(RGL) Tweaks

This commit is contained in:
Twinaphex 2012-08-06 05:05:16 +02:00
parent fd3ee5a9dd
commit 7fb25a76c8
2 changed files with 256 additions and 273 deletions

View File

@ -107,7 +107,7 @@ GLuint nvFenceCounter = 0;
#define DECLARE_TYPE(TYPE,CTYPE,MAXVAL) \
typedef CTYPE type_##TYPE; \
static inline type_##TYPE _RGLFloatTo_##TYPE(float v) { return (type_##TYPE)(_RGLClampf(v)*MAXVAL); } \
static inline type_##TYPE _RGLFloatTo_##TYPE(float v) { return (type_##TYPE)((MAX(MIN(v, 1.f), 0.f)) * MAXVAL); } \
static inline float _RGLFloatFrom_##TYPE(type_##TYPE v) { return ((float)v)/MAXVAL; }
DECLARE_C_TYPES
#undef DECLARE_TYPE
@ -256,6 +256,7 @@ static inline void _RGLFifoGlVertexAttribPointer
GLuint offset
)
{
switch(size)
{
case 0:
@ -282,24 +283,19 @@ static inline void _RGLFifoGlVertexAttribPointer
else
gcmType = CELL_GCM_VERTEX_UB256;
break;
case RGL_SHORT:
gcmType = normalized ? CELL_GCM_VERTEX_S1 : CELL_GCM_VERTEX_S32K;
break;
case RGL_FLOAT:
gcmType = CELL_GCM_VERTEX_F;
break;
case RGL_HALF_FLOAT:
gcmType = CELL_GCM_VERTEX_SF;
break;
case RGL_CMP:
size = 1;
gcmType = CELL_GCM_VERTEX_CMP;
break;
default:
break;
}
@ -379,8 +375,10 @@ static void _RGLAttribSetDeleteBuffer( PSGLcontext *LContext, GLuint buffName )
jsAttribSet *attribSet = bufferObject->attribSets[i];
for(GLuint j = 0; j < MAX_VERTEX_ATTRIBS; ++j)
{
if(attribSet->attribs.attrib[j].arrayBuffer == buffName)
attribSet->attribs.attrib[j].arrayBuffer = 0;
}
attribSet->dirty = GL_TRUE;
}
@ -432,6 +430,7 @@ static void _RGLFreeBufferObject( jsBufferObject *buffer )
_RGLPlatformDestroyBufferObject( buffer );
buffer->textureReferences.~Vector<jsTexture *>();
buffer->attribSets.~Vector<jsAttribSet *>();
if(buffer != NULL)
free( buffer );
}
@ -479,8 +478,10 @@ GLAPI void APIENTRY glDeleteBuffers(GLsizei n, const GLuint *buffers)
if(buffers[i] )
{
GLuint name = buffers[i];
if(LContext->ArrayBuffer == name)
LContext->ArrayBuffer = 0;
if(LContext->PixelUnpackBuffer == name)
LContext->PixelUnpackBuffer = 0;
@ -504,11 +505,6 @@ GLAPI void APIENTRY glGenBuffers(GLsizei n, GLuint *buffers)
_RGLTexNameSpaceGenNames( &LContext->bufferObjectNameSpace, n, buffers );
}
static inline jsFramebuffer *_RGLGetFramebuffer(PSGLcontext *LContext, GLuint name)
{
return (jsFramebuffer *)LContext->framebufferNameSpace.data[name];
}
static inline void _RGLTextureTouchFBOs(jsTexture *texture)
{
PSGLcontext *LContext = _CurrentContext;
@ -516,9 +512,10 @@ static inline void _RGLTextureTouchFBOs(jsTexture *texture)
return;
GLuint fbCount = texture->framebuffers.getCount();
if(fbCount > 0)
{
jsFramebuffer *contextFramebuffer = LContext->framebuffer ? _RGLGetFramebuffer( LContext, LContext->framebuffer ) : NULL;
jsFramebuffer *contextFramebuffer = LContext->framebuffer ? (jsFramebuffer *)LContext->framebufferNameSpace.data[LContext->framebuffer] : NULL;
for (GLuint i = 0; i < fbCount; ++i)
{
@ -644,6 +641,7 @@ static void _RGLPlatformBufferObjectSetData( jsBufferObject* bufferObject, GLint
unsigned int dstId = jsBuffer->bufferId;
unsigned int pitch = jsBuffer->pitch;
const char *src = (const char *)data;
switch ( bufferObject->usage )
{
case GL_STREAM_DRAW:
@ -835,6 +833,7 @@ static void _RGLFramebufferGetAttachmentTexture(
jsTexture** texture)
{
PSGLcontext* LContext = _CurrentContext;
switch ( attachment->type )
{
case FRAMEBUFFER_ATTACHMENT_NONE:
@ -857,13 +856,16 @@ static GLboolean _RGLTextureIsValid( const jsTexture* texture )
return GL_FALSE;
if ( !texture->image )
return GL_FALSE;
const jsImage* image = texture->image;
int width = image->width;
int height = image->height;
GLenum format = image->format;
GLenum type = image->type;
GLenum internalFormat = image->internalFormat;
if (( internalFormat == 0 ) || ( format == 0 ) || ( type == 0 ) )
return GL_FALSE;
@ -933,29 +935,36 @@ static GLenum _RGLPlatformFramebufferCheckStatus(jsFramebuffer* framebuffer)
return GL_FRAMEBUFFER_COMPLETE_OES;
}
enum _RGLTextureStrategy {
_RGL_TEXTURE_STRATEGY_END,
_RGL_TEXTURE_STRATEGY_UNTILED_ALLOC,
_RGL_TEXTURE_STRATEGY_UNTILED_CLEAR,
enum _RGLTextureStrategy
{
TEXTURE_STRATEGY_END,
TEXTURE_STRATEGY_UNTILED_ALLOC,
TEXTURE_STRATEGY_UNTILED_CLEAR,
};
static enum _RGLTextureStrategy linearGPUStrategy[] =
{
_RGL_TEXTURE_STRATEGY_UNTILED_ALLOC,
_RGL_TEXTURE_STRATEGY_UNTILED_CLEAR,
_RGL_TEXTURE_STRATEGY_UNTILED_ALLOC,
_RGL_TEXTURE_STRATEGY_END,
TEXTURE_STRATEGY_UNTILED_ALLOC,
TEXTURE_STRATEGY_UNTILED_CLEAR,
TEXTURE_STRATEGY_UNTILED_ALLOC,
TEXTURE_STRATEGY_END,
};
static void _RGLImageAllocCPUStorage( jsImage *image )
{
if((image->storageSize > image->mallocStorageSize) || (!image->mallocData))
{
if ( image->mallocData ) free( image->mallocData );
if(image->mallocData)
free( image->mallocData );
image->mallocData = ( char * )malloc( image->storageSize + 128 );
image->mallocStorageSize = image->storageSize;
}
image->data = _RGLPadPtr( image->mallocData, 128 );
intptr_t x = (intptr_t)image->mallocData;
x = ( x + 128 - 1 ) / 128 * 128;
image->data = (char*)x;
}
static inline int _RGLGetComponentCount( GLenum format )
@ -1272,11 +1281,6 @@ int _RGLGetPixelSize( GLenum format, GLenum type )
return _RGLGetComponentCount( format )*componentSize;
}
static inline int _RGLGetStorageSize( GLenum format, GLenum type, GLsizei width, GLsizei height, GLsizei depth )
{
return _RGLGetPixelSize( format, type )*width*height*depth;
}
static void _RGLPlatformChooseGPUFormatAndLayout(
const jsTexture* texture,
GLboolean forceLinear,
@ -1289,18 +1293,20 @@ static void _RGLPlatformChooseGPUFormatAndLayout(
newLayout->baseHeight = image->height;
newLayout->internalFormat = ( RGLEnum )image->internalFormat;
newLayout->pixelBits = _RGLPlatformGetBitsPerPixel( newLayout->internalFormat );
newLayout->pitch = pitch ? pitch : _RGLPad( _RGLGetStorageSize( texture->image->format, texture->image->type, texture->image->width, 1, 1 ), 64 );
newLayout->pitch = pitch ? pitch : _RGLPad( _RGLGetPixelSize( texture->image->format, texture->image->type )*texture->image->width, 64);
}
void _RGLPlatformDropUnboundTextures(GLenum pool)
{
PSGLcontext* LContext = _CurrentContext;
GLuint i, j;
for (i = 0; i < LContext->textureNameSpace.capacity; ++i)
{
GLboolean bound = GL_FALSE;
jsTexture *texture = ( jsTexture * )LContext->textureNameSpace.data[i];
if ( !texture || ( texture->referenceBuffer != 0 ) ) continue;
if(!texture || (texture->referenceBuffer != 0))
continue;
for (j = 0; j < _RGL_MAX_VERTEX_TEXTURE_IMAGE_UNITS; ++j)
{
@ -1355,7 +1361,7 @@ static void _RGLPlatformReallocateGcmTexture( jsTexture* texture )
switch ( *step++ )
{
case _RGL_TEXTURE_STRATEGY_UNTILED_ALLOC:
case TEXTURE_STRATEGY_UNTILED_ALLOC:
_RGLPlatformChooseGPUFormatAndLayout( texture, GL_TRUE, 0, &newLayout );
size = _RGLPad( newLayout.baseHeight * newLayout.pitch, 1);
@ -1388,10 +1394,10 @@ static void _RGLPlatformReallocateGcmTexture( jsTexture* texture )
}
}
break;
case _RGL_TEXTURE_STRATEGY_UNTILED_CLEAR:
case TEXTURE_STRATEGY_UNTILED_CLEAR:
_RGLPlatformDropUnboundTextures(SURFACE_POOL_LINEAR);
break;
case _RGL_TEXTURE_STRATEGY_END:
case TEXTURE_STRATEGY_END:
_RGLSetError( GL_OUT_OF_MEMORY );
done = GL_TRUE;
break;
@ -1558,6 +1564,7 @@ static void jsPlatformFramebuffer_validate( jsPlatformFramebuffer * fb, PSGLcont
GLuint defaultPitch = 0;
GLuint defaultId = GMM_ERROR;
GLuint defaultIdOffset = 0;
for(int i = 0; i < RGL_SETRENDERTARGET_MAXCOUNT; ++i)
{
jsTexture* colorTexture = NULL;
@ -1618,7 +1625,7 @@ static void _RGLValidateFramebuffer( void )
if(LContext->framebuffer)
{
jsPlatformFramebuffer* framebuffer = static_cast<jsPlatformFramebuffer *>( _RGLGetFramebuffer( LContext, LContext->framebuffer ) );
jsPlatformFramebuffer* framebuffer = static_cast<jsPlatformFramebuffer *>((jsFramebuffer *)LContext->framebufferNameSpace.data[LContext->framebuffer] );
if ( framebuffer->needValidate )
jsPlatformFramebuffer_validate( framebuffer, LContext );
@ -1721,10 +1728,10 @@ GLAPI void APIENTRY glClearColor( GLclampf red, GLclampf green, GLclampf blue, G
{
PSGLcontext *LContext = _CurrentContext;
LContext->ClearColor.R = _RGLClampf( red );
LContext->ClearColor.G = _RGLClampf( green );
LContext->ClearColor.B = _RGLClampf( blue );
LContext->ClearColor.A = _RGLClampf( alpha );
LContext->ClearColor.R = MAX(MIN(red, 1.f), 0.f);
LContext->ClearColor.G = MAX(MIN(green, 1.f), 0.f);
LContext->ClearColor.B = MAX(MIN(blue, 1.f), 0.f);
LContext->ClearColor.A = MAX(MIN(alpha, 1.f), 0.f);
}
GLAPI void APIENTRY glBlendEquation( GLenum mode )
@ -1738,15 +1745,14 @@ GLAPI void APIENTRY glBlendEquation( GLenum mode )
GLAPI void APIENTRY glBlendColor(GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha)
{
PSGLcontext* LContext = _CurrentContext;
LContext->BlendColor.R = _RGLClampf( red );
LContext->BlendColor.G = _RGLClampf( green );
LContext->BlendColor.B = _RGLClampf( blue );
LContext->BlendColor.A = _RGLClampf( alpha );
LContext->BlendColor.R = MAX(MIN(red, 1.f), 0.f);
LContext->BlendColor.G = MAX(MIN(green, 1.f), 0.f);
LContext->BlendColor.B = MAX(MIN(blue, 1.f), 0.f);
LContext->BlendColor.A = MAX(MIN(alpha, 1.f), 0.f);
LContext->needValidate |= PSGL_VALIDATE_BLENDING;
}
GLAPI void APIENTRY glBlendFunc(GLenum sfactor, GLenum dfactor)
{
PSGLcontext *LContext = _CurrentContext;
@ -1814,7 +1820,7 @@ GLAPI GLenum APIENTRY glCheckFramebufferStatusOES( GLenum target )
if ( LContext->framebuffer )
{
jsFramebuffer* framebuffer = _RGLGetFramebuffer( LContext, LContext->framebuffer );
jsFramebuffer* framebuffer = (jsFramebuffer *)LContext->framebufferNameSpace.data[LContext->framebuffer];
return _RGLPlatformFramebufferCheckStatus( framebuffer );
}
@ -1825,7 +1831,7 @@ GLAPI void APIENTRY glFramebufferTexture2DOES( GLenum target, GLenum attachment,
{
PSGLcontext* LContext = _CurrentContext;
jsFramebuffer* framebuffer = _RGLGetFramebuffer( LContext, LContext->framebuffer );
jsFramebuffer* framebuffer = (jsFramebuffer *)LContext->framebufferNameSpace.data[LContext->framebuffer];
jsFramebufferAttachment* attach = _RGLFramebufferGetAttachment( framebuffer, GL_COLOR_ATTACHMENT0_EXT );
@ -2095,10 +2101,7 @@ static uint32_t gmmInitFixedAllocator (void)
return CELL_OK;
}
static void gmmRemovePendingFree(
GmmAllocator *pAllocator,
GmmBlock *pBlock
)
static void gmmRemovePendingFree(GmmAllocator *pAllocator, GmmBlock *pBlock)
{
if (pBlock == pAllocator->pPendingFreeHead)
pAllocator->pPendingFreeHead = pBlock->pNextFree;
@ -3268,10 +3271,10 @@ template<int SIZE> inline static void swapandsetfp( int ucodeSize, unsigned int
unsigned short count = *( ec++ );
for ( unsigned long offsetIndex = 0; offsetIndex < count; ++offsetIndex )
{
void *pointer=NULL;
void *ptr = NULL;
const int paddedSIZE = (SIZE + 1) & ~1;
cellGcmSetInlineTransferPointerInline( &_RGLState.fifo, gmmIdToOffset( loadProgramId ) + loadProgramOffset + *(ec++), paddedSIZE, &pointer);
float *fp = (float*)pointer;
cellGcmSetInlineTransferPointerInline( &_RGLState.fifo, gmmIdToOffset( loadProgramId ) + loadProgramOffset + *(ec++), paddedSIZE, &ptr);
float *fp = (float*)ptr;
float *src = (float*)v;
for (uint32_t j=0; j<SIZE;j++)
{
@ -3294,6 +3297,7 @@ template<int SIZE> static void setVectorTypefp( CgRuntimeParameter* __restrict p
CgParameterResource *parameterResource = _RGLGetParameterResource( ptr->program, ptr->parameterEntry );
unsigned short resource = parameterResource->resource;
unsigned short *ec = ( unsigned short * )( ptr->program->resources ) + resource + 1;
if ( RGL_LIKELY( *ec ) )
{
swapandsetfp<SIZE>(program->header.instructionCount*16, program->loadProgramId, program->loadProgramOffset, ec, (unsigned int *)data);
@ -3408,10 +3412,9 @@ template <int ROWS, int COLS, int ORDER> static void setMatrixSharedvpIndex( CgR
for ( long row = 0; row < ROWS; ++row )
{
for(long col = 0; col < COLS; ++col)
{
tmp[row*4 + col] = dst[row * 4 + col] = ( ORDER == ROW_MAJOR ) ? f[row * COLS + col] : f[col * ROWS + row];
}
for ( long col = COLS; col < 4; ++col ) tmp[row*4 + col] = dst[row*4+col];
for(long col = COLS; col < 4; ++col)
tmp[row*4 + col] = dst[row*4+col];
}
cellGcmSetVertexProgramParameterBlockInline( &_RGLState.fifo, resource, ROWS, (const float*)tmp);
@ -4407,13 +4410,6 @@ static void _RGLPlatformDestroyTexture( jsTexture* texture )
_RGLTextureTouchFBOs( texture );
}
// Get size of texture in GPU layout
static inline GLuint _RGLPlatformTextureGetGPUSize( const jsTexture* texture )
{
RGLTexture *gcmTexture = ( RGLTexture * )texture->platformTexture;
return _RGLPad( gcmTexture->gpuLayout.baseHeight * gcmTexture->gpuLayout.pitch, 1);
}
#include <cell/gcm/gcm_method_data.h>
static void _RGLPlatformValidateTextureStage( int unit, jsTexture* texture )
@ -4573,9 +4569,7 @@ static GLenum _RGLPlatformChooseInternalStorage( jsImage* image, GLenum internal
image->internalFormat = platformInternalFormat;
_RGLPlatformExpandInternalFormat( platformInternalFormat, &image->format, &image->type );
image->storageSize = _RGLGetStorageSize(
image->format, image->type,
image->width, image->height, 1 );
image->storageSize = _RGLGetPixelSize(image->format, image->type) * image->width * image->height;
return GL_NO_ERROR;
}
@ -4662,9 +4656,7 @@ static GLboolean _RGLPlatformTexturePBOImage(
if ( LContext->PixelUnpackBuffer == 0 )
return GL_FALSE;
const GLuint pboPitch = _RGLPad(
_RGLGetStorageSize( format, type, width, 1, 1 ),
LContext->unpackAlignment );
const GLuint pboPitch = _RGLPad(_RGLGetPixelSize(format, type) * width, LContext->unpackAlignment );
if (( pboPitch&3 ) != 0 )
{
RARCH_WARN("PBO image pitch not a multiple of 4, using slow path.\n" );
@ -4722,14 +4714,15 @@ static GLboolean _RGLPlatformTexturePBOImage(
{
gcmTexture->gpuLayout = newLayout;
if ( gcmTexture->gpuAddressId != GMM_ERROR && gcmTexture->pbo == NULL )
{
_RGLPlatformFreeGcmTexture( texture );
}
gcmTexture->pbo = bufferObject;
gcmTexture->gpuAddressId = gpuId;
gcmTexture->gpuAddressIdOffset = gpuIdOffset;
gcmTexture->pool = SURFACE_POOL_LINEAR;
gcmTexture->gpuSize = _RGLPlatformTextureGetGPUSize( texture );
RGLTexture *gcmTexture = ( RGLTexture * )texture->platformTexture;
// Get size of texture in GPU layout
gcmTexture->gpuSize = _RGLPad( gcmTexture->gpuLayout.baseHeight * gcmTexture->gpuLayout.pitch, 1);
++bufferObject->refCount;
}
else
@ -6584,7 +6577,7 @@ static _CGprogram* _RGLCgProgramFindPrev( _CGcontext* ctx, _CGprogram* prog )
{
_CGprogram* ptr = ctx->programList;
while ( NULL != ptr && prog != ptr->next )
while ( ptr != NULL && prog != ptr->next )
ptr = ptr->next;
return ptr;
@ -6838,7 +6831,9 @@ static CGprogram _RGLCgCreateProgram( CGcontext ctx, CGprofile profile, const Cg
// success! add the program to the program list in the context.
_RGLCgProgramPushFront(prog->parentContext, prog);
if ( _cgProgramCreateHook ) _cgProgramCreateHook( prog );
if(_cgProgramCreateHook)
_cgProgramCreateHook(prog);
// everything worked.
return id;
@ -7235,7 +7230,8 @@ CG_API CGprogram cgCopyProgram( CGprogram program )
_RGLCgProgramPushFront(newprog->parentContext, newprog);
if ( _cgProgramCopyHook ) _cgProgramCopyHook( newprog, prog );
if(_cgProgramCopyHook)
_cgProgramCopyHook(newprog, prog);
return newprog->id;
}
@ -7248,7 +7244,9 @@ CG_API void cgDestroyProgram( CGprogram program )
_RGLCgRaiseError( CG_INVALID_PROGRAM_HANDLE_ERROR );
return;
}
_CGprogram* ptr = _cgGetProgPtr( program );
if ( ptr == NULL )
{
_RGLCgRaiseError( CG_INVALID_PROGRAM_HANDLE_ERROR );
@ -8109,12 +8107,6 @@ CGGL_API void cgGLEnableTextureParameter( CGparameter param )
ptr->samplerSetter( ptr, NULL, 0 );
}
static void _RGLCgContextZero( _CGcontext* p )
{
memset( p, 0, sizeof( *p ) );
p->compileType = CG_UNKNOWN;
}
static void _RGLCgContextPushFront(_CGcontext* ctx)
{
if(_CurrentContext->RGLcgContextHead)
@ -8131,7 +8123,9 @@ static void destroy_context(_CGcontext*ctx)
_cgContextDestroyHook(ctx);
_RGLEraseName( &_CurrentContext->cgContextNameSpace, ( jsName )ctx->id );
_RGLCgContextZero( ctx );
memset(ctx, 0, sizeof( *ctx ) );
ctx->compileType = CG_UNKNOWN;
free( ctx );
}
@ -8146,7 +8140,8 @@ CG_API CGcontext cgCreateContext(void)
return ( CGcontext )NULL;
}
_RGLCgContextZero( ptr );
memset( ptr, 0, sizeof( *ptr ) );
ptr->compileType = CG_UNKNOWN;
CGcontext result = ( CGcontext )_RGLCreateName( &_CurrentContext->cgContextNameSpace, ptr );
if ( !result )

View File

@ -550,11 +550,6 @@ struct PSGLcontext
#define RGL_LIKELY(COND) (COND)
#define RGL_UNLIKELY(COND) (COND)
static inline float _RGLClampf( const float value )
{
return MAX( MIN( value, 1.f ), 0.f );
}
static inline unsigned int endianSwapWord( unsigned int v )
{
return ( v&0xff ) << 24 | ( v&0xff00 ) << 8 |
@ -577,13 +572,6 @@ static inline unsigned long _RGLPad(unsigned long x, unsigned long pad)
return ( x + pad - 1 ) / pad*pad;
}
static inline char* _RGLPadPtr(const char* p, unsigned int pad)
{
intptr_t x = (intptr_t)p;
x = ( x + pad - 1 ) / pad * pad;
return ( char* )x;
}
typedef struct MemoryBlockManager_t_
{
char *memory;
@ -927,7 +915,7 @@ static inline GLuint _RGLPlatformGetBitsPerPixel( GLenum internalFormat )
}
void static inline _RGLFifoGlViewport( GLint x, GLint y, GLsizei width, GLsizei height, GLclampf zNear = 0.0f, GLclampf zFar = 1.0f )
static inline void _RGLFifoGlViewport( GLint x, GLint y, GLsizei width, GLsizei height, GLclampf zNear = 0.0f, GLclampf zFar = 1.0f )
{
RGLViewportState *vp = &_RGLState.state.viewport;
RGLRenderTarget *rt = &_RGLState.renderTarget;