-
-#include <memory.h>
-#include "dpsoftrast.h"
#include <stdio.h>
+#include <string.h>
+#define _USE_MATH_DEFINES
#include <math.h>
+#include "quakedef.h"
+#include "dpsoftrast.h"
+
+#ifdef USE_SDL
+#define USE_THREADS
+#endif
+
+#ifdef USE_THREADS
+#include <SDL.h>
+#include <SDL_thread.h>
+#endif
-#undef true
-#undef false
#ifndef __cplusplus
-typedef enum bool {false, true} bool;
+typedef qboolean bool;
#endif
-#if defined(__GNUC__) || (defined(_MSC_VER) && _MSC_VER >= 1400)
-#define RESTRICT __restrict
+#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
-#define RESTRICT
+#define ALIGN(var) var
+#define ATOMIC(var) var
+#define MEMORY_BARRIER ((void)0)
#endif
-#define GL_NONE 0
-#define GL_FRONT_LEFT 0x0400
-#define GL_FRONT_RIGHT 0x0401
-#define GL_BACK_LEFT 0x0402
-#define GL_BACK_RIGHT 0x0403
-#define GL_FRONT 0x0404
-#define GL_BACK 0x0405
-#define GL_LEFT 0x0406
-#define GL_RIGHT 0x0407
-#define GL_FRONT_AND_BACK 0x0408
-#define GL_AUX0 0x0409
-#define GL_AUX1 0x040A
-#define GL_AUX2 0x040B
-#define GL_AUX3 0x040C
-
-#define GL_NEVER 0x0200
-#define GL_LESS 0x0201
-#define GL_EQUAL 0x0202
-#define GL_LEQUAL 0x0203
-#define GL_GREATER 0x0204
-#define GL_NOTEQUAL 0x0205
-#define GL_GEQUAL 0x0206
-#define GL_ALWAYS 0x0207
-
-#define GL_ZERO 0x0
-#define GL_ONE 0x1
-#define GL_SRC_COLOR 0x0300
-#define GL_ONE_MINUS_SRC_COLOR 0x0301
-#define GL_DST_COLOR 0x0306
-#define GL_ONE_MINUS_DST_COLOR 0x0307
-#define GL_SRC_ALPHA 0x0302
-#define GL_ONE_MINUS_SRC_ALPHA 0x0303
-#define GL_DST_ALPHA 0x0304
-#define GL_ONE_MINUS_DST_ALPHA 0x0305
-#define GL_SRC_ALPHA_SATURATE 0x0308
-#define GL_CONSTANT_COLOR 0x8001
-#define GL_ONE_MINUS_CONSTANT_COLOR 0x8002
-#define GL_CONSTANT_ALPHA 0x8003
-#define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004
+#if !defined(USE_THREADS) || !defined(SSE2_PRESENT)
+#undef MEMORY_BARRIER
+#define MEMORY_BARRIER ((void)0)
+#endif
+
+#ifdef SSE2_PRESENT
+#include <emmintrin.h>
+
+#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);
+ return ptr;
+}
+
+#define MM_FREE _mm_free
+#else
+#define MM_MALLOC(size) malloc(size)
+#define MM_CALLOC(nmemb, size) calloc(nmemb, size)
+#define MM_FREE free
+#endif
typedef enum DPSOFTRAST_ARRAY_e
{
}
DPSOFTRAST_Texture;
-typedef struct DPSOFTRAST_State_User_s
+#define COMMAND_SIZE ALIGN_SIZE
+#define COMMAND_ALIGN(var) ALIGN(var)
+
+typedef COMMAND_ALIGN(struct DPSOFTRAST_Command_s
{
- int colormask[4];
- int blendfunc[2];
- int blendsubtract;
- int depthmask;
- int depthtest;
- int depthfunc;
- int scissortest;
- int cullface;
- int alphatest;
- int alphafunc;
- float alphavalue;
- int scissor[4];
- int viewport[4];
- float depthrange[2];
- float polygonoffset[2];
- float color[4];
+ int opcode;
+}
+DPSOFTRAST_Command);
+
+enum { DPSOFTRAST_OPCODE_Reset = 0 };
+
+#define DEFCOMMAND(opcodeval, name, fields) \
+ enum { DPSOFTRAST_OPCODE_##name = opcodeval }; \
+ typedef COMMAND_ALIGN(struct DPSOFTRAST_Command_##name##_s \
+ { \
+ int opcode; \
+ fields \
+ } DPSOFTRAST_Command_##name );
+
+#define DPSOFTRAST_DRAW_MAXCOMMANDPOOL 2097152
+
+typedef ATOMIC(struct DPSOFTRAST_State_Command_Pool_s
+{
+ int freecommand;
+ int usedcommands;
+ ATOMIC(unsigned char commands[DPSOFTRAST_DRAW_MAXCOMMANDPOOL]);
}
-DPSOFTRAST_State_User;
+DPSOFTRAST_State_Command_Pool);
-#define DPSOFTRAST_MAXSUBSPAN 16
+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);
+
+#define DPSOFTRAST_CALCATTRIB(triangle, span, data, slope, arrayindex) { \
+ slope = _mm_load_ps((triangle)->attribs[arrayindex][0]); \
+ data = _mm_add_ps(_mm_load_ps((triangle)->attribs[arrayindex][2]), \
+ _mm_add_ps(_mm_mul_ps(_mm_set1_ps((span)->x), slope), \
+ _mm_mul_ps(_mm_set1_ps((span)->y), _mm_load_ps((triangle)->attribs[arrayindex][1])))); \
+}
+#define DPSOFTRAST_CALCATTRIB4F(triangle, span, data, slope, arrayindex) { \
+ slope[0] = (triangle)->attribs[arrayindex][0][0]; \
+ slope[1] = (triangle)->attribs[arrayindex][0][1]; \
+ slope[2] = (triangle)->attribs[arrayindex][0][2]; \
+ slope[3] = (triangle)->attribs[arrayindex][0][3]; \
+ data[0] = (triangle)->attribs[arrayindex][2][0] + (span->x)*slope[0] + (span->y)*(triangle)->attribs[arrayindex][1][0]; \
+ data[1] = (triangle)->attribs[arrayindex][2][1] + (span->x)*slope[1] + (span->y)*(triangle)->attribs[arrayindex][1][1]; \
+ data[2] = (triangle)->attribs[arrayindex][2][2] + (span->x)*slope[2] + (span->y)*(triangle)->attribs[arrayindex][1][2]; \
+ data[3] = (triangle)->attribs[arrayindex][2][3] + (span->x)*slope[3] + (span->y)*(triangle)->attribs[arrayindex][1][3]; \
+}
+
+#define DPSOFTRAST_DRAW_MAXSUBSPAN 16
-typedef struct DPSOFTRAST_State_Draw_Span_s
+typedef ALIGN(struct DPSOFTRAST_State_Span_s
{
- int start; // pixel index
+ 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)
- int mip; // which mipmap of the texture(s) to use on this
unsigned char *pixelmask; // true for pixels that passed depth test, false for others
- // [0][n][] is start interpolant values (projected)
- // [1][n][] is end interpolant values (projected)
- // [0][DPSOFTRAST_ARRAY_TOTAL][] is start screencoord4f
- // [1][DPSOFTRAST_ARRAY_TOTAL][] is end screencoord4f
- // NOTE: screencoord4f[3] is W (basically 1/Z), useful for depthbuffer
- float data[2][DPSOFTRAST_ARRAY_TOTAL+1][4];
}
-DPSOFTRAST_State_Draw_Span;
+DPSOFTRAST_State_Span);
-#define DPSOFTRAST_DRAW_MAXSPANQUEUE 1024
+#define DPSOFTRAST_DRAW_MAXSPANS 1024
-typedef struct DPSOFTRAST_State_Draw_s
-{
- int numvertices;
- int maxvertices;
- float *in_array4f[DPSOFTRAST_ARRAY_TOTAL];
- float *post_array4f[DPSOFTRAST_ARRAY_TOTAL];
- float *screencoord4f;
+#define DPSOFTRAST_DRAW_MAXTRIANGLEPOOL 4096
+#define DPSOFTRAST_DRAW_FLUSHPROCESSTRIANGLES 64
- // spans are queued in this structure for dispatch to the pixel shader,
- // partly to improve cache locality, partly for batching purposes, spans
- // are flushed before DrawTriangles returns to caller
- int numspans;
- DPSOFTRAST_State_Draw_Span spanqueue[DPSOFTRAST_DRAW_MAXSPANQUEUE];
+typedef ATOMIC(struct DPSOFTRAST_State_Triangle_Pool_s
+{
+ int freetriangle;
+ int usedtriangles;
+ ATOMIC(DPSOFTRAST_State_Triangle triangles[DPSOFTRAST_DRAW_MAXTRIANGLEPOOL]);
}
-DPSOFTRAST_State_Draw;
+DPSOFTRAST_State_Triangle_Pool);
#define DPSOFTRAST_VALIDATE_FB 1
#define DPSOFTRAST_VALIDATE_DEPTHFUNC 2
}
DPSOFTRAST_BLENDMODE;
-typedef struct DPSOFTRAST_State_s
+typedef ATOMIC(struct DPSOFTRAST_State_Thread_s
{
+#ifdef USE_THREADS
+ SDL_Thread *thread;
+#endif
+ int index;
+
+ int colormask[4];
+ int blendfunc[2];
+ int blendsubtract;
+ int depthmask;
+ int depthtest;
+ int depthfunc;
+ int scissortest;
+ int alphatest;
+ int alphafunc;
+ float alphavalue;
+ int scissor[4];
+ int viewport[4];
+ float depthrange[2];
+ float polygonoffset[2];
+
+ int shader_mode;
+ int shader_permutation;
+
+ DPSOFTRAST_Texture *texbound[DPSOFTRAST_MAXTEXTUREUNITS];
+
+ ALIGN(float uniform4f[DPSOFTRAST_UNIFORM_TOTAL*4]);
+ int uniform1i[DPSOFTRAST_UNIFORM_TOTAL];
+
// DPSOFTRAST_VALIDATE_ flags
int validate;
+ // derived values (DPSOFTRAST_VALIDATE_FB)
int fb_colormask;
+ int fb_clearscissor[4];
+
+ // derived values (DPSOFTRAST_VALIDATE_DEPTHFUNC)
+ int fb_depthfunc;
+
+ // derived values (DPSOFTRAST_VALIDATE_BLENDFUNC)
+ int fb_blendmode;
+
+ ATOMIC(int commandoffset);
+ int triangleoffset;
+
+ bool waiting;
+#ifdef USE_THREADS
+ SDL_cond *waitcond;
+#endif
+
+ int numspans;
+ DPSOFTRAST_State_Span spans[DPSOFTRAST_DRAW_MAXSPANS];
+}
+DPSOFTRAST_State_Thread);
+
+typedef ATOMIC(struct DPSOFTRAST_State_s
+{
int fb_width;
int fb_height;
unsigned int *fb_depthpixels;
unsigned int *fb_colorpixels[4];
+ int viewport[4];
+ ALIGN(float fb_viewportcenter[4]);
+ ALIGN(float fb_viewportscale[4]);
+
+ float color[4];
+ 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;
int components_texcoord[DPSOFTRAST_MAXTEXCOORDARRAYS];
DPSOFTRAST_Texture *texbound[DPSOFTRAST_MAXTEXTUREUNITS];
+ int numvertices;
+ int maxvertices;
+ float *in_array4f[DPSOFTRAST_ARRAY_TOTAL];
+ float *post_array4f[DPSOFTRAST_ARRAY_TOTAL];
+ float *screencoord4f;
+
int shader_mode;
int shader_permutation;
- float uniform4f[DPSOFTRAST_UNIFORM_TOTAL*4];
- int uniform1i[DPSOFTRAST_UNIFORM_TOTAL];
-
- // derived values (DPSOFTRAST_VALIDATE_FB)
- int fb_clearscissor[4];
- int fb_viewport[4];
- int fb_viewportscissor[4];
- float fb_viewportcenter[2];
- float fb_viewportscale[2];
-
- // derived values (DPSOFTRAST_VALIDATE_DEPTHFUNC)
- int fb_depthfunc;
-
- // derived values (DPSOFTRAST_VALIDATE_BLENDFUNC)
- int fb_blendmode;
int texture_max;
int texture_end;
// error reporting
const char *errorstring;
- DPSOFTRAST_State_User user;
+ int numthreads;
+ DPSOFTRAST_State_Thread *threads;
+#ifdef USE_THREADS
+ SDL_mutex *trianglemutex;
+ SDL_cond *trianglecond;
+#endif
+
+ ATOMIC(int drawtriangle);
- DPSOFTRAST_State_Draw draw;
+ DPSOFTRAST_State_Command_Pool commandpool;
+ DPSOFTRAST_State_Triangle_Pool trianglepool;
}
-DPSOFTRAST_State;
+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(void)
+void DPSOFTRAST_RecalcFB(DPSOFTRAST_State_Thread *thread)
{
// calculate framebuffer scissor, viewport, viewport clipped by scissor,
// and viewport projection values
- int x1, x2, x3, x4, x5, x6;
- int y1, y2, y3, y4, y5, y6;
- x1 = dpsoftrast.user.scissor[0];
- x2 = dpsoftrast.user.scissor[0] + dpsoftrast.user.scissor[2];
- x3 = dpsoftrast.user.viewport[0];
- x4 = dpsoftrast.user.viewport[0] + dpsoftrast.user.viewport[2];
- y1 = dpsoftrast.fb_height - dpsoftrast.user.scissor[1] - dpsoftrast.user.scissor[3];
- y2 = dpsoftrast.fb_height - dpsoftrast.user.scissor[1];
- y3 = dpsoftrast.fb_height - dpsoftrast.user.viewport[1] - dpsoftrast.user.viewport[3];
- y4 = dpsoftrast.fb_height - dpsoftrast.user.viewport[1];
- if (!dpsoftrast.user.scissortest) {x1 = 0;y1 = 0;x2 = dpsoftrast.fb_width;y2 = dpsoftrast.fb_height;}
+ int x1, x2;
+ int y1, y2;
+ x1 = thread->scissor[0];
+ x2 = thread->scissor[0] + thread->scissor[2];
+ y1 = dpsoftrast.fb_height - thread->scissor[1] - thread->scissor[3];
+ y2 = dpsoftrast.fb_height - thread->scissor[1];
+ if (!thread->scissortest) {x1 = 0;y1 = 0;x2 = dpsoftrast.fb_width;y2 = dpsoftrast.fb_height;}
if (x1 < 0) x1 = 0;
if (x2 > dpsoftrast.fb_width) x2 = dpsoftrast.fb_width;
- if (x3 < 0) x1 = 0;
- if (x4 > dpsoftrast.fb_width) x4 = dpsoftrast.fb_width;
if (y1 < 0) y1 = 0;
if (y2 > dpsoftrast.fb_height) y2 = dpsoftrast.fb_height;
- if (y3 < 0) y1 = 0;
- if (y4 > dpsoftrast.fb_height) y4 = dpsoftrast.fb_height;
- x5 = x1;if (x5 < x3) x5 = x3;
- x6 = x2;if (x6 > x4) x4 = x4;
- y5 = y1;if (y5 < y3) y5 = y3;
- y6 = y2;if (y6 > y4) y6 = y4;
- dpsoftrast.fb_clearscissor[0] = x1;
- dpsoftrast.fb_clearscissor[1] = y1;
- dpsoftrast.fb_clearscissor[2] = x2 - x1;
- dpsoftrast.fb_clearscissor[3] = y2 - y1;
- dpsoftrast.fb_viewport[0] = x3;
- dpsoftrast.fb_viewport[1] = y3;
- dpsoftrast.fb_viewport[2] = x4 - x3;
- dpsoftrast.fb_viewport[3] = y4 - y3;
- dpsoftrast.fb_viewportscissor[0] = x5;
- dpsoftrast.fb_viewportscissor[1] = y5;
- dpsoftrast.fb_viewportscissor[2] = x6 - x5;
- dpsoftrast.fb_viewportscissor[3] = y6 - y5;
- dpsoftrast.fb_viewportcenter[0] = dpsoftrast.user.viewport[0] + 0.5f * dpsoftrast.user.viewport[2] - 0.5f;
- dpsoftrast.fb_viewportcenter[1] = dpsoftrast.fb_height - dpsoftrast.user.viewport[1] - 0.5f * dpsoftrast.user.viewport[3] - 0.5f;
- dpsoftrast.fb_viewportscale[0] = 0.5f * dpsoftrast.user.viewport[2];
- dpsoftrast.fb_viewportscale[1] = -0.5f * dpsoftrast.user.viewport[3];
-}
-
-void DPSOFTRAST_RecalcDepthFunc(void)
-{
- dpsoftrast.fb_depthfunc = dpsoftrast.user.depthtest ? dpsoftrast.user.depthfunc : GL_ALWAYS;
-}
-
-int blendmodetable[][4] =
-{
- {DPSOFTRAST_BLENDMODE_OPAQUE, GL_ONE, GL_ZERO, false},
- {DPSOFTRAST_BLENDMODE_ALPHA, GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, false},
- {DPSOFTRAST_BLENDMODE_ADDALPHA, GL_SRC_ALPHA, GL_ONE, false},
- {DPSOFTRAST_BLENDMODE_ADD, GL_ONE, GL_ONE, false},
- {DPSOFTRAST_BLENDMODE_INVMOD, GL_ZERO, GL_ONE_MINUS_SRC_COLOR, false},
- {DPSOFTRAST_BLENDMODE_MUL, GL_ZERO, GL_SRC_COLOR, false},
- {DPSOFTRAST_BLENDMODE_MUL, GL_DST_COLOR, GL_ZERO, false},
- {DPSOFTRAST_BLENDMODE_MUL2, GL_DST_COLOR, GL_SRC_COLOR, false},
- {DPSOFTRAST_BLENDMODE_PSEUDOALPHA, GL_ONE, GL_ONE_MINUS_SRC_ALPHA, false},
- {DPSOFTRAST_BLENDMODE_SUBALPHA, GL_SRC_COLOR, GL_ONE, true}
-};
+ thread->fb_clearscissor[0] = x1;
+ thread->fb_clearscissor[1] = y1;
+ thread->fb_clearscissor[2] = x2 - x1;
+ thread->fb_clearscissor[3] = y2 - y1;
+}
-void DPSOFTRAST_RecalcBlendFunc(void)
+void DPSOFTRAST_RecalcDepthFunc(DPSOFTRAST_State_Thread *thread)
{
- int i;
- dpsoftrast.fb_blendmode = DPSOFTRAST_BLENDMODE_OPAQUE;
- for (i = 0;i < (int)(sizeof(blendmodetable) / sizeof(blendmodetable[0]));i++)
+ thread->fb_depthfunc = thread->depthtest ? thread->depthfunc : GL_ALWAYS;
+}
+
+void DPSOFTRAST_RecalcBlendFunc(DPSOFTRAST_State_Thread *thread)
+{
+ if (thread->blendsubtract)
{
- if (dpsoftrast.user.blendfunc[0] == blendmodetable[i][1] && dpsoftrast.user.blendfunc[1] == blendmodetable[i][2] && dpsoftrast.user.blendsubtract == blendmodetable[i][3])
+ switch ((thread->blendfunc[0]<<16)|thread->blendfunc[1])
{
- dpsoftrast.fb_blendmode = blendmodetable[i][0];
- break;
+ #define BLENDFUNC(sfactor, dfactor, blendmode) \
+ case (sfactor<<16)|dfactor: thread->fb_blendmode = blendmode; break;
+ BLENDFUNC(GL_SRC_COLOR, GL_ONE, DPSOFTRAST_BLENDMODE_SUBALPHA)
+ default: thread->fb_blendmode = DPSOFTRAST_BLENDMODE_OPAQUE; break;
}
}
+ else
+ {
+ 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)
+ BLENDFUNC(GL_ONE, GL_ONE, DPSOFTRAST_BLENDMODE_ADD)
+ BLENDFUNC(GL_ZERO, GL_ONE_MINUS_SRC_COLOR, DPSOFTRAST_BLENDMODE_INVMOD)
+ BLENDFUNC(GL_ZERO, GL_SRC_COLOR, DPSOFTRAST_BLENDMODE_MUL)
+ 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)
+ default: thread->fb_blendmode = DPSOFTRAST_BLENDMODE_OPAQUE; break;
+ }
+ }
}
-#define DPSOFTRAST_ValidateQuick(f) ((dpsoftrast.validate & (f)) ? (DPSOFTRAST_Validate(f), 0) : 0)
+#define DPSOFTRAST_ValidateQuick(thread, f) ((thread->validate & (f)) ? (DPSOFTRAST_Validate(thread, f), 0) : 0)
-void DPSOFTRAST_Validate(int mask)
+void DPSOFTRAST_Validate(DPSOFTRAST_State_Thread *thread, int mask)
{
- mask &= dpsoftrast.validate;
+ mask &= thread->validate;
if (!mask)
return;
if (mask & DPSOFTRAST_VALIDATE_FB)
{
- dpsoftrast.validate &= ~DPSOFTRAST_VALIDATE_FB;
- DPSOFTRAST_RecalcFB();
+ thread->validate &= ~DPSOFTRAST_VALIDATE_FB;
+ DPSOFTRAST_RecalcFB(thread);
}
if (mask & DPSOFTRAST_VALIDATE_DEPTHFUNC)
{
- dpsoftrast.validate &= ~DPSOFTRAST_VALIDATE_DEPTHFUNC;
- DPSOFTRAST_RecalcDepthFunc();
+ thread->validate &= ~DPSOFTRAST_VALIDATE_DEPTHFUNC;
+ DPSOFTRAST_RecalcDepthFunc(thread);
}
if (mask & DPSOFTRAST_VALIDATE_BLENDFUNC)
{
- dpsoftrast.validate &= ~DPSOFTRAST_VALIDATE_BLENDFUNC;
- DPSOFTRAST_RecalcBlendFunc();
+ thread->validate &= ~DPSOFTRAST_VALIDATE_BLENDFUNC;
+ DPSOFTRAST_RecalcBlendFunc(thread);
}
}
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)
texture->size = size;
// allocate the pixels now
- texture->bytes = calloc(1, size);
+ texture->bytes = (unsigned char *)MM_CALLOC(1, size);
return texnum;
}
{
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return;
+ DPSOFTRAST_Flush();
if (texture->bytes)
- free(texture->bytes);
+ MM_FREE(texture->bytes);
texture->bytes = NULL;
memset(texture, 0, sizeof(*texture));
// adjust the free range and used range
row1 = y*2+1;
if (row1 >= texture->mipmap[i-1][3])
row1 = texture->mipmap[i-1][3]-1;
- o = texture->bytes + texture->mipmap[i ][0] + 4*(texture->mipmap[i ][3] * texture->mipmap[i ][2] * z + texture->mipmap[i ][2] * y );
- i0 = texture->bytes + texture->mipmap[i-1][0] + 4*(texture->mipmap[i-1][3] * texture->mipmap[i-1][2] * layer0 + texture->mipmap[i-1][2] * row0);
- i1 = texture->bytes + texture->mipmap[i-1][0] + 4*(texture->mipmap[i-1][3] * texture->mipmap[i-1][2] * layer0 + texture->mipmap[i-1][2] * row1);
- i2 = texture->bytes + texture->mipmap[i-1][0] + 4*(texture->mipmap[i-1][3] * texture->mipmap[i-1][2] * layer1 + texture->mipmap[i-1][2] * row0);
- i3 = texture->bytes + texture->mipmap[i-1][0] + 4*(texture->mipmap[i-1][3] * texture->mipmap[i-1][2] * layer1 + texture->mipmap[i-1][2] * row1);
+ o = texture->bytes + texture->mipmap[i ][0] + 4*((texture->mipmap[i ][3] * z + y ) * texture->mipmap[i ][2]);
+ i0 = texture->bytes + texture->mipmap[i-1][0] + 4*((texture->mipmap[i-1][3] * layer0 + row0) * texture->mipmap[i-1][2]);
+ i1 = texture->bytes + texture->mipmap[i-1][0] + 4*((texture->mipmap[i-1][3] * layer0 + row1) * texture->mipmap[i-1][2]);
+ i2 = texture->bytes + texture->mipmap[i-1][0] + 4*((texture->mipmap[i-1][3] * layer1 + row0) * texture->mipmap[i-1][2]);
+ i3 = texture->bytes + texture->mipmap[i-1][0] + 4*((texture->mipmap[i-1][3] * layer1 + row1) * texture->mipmap[i-1][2]);
w = texture->mipmap[i][2];
if (layer1 > layer0)
{
if (texture->mipmap[i-1][2] > 1)
{
// average 3D texture
- for (x = 0;x < w;x++)
+ for (x = 0;x < w;x++, o += 4, i0 += 8, i1 += 8, i2 += 8, i3 += 8)
{
o[0] = (i0[0] + i0[4] + i1[0] + i1[4] + i2[0] + i2[4] + i3[0] + i3[4] + 4) >> 3;
o[1] = (i0[1] + i0[5] + i1[1] + i1[5] + i2[1] + i2[5] + i3[1] + i3[5] + 4) >> 3;
else
{
// average 3D mipmap with parent width == 1
- for (x = 0;x < w;x++)
+ for (x = 0;x < w;x++, o += 4, i0 += 8, i1 += 8)
{
o[0] = (i0[0] + i1[0] + i2[0] + i3[0] + 2) >> 2;
o[1] = (i0[1] + i1[1] + i2[1] + i3[1] + 2) >> 2;
if (texture->mipmap[i-1][2] > 1)
{
// average 2D texture (common case)
- for (x = 0;x < w;x++)
+ for (x = 0;x < w;x++, o += 4, i0 += 8, i1 += 8)
{
o[0] = (i0[0] + i0[4] + i1[0] + i1[4] + 2) >> 2;
o[1] = (i0[1] + i0[5] + i1[1] + i1[5] + 2) >> 2;
else
{
// 2D texture with parent width == 1
- for (x = 0;x < w;x++)
- {
- o[0] = (i0[0] + i1[0] + 1) >> 1;
- o[1] = (i0[1] + i1[1] + 1) >> 1;
- o[2] = (i0[2] + i1[2] + 1) >> 1;
- o[3] = (i0[3] + i1[3] + 1) >> 1;
- }
+ o[0] = (i0[0] + i1[0] + 1) >> 1;
+ o[1] = (i0[1] + i1[1] + 1) >> 1;
+ o[2] = (i0[2] + i1[2] + 1) >> 1;
+ o[3] = (i0[3] + i1[3] + 1) >> 1;
}
}
}
void DPSOFTRAST_Texture_UpdatePartial(int index, int mip, const unsigned char *pixels, int blockx, int blocky, int blockwidth, int blockheight)
{
DPSOFTRAST_Texture *texture;
+ unsigned char *dst;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return;
-
- // FIXME IMPLEMENT
-
- dpsoftrast.errorstring = "DPSOFTRAST_Texture_UpdatePartial: Not implemented.";
+ DPSOFTRAST_Flush();
+ 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);
}
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]);
DPSOFTRAST_Texture_CalculateMipmaps(index);
}
{
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return 0;
- return texture->width;
+ return texture->mipmap[mip][2];
}
int DPSOFTRAST_Texture_GetHeight(int index, int mip)
{
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return 0;
- return texture->height;
+ return texture->mipmap[mip][3];
}
int DPSOFTRAST_Texture_GetDepth(int index, int mip)
{
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return 0;
- return texture->depth;
+ return texture->mipmap[mip][4];
}
unsigned char *DPSOFTRAST_Texture_GetPixelPointer(int index, int mip)
{
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return 0;
- return texture->bytes;
+ DPSOFTRAST_Flush();
+ return texture->bytes + texture->mipmap[mip][0];
}
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])
+ DPSOFTRAST_Flush();
dpsoftrast.fb_width = width;
dpsoftrast.fb_height = height;
dpsoftrast.fb_depthpixels = depthpixels;
dpsoftrast.fb_colorpixels[2] = colorpixels2;
dpsoftrast.fb_colorpixels[3] = colorpixels3;
}
+
+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 USE_THREADS
+ 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 USE_THREADS
+ thread = &dpsoftrast.threads[waitindex];
+ thread->waiting = true;
+ SDL_CondBroadcast(dpsoftrast.trianglecond);
+ SDL_CondWait(thread->waitcond, dpsoftrast.trianglemutex);
+ thread->waiting = false;
+#endif
+ }
+#ifdef USE_THREADS
+ SDL_UnlockMutex(dpsoftrast.trianglemutex);
+#endif
+ dpsoftrast.trianglepool.usedtriangles = usedtriangles;
+}
+
+void DPSOFTRAST_Draw_SyncCommands(void)
+{
+ DPSOFTRAST_State_Triangle *triangle;
+ if (dpsoftrast.trianglepool.usedtriangles >= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1)
+#ifdef USE_THREADS
+ 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;
+}
+
+void DPSOFTRAST_Draw_FreeCommandPool(int space)
+{
+ DPSOFTRAST_State_Thread *thread;
+ int i;
+ int freecommand = dpsoftrast.commandpool.freecommand;
+ int usedcommands = dpsoftrast.commandpool.usedcommands;
+ if (usedcommands <= DPSOFTRAST_DRAW_MAXCOMMANDPOOL-space)
+ return;
+ DPSOFTRAST_Draw_SyncCommands();
+#ifdef USE_THREADS
+ SDL_LockMutex(dpsoftrast.trianglemutex);
+#endif
+ for(;;)
+ {
+ int waitindex = -1;
+ int commandoffset;
+ usedcommands = 0;
+ for (i = 0; i < dpsoftrast.numthreads; i++)
+ {
+ thread = &dpsoftrast.threads[i];
+ commandoffset = freecommand - thread->commandoffset;
+ if (commandoffset < 0)
+ commandoffset += DPSOFTRAST_DRAW_MAXCOMMANDPOOL;
+ if (commandoffset > usedcommands)
+ {
+ waitindex = i;
+ usedcommands = commandoffset;
+ }
+ }
+ if (usedcommands <= DPSOFTRAST_DRAW_MAXCOMMANDPOOL-space || waitindex < 0)
+ break;
+#ifdef USE_THREADS
+ thread = &dpsoftrast.threads[waitindex];
+ thread->waiting = true;
+ SDL_CondBroadcast(dpsoftrast.trianglecond);
+ SDL_CondWait(thread->waitcond, dpsoftrast.trianglemutex);
+ thread->waiting = false;
+#endif
+ }
+#ifdef USE_THREADS
+ SDL_UnlockMutex(dpsoftrast.trianglemutex);
+#endif
+ dpsoftrast.commandpool.usedcommands = usedcommands;
+}
+
+#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))))
+
+static void *DPSOFTRAST_AllocateCommand(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)
+ extra += DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand;
+ if(usedcommands > DPSOFTRAST_DRAW_MAXCOMMANDPOOL - (size + extra))
+ {
+#ifdef USE_THREADS
+ DPSOFTRAST_Draw_FreeCommandPool(size + extra);
+#else
+ DPSOFTRAST_Draw_FlushThreads();
+#endif
+ freecommand = dpsoftrast.commandpool.freecommand;
+ usedcommands = dpsoftrast.commandpool.usedcommands;
+ }
+ if(DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand < size)
+ {
+ command = (DPSOFTRAST_Command *) &dpsoftrast.commandpool.commands[freecommand];
+ command->opcode = DPSOFTRAST_OPCODE_Reset;
+ usedcommands += DPSOFTRAST_DRAW_MAXCOMMANDPOOL - freecommand;
+ freecommand = 0;
+ }
+ command = (DPSOFTRAST_Command *) &dpsoftrast.commandpool.commands[freecommand];
+ freecommand += size;
+ if (freecommand >= DPSOFTRAST_DRAW_MAXCOMMANDPOOL)
+ freecommand = 0;
+
+ dpsoftrast.commandpool.freecommand = freecommand;
+ dpsoftrast.commandpool.usedcommands = usedcommands + size;
+ return command;
+}
+
+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)
+{
+ thread->viewport[0] = command->x;
+ thread->viewport[1] = command->y;
+ thread->viewport[2] = command->width;
+ thread->viewport[3] = command->height;
+ thread->validate |= DPSOFTRAST_VALIDATE_FB;
+}
void DPSOFTRAST_Viewport(int x, int y, int width, int height)
{
- dpsoftrast.user.viewport[0] = x;
- dpsoftrast.user.viewport[1] = y;
- dpsoftrast.user.viewport[2] = width;
- dpsoftrast.user.viewport[3] = height;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_FB;
+ DPSOFTRAST_Command_Viewport *command = DPSOFTRAST_ALLOCATECOMMAND(Viewport);
+ command->opcode = DPSOFTRAST_OPCODE_Viewport;
+ command->x = x;
+ command->y = y;
+ command->width = width;
+ command->height = height;
+
+ dpsoftrast.viewport[0] = x;
+ 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;
}
-void DPSOFTRAST_ClearColor(float r, float g, float b, float a)
+
+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;
+ int i, x1, y1, x2, y2, w, h, x, y, t1, t2;
unsigned int *p;
unsigned int c;
- DPSOFTRAST_Validate(DPSOFTRAST_VALIDATE_FB);
- x1 = dpsoftrast.fb_clearscissor[0];
- y1 = dpsoftrast.fb_clearscissor[1];
- x2 = dpsoftrast.fb_clearscissor[2];
- y2 = dpsoftrast.fb_clearscissor[1] + dpsoftrast.fb_clearscissor[3];
+ 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;
w = x2 - x1;
h = y2 - y1;
if (w < 1 || h < 1)
return;
- // FIXME: honor dpsoftrast.fb_colormask?
- c = DPSOFTRAST_BGRA8_FROM_RGBA32F(r,g,b,a);
+ // FIXME: honor fb_colormask?
+ c = DPSOFTRAST_BGRA8_FROM_RGBA32F(command->r,command->g,command->b,command->a);
for (i = 0;i < 4;i++)
{
if (!dpsoftrast.fb_colorpixels[i])
}
}
}
-void DPSOFTRAST_ClearDepth(float d)
+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;
+ command->a = 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;
+ int x1, y1, x2, y2, w, h, x, y, t1, t2;
unsigned int *p;
unsigned int c;
- DPSOFTRAST_Validate(DPSOFTRAST_VALIDATE_FB);
- x1 = dpsoftrast.fb_clearscissor[0];
- y1 = dpsoftrast.fb_clearscissor[1];
- x2 = dpsoftrast.fb_clearscissor[2];
- y2 = dpsoftrast.fb_clearscissor[1] + dpsoftrast.fb_clearscissor[3];
+ 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;
w = x2 - x1;
h = y2 - y1;
if (w < 1 || h < 1)
return;
- c = DPSOFTRAST_DEPTH32_FROM_DEPTH32F(d);
+ c = DPSOFTRAST_DEPTH32_FROM_DEPTH32F(command->depth);
for (y = y1;y < y2;y++)
{
p = dpsoftrast.fb_depthpixels + y * dpsoftrast.fb_width;
p[x] = c;
}
}
+void DPSOFTRAST_ClearDepth(float d)
+{
+ DPSOFTRAST_Command_ClearDepth *command = DPSOFTRAST_ALLOCATECOMMAND(ClearDepth);
+ command->opcode = DPSOFTRAST_OPCODE_ClearDepth;
+ command->depth = d;
+}
+
+DEFCOMMAND(4, ColorMask, int r; int g; int b; int a;)
+static void DPSOFTRAST_Interpret_ColorMask(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_ColorMask *command)
+{
+ thread->colormask[0] = command->r != 0;
+ thread->colormask[1] = command->g != 0;
+ thread->colormask[2] = command->b != 0;
+ thread->colormask[3] = command->a != 0;
+ thread->fb_colormask = ((-thread->colormask[0]) & 0x00FF0000) | ((-thread->colormask[1]) & 0x0000FF00) | ((-thread->colormask[2]) & 0x000000FF) | ((-thread->colormask[3]) & 0xFF000000);
+}
void DPSOFTRAST_ColorMask(int r, int g, int b, int a)
{
- dpsoftrast.user.colormask[0] = r != 0;
- dpsoftrast.user.colormask[1] = g != 0;
- dpsoftrast.user.colormask[2] = b != 0;
- dpsoftrast.user.colormask[3] = a != 0;
- dpsoftrast.fb_colormask = ((-dpsoftrast.user.colormask[0]) & 0x00FF0000) | ((-dpsoftrast.user.colormask[1]) & 0x0000FF00) | ((-dpsoftrast.user.colormask[2]) & 0x000000FF) | ((-dpsoftrast.user.colormask[3]) & 0xFF000000);
+ DPSOFTRAST_Command_ColorMask *command = DPSOFTRAST_ALLOCATECOMMAND(ColorMask);
+ command->opcode = DPSOFTRAST_OPCODE_ColorMask;
+ command->r = r;
+ command->g = g;
+ command->b = b;
+ command->a = a;
+}
+
+DEFCOMMAND(5, DepthTest, int enable;)
+static void DPSOFTRAST_Interpret_DepthTest(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_DepthTest *command)
+{
+ thread->depthtest = command->enable;
+ thread->validate |= DPSOFTRAST_VALIDATE_DEPTHFUNC;
}
void DPSOFTRAST_DepthTest(int enable)
{
- dpsoftrast.user.depthtest = enable;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_DEPTHFUNC;
+ DPSOFTRAST_Command_DepthTest *command = DPSOFTRAST_ALLOCATECOMMAND(DepthTest);
+ command->opcode = DPSOFTRAST_OPCODE_DepthTest;
+ command->enable = enable;
+}
+
+DEFCOMMAND(6, ScissorTest, int enable;)
+static void DPSOFTRAST_Interpret_ScissorTest(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_ScissorTest *command)
+{
+ thread->scissortest = command->enable;
+ thread->validate |= DPSOFTRAST_VALIDATE_FB;
}
void DPSOFTRAST_ScissorTest(int enable)
{
- dpsoftrast.user.scissortest = enable;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_FB;
+ DPSOFTRAST_Command_ScissorTest *command = DPSOFTRAST_ALLOCATECOMMAND(ScissorTest);
+ command->opcode = DPSOFTRAST_OPCODE_ScissorTest;
+ command->enable = enable;
+}
+
+DEFCOMMAND(7, Scissor, float x; float y; float width; float height;)
+static void DPSOFTRAST_Interpret_Scissor(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_Scissor *command)
+{
+ thread->scissor[0] = command->x;
+ thread->scissor[1] = command->y;
+ thread->scissor[2] = command->width;
+ thread->scissor[3] = command->height;
+ thread->validate |= DPSOFTRAST_VALIDATE_FB;
}
void DPSOFTRAST_Scissor(float x, float y, float width, float height)
{
- dpsoftrast.user.scissor[0] = x;
- dpsoftrast.user.scissor[1] = y;
- dpsoftrast.user.scissor[2] = width;
- dpsoftrast.user.scissor[3] = height;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_FB;
+ DPSOFTRAST_Command_Scissor *command = DPSOFTRAST_ALLOCATECOMMAND(Scissor);
+ command->opcode = DPSOFTRAST_OPCODE_Scissor;
+ command->x = x;
+ command->y = y;
+ command->width = width;
+ command->height = height;
+}
+
+DEFCOMMAND(8, BlendFunc, int sfactor; int dfactor;)
+static void DPSOFTRAST_Interpret_BlendFunc(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_BlendFunc *command)
+{
+ thread->blendfunc[0] = command->sfactor;
+ thread->blendfunc[1] = command->dfactor;
+ thread->validate |= DPSOFTRAST_VALIDATE_BLENDFUNC;
+}
+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;
}
-void DPSOFTRAST_BlendFunc(int smodulate, int dmodulate)
+DEFCOMMAND(9, BlendSubtract, int enable;)
+static void DPSOFTRAST_Interpret_BlendSubtract(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_BlendSubtract *command)
{
- // FIXME: validate
- dpsoftrast.user.blendfunc[0] = smodulate;
- dpsoftrast.user.blendfunc[1] = dmodulate;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_BLENDFUNC;
+ thread->blendsubtract = command->enable;
+ thread->validate |= DPSOFTRAST_VALIDATE_BLENDFUNC;
}
void DPSOFTRAST_BlendSubtract(int enable)
{
- dpsoftrast.user.blendsubtract = enable != 0;
- dpsoftrast.validate |= DPSOFTRAST_VALIDATE_BLENDFUNC;
+ DPSOFTRAST_Command_BlendSubtract *command = DPSOFTRAST_ALLOCATECOMMAND(BlendSubtract);
+ command->opcode = DPSOFTRAST_OPCODE_BlendSubtract;
+ command->enable = enable;
+}
+
+DEFCOMMAND(10, DepthMask, int enable;)
+static void DPSOFTRAST_Interpret_DepthMask(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_DepthMask *command)
+{
+ thread->depthmask = command->enable;
}
void DPSOFTRAST_DepthMask(int enable)
{
- dpsoftrast.user.depthmask = enable;
+ DPSOFTRAST_Command_DepthMask *command = DPSOFTRAST_ALLOCATECOMMAND(DepthMask);
+ command->opcode = DPSOFTRAST_OPCODE_DepthMask;
+ command->enable = enable;
+}
+
+DEFCOMMAND(11, DepthFunc, int func;)
+static void DPSOFTRAST_Interpret_DepthFunc(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_DepthFunc *command)
+{
+ thread->depthfunc = command->func;
+}
+void DPSOFTRAST_DepthFunc(int func)
+{
+ DPSOFTRAST_Command_DepthFunc *command = DPSOFTRAST_ALLOCATECOMMAND(DepthFunc);
+ command->opcode = DPSOFTRAST_OPCODE_DepthFunc;
+ command->func = func;
+}
+
+DEFCOMMAND(12, DepthRange, float nearval; float farval;)
+static void DPSOFTRAST_Interpret_DepthRange(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_DepthRange *command)
+{
+ thread->depthrange[0] = command->nearval;
+ thread->depthrange[1] = command->farval;
}
-void DPSOFTRAST_DepthFunc(int comparemode)
+void DPSOFTRAST_DepthRange(float nearval, float farval)
{
- // FIXME: validate
- dpsoftrast.user.depthfunc = comparemode;
+ DPSOFTRAST_Command_DepthRange *command = DPSOFTRAST_ALLOCATECOMMAND(DepthRange);
+ command->opcode = DPSOFTRAST_OPCODE_DepthRange;
+ command->nearval = nearval;
+ command->farval = farval;
}
-void DPSOFTRAST_DepthRange(float range0, float range1)
+
+DEFCOMMAND(13, PolygonOffset, float alongnormal; float intoview;)
+static void DPSOFTRAST_Interpret_PolygonOffset(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_PolygonOffset *command)
{
- dpsoftrast.user.depthrange[0] = range0;
- dpsoftrast.user.depthrange[1] = range1;
+ thread->polygonoffset[0] = command->alongnormal;
+ thread->polygonoffset[1] = command->intoview;
}
void DPSOFTRAST_PolygonOffset(float alongnormal, float intoview)
{
- dpsoftrast.user.polygonoffset[0] = alongnormal;
- dpsoftrast.user.polygonoffset[1] = intoview;
+ DPSOFTRAST_Command_PolygonOffset *command = DPSOFTRAST_ALLOCATECOMMAND(PolygonOffset);
+ command->opcode = DPSOFTRAST_OPCODE_PolygonOffset;
+ command->alongnormal = alongnormal;
+ command->intoview = intoview;
}
+
void DPSOFTRAST_CullFace(int mode)
{
- // FIXME: validate
- dpsoftrast.user.cullface = 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;
}
-void DPSOFTRAST_AlphaTest(float enable)
+
+DEFCOMMAND(16, AlphaFunc, int func; float ref;)
+static void DPSOFTRAST_Interpret_AlphaFunc(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_AlphaFunc *command)
{
- dpsoftrast.user.alphatest = enable;
+ thread->alphafunc = command->func;
+ thread->alphavalue = command->ref;
}
-void DPSOFTRAST_AlphaFunc(int alphafunc, float alphavalue)
+void DPSOFTRAST_AlphaFunc(int func, float ref)
{
- // FIXME: validate
- dpsoftrast.user.alphafunc = alphafunc;
- dpsoftrast.user.alphavalue = alphavalue;
+ DPSOFTRAST_Command_AlphaFunc *command = DPSOFTRAST_ALLOCATECOMMAND(AlphaFunc);
+ command->opcode = DPSOFTRAST_OPCODE_AlphaFunc;
+ command->func = func;
+ command->ref = ref;
}
+
void DPSOFTRAST_Color4f(float r, float g, float b, float a)
{
- dpsoftrast.user.color[0] = r;
- dpsoftrast.user.color[1] = g;
- dpsoftrast.user.color[2] = b;
- dpsoftrast.user.color[3] = a;
+ dpsoftrast.color[0] = r;
+ dpsoftrast.color[1] = g;
+ dpsoftrast.color[2] = b;
+ dpsoftrast.color[3] = a;
}
+
void DPSOFTRAST_GetPixelsBGRA(int blockx, int blocky, int blockwidth, int blockheight, unsigned char *outpixels)
{
int outstride = blockwidth * 4;
unsigned char *inpixels;
unsigned char *b;
unsigned char *o;
+ DPSOFTRAST_Flush();
if (bx1 < 0) bx1 = 0;
if (by1 < 0) by1 = 0;
if (bx2 > dpsoftrast.fb_width) bx2 = dpsoftrast.fb_width;
DPSOFTRAST_Texture *texture;
texture = DPSOFTRAST_Texture_GetByIndex(index);if (!texture) return;
if (mip < 0 || mip >= texture->mipmaps) return;
+ DPSOFTRAST_Flush();
spixels = dpsoftrast.fb_colorpixels[0];
swidth = dpsoftrast.fb_width;
sheight = dpsoftrast.fb_height;
if (texture->mipmaps > 1)
DPSOFTRAST_Texture_CalculateMipmaps(index);
}
+
+DEFCOMMAND(17, SetTexture, int unitnum; DPSOFTRAST_Texture *texture;)
+static void DPSOFTRAST_Interpret_SetTexture(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_SetTexture *command)
+{
+ thread->texbound[command->unitnum] = command->texture;
+}
void DPSOFTRAST_SetTexture(int unitnum, int index)
{
+ DPSOFTRAST_Command_SetTexture *command;
DPSOFTRAST_Texture *texture;
if (unitnum < 0 || unitnum >= DPSOFTRAST_MAXTEXTUREUNITS)
{
dpsoftrast.errorstring = "DPSOFTRAST_SetTexture: invalid texture handle";
return;
}
+
+ command = DPSOFTRAST_ALLOCATECOMMAND(SetTexture);
+ command->opcode = DPSOFTRAST_OPCODE_SetTexture;
+ command->unitnum = unitnum;
+ command->texture = texture;
+
dpsoftrast.texbound[unitnum] = texture;
}
dpsoftrast.stride_texcoord[unitnum] = stride;
}
-void DPSOFTRAST_SetShader(unsigned int mode, unsigned int permutation)
+DEFCOMMAND(18, SetShader, int mode; int permutation;)
+static void DPSOFTRAST_Interpret_SetShader(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_SetShader *command)
{
+ thread->shader_mode = command->mode;
+ thread->shader_permutation = command->permutation;
+}
+void DPSOFTRAST_SetShader(int mode, int permutation)
+{
+ DPSOFTRAST_Command_SetShader *command = DPSOFTRAST_ALLOCATECOMMAND(SetShader);
+ command->opcode = DPSOFTRAST_OPCODE_SetShader;
+ command->mode = mode;
+ command->permutation = permutation;
+
dpsoftrast.shader_mode = mode;
dpsoftrast.shader_permutation = permutation;
}
-void DPSOFTRAST_Uniform4fARB(DPSOFTRAST_UNIFORM index, float v0, float v1, float v2, float v3)
+
+DEFCOMMAND(19, Uniform4f, DPSOFTRAST_UNIFORM index; float val[4];)
+static void DPSOFTRAST_Interpret_Uniform4f(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_Uniform4f *command)
+{
+ memcpy(&thread->uniform4f[command->index*4], command->val, sizeof(command->val));
+}
+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;
+ command->val[2] = v2;
+ command->val[3] = v3;
+
dpsoftrast.uniform4f[index*4+0] = v0;
dpsoftrast.uniform4f[index*4+1] = v1;
dpsoftrast.uniform4f[index*4+2] = v2;
dpsoftrast.uniform4f[index*4+3] = v3;
}
-void DPSOFTRAST_Uniform4fvARB(DPSOFTRAST_UNIFORM index, const float *v)
+void DPSOFTRAST_Uniform4fv(DPSOFTRAST_UNIFORM index, const float *v)
{
- dpsoftrast.uniform4f[index*4+0] = v[0];
- dpsoftrast.uniform4f[index*4+1] = v[1];
- dpsoftrast.uniform4f[index*4+2] = v[2];
- dpsoftrast.uniform4f[index*4+3] = v[3];
+ DPSOFTRAST_Command_Uniform4f *command = DPSOFTRAST_ALLOCATECOMMAND(Uniform4f);
+ command->opcode = DPSOFTRAST_OPCODE_Uniform4f;
+ command->index = index;
+ memcpy(command->val, v, sizeof(command->val));
+
+ memcpy(&dpsoftrast.uniform4f[index*4], v, sizeof(float[4]));
}
-void DPSOFTRAST_UniformMatrix4fvARB(DPSOFTRAST_UNIFORM index, int arraysize, int transpose, const float *v)
+
+DEFCOMMAND(20, UniformMatrix4f, DPSOFTRAST_UNIFORM index; ALIGN(float val[16]);)
+static void DPSOFTRAST_Interpret_UniformMatrix4f(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_UniformMatrix4f *command)
{
- dpsoftrast.uniform4f[index*4+0] = v[0];
- dpsoftrast.uniform4f[index*4+1] = v[1];
- dpsoftrast.uniform4f[index*4+2] = v[2];
- dpsoftrast.uniform4f[index*4+3] = v[3];
- dpsoftrast.uniform4f[index*4+4] = v[4];
- dpsoftrast.uniform4f[index*4+5] = v[5];
- dpsoftrast.uniform4f[index*4+6] = v[6];
- dpsoftrast.uniform4f[index*4+7] = v[7];
- dpsoftrast.uniform4f[index*4+8] = v[8];
- dpsoftrast.uniform4f[index*4+9] = v[9];
- dpsoftrast.uniform4f[index*4+10] = v[10];
- dpsoftrast.uniform4f[index*4+11] = v[11];
- dpsoftrast.uniform4f[index*4+12] = v[12];
- dpsoftrast.uniform4f[index*4+13] = v[13];
- dpsoftrast.uniform4f[index*4+14] = v[14];
- dpsoftrast.uniform4f[index*4+15] = v[15];
+ memcpy(&thread->uniform4f[command->index*4], command->val, sizeof(command->val));
}
-void DPSOFTRAST_Uniform1iARB(DPSOFTRAST_UNIFORM index, int i0)
+void DPSOFTRAST_UniformMatrix4fv(DPSOFTRAST_UNIFORM uniform, int arraysize, int transpose, const float *v)
{
- dpsoftrast.uniform1i[index] = i0;
+#ifdef SSE2_PRESENT
+ 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;
+ if (((size_t)v)&(ALIGN_SIZE-1))
+ {
+ m0 = _mm_loadu_ps(v);
+ m1 = _mm_loadu_ps(v+4);
+ m2 = _mm_loadu_ps(v+8);
+ m3 = _mm_loadu_ps(v+12);
+ }
+ else
+ {
+ m0 = _mm_load_ps(v);
+ m1 = _mm_load_ps(v+4);
+ m2 = _mm_load_ps(v+8);
+ m3 = _mm_load_ps(v+12);
+ }
+ if (transpose)
+ {
+ __m128 t0, t1, t2, t3;
+ t0 = _mm_unpacklo_ps(m0, m1);
+ t1 = _mm_unpacklo_ps(m2, m3);
+ t2 = _mm_unpackhi_ps(m0, m1);
+ t3 = _mm_unpackhi_ps(m2, m3);
+ m0 = _mm_movelh_ps(t0, t1);
+ m1 = _mm_movehl_ps(t1, t0);
+ m2 = _mm_movelh_ps(t2, t3);
+ m3 = _mm_movehl_ps(t3, t2);
+ }
+ _mm_store_ps(command->val, m0);
+ _mm_store_ps(command->val+4, m1);
+ _mm_store_ps(command->val+8, m2);
+ _mm_store_ps(command->val+12, m3);
+ _mm_store_ps(&dpsoftrast.uniform4f[index*4+0], m0);
+ _mm_store_ps(&dpsoftrast.uniform4f[index*4+4], m1);
+ _mm_store_ps(&dpsoftrast.uniform4f[index*4+8], m2);
+ _mm_store_ps(&dpsoftrast.uniform4f[index*4+12], m3);
+ }
+#endif
}
-void DPSOFTRAST_Draw_LoadVertices(int firstvertex, int numvertices, bool needcolors)
+DEFCOMMAND(21, Uniform1i, DPSOFTRAST_UNIFORM index; int val;)
+static void DPSOFTRAST_Interpret_Uniform1i(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_Uniform1i *command)
{
- int i;
- int j;
- int stride;
- const float *v;
+ thread->uniform1i[command->index] = command->val;
+}
+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
+static void DPSOFTRAST_Load4fTo4f(float *dst, const unsigned char *src, int size, int stride)
+{
+ float *end = dst + size*4;
+ if ((((size_t)src)|stride)&(ALIGN_SIZE - 1)) // check for alignment
+ {
+ while (dst < end)
+ {
+ _mm_store_ps(dst, _mm_loadu_ps((const float *)src));
+ dst += 4;
+ src += stride;
+ }
+ }
+ else
+ {
+ while (dst < end)
+ {
+ _mm_store_ps(dst, _mm_load_ps((const float *)src));
+ dst += 4;
+ src += stride;
+ }
+ }
+}
+
+static void DPSOFTRAST_Load3fTo4f(float *dst, const unsigned char *src, int size, int stride)
+{
+ float *end = dst + size*4;
+ if (stride == sizeof(float[3]))
+ {
+ float *end4 = dst + (size&~3)*4;
+ if (((size_t)src)&(ALIGN_SIZE - 1)) // check for alignment
+ {
+ while (dst < end4)
+ {
+ __m128 v1 = _mm_loadu_ps((const float *)src), v2 = _mm_loadu_ps((const float *)src + 4), v3 = _mm_loadu_ps((const float *)src + 8), dv;
+ dv = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(2, 1, 0, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_shuffle_ps(v1, v2, _MM_SHUFFLE(1, 0, 3, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 4, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_shuffle_ps(v2, v3, _MM_SHUFFLE(0, 0, 3, 2));
+ dv = _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(2, 1, 0, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 8, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_move_ss(v3, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 12, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dst += 16;
+ src += 4*sizeof(float[3]);
+ }
+ }
+ else
+ {
+ while (dst < end4)
+ {
+ __m128 v1 = _mm_load_ps((const float *)src), v2 = _mm_load_ps((const float *)src + 4), v3 = _mm_load_ps((const float *)src + 8), dv;
+ dv = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(2, 1, 0, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_shuffle_ps(v1, v2, _MM_SHUFFLE(1, 0, 3, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 4, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_shuffle_ps(v2, v3, _MM_SHUFFLE(0, 0, 3, 2));
+ dv = _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(2, 1, 0, 3));
+ dv = _mm_move_ss(dv, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 8, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dv = _mm_move_ss(v3, _mm_set_ss(1.0f));
+ _mm_store_ps(dst + 12, _mm_shuffle_ps(dv, dv, _MM_SHUFFLE(0, 3, 2, 1)));
+ dst += 16;
+ src += 4*sizeof(float[3]);
+ }
+ }
+ }
+ if ((((size_t)src)|stride)&(ALIGN_SIZE - 1))
+ {
+ while (dst < end)
+ {
+ __m128 v = _mm_loadu_ps((const float *)src);
+ v = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 1, 0, 3));
+ v = _mm_move_ss(v, _mm_set_ss(1.0f));
+ v = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1));
+ _mm_store_ps(dst, v);
+ dst += 4;
+ src += stride;
+ }
+ }
+ else
+ {
+ while (dst < end)
+ {
+ __m128 v = _mm_load_ps((const float *)src);
+ v = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 1, 0, 3));
+ v = _mm_move_ss(v, _mm_set_ss(1.0f));
+ v = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1));
+ _mm_store_ps(dst, v);
+ dst += 4;
+ src += stride;
+ }
+ }
+}
+
+static void DPSOFTRAST_Load2fTo4f(float *dst, const unsigned char *src, int size, int stride)
+{
+ float *end = dst + size*4;
+ __m128 v2 = _mm_setr_ps(0.0f, 0.0f, 0.0f, 1.0f);
+ if (stride == sizeof(float[2]))
+ {
+ float *end2 = dst + (size&~1)*4;
+ if (((size_t)src)&(ALIGN_SIZE - 1)) // check for alignment
+ {
+ while (dst < end2)
+ {
+ __m128 v = _mm_loadu_ps((const float *)src);
+ _mm_store_ps(dst, _mm_shuffle_ps(v, v2, _MM_SHUFFLE(3, 2, 1, 0)));
+ _mm_store_ps(dst + 4, _mm_movehl_ps(v2, v));
+ dst += 8;
+ src += 2*sizeof(float[2]);
+ }
+ }
+ else
+ {
+ while (dst < end2)
+ {
+ __m128 v = _mm_load_ps((const float *)src);
+ _mm_store_ps(dst, _mm_shuffle_ps(v, v2, _MM_SHUFFLE(3, 2, 1, 0)));
+ _mm_store_ps(dst + 4, _mm_movehl_ps(v2, v));
+ dst += 8;
+ src += 2*sizeof(float[2]);
+ }
+ }
+ }
+ while (dst < end)
+ {
+ _mm_store_ps(dst, _mm_loadl_pi(v2, (__m64 *)src));
+ dst += 4;
+ src += stride;
+ }
+}
+
+static void DPSOFTRAST_Load4bTo4f(float *dst, const unsigned char *src, int size, int stride)
+{
+ float *end = dst + size*4;
+ __m128 scale = _mm_set1_ps(1.0f/255.0f);
+ if (stride == sizeof(unsigned char[4]))
+ {
+ float *end4 = dst + (size&~3)*4;
+ if (((size_t)src)&(ALIGN_SIZE - 1)) // check for alignment
+ {
+ while (dst < end4)
+ {
+ __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));
+ 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 < end)
+ {
+ __m128i v = _mm_cvtsi32_si128(*(const int *)src);
+ _mm_store_ps(dst, _mm_mul_ps(_mm_cvtepi32_ps(_mm_unpacklo_epi16(_mm_unpacklo_epi8(v, _mm_setzero_si128()), _mm_setzero_si128())), scale));
+ dst += 4;
+ src += stride;
+ }
+}
+
+static void DPSOFTRAST_Fill4f(float *dst, const float *src, int size)
+{
+ float *end = dst + 4*size;
+ __m128 v = _mm_loadu_ps(src);
+ while (dst < end)
+ {
+ _mm_store_ps(dst, v);
+ dst += 4;
+ }
+}
+#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.draw.numvertices = numvertices;
- if (dpsoftrast.draw.maxvertices < dpsoftrast.draw.numvertices)
- {
- if (dpsoftrast.draw.maxvertices < 4096)
- dpsoftrast.draw.maxvertices = 4096;
- while (dpsoftrast.draw.maxvertices < dpsoftrast.draw.numvertices)
- dpsoftrast.draw.maxvertices *= 2;
- if (dpsoftrast.draw.in_array4f[0])
- free(dpsoftrast.draw.in_array4f[0]);
- data = calloc(1, dpsoftrast.draw.maxvertices * sizeof(float[4])*(DPSOFTRAST_ARRAY_TOTAL*2 + 1));
- for (i = 0;i < DPSOFTRAST_ARRAY_TOTAL;i++, data += dpsoftrast.draw.maxvertices * 4)
- dpsoftrast.draw.in_array4f[i] = data;
- for (i = 0;i < DPSOFTRAST_ARRAY_TOTAL;i++, data += dpsoftrast.draw.maxvertices * 4)
- dpsoftrast.draw.post_array4f[i] = data;
- dpsoftrast.draw.screencoord4f = data;
- data += dpsoftrast.draw.maxvertices * 4;
+ 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.draw.in_array4f[0];
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = v[2];
- p[3] = 1.0f;
- p += 4;
- v = (const float *)((const unsigned char *)v + 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.draw.in_array4f[1];
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = v[2];
- p[3] = v[3];
- p += 4;
- v = (const float *)((const unsigned char *)v + 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.draw.in_array4f[1];
- for (i = 0;i < numvertices;i++)
- {
- p[0] = b[0] * (1.0f / 255.0f);
- p[1] = b[1] * (1.0f / 255.0f);
- p[2] = b[2] * (1.0f / 255.0f);
- p[3] = b[3] * (1.0f / 255.0f);
- p += 4;
- b = (const unsigned char *)((const unsigned char *)b + stride);
- }
+ p = dpsoftrast.in_array4f[1];
+ DPSOFTRAST_Load4bTo4f(p, b, numvertices, stride);
}
else
{
- v = dpsoftrast.user.color;
- p = dpsoftrast.draw.in_array4f[1];
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = v[2];
- p[3] = v[3];
- p += 4;
- }
+ p = dpsoftrast.in_array4f[1];
+ DPSOFTRAST_Fill4f(p, dpsoftrast.color, numvertices);
}
}
for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL-2;j++)
{
stride = dpsoftrast.stride_texcoord[j];
v = (const float *)((const unsigned char *)dpsoftrast.pointer_texcoordf[j] + firstvertex * stride);
- p = dpsoftrast.draw.in_array4f[j+2];
+ p = dpsoftrast.in_array4f[j+2];
switch(dpsoftrast.components_texcoord[j])
{
case 2:
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = 0.0f;
- p[3] = 1.0f;
- p += 4;
- v = (const float *)((const unsigned char *)v + stride);
- }
+ DPSOFTRAST_Load2fTo4f(p, (const unsigned char *)v, numvertices, stride);
break;
case 3:
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = v[2];
- p[3] = 1.0f;
- p += 4;
- v = (const float *)((const unsigned char *)v + stride);
- }
+ DPSOFTRAST_Load3fTo4f(p, (const unsigned char *)v, numvertices, stride);
break;
case 4:
- for (i = 0;i < numvertices;i++)
- {
- p[0] = v[0];
- p[1] = v[1];
- p[2] = v[2];
- p[3] = v[3];
- p += 4;
- v = (const float *)((const unsigned char *)v + stride);
- }
+ 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)
{
- // TODO: SIMD
- float matrix[4][4];
- int i;
- memcpy(matrix, inmatrix16f, sizeof(float[16]));
- for (i = 0;i < numitems;i++, out4f += 4, in4f += 4)
+#ifdef SSE2_PRESENT
+ 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;
+ if (!memcmp(identitymatrix, inmatrix16f, sizeof(float[16])))
+ {
+ // fast case for identity matrix
+ memcpy(out4f, in4f, numitems * sizeof(float[4]));
+ return;
+ }
+ m0 = _mm_loadu_ps(inmatrix16f);
+ m1 = _mm_loadu_ps(inmatrix16f + 4);
+ m2 = _mm_loadu_ps(inmatrix16f + 8);
+ m3 = _mm_loadu_ps(inmatrix16f + 12);
+ if (((size_t)in4f)&(ALIGN_SIZE-1)) // check alignment
+ {
+ while (out4f < end)
+ {
+ __m128 v = _mm_loadu_ps(in4f);
+ _mm_store_ps(out4f,
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0)), m0),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1)), m1),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2)), m2),
+ _mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)), m3)))));
+ out4f += 4;
+ in4f += 4;
+ }
+ }
+ else
{
- out4f[0] = in4f[0] * matrix[0][0] + in4f[1] * matrix[1][0] + in4f[2] * matrix[2][0] + in4f[3] * matrix[3][0];
- out4f[1] = in4f[0] * matrix[0][1] + in4f[1] * matrix[1][1] + in4f[2] * matrix[2][1] + in4f[3] * matrix[3][1];
- out4f[2] = in4f[0] * matrix[0][2] + in4f[1] * matrix[1][2] + in4f[2] * matrix[2][2] + in4f[3] * matrix[3][2];
- out4f[3] = in4f[0] * matrix[0][3] + in4f[1] * matrix[1][3] + in4f[2] * matrix[2][3] + in4f[3] * matrix[3][3];
+ while (out4f < end)
+ {
+ __m128 v = _mm_load_ps(in4f);
+ _mm_store_ps(out4f,
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0)), m0),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1)), m1),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2)), m2),
+ _mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)), m3)))));
+ out4f += 4;
+ in4f += 4;
+ }
}
+#endif
}
void DPSOFTRAST_Array_Copy(float *out4f, const float *in4f, int numitems)
memcpy(out4f, in4f, numitems * sizeof(float[4]));
}
-void DPSOFTRAST_Draw_ProjectVertices(float *out4f, const float *in4f, int numitems)
+#ifdef SSE2_PRESENT
+static __m128 DPSOFTRAST_Draw_ProjectVertex(__m128 v)
{
- // NOTE: this is used both as a whole mesh transform function and a
- // per-triangle transform function (for clipped triangles), accordingly
- // it should not crash on divide by 0 but the result of divide by 0 is
- // unimportant...
- // TODO: SIMD
- int i;
- float w;
- float viewportcenter[4];
- float viewportscale[4];
- viewportscale[0] = dpsoftrast.fb_viewportscale[0];
- viewportscale[1] = dpsoftrast.fb_viewportscale[1];
- viewportscale[2] = 0.5f;
- viewportscale[3] = 0.0f;
- viewportcenter[0] = dpsoftrast.fb_viewportcenter[0];
- viewportcenter[1] = dpsoftrast.fb_viewportcenter[1];
- viewportcenter[2] = 0.5f;
- viewportcenter[3] = 0.0f;
- for (i = 0;i < numitems;i++)
- {
- if (!in4f[3])
- {
- out4f[0] = 0.0f;
- out4f[1] = 0.0f;
- out4f[2] = 0.0f;
- out4f[3] = 0.0f;
- continue;
- }
- w = 1.0f / in4f[3];
- out4f[0] = viewportcenter[0] + viewportscale[0] * in4f[0] * w;
- out4f[1] = viewportcenter[1] + viewportscale[1] * in4f[1] * w;
- out4f[2] = viewportcenter[2] + viewportscale[2] * in4f[2] * w;
- out4f[3] = viewportcenter[3] + viewportscale[3] * in4f[3] * w;
- out4f[3] = w;
+ __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;
+}
+#endif
+
+void DPSOFTRAST_Array_Project(float *out4f, float *screen4f, const float *in4f, int numitems)
+{
+#ifdef SSE2_PRESENT
+ float *end = out4f + numitems*4;
+ __m128 viewportcenter = _mm_load_ps(dpsoftrast.fb_viewportcenter), viewportscale = _mm_load_ps(dpsoftrast.fb_viewportscale);
+ while (out4f < end)
+ {
+ __m128 v = _mm_load_ps(in4f), w = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3));
+ _mm_store_ps(out4f, v);
+ 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));
+ _mm_store_ps(screen4f, _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1)));
in4f += 4;
out4f += 4;
+ screen4f += 4;
}
+#endif
}
-void DPSOFTRAST_Draw_DebugEdgePoints(const float *screen0, const float *screen1)
+void DPSOFTRAST_Array_TransformProject(float *out4f, float *screen4f, const float *in4f, int numitems, const float *inmatrix16f)
{
- int i;
- int x;
- int y;
- int w = dpsoftrast.fb_width;
- int bounds[4];
- float v0[2], v1[2];
- unsigned int *pixels = dpsoftrast.fb_colorpixels[0];
- //const float *c4f;
- bounds[0] = dpsoftrast.fb_viewportscissor[0];
- bounds[1] = dpsoftrast.fb_viewportscissor[1];
- bounds[2] = dpsoftrast.fb_viewportscissor[0] + dpsoftrast.fb_viewportscissor[2];
- bounds[3] = dpsoftrast.fb_viewportscissor[1] + dpsoftrast.fb_viewportscissor[3];
- v0[0] = screen0[0];
- v0[1] = screen0[1];
- v1[0] = screen1[0];
- v1[1] = screen1[1];
- for (i = 0;i <= 128;i++)
- {
- // check nearclip
- //if (dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+3] != 1.0f)
- // continue;
- x = (int)(v0[0] + (v1[0] - v0[0]) * (i/128.0f));
- y = (int)(v0[1] + (v1[1] - v0[1]) * (i/128.0f));
- if (x < bounds[0] || y < bounds[1] || x >= bounds[2] || y >= bounds[3])
- continue;
- //c4f = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_COLOR] + element0*4;
- //pixels[y*w+x] = DPSOFTRAST_BGRA8_FROM_RGBA32F(c4f[0], c4f[1], c4f[2], c4f[3]);
- pixels[y*w+x] = 0xFFFFFFFF;
+#ifdef SSE2_PRESENT
+ 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, viewportcenter, viewportscale;
+ float *end = out4f + numitems*4;
+ if (!memcmp(identitymatrix, inmatrix16f, sizeof(float[16])))
+ {
+ DPSOFTRAST_Array_Project(out4f, screen4f, in4f, numitems);
+ return;
+ }
+ viewportcenter = _mm_load_ps(dpsoftrast.fb_viewportcenter);
+ viewportscale = _mm_load_ps(dpsoftrast.fb_viewportscale);
+ m0 = _mm_loadu_ps(inmatrix16f);
+ m1 = _mm_loadu_ps(inmatrix16f + 4);
+ m2 = _mm_loadu_ps(inmatrix16f + 8);
+ m3 = _mm_loadu_ps(inmatrix16f + 12);
+ if (((size_t)in4f)&(ALIGN_SIZE-1)) // check alignment
+ {
+ while (out4f < end)
+ {
+ __m128 v = _mm_loadu_ps(in4f), w;
+ v = _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0)), m0),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1)), m1),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2)), m2),
+ _mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)), m3))));
+ _mm_store_ps(out4f, v);
+ 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));
+ _mm_store_ps(screen4f, _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1)));
+ in4f += 4;
+ out4f += 4;
+ screen4f += 4;
+ }
+ }
+ else
+ {
+ while (out4f < end)
+ {
+ __m128 v = _mm_load_ps(in4f), w;
+ v = _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0)), m0),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1)), m1),
+ _mm_add_ps(_mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2)), m2),
+ _mm_mul_ps(_mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)), m3))));
+ _mm_store_ps(out4f, v);
+ 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));
+ _mm_store_ps(screen4f, _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 3, 2, 1)));
+ in4f += 4;
+ out4f += 4;
+ screen4f += 4;
+ }
}
+#endif
}
-void DPSOFTRAST_Draw_Span_Begin(const DPSOFTRAST_State_Draw_Span *span, float *zf)
+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 w = span->data[0][DPSOFTRAST_ARRAY_TOTAL][3];
- float wslope = span->data[1][DPSOFTRAST_ARRAY_TOTAL][3];
+ 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);
for (x = startx;x < endx;)
- {
- int endsub = x + DPSOFTRAST_MAXSUBSPAN-1;
- float z = 1.0f / (w + wslope * x), dz;
- if (endsub >= endx)
- {
- endsub = endx-1;
- dz = endsub > x ? (1.0f / (w + wslope * endsub) - z) / (endsub - x) : 0.0f;
- }
- else
- {
- dz = (1.0f / (w + wslope * endsub) - z) * (1.0f / (DPSOFTRAST_MAXSUBSPAN-1));
- }
- for (; x <= endsub; x++, z += dz)
- zf[x] = z;
- }
-}
-
-void DPSOFTRAST_Draw_Span_Finish(const DPSOFTRAST_State_Draw_Span *span, const float * RESTRICT in4f)
+ {
+ int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1;
+ float z = endz, dz;
+ 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)
+ zf[x] = z;
+ }
+}
+
+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)
{
int x;
int startx = span->startx;
unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0];
if (!pixel)
return;
- pixel += span->start * 4;
+ pixel += (span->y * dpsoftrast.fb_width + span->x) * 4;
// handle alphatest now (this affects depth writes too)
- if (dpsoftrast.user.alphatest)
+ if (thread->alphatest)
for (x = startx;x < endx;x++)
if (in4f[x*4+3] < 0.5f)
pixelmask[x] = false;
// FIXME: this does not handle bigendian
- switch(dpsoftrast.fb_blendmode)
+ switch(thread->fb_blendmode)
{
case DPSOFTRAST_BLENDMODE_OPAQUE:
for (x = startx;x < endx;x++)
}
}
-void DPSOFTRAST_Draw_Span_Texture2DVarying(const DPSOFTRAST_State_Draw_Span *span, float * RESTRICT out4f, int texunitindex, int arrayindex, const float * RESTRICT zf)
+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;)
+ {
+ if (*(const unsigned int *)&pixelmask[x] == 0x01010101)
+ {
+ _mm_storeu_si128((__m128i *)&pixeli[x], _mm_loadu_si128((const __m128i *)&ini[x]));
+ x += 4;
+ }
+ else
+ {
+ if (pixelmask[x])
+ pixeli[x] = ini[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]) \
+ { \
+ case 0x0101: \
+ 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: \
+ 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)); \
+ }
+
+ 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
+}
+
+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;
int startx = span->startx;
float c[4];
float data[4];
float slope[4];
- float tc[2];
+ float tc[2], endtc[2];
float tcscale[2];
unsigned int tci[2];
unsigned int tci1[2];
int mip;
const unsigned char * RESTRICT pixelbase;
const unsigned char * RESTRICT pixel[4];
- DPSOFTRAST_Texture *texture = dpsoftrast.texbound[texunitindex];
+ DPSOFTRAST_Texture *texture = thread->texbound[texunitindex];
// if no texture is bound, just fill it with white
if (!texture)
{
}
return;
}
+ mip = triangle->mip[texunitindex];
+ pixelbase = (unsigned char *)texture->bytes + texture->mipmap[mip][0];
// if this mipmap of the texture is 1 pixel, just fill it with that color
- mip = span->mip;
- if (mip >= texture->mipmaps)
- mip = texture->mipmaps - 1;
if (texture->mipmap[mip][1] == 4)
{
c[0] = texture->bytes[2] * (1.0f/255.0f);
return;
}
filter = texture->filter & DPSOFTRAST_TEXTURE_FILTER_LINEAR;
- data[0] = span->data[0][arrayindex][0];
- data[1] = span->data[0][arrayindex][1];
- data[2] = span->data[0][arrayindex][2];
- data[3] = span->data[0][arrayindex][3];
- slope[0] = span->data[1][arrayindex][0];
- slope[1] = span->data[1][arrayindex][1];
- slope[2] = span->data[1][arrayindex][2];
- slope[3] = span->data[1][arrayindex][3];
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, data, slope, arrayindex);
flags = texture->flags;
- pixelbase = (unsigned char *)texture->bytes + texture->mipmap[mip][0];
tcscale[0] = texture->mipmap[mip][2];
tcscale[1] = texture->mipmap[mip][3];
tciwidth = texture->mipmap[mip][2];
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;
for (x = startx;x < endx;)
{
- float endtc[2];
unsigned int subtc[2];
unsigned int substep[2];
- int endsub = x + DPSOFTRAST_MAXSUBSPAN-1;
- float subscale = 4096.0f/(DPSOFTRAST_MAXSUBSPAN-1);
- if (endsub >= endx)
+ float subscale = 65536.0f/DPSOFTRAST_DRAW_MAXSUBSPAN;
+ int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1;
+ if(nextsub >= endx)
{
- endsub = endx-1;
- subscale = endsub > x ? 4096.0f / (endsub - x) : 1.0f;
+ nextsub = endsub = endx-1;
+ if(x < nextsub) subscale = 65536.0f / (nextsub - x);
}
- tc[0] = (data[0] + slope[0]*x) * zf[x] * tcscale[0] - 0.5f;
- tc[1] = (data[1] + slope[1]*x) * zf[x] * tcscale[1] - 0.5f;
- endtc[0] = (data[0] + slope[0]*endsub) * zf[endsub] * tcscale[0] - 0.5f;
- endtc[1] = (data[1] + slope[1]*endsub) * zf[endsub] * tcscale[1] - 0.5f;
+ 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;
substep[0] = (endtc[0] - tc[0]) * subscale;
substep[1] = (endtc[1] - tc[1]) * subscale;
- subtc[0] = tc[0] * (1<<12);
- subtc[1] = tc[1] * (1<<12);
- if (!(flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE))
- {
- subtc[0] &= (tciwrapmask[0]<<12)|0xFFF;
- subtc[1] &= (tciwrapmask[1]<<12)|0xFFF;
- }
+ subtc[0] = tc[0] * (1<<16);
+ subtc[1] = tc[1] * (1<<16);
if(filter)
{
- tci[0] = (subtc[0]>>12) - tcimin[0];
- tci[1] = (subtc[1]>>12) - tcimin[0];
- tci1[0] = ((subtc[0] + (endsub - x)*substep[0])>>12) + 1;
- tci1[1] = ((subtc[1] + (endsub - x)*substep[1])>>12) + 1;
- if (tci[0] <= tcimax[0] && tci[1] <= tcimax[1] && tci1[0] <= tcimax[0] && tci1[1] <= tcimax[1])
- {
- for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1])
- {
- 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]>>12;
- tci[1] = subtc[1]>>12;
- pixel[0] = pixelbase + 4 * (tci[1]*tciwidth+tci[0]);
- pixel[1] = pixel[0] + 4 * tciwidth;
- c[0] = (pixel[0][2]*lerp[0]+pixel[0][4+2]*lerp[1]+pixel[1][2]*lerp[2]+pixel[1][4+2]*lerp[3]) * (1.0f / 0xFF000000);
- c[1] = (pixel[0][1]*lerp[0]+pixel[0][4+1]*lerp[1]+pixel[1][1]*lerp[2]+pixel[1][4+1]*lerp[3]) * (1.0f / 0xFF000000);
- c[2] = (pixel[0][0]*lerp[0]+pixel[0][4+0]*lerp[1]+pixel[1][0]*lerp[2]+pixel[1][4+0]*lerp[3]) * (1.0f / 0xFF000000);
- c[3] = (pixel[0][3]*lerp[0]+pixel[0][4+3]*lerp[1]+pixel[1][3]*lerp[2]+pixel[1][4+3]*lerp[3]) * (1.0f / 0xFF000000);
- out4f[x*4+0] = c[0];
- out4f[x*4+1] = c[1];
- out4f[x*4+2] = c[2];
- out4f[x*4+3] = c[3];
- }
- }
- else if (flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE)
+ if (flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE)
{
for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1])
{
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]>>12;
- tci[1] = subtc[1]>>12;
+ tci[0] = subtc[0]>>16;
+ tci[1] = subtc[1]>>16;
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];
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]>>12;
- tci[1] = subtc[1]>>12;
+ tci[0] = subtc[0]>>16;
+ tci[1] = subtc[1]>>16;
tci1[0] = tci[0] + 1;
tci1[1] = tci[1] + 1;
tci[0] &= tciwrapmask[0];
{
for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1])
{
- tci[0] = subtc[0]>>12;
- tci[1] = subtc[1]>>12;
+ tci[0] = subtc[0]>>16;
+ tci[1] = subtc[1]>>16;
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]);
{
for (; x <= endsub; x++, subtc[0] += substep[0], subtc[1] += substep[1])
{
- tci[0] = subtc[0]>>12;
- tci[1] = subtc[1]>>12;
+ tci[0] = subtc[0]>>16;
+ tci[1] = subtc[1]>>16;
tci[0] &= tciwrapmask[0];
tci[1] &= tciwrapmask[1];
pixel[0] = pixelbase + 4 * (tci[1]*tciwidth+tci[0]);
}
}
-void DPSOFTRAST_Draw_Span_MultiplyVarying(const DPSOFTRAST_State_Draw_Span *span, float *out4f, const float *in4f, int arrayindex, const float *zf)
+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
int x;
int startx = span->startx;
int endx = span->endx;
- float c[4];
- float data[4];
- float slope[4];
- float z;
- data[0] = span->data[0][arrayindex][0];
- data[1] = span->data[0][arrayindex][1];
- data[2] = span->data[0][arrayindex][2];
- data[3] = span->data[0][arrayindex][3];
- slope[0] = span->data[1][arrayindex][0];
- slope[1] = span->data[1][arrayindex][1];
- slope[2] = span->data[1][arrayindex][2];
- slope[3] = span->data[1][arrayindex][3];
- for (x = startx;x < endx;x++)
+ int flags;
+ __m128 data, slope, tcscale;
+ __m128i tcsize, tcmask, tcoffset, tcmax;
+ __m128 tc, endtc;
+ __m128i subtc, substep, endsubtc;
+ int filter;
+ int mip;
+ unsigned int *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)
{
- z = zf[x];
- c[0] = (data[0] + slope[0]*x) * z;
- c[1] = (data[1] + slope[1]*x) * z;
- c[2] = (data[2] + slope[2]*x) * z;
- c[3] = (data[3] + slope[3]*x) * z;
- out4f[x*4+0] = in4f[x*4+0] * c[0];
- out4f[x*4+1] = in4f[x*4+1] * c[1];
- out4f[x*4+2] = in4f[x*4+2] * c[2];
- out4f[x*4+3] = in4f[x*4+3] * c[3];
+ memset(out4ub + startx*4, 255, span->length*4);
+ return;
}
-}
-
-void DPSOFTRAST_Draw_Span_AddBloom(const DPSOFTRAST_State_Draw_Span *span, float *out4f, const float *ina4f, const float *inb4f, const float *subcolor)
-{
- int x, startx = span->startx, endx = span->endx;
- float c[4], localcolor[4];
- localcolor[0] = subcolor[0];
- localcolor[1] = subcolor[1];
- localcolor[2] = subcolor[2];
- localcolor[3] = subcolor[3];
- for (x = startx;x < endx;x++)
+ mip = triangle->mip[texunitindex];
+ pixelbase = (const unsigned char *)texture->bytes + texture->mipmap[mip][0];
+ // if this mipmap of the texture is 1 pixel, just fill it with that color
+ if (texture->mipmap[mip][1] == 4)
{
- c[0] = inb4f[x*4+0] - localcolor[0];if (c[0] < 0.0f) c[0] = 0.0f;
- c[1] = inb4f[x*4+1] - localcolor[1];if (c[1] < 0.0f) c[1] = 0.0f;
- c[2] = inb4f[x*4+2] - localcolor[2];if (c[2] < 0.0f) c[2] = 0.0f;
- c[3] = inb4f[x*4+3] - localcolor[3];if (c[3] < 0.0f) c[3] = 0.0f;
- out4f[x*4+0] = ina4f[x*4+0] + c[0];
- out4f[x*4+1] = ina4f[x*4+1] + c[1];
- out4f[x*4+2] = ina4f[x*4+2] + c[2];
- out4f[x*4+3] = ina4f[x*4+3] + c[3];
+ unsigned int k = *((const unsigned int *)pixelbase);
+ for (x = startx;x < endx;x++)
+ outi[x] = k;
+ return;
}
-}
-
-void DPSOFTRAST_Draw_Span_MixUniformColor(const DPSOFTRAST_State_Draw_Span *span, float *out4f, const float *in4f, const float *color)
-{
- int x, startx = span->startx, endx = span->endx;
- float localcolor[4], ilerp, lerp;
- localcolor[0] = color[0];
- localcolor[1] = color[1];
- localcolor[2] = color[2];
- localcolor[3] = color[3];
- ilerp = 1.0f - localcolor[3];
- lerp = localcolor[3];
- for (x = startx;x < endx;x++)
+ filter = texture->filter & DPSOFTRAST_TEXTURE_FILTER_LINEAR;
+ DPSOFTRAST_CALCATTRIB(triangle, span, data, slope, arrayindex);
+ flags = texture->flags;
+ tcsize = _mm_shuffle_epi32(_mm_loadu_si128((const __m128i *)&texture->mipmap[mip][0]), _MM_SHUFFLE(3, 2, 3, 2));
+ tcmask = _mm_sub_epi32(tcsize, _mm_set1_epi32(1));
+ 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));
+ 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);
+ for (x = startx;x < endx;)
{
- out4f[x*4+0] = in4f[x*4+0] * ilerp + localcolor[0] * lerp;
- out4f[x*4+1] = in4f[x*4+1] * ilerp + localcolor[1] * lerp;
- out4f[x*4+2] = in4f[x*4+2] * ilerp + localcolor[2] * lerp;
- out4f[x*4+3] = in4f[x*4+3] * ilerp + localcolor[3] * lerp;
+ int nextsub = x + DPSOFTRAST_DRAW_MAXSUBSPAN, endsub = nextsub - 1;
+ __m128 subscale = _mm_set1_ps(65536.0f/DPSOFTRAST_DRAW_MAXSUBSPAN);
+ if(nextsub >= endx)
+ {
+ nextsub = endsub = endx-1;
+ 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));
+ 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));
+ substep = _mm_slli_epi32(substep, 1);
+ if (filter)
+ {
+ __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)
+ {
+ for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep))
+ {
+ __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());
+ fracm = _mm_srli_epi16(subtc, 1);
+ pix1 = _mm_add_epi16(pix1,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix3 = _mm_add_epi16(pix3,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix3), 1),
+ _mm_shuffle_epi32(_mm_shufflehi_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(3, 2, 3, 2))));
+ pix2 = _mm_unpacklo_epi64(pix1, pix3);
+ pix4 = _mm_unpackhi_epi64(pix1, pix3);
+ pix2 = _mm_add_epi16(pix2,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix2), 1),
+ _mm_shufflehi_epi16(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0)), _MM_SHUFFLE(0, 0, 0, 0))));
+ _mm_storel_epi64((__m128i *)&outi[x], _mm_packus_epi16(pix2, _mm_shufflelo_epi16(pix2, _MM_SHUFFLE(3, 2, 3, 2))));
+ }
+ if (x <= endsub)
+ {
+ __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());
+ fracm = _mm_srli_epi16(subtc, 1);
+ pix1 = _mm_add_epi16(pix1,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix2 = _mm_shuffle_epi32(pix1, _MM_SHUFFLE(3, 2, 3, 2));
+ pix1 = _mm_add_epi16(pix1,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1),
+ _mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0))));
+ outi[x] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ x++;
+ }
+ }
+ else if (flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE)
+ {
+ for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep))
+ {
+ __m128i tci = _mm_shuffle_epi32(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(1, 0, 1, 0)), pix1, pix2, pix3, pix4, fracm;
+ tci = _mm_min_epi16(_mm_max_epi16(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), _mm_setzero_si128()), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix1 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix2 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _mm_setzero_si128());
+ tci = _mm_shuffle_epi32(_mm_shufflehi_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 2, 3, 2));
+ tci = _mm_and_si128(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix3 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix4 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _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),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix3 = _mm_add_epi16(pix3,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix3), 1),
+ _mm_shuffle_epi32(_mm_shufflehi_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(3, 2, 3, 2))));
+ pix2 = _mm_unpacklo_epi64(pix1, pix3);
+ pix4 = _mm_unpackhi_epi64(pix1, pix3);
+ pix2 = _mm_add_epi16(pix2,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix2), 1),
+ _mm_shufflehi_epi16(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0)), _MM_SHUFFLE(0, 0, 0, 0))));
+ _mm_storel_epi64((__m128i *)&outi[x], _mm_packus_epi16(pix2, _mm_shufflelo_epi16(pix2, _MM_SHUFFLE(3, 2, 3, 2))));
+ }
+ if (x <= endsub)
+ {
+ __m128i tci = _mm_shuffle_epi32(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(1, 0, 1, 0)), pix1, pix2, fracm;
+ tci = _mm_min_epi16(_mm_max_epi16(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), _mm_setzero_si128()), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix1 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix2 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _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),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix2 = _mm_shuffle_epi32(pix1, _MM_SHUFFLE(3, 2, 3, 2));
+ pix1 = _mm_add_epi16(pix1,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1),
+ _mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0))));
+ outi[x] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ x++;
+ }
+ }
+ else
+ {
+ for (; x + 1 <= endsub; x += 2, subtc = _mm_add_epi32(subtc, substep))
+ {
+ __m128i tci = _mm_shuffle_epi32(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(1, 0, 1, 0)), pix1, pix2, pix3, pix4, fracm;
+ tci = _mm_and_si128(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix1 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix2 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _mm_setzero_si128());
+ tci = _mm_shuffle_epi32(_mm_shufflehi_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(3, 2, 3, 2));
+ tci = _mm_and_si128(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix3 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix4 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _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),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix3 = _mm_add_epi16(pix3,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix3), 1),
+ _mm_shuffle_epi32(_mm_shufflehi_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(3, 2, 3, 2))));
+ pix2 = _mm_unpacklo_epi64(pix1, pix3);
+ pix4 = _mm_unpackhi_epi64(pix1, pix3);
+ pix2 = _mm_add_epi16(pix2,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix4, pix2), 1),
+ _mm_shufflehi_epi16(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0)), _MM_SHUFFLE(0, 0, 0, 0))));
+ _mm_storel_epi64((__m128i *)&outi[x], _mm_packus_epi16(pix2, _mm_shufflelo_epi16(pix2, _MM_SHUFFLE(3, 2, 3, 2))));
+ }
+ if (x <= endsub)
+ {
+ __m128i tci = _mm_shuffle_epi32(_mm_shufflelo_epi16(subtc, _MM_SHUFFLE(3, 1, 3, 1)), _MM_SHUFFLE(1, 0, 1, 0)), pix1, pix2, fracm;
+ tci = _mm_and_si128(_mm_add_epi16(tci, _mm_setr_epi32(0, 1, 0x10000, 0x10001)), tcmax);
+ tci = _mm_madd_epi16(tci, tcoffset);
+ pix1 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(tci)]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(1, 1, 1, 1)))])),
+ _mm_setzero_si128());
+ pix2 = _mm_unpacklo_epi8(_mm_unpacklo_epi32(_mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(2, 2, 2, 2)))]),
+ _mm_cvtsi32_si128(*(const int *)&pixelbase[_mm_cvtsi128_si32(_mm_shuffle_epi32(tci, _MM_SHUFFLE(3, 3, 3, 3)))])),
+ _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),
+ _mm_shuffle_epi32(_mm_shufflelo_epi16(fracm, _MM_SHUFFLE(2, 2, 2, 2)), _MM_SHUFFLE(1, 0, 1, 0))));
+ pix2 = _mm_shuffle_epi32(pix1, _MM_SHUFFLE(3, 2, 3, 2));
+ pix1 = _mm_add_epi16(pix1,
+ _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 1),
+ _mm_shufflelo_epi16(fracm, _MM_SHUFFLE(0, 0, 0, 0))));
+ outi[x] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ x++;
+ }
+ }
+ }
+ else
+ {
+ if (flags & DPSOFTRAST_TEXTURE_FLAG_CLAMPTOEDGE)
+ {
+ 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));
+ 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)))];
+ }
+ 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));
+ tci = _mm_madd_epi16(tci, tcoffset);
+ outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)];
+ x++;
+ }
+ }
+ else
+ {
+ 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));
+ 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)))];
+ }
+ if (x <= endsub)
+ {
+ __m128i tci = _mm_and_si128(subtc, tcmax);
+ tci = _mm_shufflelo_epi16(tci, _MM_SHUFFLE(3, 1, 3, 1));
+ tci = _mm_madd_epi16(tci, tcoffset);
+ outi[x] = *(const int *)&pixelbase[_mm_cvtsi128_si32(tci)];
+ x++;
+ }
+ }
+ }
}
+#endif
}
-void DPSOFTRAST_Draw_Span_Lightmap(const DPSOFTRAST_State_Draw_Span *span, float * RESTRICT out4f, const float * RESTRICT diffuse, const float * RESTRICT lightmap)
+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)
{
- int x, startx = span->startx, endx = span->endx;
- float Color_Ambient[4], Color_Diffuse[4];
- Color_Ambient[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0];
- Color_Ambient[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1];
- Color_Ambient[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2];
- Color_Ambient[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0];
- Color_Diffuse[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0];
- Color_Diffuse[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1];
- Color_Diffuse[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2];
- Color_Diffuse[3] = 0.0f;
- for (x = startx;x < endx;x++)
- {
- out4f[x*4+0] = diffuse[x*4+0] * (Color_Ambient[0] + lightmap[x*4+0] * Color_Diffuse[0]);
- out4f[x*4+1] = diffuse[x*4+1] * (Color_Ambient[1] + lightmap[x*4+1] * Color_Diffuse[1]);
- out4f[x*4+2] = diffuse[x*4+2] * (Color_Ambient[2] + lightmap[x*4+2] * Color_Diffuse[2]);
- out4f[x*4+3] = diffuse[x*4+3] * (Color_Ambient[3] + lightmap[x*4+3] * Color_Diffuse[3]);
- }
+ // TODO: IMPLEMENT
+ memset(out4ub, 255, span->length*4);
}
-void DPSOFTRAST_Draw_Span_Lightmap_Finish(const DPSOFTRAST_State_Draw_Span *span, const float * RESTRICT diffuse, const float * RESTRICT lightmap)
+float DPSOFTRAST_SampleShadowmap(const float *vector)
{
- int x, startx = span->startx, endx = span->endx;
- int d[4];
- float Color_Ambient[4], Color_Diffuse[4];
- unsigned char * RESTRICT pixelmask = span->pixelmask;
- unsigned char * RESTRICT pixel = (unsigned char *)dpsoftrast.fb_colorpixels[0];
- if (!pixel)
- return;
- pixel += span->start * 4;
- Color_Ambient[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0]*255.0f;
- Color_Ambient[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1]*255.0f;
- Color_Ambient[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2]*255.0f;
- Color_Ambient[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0]*255.0f;
- Color_Diffuse[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0]*255.0f;
- Color_Diffuse[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1]*255.0f;
- Color_Diffuse[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2]*255.0f;
- Color_Diffuse[3] = 0.0f;
+ // TODO: IMPLEMENT
+ return 1.0f;
+}
+
+void DPSOFTRAST_Draw_Span_MultiplyVarying(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *in4f, int arrayindex, const float *zf)
+{
+ int x;
+ int startx = span->startx;
+ int endx = span->endx;
+ float c[4];
+ float data[4];
+ float slope[4];
+ float z;
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, data, slope, arrayindex);
for (x = startx;x < endx;x++)
{
- if (!pixelmask[x])
- continue;
- d[0] = diffuse[x*4+0] * (Color_Ambient[0] + lightmap[x*4+0] * Color_Diffuse[0]);if (d[0] > 255) d[0] = 255;
- d[1] = diffuse[x*4+1] * (Color_Ambient[1] + lightmap[x*4+1] * Color_Diffuse[1]);if (d[1] > 255) d[1] = 255;
- d[2] = diffuse[x*4+2] * (Color_Ambient[2] + lightmap[x*4+2] * Color_Diffuse[2]);if (d[2] > 255) d[2] = 255;
- d[3] = diffuse[x*4+3] * (Color_Ambient[3] + lightmap[x*4+3] * Color_Diffuse[3]);if (d[3] > 255) d[3] = 255;
- pixel[x*4+0] = d[2];
- pixel[x*4+1] = d[1];
- pixel[x*4+2] = d[0];
- pixel[x*4+3] = d[3];
+ z = zf[x];
+ c[0] = (data[0] + slope[0]*x) * z;
+ c[1] = (data[1] + slope[1]*x) * z;
+ c[2] = (data[2] + slope[2]*x) * z;
+ c[3] = (data[3] + slope[3]*x) * z;
+ out4f[x*4+0] = in4f[x*4+0] * c[0];
+ out4f[x*4+1] = in4f[x*4+1] * c[1];
+ out4f[x*4+2] = in4f[x*4+2] * c[2];
+ out4f[x*4+3] = in4f[x*4+3] * c[3];
}
}
-void DPSOFTRAST_Draw_Span_VertexColor(const DPSOFTRAST_State_Draw_Span *span, float *out4f, const float *diffuse, const float *zf)
+void DPSOFTRAST_Draw_Span_Varying(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, int arrayindex, const float *zf)
{
- int x, startx = span->startx, endx = span->endx;
- float Color_Ambient[4], Color_Diffuse[4];
+ int x;
+ int startx = span->startx;
+ int endx = span->endx;
float c[4];
float data[4];
float slope[4];
float z;
- int arrayindex = DPSOFTRAST_ARRAY_COLOR;
- data[0] = span->data[0][arrayindex][0];
- data[1] = span->data[0][arrayindex][1];
- data[2] = span->data[0][arrayindex][2];
- data[3] = span->data[0][arrayindex][3];
- slope[0] = span->data[1][arrayindex][0];
- slope[1] = span->data[1][arrayindex][1];
- slope[2] = span->data[1][arrayindex][2];
- slope[3] = span->data[1][arrayindex][3];
- Color_Ambient[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0];
- Color_Ambient[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1];
- Color_Ambient[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2];
- Color_Ambient[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0];
- Color_Diffuse[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0];
- Color_Diffuse[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1];
- Color_Diffuse[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2];
- Color_Diffuse[3] = 0.0f;
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, data, slope, arrayindex);
for (x = startx;x < endx;x++)
{
z = zf[x];
c[1] = (data[1] + slope[1]*x) * z;
c[2] = (data[2] + slope[2]*x) * z;
c[3] = (data[3] + slope[3]*x) * z;
- out4f[x*4+0] = diffuse[x*4+0] * (Color_Ambient[0] + c[0] * Color_Diffuse[0]);
- out4f[x*4+1] = diffuse[x*4+1] * (Color_Ambient[1] + c[1] * Color_Diffuse[1]);
- out4f[x*4+2] = diffuse[x*4+2] * (Color_Ambient[2] + c[2] * Color_Diffuse[2]);
- out4f[x*4+3] = diffuse[x*4+3] * (Color_Ambient[3] + c[3] * Color_Diffuse[3]);
+ out4f[x*4+0] = c[0];
+ out4f[x*4+1] = c[1];
+ out4f[x*4+2] = c[2];
+ out4f[x*4+3] = c[3];
+ }
+}
+
+void DPSOFTRAST_Draw_Span_AddBloom(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *ina4f, const float *inb4f, const float *subcolor)
+{
+ int x, startx = span->startx, endx = span->endx;
+ float c[4], localcolor[4];
+ localcolor[0] = subcolor[0];
+ localcolor[1] = subcolor[1];
+ localcolor[2] = subcolor[2];
+ localcolor[3] = subcolor[3];
+ for (x = startx;x < endx;x++)
+ {
+ c[0] = inb4f[x*4+0] - localcolor[0];if (c[0] < 0.0f) c[0] = 0.0f;
+ c[1] = inb4f[x*4+1] - localcolor[1];if (c[1] < 0.0f) c[1] = 0.0f;
+ c[2] = inb4f[x*4+2] - localcolor[2];if (c[2] < 0.0f) c[2] = 0.0f;
+ c[3] = inb4f[x*4+3] - localcolor[3];if (c[3] < 0.0f) c[3] = 0.0f;
+ out4f[x*4+0] = ina4f[x*4+0] + c[0];
+ out4f[x*4+1] = ina4f[x*4+1] + c[1];
+ out4f[x*4+2] = ina4f[x*4+2] + c[2];
+ out4f[x*4+3] = ina4f[x*4+3] + c[3];
}
}
-void DPSOFTRAST_Draw_Span_FlatColor(const DPSOFTRAST_State_Draw_Span *span, float *out4f, const float *diffuse)
+void DPSOFTRAST_Draw_Span_MultiplyBuffers(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *ina4f, const float *inb4f)
{
int x, startx = span->startx, endx = span->endx;
- float Color_Ambient[4];
- Color_Ambient[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0];
- Color_Ambient[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1];
- Color_Ambient[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2];
- Color_Ambient[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0];
for (x = startx;x < endx;x++)
{
- out4f[x*4+0] = diffuse[x*4+0] * Color_Ambient[0];
- out4f[x*4+1] = diffuse[x*4+1] * Color_Ambient[1];
- out4f[x*4+2] = diffuse[x*4+2] * Color_Ambient[2];
- out4f[x*4+3] = diffuse[x*4+3] * Color_Ambient[3];
+ out4f[x*4+0] = ina4f[x*4+0] * inb4f[x*4+0];
+ out4f[x*4+1] = ina4f[x*4+1] * inb4f[x*4+1];
+ out4f[x*4+2] = ina4f[x*4+2] * inb4f[x*4+2];
+ out4f[x*4+3] = ina4f[x*4+3] * inb4f[x*4+3];
}
}
-void DPSOFTRAST_Draw_VertexShader(void)
+void DPSOFTRAST_Draw_Span_AddBuffers(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *ina4f, const float *inb4f)
{
- DPSOFTRAST_Array_Transform(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_POSITION], dpsoftrast.draw.numvertices, dpsoftrast.uniform4f + 4*DPSOFTRAST_UNIFORM_ModelViewProjectionMatrixM1);
- switch(dpsoftrast.shader_mode)
+ int x, startx = span->startx, endx = span->endx;
+ for (x = startx;x < endx;x++)
{
- case SHADERMODE_GENERIC: ///< (particles/HUD/etc) vertex color: optionally multiplied by one texture
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.draw.numvertices);
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.numvertices);
- break;
- case SHADERMODE_POSTPROCESS: ///< postprocessing shader (r_glsl_postprocess)
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.numvertices);
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.draw.numvertices);
- break;
- case SHADERMODE_DEPTH_OR_SHADOW: ///< (depthfirst/shadows) vertex shader only
- break;
- case SHADERMODE_FLATCOLOR: ///< (lightmap) modulate texture by uniform color (q1bsp: q3bsp)
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.numvertices);
- break;
- case SHADERMODE_VERTEXCOLOR: ///< (lightmap) modulate texture by vertex colors (q3bsp)
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_COLOR], dpsoftrast.draw.numvertices);
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.numvertices);
- break;
- case SHADERMODE_LIGHTMAP: ///< (lightmap) modulate texture by lightmap texture (q1bsp: q3bsp)
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD0], dpsoftrast.draw.numvertices);
- DPSOFTRAST_Array_Copy(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.draw.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD4], dpsoftrast.draw.numvertices);
- break;
- case SHADERMODE_FAKELIGHT: ///< (fakelight) modulate texture by "fake" lighting (no lightmaps: no nothing)
- break;
- case SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE: ///< (lightmap) use directional pixel shading from texture containing modelspace light directions (q3bsp deluxemap)
- break;
- case SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE: ///< (lightmap) use directional pixel shading from texture containing tangentspace light directions (q1bsp deluxemap)
- break;
- case SHADERMODE_LIGHTDIRECTION: ///< (lightmap) use directional pixel shading from fixed light direction (q3bsp)
- break;
- case SHADERMODE_LIGHTSOURCE: ///< (lightsource) use directional pixel shading from light source (rtlight)
- break;
- case SHADERMODE_REFRACTION: ///< refract background (the material is rendered normally after this pass)
- break;
- case SHADERMODE_WATER: ///< refract background and reflection (the material is rendered normally after this pass)
- break;
- case SHADERMODE_SHOWDEPTH: ///< (debugging) renders depth as color
- break;
- case SHADERMODE_DEFERREDGEOMETRY: ///< (deferred) render material properties to screenspace geometry buffers
- break;
- case SHADERMODE_DEFERREDLIGHTSOURCE: ///< (deferred) use directional pixel shading from light source (rtlight) on screenspace geometry buffers
- break;
- case SHADERMODE_COUNT:
- break;
+ out4f[x*4+0] = ina4f[x*4+0] + inb4f[x*4+0];
+ out4f[x*4+1] = ina4f[x*4+1] + inb4f[x*4+1];
+ out4f[x*4+2] = ina4f[x*4+2] + inb4f[x*4+2];
+ out4f[x*4+3] = ina4f[x*4+3] + inb4f[x*4+3];
+ }
+}
+
+void DPSOFTRAST_Draw_Span_MixBuffers(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *ina4f, const float *inb4f)
+{
+ int x, startx = span->startx, endx = span->endx;
+ float a, b;
+ for (x = startx;x < endx;x++)
+ {
+ a = 1.0f - inb4f[x*4+3];
+ b = inb4f[x*4+3];
+ out4f[x*4+0] = ina4f[x*4+0] * a + inb4f[x*4+0] * b;
+ out4f[x*4+1] = ina4f[x*4+1] * a + inb4f[x*4+1] * b;
+ out4f[x*4+2] = ina4f[x*4+2] * a + inb4f[x*4+2] * b;
+ out4f[x*4+3] = ina4f[x*4+3] * a + inb4f[x*4+3] * b;
+ }
+}
+
+void DPSOFTRAST_Draw_Span_MixUniformColor(const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span, float *out4f, const float *in4f, const float *color)
+{
+ int x, startx = span->startx, endx = span->endx;
+ float localcolor[4], ilerp, lerp;
+ localcolor[0] = color[0];
+ localcolor[1] = color[1];
+ localcolor[2] = color[2];
+ localcolor[3] = color[3];
+ ilerp = 1.0f - localcolor[3];
+ lerp = localcolor[3];
+ for (x = startx;x < endx;x++)
+ {
+ out4f[x*4+0] = in4f[x*4+0] * ilerp + localcolor[0] * lerp;
+ out4f[x*4+1] = in4f[x*4+1] * ilerp + localcolor[1] * lerp;
+ out4f[x*4+2] = in4f[x*4+2] * ilerp + localcolor[2] * lerp;
+ out4f[x*4+3] = in4f[x*4+3] * ilerp + localcolor[3] * lerp;
+ }
+}
+
+
+
+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
+ int x;
+ int startx = span->startx;
+ int endx = span->endx;
+ __m128 data, slope;
+ 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))
+ {
+ __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
+}
+
+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
+ int x;
+ int startx = span->startx;
+ int endx = span->endx;
+ __m128 data, slope;
+ 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))
+ {
+ __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));
+ }
+#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
+ 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));
+ 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));
+ _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1));
+ }
+ 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));
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ }
+#endif
+}
+
+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
+ int x, startx = span->startx, endx = span->endx;
+ 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_setzero_si128(), _mm_loadl_epi64((const __m128i *)&inb4ub[x*4]));
+ pix1 = _mm_mulhi_epu16(pix1, pix2);
+ _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1));
+ }
+ 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]));
+ pix1 = _mm_mulhi_epu16(pix1, pix2);
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ }
+#endif
+}
+
+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
+ int x, startx = span->startx, endx = span->endx;
+ 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, pix2);
+ _mm_storel_epi64((__m128i *)&out4ub[x*4], _mm_packus_epi16(pix1, pix1));
+ }
+ 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, pix2);
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ }
+#endif
+}
+
+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
+ 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));
+ 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_setzero_si128(), _mm_loadl_epi64((const __m128i *)&inb4ub[x*4]));
+ 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)
+ {
+ __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]));
+ pix1 = _mm_add_epi16(pix1, _mm_mulhi_epu16(tint, pix2));
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ }
+#endif
+}
+
+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
+ int x, startx = span->startx, endx = span->endx;
+ 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());
+ __m128i blend = _mm_shufflehi_epi16(_mm_shufflelo_epi16(pix2, _MM_SHUFFLE(3, 3, 3, 3)), _MM_SHUFFLE(3, 3, 3, 3));
+ 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)
+ {
+ __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());
+ __m128i blend = _mm_shufflelo_epi16(pix2, _MM_SHUFFLE(3, 3, 3, 3));
+ pix1 = _mm_add_epi16(pix1, _mm_mulhi_epi16(_mm_slli_epi16(_mm_sub_epi16(pix2, pix1), 4), _mm_slli_epi16(blend, 4)));
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix1, pix1));
+ }
+#endif
+}
+
+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
+ 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));
+ 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)
+ {
+ __m128i pix = _mm_unpacklo_epi8(_mm_loadl_epi64((const __m128i *)&in4ub[x*4]), _mm_setzero_si128());
+ 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)
+ {
+ __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));
+ *(int *)&out4ub[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix));
+ }
+#endif
+}
+
+
+
+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);
+ if (dpsoftrast.shader_permutation & SHADERPERMUTATION_SPECULAR)
+ DPSOFTRAST_Array_Copy(dpsoftrast.post_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.in_array4f[DPSOFTRAST_ARRAY_TEXCOORD1], dpsoftrast.numvertices);
+}
+
+void DPSOFTRAST_PixelShader_Generic(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_lightmapbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_DIFFUSE)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_FIRST, 2, buffer_z);
+ DPSOFTRAST_Draw_Span_MultiplyVaryingBGRA8(triangle, span, buffer_FragColorbgra8, buffer_texture_colorbgra8, 1, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_SPECULAR)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_SECOND, 2, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ // multiply
+ DPSOFTRAST_Draw_Span_MultiplyBuffersBGRA8(triangle, span, buffer_FragColorbgra8, buffer_FragColorbgra8, buffer_texture_lightmapbgra8);
+ }
+ else if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ // add
+ DPSOFTRAST_Draw_Span_AddBuffersBGRA8(triangle, span, buffer_FragColorbgra8, buffer_FragColorbgra8, buffer_texture_lightmapbgra8);
+ }
+ else if (thread->shader_permutation & SHADERPERMUTATION_VERTEXTEXTUREBLEND)
+ {
+ // alphablend
+ DPSOFTRAST_Draw_Span_MixBuffersBGRA8(triangle, span, buffer_FragColorbgra8, buffer_FragColorbgra8, buffer_texture_lightmapbgra8);
+ }
+ }
}
+ else
+ DPSOFTRAST_Draw_Span_VaryingBGRA8(triangle, span, buffer_FragColorbgra8, 1, buffer_z);
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_PostProcess(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+ // TODO: optimize!! at the very least there is no reason to use texture sampling on the frame texture
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z);
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_FragColorbgra8, GL20TU_FIRST, 2, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_BLOOM)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_SECOND, 3, buffer_z);
+ DPSOFTRAST_Draw_Span_AddBloomBGRA8(triangle, span, buffer_FragColorbgra8, buffer_FragColorbgra8, buffer_texture_colorbgra8, thread->uniform4f + DPSOFTRAST_UNIFORM_BloomColorSubtract * 4);
+ }
+ DPSOFTRAST_Draw_Span_MixUniformColorBGRA8(triangle, span, buffer_FragColorbgra8, buffer_FragColorbgra8, thread->uniform4f + DPSOFTRAST_UNIFORM_ViewTintColor * 4);
+ if (thread->shader_permutation & SHADERPERMUTATION_SATURATION)
+ {
+ // TODO: implement saturation
+ }
+ if (thread->shader_permutation & SHADERPERMUTATION_GAMMARAMPS)
+ {
+ // TODO: implement gammaramps
+ }
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_Depth_Or_Shadow(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+ // this is never called (because colormask is off when this shader is used)
+ 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);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_FlatColor(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+ int x, startx = span->startx, endx = span->endx;
+ int Color_Ambienti[4];
+ 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);
+ 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;
+ }
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_VertexColor(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+#ifdef SSE2_PRESENT
+ 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;
+ __m128i Color_Ambientm, Color_Diffusem;
+ __m128 data, slope;
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ 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)
+ 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);
+ Color_Diffusem = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4]), _mm_set1_ps(4096.0f))), _MM_SHUFFLE(3, 0, 1, 2));
+ Color_Diffusem = _mm_and_si128(Color_Diffusem, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
+ Color_Diffusem = _mm_packs_epi32(Color_Diffusem, Color_Diffusem);
+ 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(4096.0f));
+ slope = _mm_mul_ps(slope, _mm_set1_ps(4096.0f));
+ for (x = startx;x < endx;x++, data = _mm_add_ps(data, slope))
+ {
+ __m128i color, mod, pix;
+ if (x + 4 <= endx && *(const unsigned int *)&pixelmask[x] == 0x01010101)
+ {
+ __m128i pix2, mod2;
+ __m128 z = _mm_loadu_ps(&buffer_z[x]);
+ color = _mm_loadu_si128((const __m128i *)&buffer_texture_colorbgra8[x*4]);
+ mod = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_shuffle_ps(z, z, _MM_SHUFFLE(0, 0, 0, 0))));
+ data = _mm_add_ps(data, slope);
+ mod = _mm_packs_epi32(mod, _mm_cvtps_epi32(_mm_mul_ps(data, _mm_shuffle_ps(z, z, _MM_SHUFFLE(1, 1, 1, 1)))));
+ data = _mm_add_ps(data, slope);
+ mod2 = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_shuffle_ps(z, z, _MM_SHUFFLE(2, 2, 2, 2))));
+ data = _mm_add_ps(data, slope);
+ mod2 = _mm_packs_epi32(mod2, _mm_cvtps_epi32(_mm_mul_ps(data, _mm_shuffle_ps(z, z, _MM_SHUFFLE(3, 3, 3, 3)))));
+ pix = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, mod), Color_Ambientm),
+ _mm_unpacklo_epi8(_mm_setzero_si128(), color));
+ pix2 = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, mod2), 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]));
+ mod = _mm_cvtps_epi32(_mm_mul_ps(data, _mm_load1_ps(&buffer_z[x])));
+ mod = _mm_packs_epi32(mod, mod);
+ 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)
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+#endif
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_Lightmap(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+#ifdef SSE2_PRESENT
+ 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;
+ __m128i Color_Ambientm, Color_Diffusem, Color_Glowm, Color_AmbientGlowm;
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_lightmapbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_glowbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_FragColorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ 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)
+ 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);
+ Color_Diffusem = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2));
+ Color_Diffusem = _mm_and_si128(Color_Diffusem, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
+ Color_Diffusem = _mm_packs_epi32(Color_Diffusem, Color_Diffusem);
+ if (thread->shader_permutation & SHADERPERMUTATION_GLOW)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_glowbgra8, GL20TU_GLOW, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ Color_Glowm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2));
+ Color_Glowm = _mm_and_si128(Color_Glowm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
+ Color_Glowm = _mm_packs_epi32(Color_Glowm, Color_Glowm);
+ Color_AmbientGlowm = _mm_unpacklo_epi64(Color_Ambientm, Color_Glowm);
+ for (x = startx;x < endx;x++)
+ {
+ __m128i color, lightmap, glow, pix;
+ if (x + 4 <= endx && *(const unsigned int *)&pixelmask[x] == 0x01010101)
+ {
+ __m128i pix2;
+ color = _mm_loadu_si128((const __m128i *)&buffer_texture_colorbgra8[x*4]);
+ lightmap = _mm_loadu_si128((const __m128i *)&buffer_texture_lightmapbgra8[x*4]);
+ glow = _mm_loadu_si128((const __m128i *)&buffer_texture_glowbgra8[x*4]);
+ pix = _mm_add_epi16(_mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, _mm_unpacklo_epi8(_mm_setzero_si128(), lightmap)), Color_Ambientm),
+ _mm_unpacklo_epi8(_mm_setzero_si128(), color)),
+ _mm_mulhi_epu16(Color_Glowm, _mm_unpacklo_epi8(_mm_setzero_si128(), glow)));
+ pix2 = _mm_add_epi16(_mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, _mm_unpackhi_epi8(_mm_setzero_si128(), lightmap)), Color_Ambientm),
+ _mm_unpackhi_epi8(_mm_setzero_si128(), color)),
+ _mm_mulhi_epu16(Color_Glowm, _mm_unpackhi_epi8(_mm_setzero_si128(), glow)));
+ _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]));
+ lightmap = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_lightmapbgra8[x*4]));
+ glow = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_glowbgra8[x*4]));
+ pix = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, lightmap), Color_AmbientGlowm), _mm_unpacklo_epi64(color, glow));
+ pix = _mm_add_epi16(pix, _mm_shuffle_epi32(pix, _MM_SHUFFLE(3, 2, 3, 2)));
+ *(int *)&pixel[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix));
+ }
+ }
+ else
+ {
+ for (x = startx;x < endx;x++)
+ {
+ __m128i color, lightmap, pix;
+ if (x + 4 <= endx && *(const unsigned int *)&pixelmask[x] == 0x01010101)
+ {
+ __m128i pix2;
+ color = _mm_loadu_si128((const __m128i *)&buffer_texture_colorbgra8[x*4]);
+ lightmap = _mm_loadu_si128((const __m128i *)&buffer_texture_lightmapbgra8[x*4]);
+ pix = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, _mm_unpacklo_epi8(_mm_setzero_si128(), lightmap)), Color_Ambientm),
+ _mm_unpacklo_epi8(_mm_setzero_si128(), color));
+ pix2 = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(Color_Diffusem, _mm_unpackhi_epi8(_mm_setzero_si128(), lightmap)), 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]));
+ lightmap = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_cvtsi32_si128(*(const int *)&buffer_texture_lightmapbgra8[x*4]));
+ pix = _mm_mulhi_epu16(_mm_add_epi16(_mm_mulhi_epu16(lightmap, Color_Diffusem), Color_Ambientm), color);
+ *(int *)&pixel[x*4] = _mm_cvtsi128_si32(_mm_packus_epi16(pix, pix));
+ }
+ }
+ if(pixel == buffer_FragColorbgra8)
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+#endif
+}
+
+
+
+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);
+}
+
+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);
+}
+
+
+
+void DPSOFTRAST_VertexShader_LightDirectionMap_ModelSpace(void)
+{
+ DPSOFTRAST_VertexShader_Lightmap();
+}
+
+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
+}
+
+
+
+void DPSOFTRAST_VertexShader_LightDirectionMap_TangentSpace(void)
+{
+ DPSOFTRAST_VertexShader_Lightmap();
+}
+
+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
+}
+
+
+
+void DPSOFTRAST_VertexShader_LightDirection(void)
+{
+ int i;
+ int numvertices = dpsoftrast.numvertices;
+ float LightDir[4];
+ float LightVector[4];
+ float EyePosition[4];
+ float EyeVectorModelSpace[4];
+ float EyeVector[4];
+ float position[4];
+ float svector[4];
+ float tvector[4];
+ float normal[4];
+ LightDir[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightDir*4+0];
+ LightDir[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightDir*4+1];
+ LightDir[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightDir*4+2];
+ LightDir[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightDir*4+3];
+ 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_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);
+ 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];
+ 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;
+ 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;
+ }
+}
+
+#define DPSOFTRAST_Min(a,b) ((a) < (b) ? (a) : (b))
+#define DPSOFTRAST_Max(a,b) ((a) > (b) ? (a) : (b))
+#define DPSOFTRAST_Vector3Dot(a,b) ((a)[0]*(b)[0]+(a)[1]*(b)[1]+(a)[2]*(b)[2])
+#define DPSOFTRAST_Vector3LengthSquared(v) (DPSOFTRAST_Vector3Dot((v),(v)))
+#define DPSOFTRAST_Vector3Length(v) (sqrt(DPSOFTRAST_Vector3LengthSquared(v)))
+#define DPSOFTRAST_Vector3Normalize(v)\
+do\
+{\
+ float len = sqrt(DPSOFTRAST_Vector3Dot(v,v));\
+ if (len)\
+ {\
+ len = 1.0f / len;\
+ v[0] *= len;\
+ v[1] *= len;\
+ v[2] *= len;\
+ }\
+}\
+while(0)
+
+void DPSOFTRAST_PixelShader_LightDirection(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_normalbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_glossbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ 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_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];
+ float LightVectordata[4];
+ float LightVectorslope[4];
+ float EyeVectordata[4];
+ float EyeVectorslope[4];
+ float z;
+ float diffusetex[4];
+ float glosstex[4];
+ float surfacenormal[4];
+ float lightnormal[4];
+ float eyenormal[4];
+ float specularnormal[4];
+ float diffuse;
+ float specular;
+ float SpecularPower;
+ int d[4];
+ Color_Glow[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+0];
+ Color_Glow[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+1];
+ Color_Glow[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+2];
+ Color_Glow[3] = 0.0f;
+ Color_Ambient[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0];
+ Color_Ambient[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1];
+ Color_Ambient[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2];
+ Color_Ambient[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0];
+ Color_Pants[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+0];
+ Color_Pants[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+1];
+ Color_Pants[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+2];
+ Color_Pants[3] = 0.0f;
+ Color_Shirt[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+0];
+ Color_Shirt[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+1];
+ Color_Shirt[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+2];
+ Color_Shirt[3] = 0.0f;
+ 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);
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_pantsbgra8, GL20TU_PANTS, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_shirtbgra8, GL20TU_SHIRT, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ }
+ if (thread->shader_permutation & SHADERPERMUTATION_GLOW)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_glowbgra8, GL20TU_GLOW, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ }
+ if (thread->shader_permutation & SHADERPERMUTATION_SPECULAR)
+ {
+ Color_Diffuse[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0];
+ Color_Diffuse[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1];
+ Color_Diffuse[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2];
+ Color_Diffuse[3] = 0.0f;
+ LightColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+0];
+ 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_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_glossbgra8, GL20TU_GLOSS, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ diffusetex[0] += buffer_texture_pantsbgra8[x*4+0] * Color_Pants[0] + buffer_texture_shirtbgra8[x*4+0] * Color_Shirt[0];
+ diffusetex[1] += buffer_texture_pantsbgra8[x*4+1] * Color_Pants[1] + buffer_texture_shirtbgra8[x*4+1] * Color_Shirt[1];
+ diffusetex[2] += buffer_texture_pantsbgra8[x*4+2] * Color_Pants[2] + buffer_texture_shirtbgra8[x*4+2] * Color_Shirt[2];
+ diffusetex[3] += buffer_texture_pantsbgra8[x*4+3] * Color_Pants[3] + buffer_texture_shirtbgra8[x*4+3] * Color_Shirt[3];
+ }
+ glosstex[0] = buffer_texture_glossbgra8[x*4+0];
+ glosstex[1] = buffer_texture_glossbgra8[x*4+1];
+ glosstex[2] = buffer_texture_glossbgra8[x*4+2];
+ glosstex[3] = buffer_texture_glossbgra8[x*4+3];
+ surfacenormal[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f;
+ surfacenormal[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f;
+ 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);
+
+ 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);
+
+ 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)
+ {
+ d[0] = (int)(buffer_texture_glowbgra8[x*4+0] * Color_Glow[0] + diffusetex[0] * Color_Ambient[0] + (diffusetex[0] * Color_Diffuse[0] * diffuse + glosstex[0] * Color_Specular[0] * specular) * LightColor[0]);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)(buffer_texture_glowbgra8[x*4+1] * Color_Glow[1] + diffusetex[1] * Color_Ambient[1] + (diffusetex[1] * Color_Diffuse[1] * diffuse + glosstex[1] * Color_Specular[1] * specular) * LightColor[1]);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)(buffer_texture_glowbgra8[x*4+2] * Color_Glow[2] + 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;
+ }
+ else
+ {
+ d[0] = (int)( diffusetex[0] * Color_Ambient[0] + (diffusetex[0] * Color_Diffuse[0] * diffuse + glosstex[0] * Color_Specular[0] * specular) * LightColor[0]);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)( diffusetex[1] * Color_Ambient[1] + (diffusetex[1] * Color_Diffuse[1] * diffuse + glosstex[1] * Color_Specular[1] * specular) * LightColor[1]);if (d[1] > 255) d[1] = 255;
+ 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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ else if (thread->shader_permutation & SHADERPERMUTATION_DIFFUSE)
+ {
+ Color_Diffuse[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0];
+ Color_Diffuse[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1];
+ Color_Diffuse[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2];
+ Color_Diffuse[3] = 0.0f;
+ LightColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+0];
+ 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);
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+ surfacenormal[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f;
+ surfacenormal[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f;
+ 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);
+
+ diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f;
+ if (thread->shader_permutation & SHADERPERMUTATION_GLOW)
+ {
+ d[0] = (int)(buffer_texture_glowbgra8[x*4+0] * Color_Glow[0] + diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse * LightColor[0]));if (d[0] > 255) d[0] = 255;
+ d[1] = (int)(buffer_texture_glowbgra8[x*4+1] * Color_Glow[1] + diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse * LightColor[1]));if (d[1] > 255) d[1] = 255;
+ d[2] = (int)(buffer_texture_glowbgra8[x*4+2] * Color_Glow[2] + diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse * LightColor[2]));if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[3] * (Color_Ambient[3] ));if (d[3] > 255) d[3] = 255;
+ }
+ else
+ {
+ d[0] = (int)( + diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse * LightColor[0]));if (d[0] > 255) d[0] = 255;
+ d[1] = (int)( + diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse * LightColor[1]));if (d[1] > 255) d[1] = 255;
+ d[2] = (int)( + diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse * 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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ else
+ {
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+
+ if (thread->shader_permutation & SHADERPERMUTATION_GLOW)
+ {
+ d[0] = (int)(buffer_texture_glowbgra8[x*4+0] * Color_Glow[0] + diffusetex[0] * Color_Ambient[0]);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)(buffer_texture_glowbgra8[x*4+1] * Color_Glow[1] + diffusetex[1] * Color_Ambient[1]);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)(buffer_texture_glowbgra8[x*4+2] * Color_Glow[2] + diffusetex[2] * Color_Ambient[2]);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[3] * Color_Ambient[3]);if (d[3] > 255) d[3] = 255;
+ }
+ else
+ {
+ d[0] = (int)( diffusetex[0] * Color_Ambient[0]);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)( diffusetex[1] * Color_Ambient[1]);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)( diffusetex[2] * Color_Ambient[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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ DPSOFTRAST_Draw_Span_FinishBGRA8(thread, triangle, span, buffer_FragColorbgra8);
+}
+
+
+
+void DPSOFTRAST_VertexShader_LightSource(void)
+{
+ int i;
+ int numvertices = dpsoftrast.numvertices;
+ float LightPosition[4];
+ float LightVector[4];
+ float LightVectorModelSpace[4];
+ float EyePosition[4];
+ float EyeVectorModelSpace[4];
+ float EyeVector[4];
+ float position[4];
+ float svector[4];
+ float tvector[4];
+ float normal[4];
+ LightPosition[0] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightPosition*4+0];
+ LightPosition[1] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightPosition*4+1];
+ LightPosition[2] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightPosition*4+2];
+ LightPosition[3] = dpsoftrast.uniform4f[DPSOFTRAST_UNIFORM_LightPosition*4+3];
+ 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_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);
+ 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];
+ LightVectorModelSpace[0] = LightPosition[0] - position[0];
+ LightVectorModelSpace[1] = LightPosition[1] - position[1];
+ LightVectorModelSpace[2] = LightPosition[2] - position[2];
+ LightVector[0] = svector[0] * LightVectorModelSpace[0] + svector[1] * LightVectorModelSpace[1] + svector[2] * LightVectorModelSpace[2];
+ LightVector[1] = tvector[0] * LightVectorModelSpace[0] + tvector[1] * LightVectorModelSpace[1] + tvector[2] * LightVectorModelSpace[2];
+ LightVector[2] = normal[0] * LightVectorModelSpace[0] + normal[1] * LightVectorModelSpace[1] + normal[2] * LightVectorModelSpace[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;
+ 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;
+ }
+}
+
+void DPSOFTRAST_PixelShader_LightSource(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span)
+{
+#ifdef SSE2_PRESENT
+ float buffer_z[DPSOFTRAST_DRAW_MAXSPANLENGTH];
+ unsigned char buffer_texture_colorbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_normalbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_glossbgra8[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
+ unsigned char buffer_texture_cubebgra8[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_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];
+ float CubeVectordata[4];
+ float CubeVectorslope[4];
+ float LightVectordata[4];
+ float LightVectorslope[4];
+ float EyeVectordata[4];
+ float EyeVectorslope[4];
+ float z;
+ float diffusetex[4];
+ float glosstex[4];
+ float surfacenormal[4];
+ float lightnormal[4];
+ float eyenormal[4];
+ float specularnormal[4];
+ float diffuse;
+ float specular;
+ float SpecularPower;
+ float CubeVector[4];
+ float attenuation;
+ int d[4];
+ Color_Glow[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+0];
+ Color_Glow[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+1];
+ Color_Glow[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Glow*4+2];
+ Color_Glow[3] = 0.0f;
+ Color_Ambient[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+0];
+ Color_Ambient[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+1];
+ Color_Ambient[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4+2];
+ Color_Ambient[3] = thread->uniform4f[DPSOFTRAST_UNIFORM_Alpha*4+0];
+ Color_Diffuse[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+0];
+ Color_Diffuse[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+1];
+ Color_Diffuse[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Diffuse*4+2];
+ Color_Diffuse[3] = 0.0f;
+ 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;
+ Color_Pants[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+0];
+ Color_Pants[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+1];
+ Color_Pants[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Pants*4+2];
+ Color_Pants[3] = 0.0f;
+ Color_Shirt[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+0];
+ Color_Shirt[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+1];
+ Color_Shirt[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Shirt*4+2];
+ Color_Shirt[3] = 0.0f;
+ LightColor[2] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+0];
+ LightColor[1] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+1];
+ LightColor[0] = thread->uniform4f[DPSOFTRAST_UNIFORM_LightColor*4+2];
+ LightColor[3] = 0.0f;
+ SpecularPower = thread->uniform4f[DPSOFTRAST_UNIFORM_SpecularPower*4+0] * (1.0f / 255.0f);
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, LightVectordata, LightVectorslope, DPSOFTRAST_ARRAY_TEXCOORD1);
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, EyeVectordata, EyeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD2);
+ DPSOFTRAST_CALCATTRIB4F(triangle, span, CubeVectordata, CubeVectorslope, DPSOFTRAST_ARRAY_TEXCOORD3);
+ DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z);
+ memset(buffer_FragColorbgra8 + startx*4, 0, (endx-startx)*4); // clear first, because we skip writing black pixels, and there are a LOT of them...
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_pantsbgra8, GL20TU_PANTS, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_shirtbgra8, GL20TU_SHIRT, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ }
+ if (thread->shader_permutation & SHADERPERMUTATION_CUBEFILTER)
+ DPSOFTRAST_Draw_Span_TextureCubeVaryingBGRA8(triangle, span, buffer_texture_cubebgra8, GL20TU_CUBE, DPSOFTRAST_ARRAY_TEXCOORD3, buffer_z);
+ if (thread->shader_permutation & SHADERPERMUTATION_SPECULAR)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_glossbgra8, GL20TU_GLOSS, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ CubeVector[0] = (CubeVectordata[0] + CubeVectorslope[0]*x) * z;
+ CubeVector[1] = (CubeVectordata[1] + CubeVectorslope[1]*x) * z;
+ CubeVector[2] = (CubeVectordata[2] + CubeVectorslope[2]*x) * z;
+ attenuation = 1.0f - DPSOFTRAST_Vector3LengthSquared(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ if (thread->shader_permutation & SHADERPERMUTATION_SHADOWMAP2D)
+ {
+ attenuation *= DPSOFTRAST_SampleShadowmap(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ }
+
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ diffusetex[0] += buffer_texture_pantsbgra8[x*4+0] * Color_Pants[0] + buffer_texture_shirtbgra8[x*4+0] * Color_Shirt[0];
+ diffusetex[1] += buffer_texture_pantsbgra8[x*4+1] * Color_Pants[1] + buffer_texture_shirtbgra8[x*4+1] * Color_Shirt[1];
+ diffusetex[2] += buffer_texture_pantsbgra8[x*4+2] * Color_Pants[2] + buffer_texture_shirtbgra8[x*4+2] * Color_Shirt[2];
+ diffusetex[3] += buffer_texture_pantsbgra8[x*4+3] * Color_Pants[3] + buffer_texture_shirtbgra8[x*4+3] * Color_Shirt[3];
+ }
+ glosstex[0] = buffer_texture_glossbgra8[x*4+0];
+ glosstex[1] = buffer_texture_glossbgra8[x*4+1];
+ glosstex[2] = buffer_texture_glossbgra8[x*4+2];
+ glosstex[3] = buffer_texture_glossbgra8[x*4+3];
+ surfacenormal[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f;
+ surfacenormal[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f;
+ 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);
+
+ 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);
+
+ 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_CUBEFILTER)
+ {
+ // scale down the attenuation to account for the cubefilter multiplying everything by 255
+ attenuation *= (1.0f / 255.0f);
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse) + glosstex[0] * Color_Specular[0] * specular) * LightColor[0] * buffer_texture_cubebgra8[x*4+0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse) + glosstex[1] * Color_Specular[1] * specular) * LightColor[1] * buffer_texture_cubebgra8[x*4+1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse) + glosstex[2] * Color_Specular[2] * specular) * LightColor[2] * buffer_texture_cubebgra8[x*4+2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[3] );if (d[3] > 255) d[3] = 255;
+ }
+ else
+ {
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse) + glosstex[0] * Color_Specular[0] * specular) * LightColor[0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse) + glosstex[1] * Color_Specular[1] * specular) * LightColor[1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse) + glosstex[2] * Color_Specular[2] * specular) * LightColor[2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ else if (thread->shader_permutation & SHADERPERMUTATION_DIFFUSE)
+ {
+ DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_normalbgra8, GL20TU_NORMAL, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ CubeVector[0] = (CubeVectordata[0] + CubeVectorslope[0]*x) * z;
+ CubeVector[1] = (CubeVectordata[1] + CubeVectorslope[1]*x) * z;
+ CubeVector[2] = (CubeVectordata[2] + CubeVectorslope[2]*x) * z;
+ attenuation = 1.0f - DPSOFTRAST_Vector3LengthSquared(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ if (thread->shader_permutation & SHADERPERMUTATION_SHADOWMAP2D)
+ {
+ attenuation *= DPSOFTRAST_SampleShadowmap(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ }
+
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ diffusetex[0] += buffer_texture_pantsbgra8[x*4+0] * Color_Pants[0] + buffer_texture_shirtbgra8[x*4+0] * Color_Shirt[0];
+ diffusetex[1] += buffer_texture_pantsbgra8[x*4+1] * Color_Pants[1] + buffer_texture_shirtbgra8[x*4+1] * Color_Shirt[1];
+ diffusetex[2] += buffer_texture_pantsbgra8[x*4+2] * Color_Pants[2] + buffer_texture_shirtbgra8[x*4+2] * Color_Shirt[2];
+ diffusetex[3] += buffer_texture_pantsbgra8[x*4+3] * Color_Pants[3] + buffer_texture_shirtbgra8[x*4+3] * Color_Shirt[3];
+ }
+ surfacenormal[0] = buffer_texture_normalbgra8[x*4+2] * (1.0f / 128.0f) - 1.0f;
+ surfacenormal[1] = buffer_texture_normalbgra8[x*4+1] * (1.0f / 128.0f) - 1.0f;
+ 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);
+
+ diffuse = DPSOFTRAST_Vector3Dot(surfacenormal, lightnormal);if (diffuse < 0.0f) diffuse = 0.0f;
+ if (thread->shader_permutation & SHADERPERMUTATION_CUBEFILTER)
+ {
+ // scale down the attenuation to account for the cubefilter multiplying everything by 255
+ attenuation *= (1.0f / 255.0f);
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse)) * LightColor[0] * buffer_texture_cubebgra8[x*4+0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse)) * LightColor[1] * buffer_texture_cubebgra8[x*4+1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse)) * LightColor[2] * buffer_texture_cubebgra8[x*4+2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[3] );if (d[3] > 255) d[3] = 255;
+ }
+ else
+ {
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0] + Color_Diffuse[0] * diffuse)) * LightColor[0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1] + Color_Diffuse[1] * diffuse)) * LightColor[1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2] + Color_Diffuse[2] * diffuse)) * LightColor[2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ else
+ {
+ for (x = startx;x < endx;x++)
+ {
+ z = buffer_z[x];
+ CubeVector[0] = (CubeVectordata[0] + CubeVectorslope[0]*x) * z;
+ CubeVector[1] = (CubeVectordata[1] + CubeVectorslope[1]*x) * z;
+ CubeVector[2] = (CubeVectordata[2] + CubeVectorslope[2]*x) * z;
+ attenuation = 1.0f - DPSOFTRAST_Vector3LengthSquared(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ if (thread->shader_permutation & SHADERPERMUTATION_SHADOWMAP2D)
+ {
+ attenuation *= DPSOFTRAST_SampleShadowmap(CubeVector);
+ if (attenuation < 0.01f)
+ continue;
+ }
+
+ diffusetex[0] = buffer_texture_colorbgra8[x*4+0];
+ diffusetex[1] = buffer_texture_colorbgra8[x*4+1];
+ diffusetex[2] = buffer_texture_colorbgra8[x*4+2];
+ diffusetex[3] = buffer_texture_colorbgra8[x*4+3];
+ if (thread->shader_permutation & SHADERPERMUTATION_COLORMAPPING)
+ {
+ diffusetex[0] += buffer_texture_pantsbgra8[x*4+0] * Color_Pants[0] + buffer_texture_shirtbgra8[x*4+0] * Color_Shirt[0];
+ diffusetex[1] += buffer_texture_pantsbgra8[x*4+1] * Color_Pants[1] + buffer_texture_shirtbgra8[x*4+1] * Color_Shirt[1];
+ diffusetex[2] += buffer_texture_pantsbgra8[x*4+2] * Color_Pants[2] + buffer_texture_shirtbgra8[x*4+2] * Color_Shirt[2];
+ diffusetex[3] += buffer_texture_pantsbgra8[x*4+3] * Color_Pants[3] + buffer_texture_shirtbgra8[x*4+3] * Color_Shirt[3];
+ }
+ if (thread->shader_permutation & SHADERPERMUTATION_CUBEFILTER)
+ {
+ // scale down the attenuation to account for the cubefilter multiplying everything by 255
+ attenuation *= (1.0f / 255.0f);
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0])) * LightColor[0] * buffer_texture_cubebgra8[x*4+0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1])) * LightColor[1] * buffer_texture_cubebgra8[x*4+1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2])) * LightColor[2] * buffer_texture_cubebgra8[x*4+2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[3] );if (d[3] > 255) d[3] = 255;
+ }
+ else
+ {
+ d[0] = (int)((diffusetex[0] * (Color_Ambient[0])) * LightColor[0] * attenuation);if (d[0] > 255) d[0] = 255;
+ d[1] = (int)((diffusetex[1] * (Color_Ambient[1])) * LightColor[1] * attenuation);if (d[1] > 255) d[1] = 255;
+ d[2] = (int)((diffusetex[2] * (Color_Ambient[2])) * LightColor[2] * attenuation);if (d[2] > 255) d[2] = 255;
+ d[3] = (int)( diffusetex[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];
+ buffer_FragColorbgra8[x*4+3] = d[3];
+ }
+ }
+ 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);
+}
+
+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];
+ 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);
+}
+
+
+
+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);
+}
+
+
+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];
+ 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);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_ShowDepth(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);
+}
+
+
+
+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);
+}
+
+void DPSOFTRAST_PixelShader_DeferredGeometry(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);
+}
+
+
+
+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);
}
-void DPSOFTRAST_Draw_PixelShaderSpan(const DPSOFTRAST_State_Draw_Span *span)
+void DPSOFTRAST_PixelShader_DeferredLightSource(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 buffer_texture_color[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
- float buffer_texture_lightmap[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
- float buffer_FragColor[DPSOFTRAST_DRAW_MAXSPANLENGTH*4];
- switch(dpsoftrast.shader_mode)
- {
- case SHADERMODE_GENERIC: ///< (particles/HUD/etc) vertex color: optionally multiplied by one texture
- DPSOFTRAST_Draw_Span_Begin(span, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_color, GL20TU_FIRST, 2, buffer_z);
- DPSOFTRAST_Draw_Span_MultiplyVarying(span, buffer_FragColor, buffer_texture_color, 1, buffer_z);
- DPSOFTRAST_Draw_Span_Finish(span, buffer_FragColor);
- break;
- case SHADERMODE_POSTPROCESS: ///< postprocessing shader (r_glsl_postprocess)
- // TODO: optimize!! at the very least there is no reason to use texture sampling on the frame texture
- DPSOFTRAST_Draw_Span_Begin(span, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_FragColor, GL20TU_FIRST, 2, buffer_z);
- if (dpsoftrast.shader_permutation & SHADERPERMUTATION_BLOOM)
- {
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_color, GL20TU_SECOND, 3, buffer_z);
- DPSOFTRAST_Draw_Span_AddBloom(span, buffer_FragColor, buffer_FragColor, buffer_texture_color, dpsoftrast.uniform4f + DPSOFTRAST_UNIFORM_BloomColorSubtract * 4);
- }
- DPSOFTRAST_Draw_Span_MixUniformColor(span, buffer_FragColor, buffer_FragColor, dpsoftrast.uniform4f + DPSOFTRAST_UNIFORM_ViewTintColor * 4);
- if (dpsoftrast.shader_permutation & SHADERPERMUTATION_SATURATION)
- {
- // TODO: implement saturation
- }
- if (dpsoftrast.shader_permutation & SHADERPERMUTATION_GAMMARAMPS)
- {
- // TODO: implement gammaramps
- }
- DPSOFTRAST_Draw_Span_Finish(span, buffer_FragColor);
- break;
- case SHADERMODE_DEPTH_OR_SHADOW: ///< (depthfirst/shadows) vertex shader only
- break;
- case SHADERMODE_FLATCOLOR: ///< (lightmap) modulate texture by uniform color (q1bsp: q3bsp)
- DPSOFTRAST_Draw_Span_Begin(span, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_color, GL20TU_COLOR, 2, buffer_z);
- DPSOFTRAST_Draw_Span_FlatColor(span, buffer_FragColor, buffer_texture_color);
- DPSOFTRAST_Draw_Span_Finish(span, buffer_FragColor);
- break;
- case SHADERMODE_VERTEXCOLOR: ///< (lightmap) modulate texture by vertex colors (q3bsp)
- DPSOFTRAST_Draw_Span_Begin(span, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_color, GL20TU_COLOR, 2, buffer_z);
- DPSOFTRAST_Draw_Span_VertexColor(span, buffer_FragColor, buffer_texture_color, buffer_z);
- DPSOFTRAST_Draw_Span_Finish(span, buffer_FragColor);
- break;
- case SHADERMODE_LIGHTMAP: ///< (lightmap) modulate texture by lightmap texture (q1bsp: q3bsp)
- DPSOFTRAST_Draw_Span_Begin(span, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_color, GL20TU_COLOR, 2, buffer_z);
- DPSOFTRAST_Draw_Span_Texture2DVarying(span, buffer_texture_lightmap, GL20TU_LIGHTMAP, 6, buffer_z);
- if(!dpsoftrast.user.alphatest && dpsoftrast.fb_blendmode == DPSOFTRAST_BLENDMODE_OPAQUE)
- {
- DPSOFTRAST_Draw_Span_Lightmap_Finish(span, buffer_texture_color, buffer_texture_lightmap);
- }
- else
+ 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);
+}
+
+
+
+typedef struct DPSOFTRAST_ShaderModeInfo_s
+{
+ int lodarrayindex;
+ void (*Vertex)(void);
+ void (*Span)(DPSOFTRAST_State_Thread *thread, const DPSOFTRAST_State_Triangle * RESTRICT triangle, const DPSOFTRAST_State_Span * RESTRICT span);
+ unsigned char arrays[DPSOFTRAST_ARRAY_TOTAL];
+ unsigned char texunits[DPSOFTRAST_MAXTEXTUREUNITS];
+}
+DPSOFTRAST_ShaderModeInfo;
+
+static const DPSOFTRAST_ShaderModeInfo DPSOFTRAST_ShaderModeTable[SHADERMODE_COUNT] =
+{
+ {2, DPSOFTRAST_VertexShader_Generic, DPSOFTRAST_PixelShader_Generic, {DPSOFTRAST_ARRAY_COLOR, DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, ~0}, {GL20TU_FIRST, GL20TU_SECOND, ~0}},
+ {2, DPSOFTRAST_VertexShader_PostProcess, DPSOFTRAST_PixelShader_PostProcess, {DPSOFTRAST_ARRAY_TEXCOORD0, DPSOFTRAST_ARRAY_TEXCOORD1, ~0}, {GL20TU_FIRST, GL20TU_SECOND, ~0}},
+ {2, DPSOFTRAST_VertexShader_Depth_Or_Shadow, DPSOFTRAST_PixelShader_Depth_Or_Shadow, {~0}, {~0}},
+ {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_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_ShowDepth, DPSOFTRAST_PixelShader_ShowDepth, {~0}},
+ {2, DPSOFTRAST_VertexShader_DeferredGeometry, DPSOFTRAST_PixelShader_DeferredGeometry, {~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)
{
- DPSOFTRAST_Draw_Span_Lightmap(span, buffer_FragColor, buffer_texture_color, buffer_texture_lightmap);
- DPSOFTRAST_Draw_Span_Finish(span, buffer_FragColor);
+#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;
}
- break;
- case SHADERMODE_FAKELIGHT: ///< (fakelight) modulate texture by "fake" lighting (no lightmaps: no nothing)
- break;
- case SHADERMODE_LIGHTDIRECTIONMAP_MODELSPACE: ///< (lightmap) use directional pixel shading from texture containing modelspace light directions (q3bsp deluxemap)
- break;
- case SHADERMODE_LIGHTDIRECTIONMAP_TANGENTSPACE: ///< (lightmap) use directional pixel shading from texture containing tangentspace light directions (q1bsp deluxemap)
- break;
- case SHADERMODE_LIGHTDIRECTION: ///< (lightmap) use directional pixel shading from fixed light direction (q3bsp)
- break;
- case SHADERMODE_LIGHTSOURCE: ///< (lightsource) use directional pixel shading from light source (rtlight)
- break;
- case SHADERMODE_REFRACTION: ///< refract background (the material is rendered normally after this pass)
- break;
- case SHADERMODE_WATER: ///< refract background and reflection (the material is rendered normally after this pass)
- break;
- case SHADERMODE_SHOWDEPTH: ///< (debugging) renders depth as color
- break;
- case SHADERMODE_DEFERREDGEOMETRY: ///< (deferred) render material properties to screenspace geometry buffers
- break;
- case SHADERMODE_DEFERREDLIGHTSOURCE: ///< (deferred) use directional pixel shading from light source (rtlight) on screenspace geometry buffers
- break;
- case SHADERMODE_COUNT:
- break;
}
+ return commandoffset;
}
-
-void DPSOFTRAST_Draw_ProcessSpans(void)
+
+int DPSOFTRAST_Draw_ProcessSpans(DPSOFTRAST_State_Thread *thread, int commandoffset)
{
int i;
int x;
int startx;
int endx;
- int numspans = dpsoftrast.draw.numspans;
// unsigned int c;
// unsigned int *colorpixel;
unsigned int *depthpixel;
int depth;
int depthslope;
unsigned int d;
- DPSOFTRAST_State_Draw_Span *span = dpsoftrast.draw.spanqueue;
+ DPSOFTRAST_State_Triangle *triangle;
+ DPSOFTRAST_State_Span *span;
unsigned char pixelmask[DPSOFTRAST_DRAW_MAXSPANLENGTH];
- for (i = 0;i < numspans;i++, span++)
+ for (i = 0; i < thread->numspans; i++)
{
- w = span->data[0][DPSOFTRAST_ARRAY_TOTAL][3];
- wslope = span->data[1][DPSOFTRAST_ARRAY_TOTAL][3];
- if (dpsoftrast.user.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)
{
- depth = (int)(w*DPSOFTRAST_DEPTHSCALE);
+ wslope = triangle->w[0];
+ w = triangle->w[2] + span->x*wslope + span->y*triangle->w[1];
depthslope = (int)(wslope*DPSOFTRAST_DEPTHSCALE);
- depthpixel = dpsoftrast.fb_depthpixels + span->start;
- switch(dpsoftrast.fb_depthfunc)
+ 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_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->start;
+ //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
// 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] && dpsoftrast.fb_colormask)
- DPSOFTRAST_Draw_PixelShaderSpan(span);
- if (dpsoftrast.user.depthmask)
- for (x = 0, d = depth;x < span->length;x++, d += depthslope)
+ 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;
}
{
// no depth testing means we're just dealing with color...
// if there is no color buffer, skip pixel shader
- if (dpsoftrast.fb_colorpixels[0] && dpsoftrast.fb_colormask)
+ if (dpsoftrast.fb_colorpixels[0] && thread->fb_colormask)
{
memset(pixelmask, 1, span->length);
span->pixelmask = pixelmask;
span->startx = 0;
span->endx = span->length;
- DPSOFTRAST_Draw_PixelShaderSpan(span);
+ DPSOFTRAST_ShaderModeTable[thread->shader_mode].Span(thread, triangle, span);
+ }
+ }
+ }
+ thread->numspans = 0;
+ return commandoffset;
+}
+
+void DPSOFTRAST_Draw_GenerateSpans(DPSOFTRAST_State_Thread *thread, int freetriangle)
+{
+#ifdef SSE2_PRESENT
+ 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)
+ {
+ commandoffset = DPSOFTRAST_Draw_InterpretCommands(thread, commandoffset, triangle->commandoffset);
+ DPSOFTRAST_ValidateQuick(thread, DPSOFTRAST_VALIDATE_DRAW);
+ }
+
+ MEMORY_BARRIER;
+
+ thread->commandoffset = commandoffset;
+ thread->triangleoffset = triangleoffset;
+#endif
+}
+
+void DPSOFTRAST_Draw_FlushThreads(void)
+{
+ DPSOFTRAST_State_Thread *thread;
+ int i;
+ if(dpsoftrast.drawtriangle != dpsoftrast.trianglepool.freetriangle)
+ {
+ MEMORY_BARRIER;
+ dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle;
+ }
+#ifdef USE_THREADS
+ SDL_LockMutex(dpsoftrast.trianglemutex);
+#endif
+ for (i = 0; i < dpsoftrast.numthreads; i++)
+ {
+ thread = &dpsoftrast.threads[i];
+#ifdef USE_THREADS
+ 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
}
+#ifdef USE_THREADS
+ SDL_UnlockMutex(dpsoftrast.trianglemutex);
+#endif
+ dpsoftrast.trianglepool.usedtriangles = 0;
+ dpsoftrast.commandpool.usedcommands = 0;
+}
+
+#ifdef USE_THREADS
+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
-void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s, unsigned char *arraymask)
+void DPSOFTRAST_Draw_ProcessTriangles(int firstvertex, int numtriangles, const int *element3i, const unsigned short *element3s, unsigned char *arraymask, int numarrays)
{
- int cullface = dpsoftrast.user.cullface;
+#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;
int i;
int j;
int k;
int y;
- int mip;
int e[3];
- int screenx[4];
- int screeny[4];
- int screenyless[4];
+ __m128i screeny;
+ int starty, endy;
int numpoints;
- int clipflags;
- int edge0p;
- int edge0n;
- int edge1p;
- int edge1n;
- int extent[4];
- int startx;
- int endx;
- float startxf;
- float endxf;
- float edge0ylerp;
- float edge0yilerp;
- float edge1ylerp;
- float edge1yilerp;
- float edge0xf;
- float edge1xf;
- float spanilength;
- float startxlerp;
- float yc;
- float w;
- float frac;
- float ifrac;
- float trianglearea2;
- float triangleedge[2][4];
- float trianglenormal[4];
+ int clipcase;
float clipdist[4];
- float clipped[DPSOFTRAST_ARRAY_TOTAL][4][4];
- float screen[4][4];
- float proj[DPSOFTRAST_ARRAY_TOTAL][4][4];
- DPSOFTRAST_State_Draw_Span *span;
- DPSOFTRAST_State_Draw_Span *oldspan;
+ __m128 triangleedge1, triangleedge2, trianglenormal;
+ __m128 clipfrac[3];
+ __m128 screen[4];
+ DPSOFTRAST_Texture *texture;
+ screen[3] = _mm_setzero_ps();
+ clipfrac[0] = clipfrac[1] = clipfrac[2] = _mm_setzero_ps();
for (i = 0;i < numtriangles;i++)
{
// generate the 3 edges of this triangle
e[1] = i*3+1;
e[2] = i*3+2;
}
- triangleedge[0][0] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+0] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+0];
- triangleedge[0][1] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+1] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+1];
- triangleedge[0][2] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+2] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+2];
- triangleedge[1][0] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+0] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+0];
- triangleedge[1][1] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+1] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+1];
- triangleedge[1][2] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+2] - dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+2];
- 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];
- trianglenormal[2] = triangleedge[0][0] * triangleedge[1][1] - triangleedge[0][1] * triangleedge[1][0];
- trianglearea2 = trianglenormal[0] * trianglenormal[0] + trianglenormal[1] * trianglenormal[1] + trianglenormal[2] * trianglenormal[2];
- // skip degenerate triangles, nothing good can come from them...
- if (trianglearea2 == 0.0f)
- continue;
- // apply current cullface mode (this culls many triangles)
- switch(cullface)
- {
- case GL_BACK:
- if (trianglenormal[2] < 0)
- continue;
- break;
- case GL_FRONT:
- if (trianglenormal[2] > 0)
- continue;
- break;
+
+#define SKIPBACKFACE \
+ triangleedge1 = _mm_sub_ps(screen[0], screen[1]); \
+ triangleedge2 = _mm_sub_ps(screen[2], screen[1]); \
+ /* store normal in 2, 0, 1 order instead of 0, 1, 2 as it requires fewer shuffles and leaves z component accessible as scalar */ \
+ trianglenormal = _mm_sub_ss(_mm_mul_ss(triangleedge1, _mm_shuffle_ps(triangleedge2, triangleedge2, _MM_SHUFFLE(3, 0, 2, 1))), \
+ _mm_mul_ss(_mm_shuffle_ps(triangleedge1, triangleedge1, _MM_SHUFFLE(3, 0, 2, 1)), triangleedge2)); \
+ switch(cullface) \
+ { \
+ case GL_BACK: \
+ if (_mm_ucomilt_ss(trianglenormal, _mm_setzero_ps())) \
+ continue; \
+ break; \
+ case GL_FRONT: \
+ if (_mm_ucomigt_ss(trianglenormal, _mm_setzero_ps())) \
+ continue; \
+ break; \
}
- // calculate distance from nearplane
- clipdist[0] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[0]*4+2] + 1.0f;
- clipdist[1] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[1]*4+2] + 1.0f;
- clipdist[2] = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][e[2]*4+2] + 1.0f;
- clipflags = 0;
- if (clipdist[0] < 0.0f)
- clipflags |= 1;
- if (clipdist[1] < 0.0f)
- clipflags |= 2;
- if (clipdist[2] < 0.0f)
- clipflags |= 4;
- // clip triangle if necessary
- switch(clipflags)
- {
- case 0: /*000*/
- // triangle is entirely in front of nearplane
+ //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) \
- frac = clipdist[p1] / (clipdist[p1] - clipdist[p2]);\
- ifrac = 1.0f - frac;\
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)\
- {\
- if (arraymask[j])\
- {\
- clipped[j][k][0] = dpsoftrast.draw.post_array4f[j][e[p1]*4+0]*ifrac+dpsoftrast.draw.post_array4f[j][e[p2]*4+0]*frac;\
- clipped[j][k][1] = dpsoftrast.draw.post_array4f[j][e[p1]*4+1]*ifrac+dpsoftrast.draw.post_array4f[j][e[p2]*4+1]*frac;\
- clipped[j][k][2] = dpsoftrast.draw.post_array4f[j][e[p1]*4+2]*ifrac+dpsoftrast.draw.post_array4f[j][e[p2]*4+2]*frac;\
- clipped[j][k][3] = dpsoftrast.draw.post_array4f[j][e[p1]*4+3]*ifrac+dpsoftrast.draw.post_array4f[j][e[p2]*4+3]*frac;\
- }\
- }\
- DPSOFTRAST_Draw_ProjectVertices(screen[k], clipped[DPSOFTRAST_ARRAY_POSITION][k], 1)
+
+#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]))); \
+ }
#define CLIPPEDVERTEXCOPY(k,p1) \
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)\
- {\
- if (arraymask[j])\
- {\
- clipped[j][k][0] = dpsoftrast.draw.post_array4f[j][e[p1]*4+0];\
- clipped[j][k][1] = dpsoftrast.draw.post_array4f[j][e[p1]*4+1];\
- clipped[j][k][2] = dpsoftrast.draw.post_array4f[j][e[p1]*4+2];\
- clipped[j][k][3] = dpsoftrast.draw.post_array4f[j][e[p1]*4+3];\
- }\
- }\
- screen[k][0] = dpsoftrast.draw.screencoord4f[e[p1]*4+0];\
- screen[k][1] = dpsoftrast.draw.screencoord4f[e[p1]*4+1];\
- screen[k][2] = dpsoftrast.draw.screencoord4f[e[p1]*4+2];\
- screen[k][3] = dpsoftrast.draw.screencoord4f[e[p1]*4+3];
-
- CLIPPEDVERTEXCOPY(0,0);
- CLIPPEDVERTEXCOPY(1,1);
- CLIPPEDVERTEXCOPY(2,2);
- numpoints = 3;
- break;
- case 1: /*100*/
- CLIPPEDVERTEXLERP(0,0,1);
- CLIPPEDVERTEXCOPY(1,1);
- CLIPPEDVERTEXCOPY(2,2);
- CLIPPEDVERTEXLERP(3,2,0);
- numpoints = 4;
- break;
- case 2: /*010*/
- CLIPPEDVERTEXCOPY(0,0);
- CLIPPEDVERTEXLERP(1,0,1);
- CLIPPEDVERTEXLERP(2,1,2);
- CLIPPEDVERTEXCOPY(3,2);
- numpoints = 4;
- break;
- case 3: /*110*/
- CLIPPEDVERTEXLERP(0,1,2);
- CLIPPEDVERTEXCOPY(1,2);
- CLIPPEDVERTEXLERP(2,2,0);
- numpoints = 3;
- break;
- case 4: /*001*/
- CLIPPEDVERTEXCOPY(0,0);
- CLIPPEDVERTEXCOPY(1,1);
- CLIPPEDVERTEXLERP(2,1,2);
- CLIPPEDVERTEXLERP(3,2,0);
- numpoints = 4;
- break;
- case 5: /*101*/
- CLIPPEDVERTEXLERP(0,0,1);
- CLIPPEDVERTEXCOPY(1,1);
- CLIPPEDVERTEXLERP(2,1,2);
- numpoints = 3;
- break;
- case 6: /*011*/
- CLIPPEDVERTEXCOPY(0,0);
- CLIPPEDVERTEXLERP(1,0,1);
- CLIPPEDVERTEXLERP(2,2,0);
- numpoints = 3;
- break;
- case 7: /*111*/
- // triangle is entirely behind nearplane
- continue;
+ screen[k] = _mm_load_ps(&dpsoftrast.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) \
+ { \
+ __m128 v1 = _mm_load_ps(&dpsoftrast.post_array4f[j][e[p1]*4]), v2 = _mm_load_ps(&dpsoftrast.post_array4f[j][e[p2]*4]); \
+ attrib = _mm_add_ps(v1, _mm_mul_ps(_mm_sub_ps(v2, v1), clipfrac[p1])); \
}
- // calculate integer y coords for triangle points
- screenx[0] = (int)(screen[0][0]);
- screeny[0] = (int)(screen[0][1]);
- screenx[1] = (int)(screen[1][0]);
- screeny[1] = (int)(screen[1][1]);
- screenx[2] = (int)(screen[2][0]);
- screeny[2] = (int)(screen[2][1]);
- screenx[3] = (int)(screen[3][0]);
- screeny[3] = (int)(screen[3][1]);
- // figure out the extents (bounding box) of the triangle
- extent[0] = screenx[0];
- extent[1] = screeny[0];
- extent[2] = screenx[0];
- extent[3] = screeny[0];
- for (j = 1;j < numpoints;j++)
- {
- if (extent[0] > screenx[j]) extent[0] = screenx[j];
- if (extent[1] > screeny[j]) extent[1] = screeny[j];
- if (extent[2] < screenx[j]) extent[2] = screenx[j];
- if (extent[3] < screeny[j]) extent[3] = screeny[j];
- }
- //extent[0]--;
- //extent[1]--;
- extent[2]++;
- extent[3]++;
- if (extent[0] < 0)
- extent[0] = 0;
- if (extent[1] < 0)
- extent[1] = 0;
- if (extent[2] > width)
- extent[2] = width;
- if (extent[3] > height)
- extent[3] = height;
- // skip offscreen triangles
- if (extent[2] <= extent[0] || extent[3] <= extent[1])
- continue;
- // TODO: adjust texture LOD by texture density
- mip = 0;
- // okay, this triangle is going to produce spans, we'd better project
- // the interpolants now (this is what gives perspective texturing),
- // this consists of simply multiplying all arrays by the W coord
- // (which is basically 1/Z), which will be undone per-pixel
- // (multiplying by Z again) to get the perspective-correct array
- // values
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
- {
- if (arraymask[j])
- {
- for (k = 0;k < numpoints;k++)
- {
- w = screen[k][3];
- proj[j][k][0] = clipped[j][k][0] * w;
- proj[j][k][1] = clipped[j][k][1] * w;
- proj[j][k][2] = clipped[j][k][2] * w;
- proj[j][k][3] = clipped[j][k][3] * w;
- }
- }
+#define GENATTRIBS(j, 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; \
}
- // iterate potential spans
- // TODO: optimize? if we figured out the edge order beforehand, this
- // could do loops over the edges in the proper order rather than
- // selecting them for each span
- // TODO: optimize? the edges could have data slopes calculated
- // TODO: optimize? the data slopes could be calculated as a plane
- // (2D slopes) to avoid any interpolation along edges at all
- for (y = extent[1];y < extent[3];y++)
- {
- // get center of pixel y
- yc = y;
- // do the compares all at once
- screenyless[0] = y <= screeny[0];
- screenyless[1] = y <= screeny[1];
- screenyless[2] = y <= screeny[2];
- screenyless[3] = y <= screeny[3];
- if (numpoints == 4)
+
+ // 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];
+ if (clipdist[0] >= 0.0f)
+ {
+ if (clipdist[1] >= 0.0f)
{
- switch(screenyless[0] + screenyless[1] * 2 + screenyless[2] * 4 + screenyless[3] * 8)
+ if (clipdist[2] >= 0.0f)
{
- case 0: /*0000*/ continue;
- case 1: /*1000*/ edge0p = 3;edge0n = 0;edge1p = 0;edge1n = 1;break;
- case 2: /*0100*/ edge0p = 0;edge0n = 1;edge1p = 1;edge1n = 2;break;
- case 3: /*1100*/ edge0p = 3;edge0n = 0;edge1p = 1;edge1n = 2;break;
- case 4: /*0010*/ edge0p = 1;edge0n = 2;edge1p = 2;edge1n = 3;break;
- case 5: /*1010*/ edge0p = 1;edge0n = 2;edge1p = 2;edge1n = 3;break; // concave - nonsense
- case 6: /*0110*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 3;break;
- case 7: /*1110*/ edge0p = 3;edge0n = 0;edge1p = 2;edge1n = 3;break;
- case 8: /*0001*/ edge0p = 2;edge0n = 3;edge1p = 3;edge1n = 0;break;
- case 9: /*1001*/ edge0p = 2;edge0n = 3;edge1p = 0;edge1n = 1;break;
- case 10: /*0101*/ edge0p = 2;edge0n = 3;edge1p = 1;edge1n = 2;break; // concave - nonsense
- case 11: /*1101*/ edge0p = 2;edge0n = 3;edge1p = 1;edge1n = 2;break;
- case 12: /*0011*/ edge0p = 1;edge0n = 2;edge1p = 3;edge1n = 0;break;
- case 13: /*1011*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 1;break;
- case 14: /*0111*/ edge0p = 0;edge0n = 1;edge1p = 3;edge1n = 0;break;
- case 15: /*1111*/ continue;
+ // triangle is entirely in front of nearplane
+ CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXCOPY(1,1); CLIPPEDVERTEXCOPY(2,2);
+ SKIPBACKFACE;
+ numpoints = 3;
+ clipcase = 0;
}
- }
- else
- {
- switch(screenyless[0] + screenyless[1] * 2 + screenyless[2] * 4)
+ else
{
- case 0: /*000*/ continue;
- case 1: /*100*/ edge0p = 2;edge0n = 0;edge1p = 0;edge1n = 1;break;
- case 2: /*010*/ edge0p = 0;edge0n = 1;edge1p = 1;edge1n = 2;break;
- case 3: /*110*/ edge0p = 2;edge0n = 0;edge1p = 1;edge1n = 2;break;
- case 4: /*001*/ edge0p = 1;edge0n = 2;edge1p = 2;edge1n = 0;break;
- case 5: /*101*/ edge0p = 1;edge0n = 2;edge1p = 0;edge1n = 1;break;
- case 6: /*011*/ edge0p = 0;edge0n = 1;edge1p = 2;edge1n = 0;break;
- case 7: /*111*/ continue;
+ CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXCOPY(1,1); CLIPPEDVERTEXLERP(2,1,2); CLIPPEDVERTEXLERP(3,2,0);
+ SKIPBACKFACE;
+ numpoints = 4;
+ clipcase = 1;
}
}
-#if 0
- {
- int foundedges = 0;
- int cedge0p = 0;
- int cedge0n = 0;
- int cedge1p = 0;
- int cedge1n = 0;
- for (j = 0, k = numpoints-1;j < numpoints;k = j, j++)
+ else
{
- if (screenyless[k] && !screenyless[j])
+ if (clipdist[2] >= 0.0f)
{
- cedge1p = k;
- cedge1n = j;
- foundedges |= 1;
+ CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXLERP(1,0,1); CLIPPEDVERTEXLERP(2,1,2); CLIPPEDVERTEXCOPY(3,2);
+ SKIPBACKFACE;
+ numpoints = 4;
+ clipcase = 2;
}
- else if (screenyless[j] && !screenyless[k])
+ else
{
- cedge0p = k;
- cedge0n = j;
- foundedges |= 2;
+ CLIPPEDVERTEXCOPY(0,0); CLIPPEDVERTEXLERP(1,0,1); CLIPPEDVERTEXLERP(2,2,0);
+ SKIPBACKFACE;
+ numpoints = 3;
+ clipcase = 3;
}
}
- if (foundedges != 3)
- continue;
- if (cedge0p != edge0p || cedge0n != edge0n || cedge1p != edge1p || cedge1n != edge1n)
- {
- if (numpoints == 4)
- printf("case %i%i%i%i is broken %i %i %i %i != %i %i %i %i\n", screenyless[0], screenyless[1], screenyless[2], screenyless[3], cedge0p, cedge0n, cedge1p, cedge1n, edge0p, edge0n, edge1p, edge1n);
- else
- printf("case %i%i%i is broken %i %i %i %i != %i %i %i %i\n", screenyless[0], screenyless[1], screenyless[2], cedge0p, cedge0n, cedge1p, cedge1n, edge0p, edge0n, edge1p, edge1n);
- }
- }
-#endif
- edge0ylerp = (yc - screen[edge0p][1]) / (screen[edge0n][1] - screen[edge0p][1]);
- edge1ylerp = (yc - screen[edge1p][1]) / (screen[edge1n][1] - screen[edge1p][1]);
- if (edge0ylerp < 0 || edge0ylerp > 1 || edge1ylerp < 0 || edge1ylerp > 1)
- continue;
- edge0yilerp = 1.0f - edge0ylerp;
- edge1yilerp = 1.0f - edge1ylerp;
- edge0xf = screen[edge0p][0] * edge0yilerp + screen[edge0n][0] * edge0ylerp;
- edge1xf = screen[edge1p][0] * edge1yilerp + screen[edge1n][0] * edge1ylerp;
- if (edge0xf < edge1xf)
+ }
+ else if (clipdist[1] >= 0.0f)
+ {
+ if (clipdist[2] >= 0.0f)
{
- startxf = edge0xf;
- endxf = edge1xf;
+ CLIPPEDVERTEXLERP(0,0,1); CLIPPEDVERTEXCOPY(1,1); CLIPPEDVERTEXCOPY(2,2); CLIPPEDVERTEXLERP(3,2,0);
+ SKIPBACKFACE;
+ numpoints = 4;
+ clipcase = 4;
}
else
{
- startxf = edge1xf;
- endxf = edge0xf;
+ CLIPPEDVERTEXLERP(0,0,1); CLIPPEDVERTEXCOPY(1,1); CLIPPEDVERTEXLERP(2,1,2);
+ SKIPBACKFACE;
+ numpoints = 3;
+ clipcase = 5;
}
- startx = (int)ceil(startxf);
- endx = (int)ceil(endxf);
- if (startx < 0)
- startx = 0;
- if (endx > width)
- endx = width;
- if (startx >= endx)
+ }
+ else if (clipdist[2] >= 0.0f)
+ {
+ CLIPPEDVERTEXLERP(0,1,2); CLIPPEDVERTEXCOPY(1,2); CLIPPEDVERTEXLERP(2,2,0);
+ SKIPBACKFACE;
+ numpoints = 3;
+ clipcase = 6;
+ }
+ else continue; // triangle is entirely behind nearplane
+
+ {
+ // 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),
+ 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());
+ screenmax = _mm_min_epi16(screenmax, fbmax);
+ // skip offscreen triangles
+ if (_mm_cvtsi128_si32(_mm_cmplt_epi16(screenmax, screenmin)))
continue;
- if (startxf > startx || endxf < endx-1) { printf("%s:%i X wrong (%i to %i is outside %f to %f)\n", __FILE__, __LINE__, startx, endx, startxf, endxf); }
- spanilength = 1.0f / (endxf - startxf);
- startxlerp = startx - startxf;
- span = &dpsoftrast.draw.spanqueue[dpsoftrast.draw.numspans++];
- span->mip = mip;
- span->start = y * width + startx;
- span->length = endx - startx;
- j = DPSOFTRAST_ARRAY_TOTAL;
- if (edge0xf < edge1xf)
- {
- span->data[0][j][0] = screen[edge0p][0] * edge0yilerp + screen[edge0n][0] * edge0ylerp;
- span->data[0][j][1] = screen[edge0p][1] * edge0yilerp + screen[edge0n][1] * edge0ylerp;
- span->data[0][j][2] = screen[edge0p][2] * edge0yilerp + screen[edge0n][2] * edge0ylerp;
- span->data[0][j][3] = screen[edge0p][3] * edge0yilerp + screen[edge0n][3] * edge0ylerp;
- span->data[1][j][0] = screen[edge1p][0] * edge1yilerp + screen[edge1n][0] * edge1ylerp;
- span->data[1][j][1] = screen[edge1p][1] * edge1yilerp + screen[edge1n][1] * edge1ylerp;
- span->data[1][j][2] = screen[edge1p][2] * edge1yilerp + screen[edge1n][2] * edge1ylerp;
- span->data[1][j][3] = screen[edge1p][3] * edge1yilerp + screen[edge1n][3] * edge1ylerp;
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
- {
- if (arraymask[j])
- {
- span->data[0][j][0] = proj[j][edge0p][0] * edge0yilerp + proj[j][edge0n][0] * edge0ylerp;
- span->data[0][j][1] = proj[j][edge0p][1] * edge0yilerp + proj[j][edge0n][1] * edge0ylerp;
- span->data[0][j][2] = proj[j][edge0p][2] * edge0yilerp + proj[j][edge0n][2] * edge0ylerp;
- span->data[0][j][3] = proj[j][edge0p][3] * edge0yilerp + proj[j][edge0n][3] * edge0ylerp;
- span->data[1][j][0] = proj[j][edge1p][0] * edge1yilerp + proj[j][edge1n][0] * edge1ylerp;
- span->data[1][j][1] = proj[j][edge1p][1] * edge1yilerp + proj[j][edge1n][1] * edge1ylerp;
- span->data[1][j][2] = proj[j][edge1p][2] * edge1yilerp + proj[j][edge1n][2] * edge1ylerp;
- span->data[1][j][3] = proj[j][edge1p][3] * edge1yilerp + proj[j][edge1n][3] * edge1ylerp;
- }
- }
- }
- else
- {
- span->data[0][j][0] = screen[edge1p][0] * edge1yilerp + screen[edge1n][0] * edge1ylerp;
- span->data[0][j][1] = screen[edge1p][1] * edge1yilerp + screen[edge1n][1] * edge1ylerp;
- span->data[0][j][2] = screen[edge1p][2] * edge1yilerp + screen[edge1n][2] * edge1ylerp;
- span->data[0][j][3] = screen[edge1p][3] * edge1yilerp + screen[edge1n][3] * edge1ylerp;
- span->data[1][j][0] = screen[edge0p][0] * edge0yilerp + screen[edge0n][0] * edge0ylerp;
- span->data[1][j][1] = screen[edge0p][1] * edge0yilerp + screen[edge0n][1] * edge0ylerp;
- span->data[1][j][2] = screen[edge0p][2] * edge0yilerp + screen[edge0n][2] * edge0ylerp;
- span->data[1][j][3] = screen[edge0p][3] * edge0yilerp + screen[edge0n][3] * edge0ylerp;
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
- {
- if (arraymask[j])
- {
- span->data[0][j][0] = proj[j][edge1p][0] * edge1yilerp + proj[j][edge1n][0] * edge1ylerp;
- span->data[0][j][1] = proj[j][edge1p][1] * edge1yilerp + proj[j][edge1n][1] * edge1ylerp;
- span->data[0][j][2] = proj[j][edge1p][2] * edge1yilerp + proj[j][edge1n][2] * edge1ylerp;
- span->data[0][j][3] = proj[j][edge1p][3] * edge1yilerp + proj[j][edge1n][3] * edge1ylerp;
- span->data[1][j][0] = proj[j][edge0p][0] * edge0yilerp + proj[j][edge0n][0] * edge0ylerp;
- span->data[1][j][1] = proj[j][edge0p][1] * edge0yilerp + proj[j][edge0n][1] * edge0ylerp;
- span->data[1][j][2] = proj[j][edge0p][2] * edge0yilerp + proj[j][edge0n][2] * edge0ylerp;
- span->data[1][j][3] = proj[j][edge0p][3] * edge0yilerp + proj[j][edge0n][3] * edge0ylerp;
- }
- }
- }
- // change data[1][n][] to be a data slope
- j = DPSOFTRAST_ARRAY_TOTAL;
- span->data[1][j][0] = (span->data[1][j][0] - span->data[0][j][0]) * spanilength;
- span->data[1][j][1] = (span->data[1][j][1] - span->data[0][j][1]) * spanilength;
- span->data[1][j][2] = (span->data[1][j][2] - span->data[0][j][2]) * spanilength;
- span->data[1][j][3] = (span->data[1][j][3] - span->data[0][j][3]) * spanilength;
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
- {
- if (arraymask[j])
- {
- span->data[1][j][0] = (span->data[1][j][0] - span->data[0][j][0]) * spanilength;
- span->data[1][j][1] = (span->data[1][j][1] - span->data[0][j][1]) * spanilength;
- span->data[1][j][2] = (span->data[1][j][2] - span->data[0][j][2]) * spanilength;
- span->data[1][j][3] = (span->data[1][j][3] - span->data[0][j][3]) * spanilength;
- }
- }
- // adjust the data[0][n][] to be correct for the pixel centers
- // this also handles horizontal clipping where a major part of the
- // span may be off the left side of the screen
- j = DPSOFTRAST_ARRAY_TOTAL;
- span->data[0][j][0] += span->data[1][j][0] * startxlerp;
- span->data[0][j][1] += span->data[1][j][1] * startxlerp;
- span->data[0][j][2] += span->data[1][j][2] * startxlerp;
- span->data[0][j][3] += span->data[1][j][3] * startxlerp;
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
+ starty = _mm_extract_epi16(screenmin, 1);
+ endy = _mm_extract_epi16(screenmax, 1)+1;
+ screeny = _mm_srai_epi32(screeni, 16);
+ }
+
+ if (dpsoftrast.trianglepool.usedtriangles >= DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1)
+#ifdef USE_THREADS
+ 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);
+
+ // calculate attribute plans for triangle data...
+ // okay, this triangle is going to produce spans, we'd better project
+ // the interpolants now (this is what gives perspective texturing),
+ // this consists of simply multiplying all arrays by the W coord
+ // (which is basically 1/Z), which will be undone per-pixel
+ // (multiplying by Z again) to get the perspective-correct array
+ // values
+ {
+ __m128 attribuvslope, attribuxslope, attribuyslope, attribvxslope, attribvyslope, attriborigin, attribedge1, attribedge2, attribxslope, attribyslope, w0, w1, w2, x1, y1;
+ 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));
+ 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));
+ 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++)
{
if (arraymask[j])
{
- span->data[0][j][0] += span->data[1][j][0] * startxlerp;
- span->data[0][j][1] += span->data[1][j][1] * startxlerp;
- span->data[0][j][2] += span->data[1][j][2] * startxlerp;
- span->data[0][j][3] += span->data[1][j][3] * startxlerp;
- }
- }
- // to keep the shader routines from needing more than a small
- // buffer for pixel intermediate data, we split long spans...
- while (span->length > DPSOFTRAST_DRAW_MAXSPANLENGTH)
+ __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;
+ 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++)
{
- span->length = DPSOFTRAST_DRAW_MAXSPANLENGTH;
- if (dpsoftrast.draw.numspans >= DPSOFTRAST_DRAW_MAXSPANQUEUE)
+ int texunit = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].texunits[j];
+ if (texunit >= DPSOFTRAST_MAXTEXTUREUNITS)
+ break;
+ texture = dpsoftrast.texbound[texunit];
+ if (texture && texture->filter > DPSOFTRAST_TEXTURE_FILTER_LINEAR)
{
- DPSOFTRAST_Draw_ProcessSpans();
- dpsoftrast.draw.numspans = 0;
- }
- oldspan = span;
- span = &dpsoftrast.draw.spanqueue[dpsoftrast.draw.numspans++];
- *span = *oldspan;
- startx += DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->start = y * width + startx;
- span->length = endx - startx;
- j = DPSOFTRAST_ARRAY_TOTAL;
- span->data[0][j][0] += span->data[1][j][0] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][1] += span->data[1][j][1] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][2] += span->data[1][j][2] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][3] += span->data[1][j][3] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- for (j = 0;j < DPSOFTRAST_ARRAY_TOTAL;j++)
- {
- if (arraymask[j])
+ 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(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)));
+ // this will be multiplied in the texturing routine by the texture resolution
+ y = _mm_cvtss_si32(mipdensity);
+ if (y > 0)
{
- span->data[0][j][0] += span->data[1][j][0] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][1] += span->data[1][j][1] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][2] += span->data[1][j][2] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
- span->data[0][j][3] += span->data[1][j][3] * DPSOFTRAST_DRAW_MAXSPANLENGTH;
+ y = (int)(log((float)y)*0.5f/M_LN2);
+ if (y > texture->mipmaps - 1)
+ y = texture->mipmaps - 1;
+ triangle->mip[texunit] = y;
}
}
}
- // after all that, we have a span suitable for the pixel shader...
- if (dpsoftrast.draw.numspans >= DPSOFTRAST_DRAW_MAXSPANQUEUE)
- {
- DPSOFTRAST_Draw_ProcessSpans();
- dpsoftrast.draw.numspans = 0;
- }
}
- // draw outlines over triangle for debugging
- // for (j = 0, k = numpoints-1;j < numpoints;k = j, j++)
- // DPSOFTRAST_Draw_DebugEdgePoints(screen[k], screen[j]);
+
+ dpsoftrast.trianglepool.freetriangle = dpsoftrast.trianglepool.freetriangle < DPSOFTRAST_DRAW_MAXTRIANGLEPOOL-1 ? dpsoftrast.trianglepool.freetriangle + 1 : 0;
+ dpsoftrast.trianglepool.usedtriangles++;
+
+ numqueued++;
+ if (numqueued >= DPSOFTRAST_DRAW_FLUSHPROCESSTRIANGLES)
+ {
+ MEMORY_BARRIER;
+ dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle;
+
+#ifdef USE_THREADS
+ SDL_LockMutex(dpsoftrast.trianglemutex);
+ SDL_CondBroadcast(dpsoftrast.trianglecond);
+ SDL_UnlockMutex(dpsoftrast.trianglemutex);
+#else
+ DPSOFTRAST_Draw_FlushThreads();
+#endif
+ numqueued = 0;
+ }
}
- if (dpsoftrast.draw.numspans)
+ if (numqueued > 0)
{
- DPSOFTRAST_Draw_ProcessSpans();
- dpsoftrast.draw.numspans = 0;
+ MEMORY_BARRIER;
+ dpsoftrast.drawtriangle = dpsoftrast.trianglepool.freetriangle;
+
+#ifdef USE_THREADS
+ SDL_LockMutex(dpsoftrast.trianglemutex);
+ SDL_CondBroadcast(dpsoftrast.trianglecond);
+ SDL_UnlockMutex(dpsoftrast.trianglemutex);
+#else
+ DPSOFTRAST_Draw_FlushThreads();
+#endif
}
+#endif
}
-void DPSOFTRAST_Draw_DebugPoints(void)
+void DPSOFTRAST_DrawTriangles(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s)
{
int i;
- int x;
- int y;
- int numvertices = dpsoftrast.draw.numvertices;
- int w = dpsoftrast.fb_width;
- int bounds[4];
- unsigned int *pixels = dpsoftrast.fb_colorpixels[0];
- const float *c4f;
- bounds[0] = dpsoftrast.fb_viewportscissor[0];
- bounds[1] = dpsoftrast.fb_viewportscissor[1];
- bounds[2] = dpsoftrast.fb_viewportscissor[0] + dpsoftrast.fb_viewportscissor[2];
- bounds[3] = dpsoftrast.fb_viewportscissor[1] + dpsoftrast.fb_viewportscissor[3];
- for (i = 0;i < numvertices;i++)
+ int lastarray = DPSOFTRAST_ARRAY_POSITION;
+ unsigned char arraymask[DPSOFTRAST_ARRAY_TOTAL];
+ memset(arraymask, false, sizeof(arraymask));
+ arraymask[DPSOFTRAST_ARRAY_POSITION] = true;
+ for (i = 0; i < DPSOFTRAST_ARRAY_TOTAL; i++)
{
- // check nearclip
- //if (dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+3] != 1.0f)
- // continue;
- x = (int)(dpsoftrast.draw.screencoord4f[i*4+0]);
- y = (int)(dpsoftrast.draw.screencoord4f[i*4+1]);
- //x = (int)(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0] + 0.5f);
- //y = (int)(dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1] + 0.5f);
- //x = (int)((dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+0] + 1.0f) * dpsoftrast.fb_width * 0.5f + 0.5f);
- //y = (int)((dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION][i*4+1] + 1.0f) * dpsoftrast.fb_height * 0.5f + 0.5f);
- if (x < bounds[0] || y < bounds[1] || x >= bounds[2] || y >= bounds[3])
- continue;
- c4f = dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_COLOR] + i*4;
- pixels[y*w+x] = DPSOFTRAST_BGRA8_FROM_RGBA32F(c4f[0], c4f[1], c4f[2], c4f[3]);
+ int arrayindex = DPSOFTRAST_ShaderModeTable[dpsoftrast.shader_mode].arrays[i];
+ if (arrayindex >= DPSOFTRAST_ARRAY_TOTAL)
+ break;
+ switch (arrayindex)
+ {
+ case DPSOFTRAST_ARRAY_POSITION:
+ case DPSOFTRAST_ARRAY_COLOR:
+ break;
+ default:
+ if (dpsoftrast.pointer_texcoordf[arrayindex-DPSOFTRAST_ARRAY_TEXCOORD0] == NULL)
+ continue;
+ break;
+ }
+ 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);
}
-void DPSOFTRAST_DrawTriangles(int firstvertex, int numvertices, int numtriangles, const int *element3i, const unsigned short *element3s)
+void DPSOFTRAST_Flush(void)
{
- unsigned char arraymask[DPSOFTRAST_ARRAY_TOTAL];
- arraymask[0] = true;
- arraymask[1] = dpsoftrast.fb_colorpixels[0] != NULL; // TODO: optimize (decide based on shadermode)
- arraymask[2] = dpsoftrast.pointer_texcoordf[0] != NULL;
- arraymask[3] = dpsoftrast.pointer_texcoordf[1] != NULL;
- arraymask[4] = dpsoftrast.pointer_texcoordf[2] != NULL;
- arraymask[5] = dpsoftrast.pointer_texcoordf[3] != NULL;
- arraymask[6] = dpsoftrast.pointer_texcoordf[4] != NULL;
- arraymask[7] = dpsoftrast.pointer_texcoordf[5] != NULL;
- arraymask[8] = dpsoftrast.pointer_texcoordf[6] != NULL;
- arraymask[9] = dpsoftrast.pointer_texcoordf[7] != NULL;
- DPSOFTRAST_Validate(DPSOFTRAST_VALIDATE_DRAW);
- DPSOFTRAST_Draw_LoadVertices(firstvertex, numvertices, true);
- DPSOFTRAST_Draw_VertexShader();
- DPSOFTRAST_Draw_ProjectVertices(dpsoftrast.draw.screencoord4f, dpsoftrast.draw.post_array4f[DPSOFTRAST_ARRAY_POSITION], numvertices);
- DPSOFTRAST_Draw_ProcessTriangles(firstvertex, numvertices, numtriangles, element3i, element3s, arraymask);
-}
-
-void DPSOFTRAST_Init(int width, int height, unsigned int *colorpixels, unsigned int *depthpixels)
+ DPSOFTRAST_Draw_SyncCommands();
+ DPSOFTRAST_Draw_FlushThreads();
+}
+
+void DPSOFTRAST_Init(int width, int height, int numthreads, unsigned int *colorpixels, unsigned int *depthpixels)
{
+ int i;
union
{
int i;
dpsoftrast.texture_firstfree = 1;
dpsoftrast.texture_end = 1;
dpsoftrast.texture_max = 0;
- dpsoftrast.user.colormask[0] = 1;
- dpsoftrast.user.colormask[1] = 1;
- dpsoftrast.user.colormask[2] = 1;
- dpsoftrast.user.colormask[3] = 1;
- dpsoftrast.user.blendfunc[0] = GL_ONE;
- dpsoftrast.user.blendfunc[1] = GL_ZERO;
- dpsoftrast.user.depthmask = true;
- dpsoftrast.user.depthtest = true;
- dpsoftrast.user.depthfunc = GL_LEQUAL;
- dpsoftrast.user.scissortest = false;
- dpsoftrast.user.cullface = GL_BACK;
- dpsoftrast.user.alphatest = false;
- dpsoftrast.user.alphafunc = GL_GREATER;
- dpsoftrast.user.alphavalue = 0.5f;
- dpsoftrast.user.scissor[0] = 0;
- dpsoftrast.user.scissor[1] = 0;
- dpsoftrast.user.scissor[2] = dpsoftrast.fb_width;
- dpsoftrast.user.scissor[3] = dpsoftrast.fb_height;
- dpsoftrast.user.viewport[0] = 0;
- dpsoftrast.user.viewport[1] = 0;
- dpsoftrast.user.viewport[2] = dpsoftrast.fb_width;
- dpsoftrast.user.viewport[3] = dpsoftrast.fb_height;
- dpsoftrast.user.depthrange[0] = 0;
- dpsoftrast.user.depthrange[1] = 1;
- dpsoftrast.user.polygonoffset[0] = 0;
- dpsoftrast.user.polygonoffset[1] = 0;
- dpsoftrast.user.color[0] = 1;
- dpsoftrast.user.color[1] = 1;
- dpsoftrast.user.color[2] = 1;
- dpsoftrast.user.color[3] = 1;
- dpsoftrast.validate = -1;
- DPSOFTRAST_Validate(-1);
- dpsoftrast.validate = 0;
+ dpsoftrast.viewport[0] = 0;
+ dpsoftrast.viewport[1] = 0;
+ dpsoftrast.viewport[2] = dpsoftrast.fb_width;
+ dpsoftrast.viewport[3] = dpsoftrast.fb_height;
+ dpsoftrast.color[0] = 1;
+ dpsoftrast.color[1] = 1;
+ dpsoftrast.color[2] = 1;
+ dpsoftrast.color[3] = 1;
+ dpsoftrast.cullface = GL_BACK;
+#ifdef USE_THREADS
+ dpsoftrast.numthreads = bound(1, numthreads, 64);
+ dpsoftrast.trianglemutex = SDL_CreateMutex();
+ dpsoftrast.trianglecond = SDL_CreateCond();
+#else
+ dpsoftrast.numthreads = 1;
+#endif
+ 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->colormask[1] = 1;
+ thread->colormask[2] = 1;
+ thread->colormask[3] = 1;
+ thread->blendfunc[0] = GL_ONE;
+ thread->blendfunc[1] = GL_ZERO;
+ thread->depthmask = true;
+ thread->depthtest = true;
+ thread->depthfunc = GL_LEQUAL;
+ thread->scissortest = false;
+ thread->alphatest = false;
+ thread->alphafunc = GL_GREATER;
+ thread->alphavalue = 0.5f;
+ thread->scissor[0] = 0;
+ thread->scissor[1] = 0;
+ thread->scissor[2] = dpsoftrast.fb_width;
+ thread->scissor[3] = dpsoftrast.fb_height;
+ thread->depthrange[0] = 0;
+ thread->depthrange[1] = 1;
+ thread->polygonoffset[0] = 0;
+ thread->polygonoffset[1] = 0;
+
+ thread->numspans = 0;
+ thread->triangleoffset = 0;
+ thread->commandoffset = 0;
+ thread->waiting = false;
+#ifdef USE_THREADS
+ thread->waitcond = SDL_CreateCond();
+#endif
+
+ thread->validate = -1;
+ DPSOFTRAST_Validate(thread, -1);
+#ifdef USE_THREADS
+ thread->thread = SDL_CreateThread(DPSOFTRAST_Draw_Thread, thread);
+#endif
+ }
}
void DPSOFTRAST_Shutdown(void)
{
int i;
+#ifdef USE_THREADS
+ if(dpsoftrast.numthreads > 0)
+ {
+ DPSOFTRAST_State_Thread *thread;
+ SDL_LockMutex(dpsoftrast.trianglemutex);
+ for (i = 0; i < dpsoftrast.numthreads; i++)
+ {
+ thread = &dpsoftrast.threads[i];
+ thread->index = -1;
+ }
+ 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)
- free(dpsoftrast.texture[i].bytes);
+ MM_FREE(dpsoftrast.texture[i].bytes);
if (dpsoftrast.texture)
free(dpsoftrast.texture);
+ if (dpsoftrast.threads)
+ MM_FREE(dpsoftrast.threads);
memset(&dpsoftrast, 0, sizeof(dpsoftrast));
}
+
projects / xonotic / darkplaces.git / blobdiff