdef USETHREADS - DPSOFTRAST_Draw_FreeTrianglePool(DPSOFTRAST_DRAW_MAXTRIANGLEPOOL/8); -#else - DPSOFTRAST_Draw_FlushThreads(); -#endif - - triangle = &dpsoftrast.trianglepool.triangles[dpsoftrast.trianglepool.freetriangle]; - triangle->commandoffset = dpsoftrast.commandpool.freecommand; - triangle->starty = starty; - triangle->endy = endy; - triangle->numpoints = numpoints; - _mm_store_ps(triangle->coords[0], screen[0]); - _mm_store_ps(triangle->coords[1], screen[1]); - _mm_store_ps(triangle->coords[2], screen[2]); - _mm_store_ps(triangle->coords[3], numpoints > 3 ? screen[3] : screen[2]); - _mm_store_si128((__m128i *)triangle->ycoords, screeny); + triangle = &thread->triangles[thread->numtriangles]; // calculate attribute plans for triangle data... // okay, this triangle is going to produce spans, we'd better project @@ -4582,63 +5068,99 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i // values { __m128 attribuvslope, attribuxslope, attribuyslope, attribvxslope, attribvyslope, attriborigin, attribedge1, attribedge2, attribxslope, attribyslope, w0, w1, w2, x1, y1; + __m128 mipedgescale, mipdensity; attribuvslope = _mm_div_ps(_mm_movelh_ps(triangleedge1, triangleedge2), _mm_shuffle_ps(trianglenormal, trianglenormal, _MM_SHUFFLE(0, 0, 0, 0))); - attribuxslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(3, 3, 3, 3)); - attribuyslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(2, 2, 2, 2)); - attribvxslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(1, 1, 1, 1)); - attribvyslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(0, 0, 0, 0)); - w0 = _mm_shuffle_ps(screen[0], screen[0], _MM_SHUFFLE(3, 3, 3, 3)); - w1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(3, 3, 3, 3)); - w2 = _mm_shuffle_ps(screen[2], screen[2], _MM_SHUFFLE(3, 3, 3, 3)); + attribuxslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(3, 3, 3, 3)); + attribuyslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(2, 2, 2, 2)); + attribvxslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(1, 1, 1, 1)); + attribvyslope = _mm_shuffle_ps(attribuvslope, attribuvslope, _MM_SHUFFLE(0, 0, 0, 0)); + w0 = _mm_shuffle_ps(screen[0], screen[0], _MM_SHUFFLE(3, 3, 3, 3)); + w1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(3, 3, 3, 3)); + w2 = _mm_shuffle_ps(screen[2], screen[2], _MM_SHUFFLE(3, 3, 3, 3)); attribedge1 = _mm_sub_ss(w0, w1); attribedge2 = _mm_sub_ss(w2, w1); attribxslope = _mm_sub_ss(_mm_mul_ss(attribuxslope, attribedge1), _mm_mul_ss(attribvxslope, attribedge2)); attribyslope = _mm_sub_ss(_mm_mul_ss(attribvyslope, attribedge2), _mm_mul_ss(attribuyslope, attribedge1)); - x1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(0, 0, 0, 0)); - y1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(1, 1, 1, 1)); + x1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(0, 0, 0, 0)); + y1 = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(1, 1, 1, 1)); attriborigin = _mm_sub_ss(w1, _mm_add_ss(_mm_mul_ss(attribxslope, x1), _mm_mul_ss(attribyslope, y1))); _mm_store_ss(&triangle->w[0], attribxslope); _mm_store_ss(&triangle->w[1], attribyslope); _mm_store_ss(&triangle->w[2], attriborigin); - for (j = 0;j < numarrays;j++) + + clip0origin = 0; + clip0slope = 0; + clip0dir = 0; + if(thread->fb_clipplane[0] || thread->fb_clipplane[1] || thread->fb_clipplane[2]) { - if (arraymask[j]) + float cliporigin, clipxslope, clipyslope; + attriborigin = _mm_shuffle_ps(screen[1], screen[1], _MM_SHUFFLE(2, 2, 2, 2)); + attribedge1 = _mm_sub_ss(_mm_shuffle_ps(screen[0], screen[0], _MM_SHUFFLE(2, 2, 2, 2)), attriborigin); + attribedge2 = _mm_sub_ss(_mm_shuffle_ps(screen[2], screen[2], _MM_SHUFFLE(2, 2, 2, 2)), attriborigin); + attribxslope = _mm_sub_ss(_mm_mul_ss(attribuxslope, attribedge1), _mm_mul_ss(attribvxslope, attribedge2)); + attribyslope = _mm_sub_ss(_mm_mul_ss(attribvyslope, attribedge2), _mm_mul_ss(attribuyslope, attribedge1)); + attriborigin = _mm_sub_ss(attriborigin, _mm_add_ss(_mm_mul_ss(attribxslope, x1), _mm_mul_ss(attribyslope, y1))); + cliporigin = _mm_cvtss_f32(attriborigin)*thread->fb_clipplane[2] + thread->fb_clipplane[3]; + clipxslope = thread->fb_clipplane[0] + _mm_cvtss_f32(attribxslope)*thread->fb_clipplane[2]; + clipyslope = thread->fb_clipplane[1] + _mm_cvtss_f32(attribyslope)*thread->fb_clipplane[2]; + if(clipxslope != 0) { - __m128 attrib0, attrib1, attrib2; - GENATTRIBS(j, attrib0, attrib1, attrib2); - attriborigin = _mm_mul_ps(attrib1, w1); - attribedge1 = _mm_sub_ps(_mm_mul_ps(attrib0, w0), attriborigin); - attribedge2 = _mm_sub_ps(_mm_mul_ps(attrib2, w2), attriborigin); - attribxslope = _mm_sub_ps(_mm_mul_ps(attribuxslope, attribedge1), _mm_mul_ps(attribvxslope, attribedge2)); - attribyslope = _mm_sub_ps(_mm_mul_ps(attribvyslope, attribedge2), _mm_mul_ps(attribuyslope, attribedge1)); - attriborigin = _mm_sub_ps(attriborigin, _mm_add_ps(_mm_mul_ps(attribxslope, x1), _mm_mul_ps(attribyslope, y1))); - _mm_stream_ps(triangle->attribs[j][0], attribxslope); - _mm_stream_ps(triangle->attribs[j][1], attribyslope); - _mm_stream_ps(triangle->attribs[j][2], attriborigin); - } - } - } - - // adjust texture LOD by texture density, in the simplest way possible... - { - __m128 mipedgescale, mipedgetc, mipdensity, attrib0, attrib1, attrib2; + clip0origin = -cliporigin/clipxslope; + clip0slope = -clipyslope/clipxslope; + clip0dir = clipxslope > 0 ? 1 : -1; + } + else if(clipyslope > 0) + { + clip0origin = dpsoftrast.fb_width*floor(cliporigin/clipyslope); + clip0slope = dpsoftrast.fb_width; + clip0dir = -1; + } + else if(clipyslope < 0) + { + clip0origin = dpsoftrast.fb_width*ceil(cliporigin/clipyslope); + clip0slope = -dpsoftrast.fb_width; + clip0dir = -1; + } + else if(clip0origin < 0) continue; + } + + mipedgescale = _mm_setzero_ps(); + for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL; j++) + { + __m128 attrib0, attrib1, attrib2; + k = DPSOFTRAST_ShaderModeTable[thread->shader_mode].arrays[j]; + if (k >= DPSOFTRAST_ARRAY_TOTAL) + break; + arrays += numvertices*4; + GENATTRIBS(attrib0, attrib1, attrib2); + attriborigin = _mm_mul_ps(attrib1, w1); + attribedge1 = _mm_sub_ps(_mm_mul_ps(attrib0, w0), attriborigin); + attribedge2 = _mm_sub_ps(_mm_mul_ps(attrib2, w2), attriborigin); + attribxslope = _mm_sub_ps(_mm_mul_ps(attribuxslope, attribedge1), _mm_mul_ps(attribvxslope, attribedge2)); + attribyslope = _mm_sub_ps(_mm_mul_ps(attribvyslope, attribedge2), _mm_mul_ps(attribuyslope, attribedge1)); + attriborigin = _mm_sub_ps(attriborigin, _mm_add_ps(_mm_mul_ps(attribxslope, x1), _mm_mul_ps(attribyslope, y1))); + _mm_storeu_ps(triangle->attribs[k][0], attribxslope); + _mm_storeu_ps(triangle->attribs[k][1], attribyslope); + _mm_storeu_ps(triangle->attribs[k][2], attriborigin); + if (k == DPSOFTRAST_ShaderModeTable[thread->shader_mode].lodarrayindex) + { + mipedgescale = _mm_movelh_ps(triangleedge1, triangleedge2); + mipedgescale = _mm_mul_ps(mipedgescale, mipedgescale); + mipedgescale = _mm_rsqrt_ps(_mm_add_ps(mipedgescale, _mm_shuffle_ps(mipedgescale, mipedgescale, _MM_SHUFFLE(2, 3, 0, 1)))); + mipedgescale = _mm_mul_ps(_mm_sub_ps(_mm_movelh_ps(attrib0, attrib2), _mm_movelh_ps(attrib1, attrib1)), mipedgescale); + } + } + memset(triangle->mip, 0, sizeof(triangle->mip)); - mipedgescale = _mm_movelh_ps(triangleedge1, triangleedge2); - mipedgescale = _mm_mul_ps(mipedgescale, mipedgescale); - mipedgescale = _mm_rsqrt_ps(_mm_add_ps(mipedgescale, _mm_shuffle_ps(mipedgescale, mipedgescale, _MM_SHUFFLE(2, 3, 0, 1)))); - k = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].lodarrayindex; - GENATTRIBS(k, attrib0, attrib1, attrib2); - mipedgetc = _mm_sub_ps(_mm_movelh_ps(attrib0, attrib2), _mm_movelh_ps(attrib1, attrib1)); - mipedgetc = _mm_mul_ps(mipedgetc, mipedgescale); for (j = 0;j < DPSOFTRAST_MAXTEXTUREUNITS;j++) { - int texunit = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].texunits[j]; + int texunit = DPSOFTRAST_ShaderModeTable[thread->shader_mode].texunits[j]; if (texunit >= DPSOFTRAST_MAXTEXTUREUNITS) break; - texture = dpsoftrast.texbound[texunit]; + texture = thread->texbound[texunit]; if (texture && texture->filter > DPSOFTRAST_TEXTURE_FILTER_LINEAR) { - mipdensity = _mm_mul_ps(mipedgetc, _mm_cvtepi32_ps(_mm_shuffle_epi32(_mm_loadl_epi64((const __m128i *)&texture->mipmap[0][2]), _MM_SHUFFLE(1, 0, 1, 0)))); + mipdensity = _mm_mul_ps(mipedgescale, _mm_cvtepi32_ps(_mm_shuffle_epi32(_mm_loadl_epi64((const __m128i *)&texture->mipmap[0][2]), _MM_SHUFFLE(1, 0, 1, 0)))); mipdensity = _mm_mul_ps(mipdensity, mipdensity); mipdensity = _mm_add_ps(mipdensity, _mm_shuffle_ps(mipdensity, mipdensity, _MM_SHUFFLE(2, 3, 0, 1))); mipdensity = _mm_min_ss(mipdensity, _mm_shuffle_ps(mipdensity, mipdensity, _MM_SHUFFLE(2, 2, 2, 2))); @@ -4654,81 +5176,395 @@ void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const i } } } + + for (y = starty, bandy = min(endy, maxy1); y < endy; bandy = min(endy, maxy2), y = max(y, miny2)) + for (; y < bandy;) + { + __m128 xcoords, xslope; + __m128i ycc = _mm_cmpgt_epi32(_mm_set1_epi32(y), screeny); + int yccmask = _mm_movemask_epi8(ycc); + int edge0p, edge0n, edge1p, edge1n; + int nexty; + float w, wslope; + float clip0; + if (numpoints == 4) + { + switch(yccmask) + { + default: + case 0xFFFF: /*0000*/ y = endy; continue; + case 0xFFF0: /*1000*/ edge0p = 3;edge0n = 0;edge1p = 1;edge1n = 0;break; + case 0xFF0F: /*0100*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 1;break; + case 0xFF00: /*1100*/ edge0p = 3;edge0n = 0;edge1p = 2;edge1n = 1;break; + case 0xF0FF: /*0010*/ edge0p = 1;edge0n = 2;edge1p = 3;edge1n = 2;break; + case 0xF0F0: /*1010*/ edge0p = 1;edge0n = 2;edge1p = 3;edge1n = 2;break; // concave - nonsense + case 0xF00F: /*0110*/ edge0p = 0;edge0n = 1;edge1p = 3;edge1n = 2;break; + case 0xF000: /*1110*/ edge0p = 3;edge0n = 0;edge1p = 3;edge1n = 2;break; + case 0x0FFF: /*0001*/ edge0p = 2;edge0n = 3;edge1p = 0;edge1n = 3;break; + case 0x0FF0: /*1001*/ edge0p = 2;edge0n = 3;edge1p = 1;edge1n = 0;break; + case 0x0F0F: /*0101*/ edge0p = 2;edge0n = 3;edge1p = 2;edge1n = 1;break; // concave - nonsense + case 0x0F00: /*1101*/ edge0p = 2;edge0n = 3;edge1p = 2;edge1n = 1;break; + case 0x00FF: /*0011*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 3;break; + case 0x00F0: /*1011*/ edge0p = 1;edge0n = 2;edge1p = 1;edge1n = 0;break; + case 0x000F: /*0111*/ edge0p = 0;edge0n = 1;edge1p = 0;edge1n = 3;break; + case 0x0000: /*1111*/ y++; continue; + } + } + else + { + switch(yccmask) + { + default: + case 0xFFFF: /*000*/ y = endy; continue; + case 0xFFF0: /*100*/ edge0p = 2;edge0n = 0;edge1p = 1;edge1n = 0;break; + case 0xFF0F: /*010*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 1;break; + case 0xFF00: /*110*/ edge0p = 2;edge0n = 0;edge1p = 2;edge1n = 1;break; + case 0x00FF: /*001*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 2;break; + case 0x00F0: /*101*/ edge0p = 1;edge0n = 2;edge1p = 1;edge1n = 0;break; + case 0x000F: /*011*/ edge0p = 0;edge0n = 1;edge1p = 0;edge1n = 2;break; + case 0x0000: /*111*/ y++; continue; + } + } + ycc = _mm_max_epi16(_mm_srli_epi16(ycc, 1), screeny); + ycc = _mm_min_epi16(ycc, _mm_shuffle_epi32(ycc, _MM_SHUFFLE(1, 0, 3, 2))); + ycc = _mm_min_epi16(ycc, _mm_shuffle_epi32(ycc, _MM_SHUFFLE(2, 3, 0, 1))); + nexty = _mm_extract_epi16(ycc, 0); + if (nexty >= bandy) nexty = bandy-1; + xslope = _mm_sub_ps(_mm_movelh_ps(screen[edge0n], screen[edge1n]), _mm_movelh_ps(screen[edge0p], screen[edge1p])); + xslope = _mm_div_ps(xslope, _mm_shuffle_ps(xslope, xslope, _MM_SHUFFLE(3, 3, 1, 1))); + xcoords = _mm_add_ps(_mm_movelh_ps(screen[edge0p], screen[edge1p]), + _mm_mul_ps(xslope, _mm_sub_ps(_mm_set1_ps(y), _mm_shuffle_ps(screen[edge0p], screen[edge1p], _MM_SHUFFLE(1, 1, 1, 1))))); + xcoords = _mm_add_ps(xcoords, _mm_set1_ps(0.5f)); + if (_mm_ucomigt_ss(xcoords, _mm_shuffle_ps(xcoords, xcoords, _MM_SHUFFLE(1, 0, 3, 2)))) + { + xcoords = _mm_shuffle_ps(xcoords, xcoords, _MM_SHUFFLE(1, 0, 3, 2)); + xslope = _mm_shuffle_ps(xslope, xslope, _MM_SHUFFLE(1, 0, 3, 2)); + } + clip0 = clip0origin + (y+0.5f)*clip0slope + 0.5f; + for(; y <= nexty; y++, xcoords = _mm_add_ps(xcoords, xslope), clip0 += clip0slope) + { + int startx, endx, offset; + startx = _mm_cvtss_si32(xcoords); + endx = _mm_cvtss_si32(_mm_movehl_ps(xcoords, xcoords)); + if (startx < minx) startx = minx; + if (endx > maxx) endx = maxx; + if (startx >= endx) continue; - dpsoftrast.trianglepool.freetriangle = dpsoftrast.trianglepool.freetriangle < DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1 ? dpsoftrast.trianglepool.freetriangle + 1 : 0; - dpsoftrast.trianglepool.usedtriangles++; + if (clip0dir) + { + if (clip0dir > 0) + { + if (startx < clip0) + { + if(endx <= clip0) continue; + startx = (int)clip0; + } + } + else if (endx > clip0) + { + if(startx >= clip0) continue; + endx = (int)clip0; + } + } + + for (offset = startx; offset < endx;offset += DPSOFTRAST_DRAW_MAXSPANLENGTH) + { + DPSOFTRAST_State_Span *span = &thread->spans[thread->numspans]; + span->triangle = thread->numtriangles; + span->x = offset; + span->y = y; + span->startx = 0; + span->endx = min(endx - offset, DPSOFTRAST_DRAW_MAXSPANLENGTH); + if (span->startx >= span->endx) + continue; + wslope = triangle->w[0]; + w = triangle->w[2] + span->x*wslope + span->y*triangle->w[1]; + span->depthslope = (int)(wslope*DPSOFTRAST_DEPTHSCALE); + span->depthbase = (int)(w*DPSOFTRAST_DEPTHSCALE - DPSOFTRAST_DEPTHOFFSET*(thread->polygonoffset[1] + fabs(wslope)*thread->polygonoffset[0])); + if (++thread->numspans >= DPSOFTRAST_DRAW_MAXSPANS) + DPSOFTRAST_Draw_ProcessSpans(thread); + } + } + } - numqueued++; - if (numqueued >= DPSOFTRAST_DRAW_FLUSHPROCESSTRIANGLES) + if (++thread->numtriangles >= DPSOFTRAST_DRAW_MAXTRIANGLES) { - MEMORY_BARRIER; - dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle; - -#ifdef USETHREADS - SDL_LockMutex(dpsoftrast.trianglemutex); - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_UnlockMutex(dpsoftrast.trianglemutex); -#else - DPSOFTRAST_Draw_FlushThreads(); -#endif - numqueued = 0; + DPSOFTRAST_Draw_ProcessSpans(thread); + thread->numtriangles = 0; } } - if (numqueued > 0) + + if (!ATOMIC_DECREMENT(command->refcount)) { - MEMORY_BARRIER; - dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle; + if (command->commandsize <= DPSOFTRAST_ALIGNCOMMAND(sizeof(DPSOFTRAST_Command_Draw))) + MM_FREE(command->arrays); + } -#ifdef USETHREADS - SDL_LockMutex(dpsoftrast.trianglemutex); - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_UnlockMutex(dpsoftrast.trianglemutex); -#else - DPSOFTRAST_Draw_FlushThreads(); -#endif + if (thread->numspans > 0 || thread->numtriangles > 0) + { + DPSOFTRAST_Draw_ProcessSpans(thread); + thread->numtriangles = 0; } #endif } -void DPSOFTRAST_DrawTriangles(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s) +static DPSOFTRAST_Command_Draw *DPSOFTRAST_Draw_AllocateDrawCommand(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s) { int i; - int lastarray = DPSOFTRAST_ARRAY_POSITION; - unsigned char arraymask[DPSOFTRAST_ARRAY_TOTAL]; - memset(arraymask, false, sizeof(arraymask)); - arraymask[DPSOFTRAST_ARRAY_POSITION] = true; + int j; + int commandsize = DPSOFTRAST_ALIGNCOMMAND(sizeof(DPSOFTRAST_Command_Draw)); + int datasize = 2*numvertices*sizeof(float[4]); + DPSOFTRAST_Command_Draw *command; + unsigned char *data; + for (i = 0; i < DPSOFTRAST_ARRAY_TOTAL; i++) + { + j = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].arrays[i]; + if (j >= DPSOFTRAST_ARRAY_TOTAL) + break; + datasize += numvertices*sizeof(float[4]); + } + if (element3s) + datasize += numtriangles*sizeof(unsigned short[3]); + else if (element3i) + datasize += numtriangles*sizeof(int[3]); + datasize = DPSOFTRAST_ALIGNCOMMAND(datasize); + if (commandsize + datasize > DPSOFTRAST_DRAW_MAXCOMMANDSIZE) + { + command = (DPSOFTRAST_Command_Draw *) DPSOFTRAST_AllocateCommand(DPSOFTRAST_OPCODE_Draw, commandsize); + data = (unsigned char *)MM_CALLOC(datasize, 1); + } + else + { + command = (DPSOFTRAST_Command_Draw *) DPSOFTRAST_AllocateCommand(DPSOFTRAST_OPCODE_Draw, commandsize + datasize); + data = (unsigned char *)command + commandsize; + } + command->firstvertex = firstvertex; + command->numvertices = numvertices; + command->numtriangles = numtriangles; + command->arrays = (float *)data; + memset(dpsoftrast.post_array4f, 0, sizeof(dpsoftrast.post_array4f)); + dpsoftrast.firstvertex = firstvertex; + dpsoftrast.numvertices = numvertices; + dpsoftrast.screencoord4f = (float *)data; + data += numvertices*sizeof(float[4]); + dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION] = (float *)data; + data += numvertices*sizeof(float[4]); for (i = 0; i < DPSOFTRAST_ARRAY_TOTAL; i++) { - int arrayindex = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].arrays[i]; - if (arrayindex >= DPSOFTRAST_ARRAY_TOTAL) + j = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].arrays[i]; + if (j >= DPSOFTRAST_ARRAY_TOTAL) break; - switch (arrayindex) + dpsoftrast.post_array4f[j] = (float *)data; + data += numvertices*sizeof(float[4]); + } + command->element3i = NULL; + command->element3s = NULL; + if (element3s) + { + command->element3s = (unsigned short *)data; + memcpy(command->element3s, element3s, numtriangles*sizeof(unsigned short[3])); + } + else if (element3i) + { + command->element3i = (int *)data; + memcpy(command->element3i, element3i, numtriangles*sizeof(int[3])); + } + return command; +} + +void DPSOFTRAST_DrawTriangles(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s) +{ + DPSOFTRAST_Command_Draw *command = DPSOFTRAST_Draw_AllocateDrawCommand(firstvertex, numvertices, numtriangles, element3i, element3s); + DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].Vertex(); + command->starty = bound(0, dpsoftrast.drawstarty, dpsoftrast.fb_height); + command->endy = bound(0, dpsoftrast.drawendy, dpsoftrast.fb_height); + if (command->starty >= command->endy) + { + if (command->commandsize <= DPSOFTRAST_ALIGNCOMMAND(sizeof(DPSOFTRAST_Command_Draw))) + MM_FREE(command->arrays); + DPSOFTRAST_UndoCommand(command->commandsize); + return; + } + command->clipped = dpsoftrast.drawclipped; + command->refcount = dpsoftrast.numthreads; + + if (dpsoftrast.usethreads) + { + int i; + DPSOFTRAST_Draw_SyncCommands(); + for (i = 0; i < dpsoftrast.numthreads; i++) { - case DPSOFTRAST_ARRAY_POSITION: - case DPSOFTRAST_ARRAY_COLOR: - break; - default: - if (dpsoftrast.pointer_texcoordf[arrayindex-DPSOFTRAST_ARRAY_TEXCOORD0] == NULL) - continue; - break; + DPSOFTRAST_State_Thread *thread = &dpsoftrast.threads[i]; + if (((command->starty < thread->maxy1 && command->endy > thread->miny1) || (command->starty < thread->maxy2 && command->endy > thread->miny2)) && thread->starving) + Thread_CondSignal(thread->drawcond); } - arraymask[arrayindex] = true; - if (arrayindex > lastarray) - lastarray = arrayindex; } - DPSOFTRAST_Draw_LoadVertices(firstvertex, numvertices, arraymask[DPSOFTRAST_ARRAY_COLOR]); - DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].Vertex(); -// DPSOFTRAST_Draw_ProjectVertices(dpsoftrast.screencoord4f, dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_POSITION], numvertices); - DPSOFTRAST_Draw_ProcessTriangles(firstvertex, numtriangles, element3i, element3s, arraymask, lastarray+1); + else + { + DPSOFTRAST_Draw_FlushThreads(); + } } -void DPSOFTRAST_Flush(void) +DEFCOMMAND(23, SetRenderTargets, int width; int height;); +static void DPSOFTRAST_Interpret_SetRenderTargets(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_Command_SetRenderTargets *command) +{ + thread->validate |= DPSOFTRAST_VALIDATE_FB; +} +void DPSOFTRAST_SetRenderTargets(int width, int height, unsigned int *depthpixels, unsigned int *colorpixels0, unsigned int *colorpixels1, unsigned int *colorpixels2, unsigned int *colorpixels3) { + DPSOFTRAST_Command_SetRenderTargets *command; + if (width != dpsoftrast.fb_width || height != dpsoftrast.fb_height || depthpixels != dpsoftrast.fb_depthpixels || + colorpixels0 != dpsoftrast.fb_colorpixels[0] || colorpixels1 != dpsoftrast.fb_colorpixels[1] || + colorpixels2 != dpsoftrast.fb_colorpixels[2] || colorpixels3 != dpsoftrast.fb_colorpixels[3]) + DPSOFTRAST_Flush(); + dpsoftrast.fb_width = width; + dpsoftrast.fb_height = height; + dpsoftrast.fb_depthpixels = depthpixels; + dpsoftrast.fb_colorpixels[0] = colorpixels0; + dpsoftrast.fb_colorpixels[1] = colorpixels1; + dpsoftrast.fb_colorpixels[2] = colorpixels2; + dpsoftrast.fb_colorpixels[3] = colorpixels3; + DPSOFTRAST_RecalcViewport(dpsoftrast.viewport, dpsoftrast.fb_viewportcenter, dpsoftrast.fb_viewportscale); + command = DPSOFTRAST_ALLOCATECOMMAND(SetRenderTargets); + command->width = width; + command->height = height; +} + +static void DPSOFTRAST_Draw_InterpretCommands(DPSOFTRAST_State_Thread *thread, int endoffset) +{ + int commandoffset = thread->commandoffset; + while (commandoffset != endoffset) + { + DPSOFTRAST_Command *command = (DPSOFTRAST_Command *)&dpsoftrast.commandpool.commands[commandoffset]; + switch (command->opcode) + { +#define INTERPCOMMAND(name) \ + case DPSOFTRAST_OPCODE_##name : \ + DPSOFTRAST_Interpret_##name (thread, (DPSOFTRAST_Command_##name *)command); \ + commandoffset += DPSOFTRAST_ALIGNCOMMAND(sizeof( DPSOFTRAST_Command_##name )); \ + if (commandoffset >= DPSOFTRAST_DRAW_MAXCOMMANDPOOL) \ + commandoffset = 0; \ + break; + INTERPCOMMAND(Viewport) + INTERPCOMMAND(ClearColor) + INTERPCOMMAND(ClearDepth) + INTERPCOMMAND(ColorMask) + INTERPCOMMAND(DepthTest) + INTERPCOMMAND(ScissorTest) + INTERPCOMMAND(Scissor) + INTERPCOMMAND(BlendFunc) + INTERPCOMMAND(BlendSubtract) + INTERPCOMMAND(DepthMask) + INTERPCOMMAND(DepthFunc) + INTERPCOMMAND(DepthRange) + INTERPCOMMAND(PolygonOffset) + INTERPCOMMAND(CullFace) + INTERPCOMMAND(SetTexture) + INTERPCOMMAND(SetShader) + INTERPCOMMAND(Uniform4f) + INTERPCOMMAND(UniformMatrix4f) + INTERPCOMMAND(Uniform1i) + INTERPCOMMAND(SetRenderTargets) + INTERPCOMMAND(ClipPlane) + + case DPSOFTRAST_OPCODE_Draw: + DPSOFTRAST_Interpret_Draw(thread, (DPSOFTRAST_Command_Draw *)command); + commandoffset += command->commandsize; + if (commandoffset >= DPSOFTRAST_DRAW_MAXCOMMANDPOOL) + commandoffset = 0; + thread->commandoffset = commandoffset; + break; + + case DPSOFTRAST_OPCODE_Reset: + commandoffset = 0; + break; + } + } + thread->commandoffset = commandoffset; +} + +static int DPSOFTRAST_Draw_Thread(void *data) +{ + DPSOFTRAST_State_Thread *thread = (DPSOFTRAST_State_Thread *)data; + while(thread->index >= 0) + { + if (thread->commandoffset != dpsoftrast.drawcommand) + { + DPSOFTRAST_Draw_InterpretCommands(thread, dpsoftrast.drawcommand); + } + else + { + Thread_LockMutex(thread->drawmutex); + if (thread->commandoffset == dpsoftrast.drawcommand && thread->index >= 0) + { + if (thread->waiting) Thread_CondSignal(thread->waitcond); + thread->starving = true; + Thread_CondWait(thread->drawcond, thread->drawmutex); + thread->starving = false; + } + Thread_UnlockMutex(thread->drawmutex); + } + } + return 0; +} + +static void DPSOFTRAST_Draw_FlushThreads(void) +{ + DPSOFTRAST_State_Thread *thread; + int i; DPSOFTRAST_Draw_SyncCommands(); + if (dpsoftrast.usethreads) + { + for (i = 0; i < dpsoftrast.numthreads; i++) + { + thread = &dpsoftrast.threads[i]; + if (thread->commandoffset != dpsoftrast.drawcommand) + { + Thread_LockMutex(thread->drawmutex); + if (thread->commandoffset != dpsoftrast.drawcommand && thread->starving) + Thread_CondSignal(thread->drawcond); + Thread_UnlockMutex(thread->drawmutex); + } + } + for (i = 0; i < dpsoftrast.numthreads; i++) + { + thread = &dpsoftrast.threads[i]; + if (thread->commandoffset != dpsoftrast.drawcommand) + { + Thread_LockMutex(thread->drawmutex); + if (thread->commandoffset != dpsoftrast.drawcommand) + { + thread->waiting = true; + Thread_CondWait(thread->waitcond, thread->drawmutex); + thread->waiting = false; + } + Thread_UnlockMutex(thread->drawmutex); + } + } + } + else + { + for (i = 0; i < dpsoftrast.numthreads; i++) + { + thread = &dpsoftrast.threads[i]; + if (thread->commandoffset != dpsoftrast.drawcommand) + DPSOFTRAST_Draw_InterpretCommands(thread, dpsoftrast.drawcommand); + } + } + dpsoftrast.commandpool.usedcommands = 0; +} + +void DPSOFTRAST_Flush(void) +{ DPSOFTRAST_Draw_FlushThreads(); } -void DPSOFTRAST_Init(int width, int height, int numthreads, unsigned int *colorpixels, unsigned int *depthpixels) +void DPSOFTRAST_Finish(void) +{ + DPSOFTRAST_Flush(); +} + +int DPSOFTRAST_Init(int width, int height, int numthreads, int interlace, unsigned int *colorpixels, unsigned int *depthpixels) { int i; union @@ -4747,30 +5583,28 @@ void DPSOFTRAST_Init(int width, int height, int numthreads, unsigned int *colorp dpsoftrast.fb_colorpixels[1] = NULL; dpsoftrast.fb_colorpixels[1] = NULL; dpsoftrast.fb_colorpixels[1] = NULL; - dpsoftrast.texture_firstfree = 1; - dpsoftrast.texture_end = 1; - dpsoftrast.texture_max = 0; dpsoftrast.viewport[0] = 0; dpsoftrast.viewport[1] = 0; dpsoftrast.viewport[2] = dpsoftrast.fb_width; dpsoftrast.viewport[3] = dpsoftrast.fb_height; + DPSOFTRAST_RecalcViewport(dpsoftrast.viewport, dpsoftrast.fb_viewportcenter, dpsoftrast.fb_viewportscale); + dpsoftrast.texture_firstfree = 1; + dpsoftrast.texture_end = 1; + dpsoftrast.texture_max = 0; dpsoftrast.color[0] = 1; dpsoftrast.color[1] = 1; dpsoftrast.color[2] = 1; dpsoftrast.color[3] = 1; - dpsoftrast.cullface = GL_BACK; -#ifdef USETHREADS - dpsoftrast.numthreads = bound(1, numthreads, 64); - dpsoftrast.trianglemutex = SDL_CreateMutex(); - dpsoftrast.trianglecond = SDL_CreateCond(); -#else - dpsoftrast.numthreads = 1; -#endif + dpsoftrast.usethreads = numthreads > 0 && Thread_HasThreads(); + dpsoftrast.interlace = dpsoftrast.usethreads ? bound(0, interlace, 1) : 0; + dpsoftrast.numthreads = dpsoftrast.usethreads ? bound(1, numthreads, 64) : 1; dpsoftrast.threads = (DPSOFTRAST_State_Thread *)MM_CALLOC(dpsoftrast.numthreads, sizeof(DPSOFTRAST_State_Thread)); for (i = 0; i < dpsoftrast.numthreads; i++) { DPSOFTRAST_State_Thread *thread = &dpsoftrast.threads[i]; thread->index = i; + thread->cullface = GL_BACK; + thread->colormask[0] = 1; thread->colormask[1] = 1; thread->colormask[2] = 1; thread->colormask[3] = 1; @@ -4780,9 +5614,10 @@ void DPSOFTRAST_Init(int width, int height, int numthreads, unsigned int *colorp thread->depthtest = true; thread->depthfunc = GL_LEQUAL; thread->scissortest = false; - thread->alphatest = false; - thread->alphafunc = GL_GREATER; - thread->alphavalue = 0.5f; + thread->viewport[0] = 0; + thread->viewport[1] = 0; + thread->viewport[2] = dpsoftrast.fb_width; + thread->viewport[3] = dpsoftrast.fb_height; thread->scissor[0] = 0; thread->scissor[1] = 0; thread->scissor[2] = dpsoftrast.fb_width; @@ -4791,48 +5626,50 @@ void DPSOFTRAST_Init(int width, int height, int numthreads, unsigned int *colorp thread->depthrange[1] = 1; thread->polygonoffset[0] = 0; thread->polygonoffset[1] = 0; - + thread->clipplane[0] = 0; + thread->clipplane[1] = 0; + thread->clipplane[2] = 0; + thread->clipplane[3] = 1; + thread->numspans = 0; - thread->triangleoffset = 0; + thread->numtriangles = 0; thread->commandoffset = 0; thread->waiting = false; -#ifdef USETHREADS - thread->waitcond = SDL_CreateCond(); -#endif - + thread->starving = false; + thread->validate = -1; DPSOFTRAST_Validate(thread, -1); -#ifdef USETHREADS - thread->thread = SDL_CreateThread(DPSOFTRAST_Draw_Thread, thread); -#endif + + if (dpsoftrast.usethreads) + { + thread->waitcond = Thread_CreateCond(); + thread->drawcond = Thread_CreateCond(); + thread->drawmutex = Thread_CreateMutex(); + thread->thread = Thread_CreateThread(DPSOFTRAST_Draw_Thread, thread); + } } + return 0; } void DPSOFTRAST_Shutdown(void) { int i; -#ifdef USETHREADS - if(dpsoftrast.numthreads > 0) + if (dpsoftrast.usethreads && dpsoftrast.numthreads > 0) { DPSOFTRAST_State_Thread *thread; - SDL_LockMutex(dpsoftrast.trianglemutex); for (i = 0; i < dpsoftrast.numthreads; i++) { thread = &dpsoftrast.threads[i]; + Thread_LockMutex(thread->drawmutex); thread->index = -1; + Thread_CondSignal(thread->drawcond); + Thread_UnlockMutex(thread->drawmutex); + Thread_WaitThread(thread->thread, 0); + Thread_DestroyCond(thread->waitcond); + Thread_DestroyCond(thread->drawcond); + Thread_DestroyMutex(thread->drawmutex); } - SDL_CondBroadcast(dpsoftrast.trianglecond); - SDL_UnlockMutex(dpsoftrast.trianglemutex); - for (i = 0; i < dpsoftrast.numthreads; i++) - { - thread = &dpsoftrast.threads[i]; - SDL_WaitThread(thread->thread, NULL); - SDL_DestroyCond(thread->waitcond); - } - SDL_DestroyMutex(dpsoftrast.trianglemutex); - SDL_DestroyCond(dpsoftrast.trianglecond); } -#endif for (i = 0;i < dpsoftrast.texture_end;i++) if (dpsoftrast.texture[i].bytes) MM_FREE(dpsoftrast.texture[i].bytes);