/**
* Called the first time stage->run is called. In effect, don't
* allocate data until the first time the stage is run.
*/
static GLboolean init_vp( GLcontext *ctx,
struct tnl_pipeline_stage *stage )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_buffer *VB = &(tnl->vb);
struct vp_stage_data *store;
const GLuint size = VB->Size;
GLuint i;
stage->privatePtr = MALLOC(sizeof(*store));
store = VP_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
/* Allocate arrays of vertex output values */
for (i = 0; i < 15; i++) {
_mesa_vector4f_alloc( &store->attribs[i], 0, size, 32 );
store->attribs[i].size = 4;
}
/* a few other misc allocations */
_mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 );
return GL_TRUE;
}
static GLboolean init_vertex_stage( struct gl_context *ctx,
struct tnl_pipeline_stage *stage )
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct vertex_stage_data *store;
GLuint size = VB->Size;
stage->privatePtr = calloc(1, sizeof(*store));
store = VERTEX_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->eye, 0, size, 32 );
_mesa_vector4f_alloc( &store->clip, 0, size, 32 );
_mesa_vector4f_alloc( &store->proj, 0, size, 32 );
store->clipmask = _mesa_align_malloc(sizeof(GLubyte)*size, 32 );
if (!store->clipmask ||
!store->eye.data ||
!store->clip.data ||
!store->proj.data)
return GL_FALSE;
return GL_TRUE;
}
static GLboolean
alloc_point_data(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct point_stage_data *store;
stage->privatePtr = malloc(sizeof(*store));
store = POINT_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->PointSize, 0, VB->Size, 32 );
return GL_TRUE;
}
/**
* Allocate stage's private data (storage for transformed normals).
*/
static GLboolean
alloc_normal_data(GLcontext *ctx, struct tnl_pipeline_stage *stage)
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct normal_stage_data *store;
stage->privatePtr = _mesa_malloc(sizeof(*store));
store = NORMAL_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->normal, 0, tnl->vb.Size, 32 );
return GL_TRUE;
}
/* Called the first time stage->run() is invoked.
*/
static GLboolean alloc_texnorm_data( GLcontext *ctx,
struct tnl_pipeline_stage *stage )
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct texnorm_stage_data *store;
GLuint i;
stage->privatePtr = CALLOC(sizeof(*store));
store = TEXNORM_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
_mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
return GL_TRUE;
}
/* Called the first time stage->run() is invoked.
*/
static GLboolean
alloc_fog_data(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct fog_stage_data *store;
stage->privatePtr = malloc(sizeof(*store));
store = FOG_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->fogcoord, 0, tnl->vb.Size, 32 );
if (!inited)
init_static_data();
return GL_TRUE;
}
static GLboolean alloc_point_data( GLcontext *ctx,
struct gl_pipeline_stage *stage )
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct point_stage_data *store;
stage->privatePtr = MALLOC(sizeof(*store));
store = POINT_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->PointSize, 0, VB->Size, 32 );
/* Now run the stage.
*/
stage->run = run_point_stage;
return stage->run( ctx, stage );
}
/* Called the first time stage->run() is invoked.
*/
static GLboolean alloc_texgen_data( struct gl_context *ctx,
struct tnl_pipeline_stage *stage )
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct texgen_stage_data *store;
stage->privatePtr = CALLOC(sizeof(*store));
store = TEXGEN_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
_mesa_vector4f_alloc( &store->texcoord, 0, VB->Size, 32 );
store->tmp_f = (GLfloat (*)[3]) MALLOC(VB->Size * sizeof(GLfloat) * 3);
store->tmp_m = (GLfloat *) MALLOC(VB->Size * sizeof(GLfloat));
return GL_TRUE;
}
/**
* Called the first time stage->run is called. In effect, don't
* allocate data until the first time the stage is run.
*/
static GLboolean
init_vp(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_buffer *VB = &(tnl->vb);
struct vp_stage_data *store;
const GLuint size = VB->Size;
stage->privatePtr = calloc(1, sizeof(*store));
store = VP_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
/* a few other misc allocations */
_mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
store->clipmask = _mesa_align_malloc(sizeof(GLubyte)*size, 32 );
return GL_TRUE;
}
/* Called the first time stage->run() is invoked.
*/
static GLboolean alloc_texgen_data( struct gl_context *ctx,
struct tnl_pipeline_stage *stage )
{
struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
struct texgen_stage_data *store;
GLuint i;
stage->privatePtr = calloc(1, sizeof(*store));
store = TEXGEN_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++)
_mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
store->tmp_f = malloc(VB->Size * sizeof(GLfloat) * 3);
store->tmp_m = malloc(VB->Size * sizeof(GLfloat));
return GL_TRUE;
}
/* Called the first time stage->run is called. In effect, don't
* allocate data until the first time the stage is run.
*/
static GLboolean init_lighting( GLcontext *ctx,
struct tnl_pipeline_stage *stage )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct light_stage_data *store;
GLuint size = tnl->vb.Size;
stage->privatePtr = MALLOC(sizeof(*store));
store = LIGHT_STAGE_DATA(stage);
if (!store)
return GL_FALSE;
/* Do onetime init.
*/
init_lighting_tables();
_mesa_vector4f_alloc( &store->Input, 0, size, 32 );
_mesa_vector4f_alloc( &store->LitColor[0], 0, size, 32 );
_mesa_vector4f_alloc( &store->LitColor[1], 0, size, 32 );
_mesa_vector4f_alloc( &store->LitSecondary[0], 0, size, 32 );
_mesa_vector4f_alloc( &store->LitSecondary[1], 0, size, 32 );
_mesa_vector4f_alloc( &store->LitIndex[0], 0, size, 32 );
_mesa_vector4f_alloc( &store->LitIndex[1], 0, size, 32 );
store->LitColor[0].size = 4;
store->LitColor[1].size = 4;
store->LitSecondary[0].size = 3;
store->LitSecondary[1].size = 3;
store->LitIndex[0].size = 1;
store->LitIndex[0].stride = sizeof(GLfloat);
store->LitIndex[1].size = 1;
store->LitIndex[1].stride = sizeof(GLfloat);
return GL_TRUE;
}
/* Create the device specific context.
*/
GLboolean
r100CreateContext( gl_api api,
const struct gl_config *glVisual,
__DRIcontext *driContextPriv,
unsigned major_version,
unsigned minor_version,
uint32_t flags,
unsigned *error,
void *sharedContextPrivate)
{
__DRIscreen *sPriv = driContextPriv->driScreenPriv;
radeonScreenPtr screen = (radeonScreenPtr)(sPriv->driverPrivate);
struct dd_function_table functions;
r100ContextPtr rmesa;
struct gl_context *ctx;
int i;
int tcl_mode, fthrottle_mode;
switch (api) {
case API_OPENGL_COMPAT:
if (major_version > 1 || minor_version > 3) {
*error = __DRI_CTX_ERROR_BAD_VERSION;
return GL_FALSE;
}
break;
case API_OPENGLES:
break;
default:
*error = __DRI_CTX_ERROR_BAD_API;
return GL_FALSE;
}
/* Flag filtering is handled in dri2CreateContextAttribs.
*/
(void) flags;
assert(glVisual);
assert(driContextPriv);
assert(screen);
/* Allocate the Radeon context */
rmesa = calloc(1, sizeof(*rmesa));
if ( !rmesa ) {
*error = __DRI_CTX_ERROR_NO_MEMORY;
return GL_FALSE;
}
rmesa->radeon.radeonScreen = screen;
r100_init_vtbl(&rmesa->radeon);
/* init exp fog table data */
radeonInitStaticFogData();
/* Parse configuration files.
* Do this here so that initialMaxAnisotropy is set before we create
* the default textures.
*/
driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
screen->driScreen->myNum, "radeon");
rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
"def_max_anisotropy");
if ( driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) {
if ( sPriv->drm_version.minor < 13 )
fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
"disabling.\n", sPriv->drm_version.minor );
else
rmesa->using_hyperz = GL_TRUE;
}
if ( sPriv->drm_version.minor >= 15 )
rmesa->texmicrotile = GL_TRUE;
/* Init default driver functions then plug in our Radeon-specific functions
* (the texture functions are especially important)
*/
_mesa_init_driver_functions( &functions );
radeonInitTextureFuncs( &rmesa->radeon, &functions );
radeonInitQueryObjFunctions(&functions);
if (!radeonInitContext(&rmesa->radeon, &functions,
glVisual, driContextPriv,
sharedContextPrivate)) {
free(rmesa);
*error = __DRI_CTX_ERROR_NO_MEMORY;
return GL_FALSE;
}
rmesa->radeon.swtcl.RenderIndex = ~0;
rmesa->radeon.hw.all_dirty = GL_TRUE;
ctx = &rmesa->radeon.glCtx;
/* Initialize the software rasterizer and helper modules.
*/
_swrast_CreateContext( ctx );
_vbo_CreateContext( ctx );
_tnl_CreateContext( ctx );
_swsetup_CreateContext( ctx );
_ae_create_context( ctx );
/* Set the maximum texture size small enough that we can guarentee that
* all texture units can bind a maximal texture and have all of them in
* texturable memory at once. Depending on the allow_large_textures driconf
* setting allow larger textures.
*/
ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
"texture_units");
ctx->Const.FragmentProgram.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
ctx->Const.MaxCombinedTextureImageUnits = ctx->Const.MaxTextureUnits;
ctx->Const.StripTextureBorder = GL_TRUE;
i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures");
/* FIXME: When no memory manager is available we should set this
* to some reasonable value based on texture memory pool size */
ctx->Const.MaxTextureLevels = 12;
ctx->Const.Max3DTextureLevels = 9;
ctx->Const.MaxCubeTextureLevels = 12;
ctx->Const.MaxTextureRectSize = 2048;
ctx->Const.MaxTextureMaxAnisotropy = 16.0;
/* No wide points.
*/
ctx->Const.MinPointSize = 1.0;
ctx->Const.MinPointSizeAA = 1.0;
ctx->Const.MaxPointSize = 1.0;
ctx->Const.MaxPointSizeAA = 1.0;
ctx->Const.MinLineWidth = 1.0;
ctx->Const.MinLineWidthAA = 1.0;
ctx->Const.MaxLineWidth = 10.0;
ctx->Const.MaxLineWidthAA = 10.0;
ctx->Const.LineWidthGranularity = 0.0625;
/* Set maxlocksize (and hence vb size) small enough to avoid
* fallbacks in radeon_tcl.c. ie. guarentee that all vertices can
* fit in a single dma buffer for indexed rendering of quad strips,
* etc.
*/
ctx->Const.MaxArrayLockSize =
MIN2( ctx->Const.MaxArrayLockSize,
RADEON_BUFFER_SIZE / RADEON_MAX_TCL_VERTSIZE );
rmesa->boxes = 0;
ctx->Const.MaxDrawBuffers = 1;
ctx->Const.MaxColorAttachments = 1;
ctx->Const.MaxRenderbufferSize = 2048;
ctx->ShaderCompilerOptions[MESA_SHADER_VERTEX].PreferDP4 = true;
/* Install the customized pipeline:
*/
_tnl_destroy_pipeline( ctx );
_tnl_install_pipeline( ctx, radeon_pipeline );
/* Try and keep materials and vertices separate:
*/
/* _tnl_isolate_materials( ctx, GL_TRUE ); */
/* Configure swrast and T&L to match hardware characteristics:
*/
_swrast_allow_pixel_fog( ctx, GL_FALSE );
_swrast_allow_vertex_fog( ctx, GL_TRUE );
_tnl_allow_pixel_fog( ctx, GL_FALSE );
_tnl_allow_vertex_fog( ctx, GL_TRUE );
for ( i = 0 ; i < RADEON_MAX_TEXTURE_UNITS ; i++ ) {
_math_matrix_ctr( &rmesa->TexGenMatrix[i] );
_math_matrix_ctr( &rmesa->tmpmat[i] );
_math_matrix_set_identity( &rmesa->TexGenMatrix[i] );
_math_matrix_set_identity( &rmesa->tmpmat[i] );
}
ctx->Extensions.ARB_texture_border_clamp = true;
ctx->Extensions.ARB_texture_env_combine = true;
ctx->Extensions.ARB_texture_env_crossbar = true;
ctx->Extensions.ARB_texture_env_dot3 = true;
ctx->Extensions.EXT_fog_coord = true;
ctx->Extensions.EXT_packed_depth_stencil = true;
ctx->Extensions.EXT_secondary_color = true;
ctx->Extensions.EXT_texture_env_dot3 = true;
ctx->Extensions.EXT_texture_filter_anisotropic = true;
ctx->Extensions.EXT_texture_mirror_clamp = true;
ctx->Extensions.ATI_texture_env_combine3 = true;
ctx->Extensions.ATI_texture_mirror_once = true;
ctx->Extensions.MESA_ycbcr_texture = true;
ctx->Extensions.NV_blend_square = true;
ctx->Extensions.OES_EGL_image = true;
ctx->Extensions.EXT_framebuffer_object = true;
ctx->Extensions.ARB_texture_cube_map = true;
if (rmesa->radeon.glCtx.Mesa_DXTn) {
ctx->Extensions.EXT_texture_compression_s3tc = true;
ctx->Extensions.ANGLE_texture_compression_dxt = true;
}
else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) {
ctx->Extensions.EXT_texture_compression_s3tc = true;
ctx->Extensions.ANGLE_texture_compression_dxt = true;
}
ctx->Extensions.NV_texture_rectangle = true;
ctx->Extensions.ARB_occlusion_query = true;
/* XXX these should really go right after _mesa_init_driver_functions() */
radeon_fbo_init(&rmesa->radeon);
radeonInitSpanFuncs( ctx );
radeonInitIoctlFuncs( ctx );
radeonInitStateFuncs( ctx );
radeonInitState( rmesa );
radeonInitSwtcl( ctx );
_mesa_vector4f_alloc( &rmesa->tcl.ObjClean, 0,
ctx->Const.MaxArrayLockSize, 32 );
fthrottle_mode = driQueryOptioni(&rmesa->radeon.optionCache, "fthrottle_mode");
rmesa->radeon.iw.irq_seq = -1;
rmesa->radeon.irqsEmitted = 0;
rmesa->radeon.do_irqs = (rmesa->radeon.radeonScreen->irq != 0 &&
fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
rmesa->radeon.do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
#if DO_DEBUG
RADEON_DEBUG = driParseDebugString( getenv( "RADEON_DEBUG" ),
debug_control );
#endif
tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) {
fprintf(stderr, "disabling 3D acceleration\n");
FALLBACK(rmesa, RADEON_FALLBACK_DISABLE, 1);
} else if (tcl_mode == DRI_CONF_TCL_SW ||
!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
fprintf(stderr, "Disabling HW TCL support\n");
}
TCL_FALLBACK(&rmesa->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
}
if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
/* _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
}
_mesa_compute_version(ctx);
/* Exec table initialization requires the version to be computed */
_mesa_initialize_dispatch_tables(ctx);
_mesa_initialize_vbo_vtxfmt(ctx);
*error = __DRI_CTX_ERROR_SUCCESS;
return GL_TRUE;
}
/* Create the device specific context.
*/
GLboolean
r100CreateContext( gl_api api,
const struct gl_config *glVisual,
__DRIcontext *driContextPriv,
const struct __DriverContextConfig *ctx_config,
unsigned *error,
void *sharedContextPrivate)
{
__DRIscreen *sPriv = driContextPriv->driScreenPriv;
radeonScreenPtr screen = (radeonScreenPtr)(sPriv->driverPrivate);
struct dd_function_table functions;
r100ContextPtr rmesa;
struct gl_context *ctx;
int i;
int tcl_mode, fthrottle_mode;
if (ctx_config->flags & ~(__DRI_CTX_FLAG_DEBUG | __DRI_CTX_FLAG_NO_ERROR)) {
*error = __DRI_CTX_ERROR_UNKNOWN_FLAG;
return false;
}
if (ctx_config->attribute_mask) {
*error = __DRI_CTX_ERROR_UNKNOWN_ATTRIBUTE;
return false;
}
assert(driContextPriv);
assert(screen);
/* Allocate the Radeon context */
rmesa = calloc(1, sizeof(*rmesa));
if ( !rmesa ) {
*error = __DRI_CTX_ERROR_NO_MEMORY;
return GL_FALSE;
}
rmesa->radeon.radeonScreen = screen;
r100_init_vtbl(&rmesa->radeon);
/* init exp fog table data */
radeonInitStaticFogData();
/* Parse configuration files.
* Do this here so that initialMaxAnisotropy is set before we create
* the default textures.
*/
driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
screen->driScreen->myNum, "radeon", NULL);
rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
"def_max_anisotropy");
if (driQueryOptionb(&rmesa->radeon.optionCache, "hyperz"))
rmesa->using_hyperz = GL_TRUE;
/* Init default driver functions then plug in our Radeon-specific functions
* (the texture functions are especially important)
*/
_mesa_init_driver_functions( &functions );
_tnl_init_driver_draw_function( &functions );
radeonInitTextureFuncs( &rmesa->radeon, &functions );
radeonInitQueryObjFunctions(&functions);
if (!radeonInitContext(&rmesa->radeon, api, &functions,
glVisual, driContextPriv,
sharedContextPrivate)) {
free(rmesa);
*error = __DRI_CTX_ERROR_NO_MEMORY;
return GL_FALSE;
}
rmesa->radeon.swtcl.RenderIndex = ~0;
rmesa->radeon.hw.all_dirty = GL_TRUE;
ctx = &rmesa->radeon.glCtx;
driContextSetFlags(ctx, ctx_config->flags);
/* Initialize the software rasterizer and helper modules.
*/
_swrast_CreateContext( ctx );
_vbo_CreateContext( ctx );
_tnl_CreateContext( ctx );
_swsetup_CreateContext( ctx );
ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
"texture_units");
ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
ctx->Const.MaxCombinedTextureImageUnits = ctx->Const.MaxTextureUnits;
ctx->Const.StripTextureBorder = GL_TRUE;
/* FIXME: When no memory manager is available we should set this
* to some reasonable value based on texture memory pool size */
ctx->Const.MaxTextureLevels = 12;
ctx->Const.Max3DTextureLevels = 9;
ctx->Const.MaxCubeTextureLevels = 12;
ctx->Const.MaxTextureRectSize = 2048;
ctx->Const.MaxTextureMaxAnisotropy = 16.0;
/* No wide points.
*/
ctx->Const.MinPointSize = 1.0;
ctx->Const.MinPointSizeAA = 1.0;
ctx->Const.MaxPointSize = 1.0;
ctx->Const.MaxPointSizeAA = 1.0;
ctx->Const.MinLineWidth = 1.0;
ctx->Const.MinLineWidthAA = 1.0;
ctx->Const.MaxLineWidth = 10.0;
ctx->Const.MaxLineWidthAA = 10.0;
ctx->Const.LineWidthGranularity = 0.0625;
/* Set maxlocksize (and hence vb size) small enough to avoid
* fallbacks in radeon_tcl.c. ie. guarentee that all vertices can
* fit in a single dma buffer for indexed rendering of quad strips,
* etc.
*/
ctx->Const.MaxArrayLockSize =
MIN2( ctx->Const.MaxArrayLockSize,
RADEON_BUFFER_SIZE / RADEON_MAX_TCL_VERTSIZE );
rmesa->boxes = 0;
ctx->Const.MaxDrawBuffers = 1;
ctx->Const.MaxColorAttachments = 1;
ctx->Const.MaxRenderbufferSize = 2048;
ctx->Const.ShaderCompilerOptions[MESA_SHADER_VERTEX].OptimizeForAOS = true;
/* Install the customized pipeline:
*/
_tnl_destroy_pipeline( ctx );
_tnl_install_pipeline( ctx, radeon_pipeline );
/* Try and keep materials and vertices separate:
*/
/* _tnl_isolate_materials( ctx, GL_TRUE ); */
/* Configure swrast and T&L to match hardware characteristics:
*/
_swrast_allow_pixel_fog( ctx, GL_FALSE );
_swrast_allow_vertex_fog( ctx, GL_TRUE );
_tnl_allow_pixel_fog( ctx, GL_FALSE );
_tnl_allow_vertex_fog( ctx, GL_TRUE );
for ( i = 0 ; i < RADEON_MAX_TEXTURE_UNITS ; i++ ) {
_math_matrix_ctr( &rmesa->TexGenMatrix[i] );
_math_matrix_ctr( &rmesa->tmpmat[i] );
_math_matrix_set_identity( &rmesa->TexGenMatrix[i] );
_math_matrix_set_identity( &rmesa->tmpmat[i] );
}
ctx->Extensions.ARB_occlusion_query = true;
ctx->Extensions.ARB_texture_border_clamp = true;
ctx->Extensions.ARB_texture_cube_map = true;
ctx->Extensions.ARB_texture_env_combine = true;
ctx->Extensions.ARB_texture_env_crossbar = true;
ctx->Extensions.ARB_texture_env_dot3 = true;
ctx->Extensions.ARB_texture_filter_anisotropic = true;
ctx->Extensions.ARB_texture_mirror_clamp_to_edge = true;
ctx->Extensions.ATI_texture_env_combine3 = true;
ctx->Extensions.ATI_texture_mirror_once = true;
ctx->Extensions.EXT_texture_env_dot3 = true;
ctx->Extensions.EXT_texture_filter_anisotropic = true;
ctx->Extensions.EXT_texture_mirror_clamp = true;
ctx->Extensions.MESA_ycbcr_texture = true;
ctx->Extensions.NV_texture_rectangle = true;
ctx->Extensions.OES_EGL_image = true;
ctx->Extensions.EXT_texture_compression_s3tc = true;
ctx->Extensions.ANGLE_texture_compression_dxt = true;
/* XXX these should really go right after _mesa_init_driver_functions() */
radeon_fbo_init(&rmesa->radeon);
radeonInitSpanFuncs( ctx );
radeonInitIoctlFuncs( ctx );
radeonInitStateFuncs( ctx );
radeonInitState( rmesa );
radeonInitSwtcl( ctx );
_mesa_vector4f_alloc( &rmesa->tcl.ObjClean, 0,
ctx->Const.MaxArrayLockSize, 32 );
fthrottle_mode = driQueryOptioni(&rmesa->radeon.optionCache, "fthrottle_mode");
rmesa->radeon.iw.irq_seq = -1;
rmesa->radeon.irqsEmitted = 0;
rmesa->radeon.do_irqs = (rmesa->radeon.radeonScreen->irq != 0 &&
fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
rmesa->radeon.do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
if (getenv("RADEON_NO_RAST")) {
fprintf(stderr, "disabling 3D acceleration\n");
FALLBACK(rmesa, RADEON_FALLBACK_DISABLE, 1);
} else if (tcl_mode == DRI_CONF_TCL_SW ||
!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
fprintf(stderr, "Disabling HW TCL support\n");
}
TCL_FALLBACK(&rmesa->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
}
if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
/* _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
}
_mesa_override_extensions(ctx);
_mesa_compute_version(ctx);
/* Exec table initialization requires the version to be computed */
_mesa_initialize_dispatch_tables(ctx);
_mesa_initialize_vbo_vtxfmt(ctx);
*error = __DRI_CTX_ERROR_SUCCESS;
return GL_TRUE;
}
/* Create the device specific context.
*/
GLboolean
radeonCreateContext( const __GLcontextModes *glVisual,
__DRIcontextPrivate *driContextPriv,
void *sharedContextPrivate)
{
__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private);
struct dd_function_table functions;
radeonContextPtr rmesa;
GLcontext *ctx, *shareCtx;
int i;
int tcl_mode, fthrottle_mode;
assert(glVisual);
assert(driContextPriv);
assert(screen);
/* Allocate the Radeon context */
rmesa = (radeonContextPtr) CALLOC( sizeof(*rmesa) );
if ( !rmesa )
return GL_FALSE;
/* init exp fog table data */
radeonInitStaticFogData();
/* Parse configuration files.
* Do this here so that initialMaxAnisotropy is set before we create
* the default textures.
*/
driParseConfigFiles (&rmesa->optionCache, &screen->optionCache,
screen->driScreen->myNum, "radeon");
rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache,
"def_max_anisotropy");
if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) {
if ( sPriv->drm_version.minor < 13 )
fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
"disabling.\n", sPriv->drm_version.minor );
else
rmesa->using_hyperz = GL_TRUE;
}
if ( sPriv->drm_version.minor >= 15 )
rmesa->texmicrotile = GL_TRUE;
/* Init default driver functions then plug in our Radeon-specific functions
* (the texture functions are especially important)
*/
_mesa_init_driver_functions( &functions );
radeonInitDriverFuncs( &functions );
radeonInitTextureFuncs( &functions );
/* Allocate the Mesa context */
if (sharedContextPrivate)
shareCtx = ((radeonContextPtr) sharedContextPrivate)->glCtx;
else
shareCtx = NULL;
rmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
&functions, (void *) rmesa);
if (!rmesa->glCtx) {
FREE(rmesa);
return GL_FALSE;
}
driContextPriv->driverPrivate = rmesa;
/* Init radeon context data */
rmesa->dri.context = driContextPriv;
rmesa->dri.screen = sPriv;
rmesa->dri.drawable = NULL;
rmesa->dri.readable = NULL;
rmesa->dri.hwContext = driContextPriv->hHWContext;
rmesa->dri.hwLock = &sPriv->pSAREA->lock;
rmesa->dri.fd = sPriv->fd;
rmesa->dri.drmMinor = sPriv->drm_version.minor;
rmesa->radeonScreen = screen;
rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA +
screen->sarea_priv_offset);
rmesa->dma.buf0_address = rmesa->radeonScreen->buffers->list[0].address;
(void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) );
make_empty_list( & rmesa->swapped );
rmesa->nr_heaps = screen->numTexHeaps;
for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa,
screen->texSize[i],
12,
RADEON_NR_TEX_REGIONS,
(drmTextureRegionPtr)rmesa->sarea->tex_list[i],
& rmesa->sarea->tex_age[i],
& rmesa->swapped,
sizeof( radeonTexObj ),
(destroy_texture_object_t *) radeonDestroyTexObj );
driSetTextureSwapCounterLocation( rmesa->texture_heaps[i],
& rmesa->c_textureSwaps );
}
rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache,
"texture_depth");
if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
rmesa->texture_depth = ( screen->cpp == 4 ) ?
DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
rmesa->swtcl.RenderIndex = ~0;
rmesa->hw.all_dirty = GL_TRUE;
/* Set the maximum texture size small enough that we can guarentee that
* all texture units can bind a maximal texture and have all of them in
* texturable memory at once. Depending on the allow_large_textures driconf
* setting allow larger textures.
*/
ctx = rmesa->glCtx;
ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache,
"texture_units");
ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures");
driCalculateMaxTextureLevels( rmesa->texture_heaps,
rmesa->nr_heaps,
& ctx->Const,
4,
11, /* max 2D texture size is 2048x2048 */
8, /* 256^3 */
9, /* \todo: max cube texture size seems to be 512x512(x6) */
11, /* max rect texture size is 2048x2048. */
12,
GL_FALSE,
i );
ctx->Const.MaxTextureMaxAnisotropy = 16.0;
/* No wide points.
*/
ctx->Const.MinPointSize = 1.0;
ctx->Const.MinPointSizeAA = 1.0;
ctx->Const.MaxPointSize = 1.0;
ctx->Const.MaxPointSizeAA = 1.0;
ctx->Const.MinLineWidth = 1.0;
ctx->Const.MinLineWidthAA = 1.0;
ctx->Const.MaxLineWidth = 10.0;
ctx->Const.MaxLineWidthAA = 10.0;
ctx->Const.LineWidthGranularity = 0.0625;
/* Set maxlocksize (and hence vb size) small enough to avoid
* fallbacks in radeon_tcl.c. ie. guarentee that all vertices can
* fit in a single dma buffer for indexed rendering of quad strips,
* etc.
*/
ctx->Const.MaxArrayLockSize =
MIN2( ctx->Const.MaxArrayLockSize,
RADEON_BUFFER_SIZE / RADEON_MAX_TCL_VERTSIZE );
rmesa->boxes = 0;
ctx->Const.MaxDrawBuffers = 1;
_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
/* Initialize the software rasterizer and helper modules.
*/
_swrast_CreateContext( ctx );
_vbo_CreateContext( ctx );
_tnl_CreateContext( ctx );
_swsetup_CreateContext( ctx );
_ae_create_context( ctx );
/* Install the customized pipeline:
*/
_tnl_destroy_pipeline( ctx );
_tnl_install_pipeline( ctx, radeon_pipeline );
/* Try and keep materials and vertices separate:
*/
/* _tnl_isolate_materials( ctx, GL_TRUE ); */
/* Configure swrast and T&L to match hardware characteristics:
*/
_swrast_allow_pixel_fog( ctx, GL_FALSE );
_swrast_allow_vertex_fog( ctx, GL_TRUE );
_tnl_allow_pixel_fog( ctx, GL_FALSE );
_tnl_allow_vertex_fog( ctx, GL_TRUE );
for ( i = 0 ; i < RADEON_MAX_TEXTURE_UNITS ; i++ ) {
_math_matrix_ctr( &rmesa->TexGenMatrix[i] );
_math_matrix_ctr( &rmesa->tmpmat[i] );
_math_matrix_set_identity( &rmesa->TexGenMatrix[i] );
_math_matrix_set_identity( &rmesa->tmpmat[i] );
}
driInitExtensions( ctx, card_extensions, GL_TRUE );
if (rmesa->radeonScreen->drmSupportsCubeMapsR100)
_mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
if (rmesa->glCtx->Mesa_DXTn) {
_mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
_mesa_enable_extension( ctx, "GL_S3_s3tc" );
}
else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) {
_mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
}
if (rmesa->dri.drmMinor >= 9)
_mesa_enable_extension( ctx, "GL_NV_texture_rectangle");
/* XXX these should really go right after _mesa_init_driver_functions() */
radeonInitIoctlFuncs( ctx );
radeonInitStateFuncs( ctx );
radeonInitSpanFuncs( ctx );
radeonInitState( rmesa );
radeonInitSwtcl( ctx );
_mesa_vector4f_alloc( &rmesa->tcl.ObjClean, 0,
ctx->Const.MaxArrayLockSize, 32 );
fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode");
rmesa->iw.irq_seq = -1;
rmesa->irqsEmitted = 0;
rmesa->do_irqs = (rmesa->radeonScreen->irq != 0 &&
fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
(*sPriv->systemTime->getUST)( & rmesa->swap_ust );
#if DO_DEBUG
RADEON_DEBUG = driParseDebugString( getenv( "RADEON_DEBUG" ),
debug_control );
#endif
tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
if (driQueryOptionb(&rmesa->optionCache, "no_rast")) {
fprintf(stderr, "disabling 3D acceleration\n");
FALLBACK(rmesa, RADEON_FALLBACK_DISABLE, 1);
} else if (tcl_mode == DRI_CONF_TCL_SW ||
!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
rmesa->radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
fprintf(stderr, "Disabling HW TCL support\n");
}
TCL_FALLBACK(rmesa->glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
}
if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
/* _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
}
return GL_TRUE;
}
/**
* This function executes vertex programs
*/
static GLboolean
run_vp( struct gl_context *ctx, struct tnl_pipeline_stage *stage )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vp_stage_data *store = VP_STAGE_DATA(stage);
struct vertex_buffer *VB = &tnl->vb;
struct gl_vertex_program *program = ctx->VertexProgram._Current;
struct gl_program_machine *machine = &store->machine;
GLuint outputs[VARYING_SLOT_MAX], numOutputs;
GLuint i, j;
if (!program)
return GL_TRUE;
/* ARB program or vertex shader */
_mesa_load_state_parameters(ctx, program->Base.Parameters);
/* make list of outputs to save some time below */
numOutputs = 0;
for (i = 0; i < VARYING_SLOT_MAX; i++) {
if (program->Base.OutputsWritten & BITFIELD64_BIT(i)) {
outputs[numOutputs++] = i;
}
}
/* Allocate result vectors. We delay this until now to avoid allocating
* memory that would never be used if we don't run the software tnl pipeline.
*/
if (!store->results[0].storage) {
for (i = 0; i < VARYING_SLOT_MAX; i++) {
assert(!store->results[i].storage);
_mesa_vector4f_alloc( &store->results[i], 0, VB->Size, 32 );
store->results[i].size = 4;
}
}
map_textures(ctx, program);
for (i = 0; i < VB->Count; i++) {
GLuint attr;
init_machine(ctx, machine, tnl->CurInstance);
#if 0
printf("Input %d: %f, %f, %f, %f\n", i,
VB->AttribPtr[0]->data[i][0],
VB->AttribPtr[0]->data[i][1],
VB->AttribPtr[0]->data[i][2],
VB->AttribPtr[0]->data[i][3]);
printf(" color: %f, %f, %f, %f\n",
VB->AttribPtr[3]->data[i][0],
VB->AttribPtr[3]->data[i][1],
VB->AttribPtr[3]->data[i][2],
VB->AttribPtr[3]->data[i][3]);
printf(" normal: %f, %f, %f, %f\n",
VB->AttribPtr[2]->data[i][0],
VB->AttribPtr[2]->data[i][1],
VB->AttribPtr[2]->data[i][2],
VB->AttribPtr[2]->data[i][3]);
#endif
/* the vertex array case */
for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
if (program->Base.InputsRead & BITFIELD64_BIT(attr)) {
const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data;
const GLuint size = VB->AttribPtr[attr]->size;
const GLuint stride = VB->AttribPtr[attr]->stride;
const GLfloat *data = (GLfloat *) (ptr + stride * i);
#ifdef NAN_CHECK
check_float(data[0]);
check_float(data[1]);
check_float(data[2]);
check_float(data[3]);
#endif
COPY_CLEAN_4V(machine->VertAttribs[attr], size, data);
}
}
/* execute the program */
_mesa_execute_program(ctx, &program->Base, machine);
/* copy the output registers into the VB->attribs arrays */
for (j = 0; j < numOutputs; j++) {
const GLuint attr = outputs[j];
#ifdef NAN_CHECK
check_float(machine->Outputs[attr][0]);
check_float(machine->Outputs[attr][1]);
check_float(machine->Outputs[attr][2]);
check_float(machine->Outputs[attr][3]);
#endif
COPY_4V(store->results[attr].data[i], machine->Outputs[attr]);
}
/* FOGC is a special case. Fragment shader expects (f,0,0,1) */
if (program->Base.OutputsWritten & BITFIELD64_BIT(VARYING_SLOT_FOGC)) {
store->results[VARYING_SLOT_FOGC].data[i][1] = 0.0;
store->results[VARYING_SLOT_FOGC].data[i][2] = 0.0;
store->results[VARYING_SLOT_FOGC].data[i][3] = 1.0;
}
#ifdef NAN_CHECK
ASSERT(machine->Outputs[0][3] != 0.0F);
#endif
#if 0
printf("HPOS: %f %f %f %f\n",
machine->Outputs[0][0],
machine->Outputs[0][1],
machine->Outputs[0][2],
machine->Outputs[0][3]);
#endif
}
unmap_textures(ctx, program);
if (program->IsPositionInvariant) {
/* We need the exact same transform as in the fixed function path here
* to guarantee invariance, depending on compiler optimization flags
* results could be different otherwise.
*/
VB->ClipPtr = TransformRaw( &store->results[0],
&ctx->_ModelProjectMatrix,
VB->AttribPtr[0] );
/* Drivers expect this to be clean to element 4...
*/
switch (VB->ClipPtr->size) {
case 1:
/* impossible */
case 2:
_mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
/* fall-through */
case 3:
_mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
/* fall-through */
case 4:
break;
}
}
else {
/* Setup the VB pointers so that the next pipeline stages get
* their data from the right place (the program output arrays).
*/
VB->ClipPtr = &store->results[VARYING_SLOT_POS];
VB->ClipPtr->size = 4;
VB->ClipPtr->count = VB->Count;
}
VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VARYING_SLOT_COL0];
VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VARYING_SLOT_COL1];
VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VARYING_SLOT_FOGC];
VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VARYING_SLOT_PSIZ];
VB->BackfaceColorPtr = &store->results[VARYING_SLOT_BFC0];
VB->BackfaceSecondaryColorPtr = &store->results[VARYING_SLOT_BFC1];
for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]
= &store->results[VARYING_SLOT_TEX0 + i];
}
for (i = 0; i < ctx->Const.MaxVarying; i++) {
if (program->Base.OutputsWritten & BITFIELD64_BIT(VARYING_SLOT_VAR0 + i)) {
/* Note: varying results get put into the generic attributes */
VB->AttribPtr[VERT_ATTRIB_GENERIC0+i]
= &store->results[VARYING_SLOT_VAR0 + i];
}
}
/* Perform NDC and cliptest operations:
*/
return do_ndc_cliptest(ctx, store);
}
