#endif
#define ALIGN_SIZE 16
-#define ATOMIC_SIZE 32
+#define ATOMIC_SIZE 4
#ifdef SSE_POSSIBLE
#if defined(__APPLE__)
#include <libkern/OSAtomic.h>
#define ALIGN(var) var __attribute__((__aligned__(16)))
- #define ATOMIC(var) var __attribute__((__aligned__(32)))
+ #define ATOMIC(var) var __attribute__((__aligned__(4)))
#define MEMORY_BARRIER (_mm_sfence())
#define ATOMIC_COUNTER volatile int32_t
#define ATOMIC_INCREMENT(counter) (OSAtomicIncrement32Barrier(&(counter)))
#define ATOMIC_DECREMENT(counter) (OSAtomicDecrement32Barrier(&(counter)))
#define ATOMIC_ADD(counter, val) ((void)OSAtomicAdd32Barrier((val), &(counter)))
+ #elif defined(__GNUC__) && defined(WIN32)
+ #define ALIGN(var) var __attribute__((__aligned__(16)))
+ #define ATOMIC(var) var __attribute__((__aligned__(4)))
+ #define MEMORY_BARRIER (_mm_sfence())
+ //(__sync_synchronize())
+ #define ATOMIC_COUNTER volatile LONG
+ // this LONG * cast serves to fix an issue with broken mingw
+ // packages on Ubuntu; these only declare the function to take
+ // a LONG *, causing a compile error here. This seems to be
+ // error- and warn-free on platforms that DO declare
+ // InterlockedIncrement correctly, like mingw on Windows.
+ #define ATOMIC_INCREMENT(counter) (InterlockedIncrement((LONG *) &(counter)))
+ #define ATOMIC_DECREMENT(counter) (InterlockedDecrement((LONG *) &(counter)))
+ #define ATOMIC_ADD(counter, val) ((void)InterlockedExchangeAdd((LONG *) &(counter), (val)))
#elif defined(__GNUC__)
#define ALIGN(var) var __attribute__((__aligned__(16)))
- #define ATOMIC(var) var __attribute__((__aligned__(32)))
+ #define ATOMIC(var) var __attribute__((__aligned__(4)))
#define MEMORY_BARRIER (_mm_sfence())
//(__sync_synchronize())
#define ATOMIC_COUNTER volatile int
#define ATOMIC_ADD(counter, val) ((void)__sync_fetch_and_add(&(counter), (val)))
#elif defined(_MSC_VER)
#define ALIGN(var) __declspec(align(16)) var
- #define ATOMIC(var) __declspec(align(32)) var
+ #define ATOMIC(var) __declspec(align(4)) var
#define MEMORY_BARRIER (_mm_sfence())
//(MemoryBarrier())
#define ATOMIC_COUNTER volatile LONG
#define _mm_cvtss_f32(val) (__builtin_ia32_vec_ext_v4sf ((__v4sf)(val), 0))
#endif
-#define MM_MALLOC(size) _mm_malloc(size, ATOMIC_SIZE)
+#define MM_MALLOC(size) _mm_malloc(size, ALIGN_SIZE)
static void *MM_CALLOC(size_t nmemb, size_t size)
{
- void *ptr = _mm_malloc(nmemb*size, ATOMIC_SIZE);
+ void *ptr = _mm_malloc(nmemb*size, ALIGN_SIZE);
if (ptr != NULL) memset(ptr, 0, nmemb*size);
return ptr;
}
#define DPSOFTRAST_DRAW_MAXCOMMANDPOOL 2097152
#define DPSOFTRAST_DRAW_MAXCOMMANDSIZE 16384
-typedef ATOMIC(struct DPSOFTRAST_State_Command_Pool_s
+typedef ALIGN(struct DPSOFTRAST_State_Command_Pool_s
{
int freecommand;
int usedcommands;
- ATOMIC(unsigned char commands[DPSOFTRAST_DRAW_MAXCOMMANDPOOL]);
+ ALIGN(unsigned char commands[DPSOFTRAST_DRAW_MAXCOMMANDPOOL]);
}
DPSOFTRAST_State_Command_Pool);
-typedef ATOMIC(struct DPSOFTRAST_State_Triangle_s
+typedef ALIGN(struct DPSOFTRAST_State_Triangle_s
{
unsigned char mip[DPSOFTRAST_MAXTEXTUREUNITS]; // texcoord to screen space density values (for picking mipmap of textures)
float w[3];
}
DPSOFTRAST_BLENDMODE;
-typedef ATOMIC(struct DPSOFTRAST_State_Thread_s
+typedef ALIGN(struct DPSOFTRAST_State_Thread_s
{
void *thread;
int index;
int depthtest;
int depthfunc;
int scissortest;
- int alphatest;
- int alphafunc;
- float alphavalue;
int viewport[4];
int scissor[4];
float depthrange[2];
}
DPSOFTRAST_State_Thread);
-typedef ATOMIC(struct DPSOFTRAST_State_s
+typedef ALIGN(struct DPSOFTRAST_State_s
{
int fb_width;
int fb_height;
command->mode = 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->enable = enable;
-}
-
-DEFCOMMAND(16, AlphaFunc, int func; float ref;)
-static void DPSOFTRAST_Interpret_AlphaFunc(DPSOFTRAST_State_Thread *thread, DPSOFTRAST_Command_AlphaFunc *command)
-{
- thread->alphafunc = command->func;
- thread->alphavalue = command->ref;
-}
-void DPSOFTRAST_AlphaFunc(int func, float ref)
-{
- DPSOFTRAST_Command_AlphaFunc *command = DPSOFTRAST_ALLOCATECOMMAND(AlphaFunc);
- command->func = func;
- command->ref = ref;
-}
-
void DPSOFTRAST_Color4f(float r, float g, float b, float a)
{
dpsoftrast.color[0] = r;
pixel += (span->y * dpsoftrast.fb_width + span->x) * 4;
pixeli += span->y * dpsoftrast.fb_width + span->x;
// handle alphatest now (this affects depth writes too)
- if (thread->alphatest)
+ if (thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL)
for (x = startx;x < endx;x++)
if (in4ub[x*4+3] < 128)
pixelmask[x] = false;
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, 2, buffer_z);
- if (thread->alphatest || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
+ if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
pixel = buffer_FragColorbgra8;
Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2));
Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
int arrayindex = DPSOFTRAST_ARRAY_COLOR;
DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z);
DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, 2, buffer_z);
- if (thread->alphatest || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
+ if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
pixel = buffer_FragColorbgra8;
Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2));
Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
DPSOFTRAST_Draw_Span_Begin(thread, triangle, span, buffer_z);
DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_colorbgra8, GL20TU_COLOR, DPSOFTRAST_ARRAY_TEXCOORD0, buffer_z);
DPSOFTRAST_Draw_Span_Texture2DVaryingBGRA8(thread, triangle, span, buffer_texture_lightmapbgra8, GL20TU_LIGHTMAP, DPSOFTRAST_ARRAY_TEXCOORD4, buffer_z);
- if (thread->alphatest || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
+ if ((thread->shader_permutation & SHADERPERMUTATION_ALPHAKILL) || thread->fb_blendmode != DPSOFTRAST_BLENDMODE_OPAQUE)
pixel = buffer_FragColorbgra8;
Color_Ambientm = _mm_shuffle_epi32(_mm_cvtps_epi32(_mm_mul_ps(_mm_load_ps(&thread->uniform4f[DPSOFTRAST_UNIFORM_Color_Ambient*4]), _mm_set1_ps(256.0f))), _MM_SHUFFLE(3, 0, 1, 2));
Color_Ambientm = _mm_and_si128(Color_Ambientm, _mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0));
float EyeVectorslope[4];
// uniforms
- float ScreenScaleRefractReflect[2];
- float ScreenCenterRefractReflect[2];
- float DistortScaleRefractReflect[2];
+ float ScreenScaleRefractReflect[4];
+ float ScreenCenterRefractReflect[4];
+ float DistortScaleRefractReflect[4];
float RefractColor[4];
float ReflectColor[4];
float ReflectFactor;
INTERPCOMMAND(DepthRange)
INTERPCOMMAND(PolygonOffset)
INTERPCOMMAND(CullFace)
- INTERPCOMMAND(AlphaTest)
- INTERPCOMMAND(AlphaFunc)
INTERPCOMMAND(SetTexture)
INTERPCOMMAND(SetShader)
INTERPCOMMAND(Uniform4f)
thread->depthtest = true;
thread->depthfunc = GL_LEQUAL;
thread->scissortest = false;
- thread->alphatest = false;
- thread->alphafunc = GL_GREATER;
- thread->alphavalue = 0.5f;
thread->viewport[0] = 0;
thread->viewport[1] = 0;
thread->viewport[2] = dpsoftrast.fb_width;