void
_mesa_update_fetch_functions(struct gl_context *ctx, GLuint unit)
{
struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
struct gl_sampler_object *samp;
GLuint face, i;
GLuint dims;
if (!texObj)
return;
samp = _mesa_get_samplerobj(ctx, unit);
dims = _mesa_get_texture_dimensions(texObj->Target);
for (face = 0; face < 6; face++) {
for (i = 0; i < MAX_TEXTURE_LEVELS; i++) {
if (texObj->Image[face][i]) {
set_fetch_functions(samp,
swrast_texture_image(texObj->Image[face][i]),
dims);
}
}
}
}
static GLboolean
update_single_texture(struct st_context *st,
struct pipe_sampler_view **sampler_view,
GLuint texUnit)
{
struct pipe_context *pipe = st->pipe;
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *samp;
struct gl_texture_object *texObj;
struct st_texture_object *stObj;
enum pipe_format view_format;
GLboolean retval;
samp = _mesa_get_samplerobj(ctx, texUnit);
texObj = ctx->Texture.Unit[texUnit]._Current;
if (!texObj) {
texObj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
samp = &texObj->Sampler;
}
stObj = st_texture_object(texObj);
retval = st_finalize_texture(ctx, st->pipe, texObj);
if (!retval) {
/* out of mem */
return GL_FALSE;
}
/* Determine the format of the texture sampler view */
if (texObj->Target == GL_TEXTURE_BUFFER) {
view_format =
st_mesa_format_to_pipe_format(stObj->base._BufferObjectFormat);
}
else {
view_format = stObj->pt->format;
/* If sRGB decoding is off, use the linear format */
if (samp->sRGBDecode == GL_SKIP_DECODE_EXT) {
view_format = util_format_linear(view_format);
}
}
/* if sampler view has changed dereference it */
if (stObj->sampler_view) {
if (check_sampler_swizzle(stObj->sampler_view,
stObj->base._Swizzle,
stObj->base.DepthMode) ||
(view_format != stObj->sampler_view->format) ||
stObj->base.BaseLevel != stObj->sampler_view->u.tex.first_level) {
pipe_sampler_view_reference(&stObj->sampler_view, NULL);
}
}
*sampler_view = st_get_texture_sampler_view_from_stobj(stObj, pipe,
samp,
view_format);
return GL_TRUE;
}
void
brw_populate_sampler_prog_key_data(struct gl_context *ctx,
const struct gl_program *prog,
unsigned sampler_count,
struct brw_sampler_prog_key_data *key)
{
struct brw_context *brw = brw_context(ctx);
for (int s = 0; s < sampler_count; s++) {
key->swizzles[s] = SWIZZLE_NOOP;
if (!(prog->SamplersUsed & (1 << s)))
continue;
int unit_id = prog->SamplerUnits[s];
const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id];
if (unit->_ReallyEnabled && unit->_Current->Target != GL_TEXTURE_BUFFER) {
const struct gl_texture_object *t = unit->_Current;
const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id);
const bool alpha_depth = t->DepthMode == GL_ALPHA &&
(img->_BaseFormat == GL_DEPTH_COMPONENT ||
img->_BaseFormat == GL_DEPTH_STENCIL);
/* Haswell handles texture swizzling as surface format overrides
* (except for GL_ALPHA); all other platforms need MOVs in the shader.
*/
if (!brw->is_haswell || alpha_depth)
key->swizzles[s] = brw_get_texture_swizzle(ctx, t);
if (img->InternalFormat == GL_YCBCR_MESA) {
key->yuvtex_mask |= 1 << s;
if (img->TexFormat == MESA_FORMAT_YCBCR)
key->yuvtex_swap_mask |= 1 << s;
}
if (sampler->MinFilter != GL_NEAREST &&
sampler->MagFilter != GL_NEAREST) {
if (sampler->WrapS == GL_CLAMP)
key->gl_clamp_mask[0] |= 1 << s;
if (sampler->WrapT == GL_CLAMP)
key->gl_clamp_mask[1] |= 1 << s;
if (sampler->WrapR == GL_CLAMP)
key->gl_clamp_mask[2] |= 1 << s;
}
/* gather4's channel select for green from RG32F is broken;
* requires a shader w/a on IVB; fixable with just SCS on HSW. */
if (brw->gen >= 7 && !brw->is_haswell && prog->UsesGather) {
if (img->InternalFormat == GL_RG32F)
key->gather_channel_quirk_mask |= 1 << s;
}
}
}
}
void r200TexUpdateParameters(struct gl_context *ctx, GLuint unit)
{
struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit);
radeonTexObj* t = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
r200SetTexMaxAnisotropy(t , samp->MaxAnisotropy);
r200SetTexFilter(t, samp->MinFilter, samp->MagFilter);
r200SetTexWrap(t, samp->WrapS, samp->WrapT, samp->WrapR);
r200SetTexBorderColor(t, samp->BorderColor.f);
}
/**
* Fetch a texel.
*/
static void
fetch_texel(struct gl_context * ctx, const GLfloat texcoord[4], GLfloat lambda,
GLuint unit, GLfloat color[4])
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
/* XXX use a float-valued TextureSample routine here!!! */
swrast->TextureSample[unit](ctx, _mesa_get_samplerobj(ctx, unit),
ctx->Texture.Unit[unit]._Current,
1, (const GLfloat(*)[4]) texcoord,
&lambda, (GLfloat (*)[4]) color);
}
void
brw_populate_sampler_prog_key_data(struct gl_context *ctx,
const struct gl_program *prog,
struct brw_sampler_prog_key_data *key)
{
struct intel_context *intel = intel_context(ctx);
for (int s = 0; s < MAX_SAMPLERS; s++) {
key->swizzles[s] = SWIZZLE_NOOP;
if (!(prog->SamplersUsed & (1 << s)))
continue;
int unit_id = prog->SamplerUnits[s];
const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id];
if (unit->_ReallyEnabled && unit->_Current->Target != GL_TEXTURE_BUFFER) {
const struct gl_texture_object *t = unit->_Current;
const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id);
const bool alpha_depth = t->DepthMode == GL_ALPHA &&
(img->_BaseFormat == GL_DEPTH_COMPONENT ||
img->_BaseFormat == GL_DEPTH_STENCIL);
/* Haswell handles texture swizzling as surface format overrides
* (except for GL_ALPHA); all other platforms need MOVs in the shader.
*/
if (!intel->is_haswell || alpha_depth)
key->swizzles[s] = brw_get_texture_swizzle(ctx, t);
if (img->InternalFormat == GL_YCBCR_MESA) {
key->yuvtex_mask |= 1 << s;
if (img->TexFormat == MESA_FORMAT_YCBCR)
key->yuvtex_swap_mask |= 1 << s;
}
if (sampler->MinFilter != GL_NEAREST &&
sampler->MagFilter != GL_NEAREST) {
if (sampler->WrapS == GL_CLAMP)
key->gl_clamp_mask[0] |= 1 << s;
if (sampler->WrapT == GL_CLAMP)
key->gl_clamp_mask[1] |= 1 << s;
if (sampler->WrapR == GL_CLAMP)
key->gl_clamp_mask[2] |= 1 << s;
}
}
}
}
static GLboolean
update_single_texture(struct st_context *st,
struct pipe_sampler_view **sampler_view,
GLuint texUnit, unsigned glsl_version)
{
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *samp;
struct gl_texture_object *texObj;
struct st_texture_object *stObj;
enum pipe_format view_format;
GLboolean retval;
samp = _mesa_get_samplerobj(ctx, texUnit);
texObj = ctx->Texture.Unit[texUnit]._Current;
if (!texObj) {
texObj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
samp = &texObj->Sampler;
}
stObj = st_texture_object(texObj);
retval = st_finalize_texture(ctx, st->pipe, texObj);
if (!retval) {
/* out of mem */
return GL_FALSE;
}
/* Determine the format of the texture sampler view */
if (texObj->Target == GL_TEXTURE_BUFFER) {
view_format =
st_mesa_format_to_pipe_format(st, stObj->base._BufferObjectFormat);
}
else {
view_format =
stObj->surface_based ? stObj->surface_format : stObj->pt->format;
/* If sRGB decoding is off, use the linear format */
if (samp->sRGBDecode == GL_SKIP_DECODE_EXT) {
view_format = util_format_linear(view_format);
}
}
*sampler_view =
st_get_texture_sampler_view_from_stobj(st, stObj, view_format,
glsl_version);
return GL_TRUE;
}
static GLboolean
update_single_texture(struct st_context *st,
struct pipe_sampler_view **sampler_view,
GLuint texUnit, unsigned glsl_version)
{
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *samp;
struct gl_texture_object *texObj;
struct st_texture_object *stObj;
GLboolean retval;
samp = _mesa_get_samplerobj(ctx, texUnit);
texObj = ctx->Texture.Unit[texUnit]._Current;
if (!texObj) {
texObj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
samp = &texObj->Sampler;
}
stObj = st_texture_object(texObj);
retval = st_finalize_texture(ctx, st->pipe, texObj);
if (!retval) {
/* out of mem */
return GL_FALSE;
}
/* Check a few pieces of state outside the texture object to see if we
* need to force revalidation.
*/
if (stObj->prev_glsl_version != glsl_version ||
stObj->prev_sRGBDecode != samp->sRGBDecode) {
st_texture_release_all_sampler_views(st, stObj);
stObj->prev_glsl_version = glsl_version;
stObj->prev_sRGBDecode = samp->sRGBDecode;
}
if (texObj->TargetIndex == TEXTURE_EXTERNAL_INDEX &&
stObj->pt->screen->resource_changed)
stObj->pt->screen->resource_changed(stObj->pt->screen, stObj->pt);
*sampler_view =
st_get_texture_sampler_view_from_stobj(st, stObj, samp, glsl_version);
return GL_TRUE;
}
void
_swrast_update_texture_samplers(struct gl_context *ctx)
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
GLuint u;
if (!swrast)
return; /* pipe hack */
for (u = 0; u < ARRAY_SIZE(swrast->TextureSample); u++) {
struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current;
/* Note: If tObj is NULL, the sample function will be a simple
* function that just returns opaque black (0,0,0,1).
*/
_mesa_update_fetch_functions(ctx, u);
swrast->TextureSample[u] =
_swrast_choose_texture_sample_func(ctx, tObj,
_mesa_get_samplerobj(ctx, u));
}
}
/**
* Fetch a texel with the given partial derivatives to compute a level
* of detail in the mipmap.
* Called via machine->FetchTexelDeriv()
* \param lodBias the lod bias which may be specified by a TXB instruction,
* otherwise zero.
*/
static void
fetch_texel_deriv( struct gl_context *ctx, const GLfloat texcoord[4],
const GLfloat texdx[4], const GLfloat texdy[4],
GLfloat lodBias, GLuint unit, GLfloat color[4] )
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
const struct gl_texture_object *texObj = texUnit->_Current;
if (texObj) {
const struct gl_texture_image *texImg =
texObj->Image[0][texObj->BaseLevel];
const struct swrast_texture_image *swImg =
swrast_texture_image_const(texImg);
const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit);
const GLfloat texW = (GLfloat) swImg->WidthScale;
const GLfloat texH = (GLfloat) swImg->HeightScale;
GLfloat lambda;
GLfloat rgba[4];
lambda = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */
texdx[1], texdy[1], /* dt/dx, dt/dy */
texdx[3], texdy[3], /* dq/dx, dq/dy */
texW, texH,
texcoord[0], texcoord[1], texcoord[3],
1.0F / texcoord[3]);
lambda += lodBias + texUnit->LodBias + samp->LodBias;
lambda = CLAMP(lambda, samp->MinLod, samp->MaxLod);
swrast->TextureSample[unit](ctx, samp, ctx->Texture.Unit[unit]._Current,
1, (const GLfloat (*)[4]) texcoord,
&lambda, &rgba);
swizzle_texel(rgba, color, texObj->_Swizzle);
}
else {
ASSIGN_4V(color, 0.0F, 0.0F, 0.0F, 1.0F);
}
}
/**
* Fetch a texel with given lod.
* Called via machine->FetchTexelLod()
*/
static void
fetch_texel_lod( struct gl_context *ctx, const GLfloat texcoord[4], GLfloat lambda,
GLuint unit, GLfloat color[4] )
{
const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
if (texObj) {
SWcontext *swrast = SWRAST_CONTEXT(ctx);
GLfloat rgba[4];
const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit);
lambda = CLAMP(lambda, samp->MinLod, samp->MaxLod);
swrast->TextureSample[unit](ctx, samp, ctx->Texture.Unit[unit]._Current,
1, (const GLfloat (*)[4]) texcoord,
&lambda, &rgba);
swizzle_texel(rgba, color, texObj->_Swizzle);
}
else {
ASSIGN_4V(color, 0.0F, 0.0F, 0.0F, 1.0F);
}
}
/**
* Get a pipe_sampler_view object from a texture unit.
*/
void
st_update_single_texture(struct st_context *st,
struct pipe_sampler_view **sampler_view,
GLuint texUnit, bool glsl130_or_later,
bool ignore_srgb_decode)
{
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *samp;
struct gl_texture_object *texObj;
struct st_texture_object *stObj;
samp = _mesa_get_samplerobj(ctx, texUnit);
texObj = ctx->Texture.Unit[texUnit]._Current;
assert(texObj);
stObj = st_texture_object(texObj);
if (unlikely(texObj->Target == GL_TEXTURE_BUFFER)) {
*sampler_view = st_get_buffer_sampler_view_from_stobj(st, stObj);
return;
}
if (!st_finalize_texture(ctx, st->pipe, texObj, 0) ||
!stObj->pt) {
/* out of mem */
*sampler_view = NULL;
return;
}
if (texObj->TargetIndex == TEXTURE_EXTERNAL_INDEX &&
stObj->pt->screen->resource_changed)
stObj->pt->screen->resource_changed(stObj->pt->screen, stObj->pt);
*sampler_view =
st_get_texture_sampler_view_from_stobj(st, stObj, samp,
glsl130_or_later,
ignore_srgb_decode);
}
void
brw_populate_sampler_prog_key_data(struct gl_context *ctx,
const struct gl_program *prog,
unsigned sampler_count,
struct brw_sampler_prog_key_data *key)
{
struct brw_context *brw = brw_context(ctx);
for (int s = 0; s < sampler_count; s++) {
key->swizzles[s] = SWIZZLE_NOOP;
if (!(prog->SamplersUsed & (1 << s)))
continue;
int unit_id = prog->SamplerUnits[s];
const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id];
if (unit->_Current && unit->_Current->Target != GL_TEXTURE_BUFFER) {
const struct gl_texture_object *t = unit->_Current;
const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id);
const bool alpha_depth = t->DepthMode == GL_ALPHA &&
(img->_BaseFormat == GL_DEPTH_COMPONENT ||
img->_BaseFormat == GL_DEPTH_STENCIL);
/* Haswell handles texture swizzling as surface format overrides
* (except for GL_ALPHA); all other platforms need MOVs in the shader.
*/
if (alpha_depth || (brw->gen < 8 && !brw->is_haswell))
key->swizzles[s] = brw_get_texture_swizzle(ctx, t);
if (brw->gen < 8 &&
sampler->MinFilter != GL_NEAREST &&
sampler->MagFilter != GL_NEAREST) {
if (sampler->WrapS == GL_CLAMP)
key->gl_clamp_mask[0] |= 1 << s;
if (sampler->WrapT == GL_CLAMP)
key->gl_clamp_mask[1] |= 1 << s;
if (sampler->WrapR == GL_CLAMP)
key->gl_clamp_mask[2] |= 1 << s;
}
/* gather4's channel select for green from RG32F is broken; requires
* a shader w/a on IVB; fixable with just SCS on HSW.
*/
if (brw->gen == 7 && !brw->is_haswell && prog->UsesGather) {
if (img->InternalFormat == GL_RG32F)
key->gather_channel_quirk_mask |= 1 << s;
}
/* Gen6's gather4 is broken for UINT/SINT; we treat them as
* UNORM/FLOAT instead and fix it in the shader.
*/
if (brw->gen == 6 && prog->UsesGather) {
key->gen6_gather_wa[s] = gen6_gather_workaround(img->InternalFormat);
}
/* If this is a multisample sampler, and uses the CMS MSAA layout,
* then we need to emit slightly different code to first sample the
* MCS surface.
*/
struct intel_texture_object *intel_tex =
intel_texture_object((struct gl_texture_object *)t);
if (brw->gen >= 7 &&
intel_tex->mt->msaa_layout == INTEL_MSAA_LAYOUT_CMS) {
key->compressed_multisample_layout_mask |= 1 << s;
if (intel_tex->mt->num_samples >= 16) {
assert(brw->gen >= 9);
key->msaa_16 |= 1 << s;
}
}
}
}
}
/**
* Apply texture mapping to a span of fragments.
*/
void
_swrast_texture_span( struct gl_context *ctx, SWspan *span )
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
float4_array primary_rgba;
GLuint unit;
if (!swrast->TexelBuffer) {
#ifdef _OPENMP
const GLint maxThreads = omp_get_max_threads();
#else
const GLint maxThreads = 1;
#endif
/* TexelBuffer is also global and normally shared by all SWspan
* instances; when running with multiple threads, create one per
* thread.
*/
swrast->TexelBuffer =
malloc(ctx->Const.FragmentProgram.MaxTextureImageUnits * maxThreads *
SWRAST_MAX_WIDTH * 4 * sizeof(GLfloat));
if (!swrast->TexelBuffer) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
return;
}
}
primary_rgba = malloc(span->end * 4 * sizeof(GLfloat));
if (!primary_rgba) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_span");
return;
}
ASSERT(span->end <= SWRAST_MAX_WIDTH);
/*
* Save copy of the incoming fragment colors (the GL_PRIMARY_COLOR)
*/
if (swrast->_TextureCombinePrimary) {
GLuint i;
for (i = 0; i < span->end; i++) {
primary_rgba[i][RCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][RCOMP]);
primary_rgba[i][GCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][GCOMP]);
primary_rgba[i][BCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][BCOMP]);
primary_rgba[i][ACOMP] = CHAN_TO_FLOAT(span->array->rgba[i][ACOMP]);
}
}
/* First must sample all bump maps */
for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
if (texUnit->_ReallyEnabled &&
texUnit->_CurrentCombine->ModeRGB == GL_BUMP_ENVMAP_ATI) {
const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
span->array->attribs[VARYING_SLOT_TEX0 + unit];
float4_array targetcoords =
span->array->attribs[VARYING_SLOT_TEX0 +
ctx->Texture.Unit[unit].BumpTarget - GL_TEXTURE0];
const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit);
GLfloat *lambda = span->array->lambda[unit];
float4_array texels = get_texel_array(swrast, unit);
GLuint i;
GLfloat rotMatrix00 = ctx->Texture.Unit[unit].RotMatrix[0];
GLfloat rotMatrix01 = ctx->Texture.Unit[unit].RotMatrix[1];
GLfloat rotMatrix10 = ctx->Texture.Unit[unit].RotMatrix[2];
GLfloat rotMatrix11 = ctx->Texture.Unit[unit].RotMatrix[3];
/* adjust texture lod (lambda) */
if (span->arrayMask & SPAN_LAMBDA) {
if (texUnit->LodBias + samp->LodBias != 0.0F) {
/* apply LOD bias, but don't clamp yet */
const GLfloat bias = CLAMP(texUnit->LodBias + samp->LodBias,
-ctx->Const.MaxTextureLodBias,
ctx->Const.MaxTextureLodBias);
GLuint i;
for (i = 0; i < span->end; i++) {
lambda[i] += bias;
}
}
if (samp->MinLod != -1000.0 ||
samp->MaxLod != 1000.0) {
/* apply LOD clamping to lambda */
const GLfloat min = samp->MinLod;
const GLfloat max = samp->MaxLod;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat l = lambda[i];
lambda[i] = CLAMP(l, min, max);
}
}
}
/* Sample the texture (span->end = number of fragments) */
swrast->TextureSample[unit]( ctx, samp,
ctx->Texture.Unit[unit]._Current,
span->end, texcoords, lambda, texels );
/* manipulate the span values of the bump target
not sure this can work correctly even ignoring
the problem that channel is unsigned */
for (i = 0; i < span->end; i++) {
targetcoords[i][0] += (texels[i][0] * rotMatrix00 + texels[i][1] *
rotMatrix01) / targetcoords[i][3];
targetcoords[i][1] += (texels[i][0] * rotMatrix10 + texels[i][1] *
rotMatrix11) / targetcoords[i][3];
}
}
}
/*
* Must do all texture sampling before combining in order to
* accomodate GL_ARB_texture_env_crossbar.
*/
for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
if (texUnit->_ReallyEnabled &&
texUnit->_CurrentCombine->ModeRGB != GL_BUMP_ENVMAP_ATI) {
const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
span->array->attribs[VARYING_SLOT_TEX0 + unit];
const struct gl_texture_object *curObj = texUnit->_Current;
const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, unit);
GLfloat *lambda = span->array->lambda[unit];
float4_array texels = get_texel_array(swrast, unit);
/* adjust texture lod (lambda) */
if (span->arrayMask & SPAN_LAMBDA) {
if (texUnit->LodBias + samp->LodBias != 0.0F) {
/* apply LOD bias, but don't clamp yet */
const GLfloat bias = CLAMP(texUnit->LodBias + samp->LodBias,
-ctx->Const.MaxTextureLodBias,
ctx->Const.MaxTextureLodBias);
GLuint i;
for (i = 0; i < span->end; i++) {
lambda[i] += bias;
}
}
if (samp->MinLod != -1000.0 ||
samp->MaxLod != 1000.0) {
/* apply LOD clamping to lambda */
const GLfloat min = samp->MinLod;
const GLfloat max = samp->MaxLod;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat l = lambda[i];
lambda[i] = CLAMP(l, min, max);
}
}
}
else if (samp->MaxAnisotropy > 1.0 &&
samp->MinFilter == GL_LINEAR_MIPMAP_LINEAR) {
/* sample_lambda_2d_aniso is beeing used as texture_sample_func,
* it requires the current SWspan *span as an additional parameter.
* In order to keep the same function signature, the unused lambda
* parameter will be modified to actually contain the SWspan pointer.
* This is a Hack. To make it right, the texture_sample_func
* signature and all implementing functions need to be modified.
*/
/* "hide" SWspan struct; cast to (GLfloat *) to suppress warning */
lambda = (GLfloat *)span;
}
/* Sample the texture (span->end = number of fragments) */
swrast->TextureSample[unit]( ctx, samp,
ctx->Texture.Unit[unit]._Current,
span->end, texcoords, lambda, texels );
/* GL_EXT_texture_swizzle */
if (curObj->_Swizzle != SWIZZLE_NOOP) {
swizzle_texels(curObj->_Swizzle, span->end, texels);
}
}
}
/*
* OK, now apply the texture (aka texture combine/blend).
* We modify the span->color.rgba values.
*/
for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
if (ctx->Texture.Unit[unit]._ReallyEnabled)
texture_combine(ctx, unit, primary_rgba, swrast->TexelBuffer, span);
}
free(primary_rgba);
}
GLuint
intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
{
struct gl_context *ctx = &intel->ctx;
struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint face, i;
GLuint nr_faces = 0;
struct intel_texture_image *firstImage;
int width, height, depth;
/* TBOs require no validation -- they always just point to their BO. */
if (tObj->Target == GL_TEXTURE_BUFFER)
return true;
/* We know/require this is true by now:
*/
assert(intelObj->base._BaseComplete);
/* What levels must the tree include at a minimum?
*/
intel_update_max_level(intelObj, sampler);
if (intelObj->mt && intelObj->mt->first_level != tObj->BaseLevel)
intelObj->needs_validate = true;
if (!intelObj->needs_validate)
return true;
firstImage = intel_texture_image(tObj->Image[0][tObj->BaseLevel]);
/* Check tree can hold all active levels. Check tree matches
* target, imageFormat, etc.
*
* For pre-gen4, we have to match first_level == tObj->BaseLevel,
* because we don't have the control that gen4 does to make min/mag
* determination happen at a nonzero (hardware) baselevel. Because
* of that, we just always relayout on baselevel change.
*/
if (intelObj->mt &&
(!intel_miptree_match_image(intelObj->mt, &firstImage->base.Base) ||
intelObj->mt->first_level != tObj->BaseLevel ||
intelObj->mt->last_level < intelObj->_MaxLevel)) {
intel_miptree_release(&intelObj->mt);
}
/* May need to create a new tree:
*/
if (!intelObj->mt) {
intel_miptree_get_dimensions_for_image(&firstImage->base.Base,
&width, &height, &depth);
perf_debug("Creating new %s %dx%dx%d %d..%d miptree to handle finalized "
"texture miptree.\n",
_mesa_get_format_name(firstImage->base.Base.TexFormat),
width, height, depth, tObj->BaseLevel, intelObj->_MaxLevel);
intelObj->mt = intel_miptree_create(intel,
intelObj->base.Target,
firstImage->base.Base.TexFormat,
tObj->BaseLevel,
intelObj->_MaxLevel,
width,
height,
depth,
true,
0 /* num_samples */,
false /* force_y_tiling */);
if (!intelObj->mt)
return false;
}
/* Pull in any images not in the object's tree:
*/
nr_faces = _mesa_num_tex_faces(intelObj->base.Target);
for (face = 0; face < nr_faces; face++) {
for (i = tObj->BaseLevel; i <= intelObj->_MaxLevel; i++) {
struct intel_texture_image *intelImage =
intel_texture_image(intelObj->base.Image[face][i]);
/* skip too small size mipmap */
if (intelImage == NULL)
break;
if (intelObj->mt != intelImage->mt) {
intel_miptree_copy_teximage(intel, intelImage, intelObj->mt,
false /* invalidate */);
}
/* After we're done, we'd better agree that our layout is
* appropriate, or we'll end up hitting this function again on the
* next draw
*/
assert(intel_miptree_match_image(intelObj->mt, &intelImage->base.Base));
}
}
intelObj->needs_validate = false;
return true;
}
GLuint
intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
{
struct gl_context *ctx = &intel->ctx;
struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint face, i;
GLuint nr_faces = 0;
struct intel_texture_image *firstImage;
/* We know/require this is true by now:
*/
assert(intelObj->base._Complete);
/* What levels must the tree include at a minimum?
*/
intel_update_max_level(intel, intelObj, sampler);
firstImage = intel_texture_image(tObj->Image[0][tObj->BaseLevel]);
/* Fallback case:
*/
if (firstImage->base.Border) {
if (intelObj->mt) {
intel_miptree_release(intel, &intelObj->mt);
}
return GL_FALSE;
}
/* Check tree can hold all active levels. Check tree matches
* target, imageFormat, etc.
*
* For pre-gen4, we have to match first_level == tObj->BaseLevel,
* because we don't have the control that gen4 does to make min/mag
* determination happen at a nonzero (hardware) baselevel. Because
* of that, we just always relayout on baselevel change.
*/
if (intelObj->mt &&
(intelObj->mt->target != intelObj->base.Target ||
intelObj->mt->format != firstImage->base.TexFormat ||
intelObj->mt->first_level != tObj->BaseLevel ||
intelObj->mt->last_level < intelObj->_MaxLevel ||
intelObj->mt->width0 != firstImage->base.Width ||
intelObj->mt->height0 != firstImage->base.Height ||
intelObj->mt->depth0 != firstImage->base.Depth)) {
intel_miptree_release(intel, &intelObj->mt);
}
/* May need to create a new tree:
*/
if (!intelObj->mt) {
intelObj->mt = intel_miptree_create(intel,
intelObj->base.Target,
firstImage->base.TexFormat,
tObj->BaseLevel,
intelObj->_MaxLevel,
firstImage->base.Width,
firstImage->base.Height,
firstImage->base.Depth,
GL_TRUE);
if (!intelObj->mt)
return GL_FALSE;
}
/* Pull in any images not in the object's tree:
*/
nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
for (face = 0; face < nr_faces; face++) {
for (i = tObj->BaseLevel; i <= intelObj->_MaxLevel; i++) {
struct intel_texture_image *intelImage =
intel_texture_image(intelObj->base.Image[face][i]);
/* skip too small size mipmap */
if (intelImage == NULL)
break;
/* Need to import images in main memory or held in other trees.
* If it's a render target, then its data isn't needed to be in
* the object tree (otherwise we'd be FBO incomplete), and we need
* to keep track of the image's MT as needing to be pulled in still,
* or we'll lose the rendering that's done to it.
*/
if (intelObj->mt != intelImage->mt &&
!intelImage->used_as_render_target) {
copy_image_data_to_tree(intel, intelObj, intelImage);
}
}
}
return GL_TRUE;
}
/**
* At rendering-from-a-texture time, make sure that the texture object has a
* miptree that can hold the entire texture based on
* BaseLevel/MaxLevel/filtering, and copy in any texture images that are
* stored in other miptrees.
*/
GLuint
intel_finalize_mipmap_tree(struct brw_context *brw, GLuint unit)
{
struct gl_context *ctx = &brw->ctx;
struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint face, i;
GLuint nr_faces = 0;
struct intel_texture_image *firstImage;
int width, height, depth;
/* TBOs require no validation -- they always just point to their BO. */
if (tObj->Target == GL_TEXTURE_BUFFER)
return true;
/* We know that this is true by now, and if it wasn't, we might have
* mismatched level sizes and the copies would fail.
*/
assert(intelObj->base._BaseComplete);
intel_update_max_level(intelObj, sampler);
/* What levels does this validated texture image require? */
int validate_first_level = tObj->BaseLevel;
int validate_last_level = intelObj->_MaxLevel;
/* Skip the loop over images in the common case of no images having
* changed. But if the GL_BASE_LEVEL or GL_MAX_LEVEL change to something we
* haven't looked at, then we do need to look at those new images.
*/
if (!intelObj->needs_validate &&
validate_first_level >= intelObj->validated_first_level &&
validate_last_level <= intelObj->validated_last_level) {
return true;
}
/* Immutable textures should not get this far -- they should have been
* created in a validated state, and nothing can invalidate them.
*/
assert(!tObj->Immutable);
firstImage = intel_texture_image(tObj->Image[0][tObj->BaseLevel]);
/* Check tree can hold all active levels. Check tree matches
* target, imageFormat, etc.
*
* For pre-gen4, we have to match first_level == tObj->BaseLevel,
* because we don't have the control that gen4 does to make min/mag
* determination happen at a nonzero (hardware) baselevel. Because
* of that, we just always relayout on baselevel change.
*/
if (intelObj->mt &&
(!intel_miptree_match_image(intelObj->mt, &firstImage->base.Base) ||
validate_first_level < intelObj->mt->first_level ||
validate_last_level > intelObj->mt->last_level)) {
intel_miptree_release(&intelObj->mt);
}
/* May need to create a new tree:
*/
if (!intelObj->mt) {
intel_miptree_get_dimensions_for_image(&firstImage->base.Base,
&width, &height, &depth);
perf_debug("Creating new %s %dx%dx%d %d-level miptree to handle "
"finalized texture miptree.\n",
_mesa_get_format_name(firstImage->base.Base.TexFormat),
width, height, depth, validate_last_level + 1);
intelObj->mt = intel_miptree_create(brw,
intelObj->base.Target,
firstImage->base.Base.TexFormat,
0, /* first_level */
validate_last_level,
width,
height,
depth,
true,
0 /* num_samples */,
INTEL_MIPTREE_TILING_ANY);
if (!intelObj->mt)
return false;
}
/* Pull in any images not in the object's tree:
*/
nr_faces = _mesa_num_tex_faces(intelObj->base.Target);
for (face = 0; face < nr_faces; face++) {
for (i = validate_first_level; i <= validate_last_level; i++) {
struct intel_texture_image *intelImage =
intel_texture_image(intelObj->base.Image[face][i]);
/* skip too small size mipmap */
if (intelImage == NULL)
break;
if (intelObj->mt != intelImage->mt) {
intel_miptree_copy_teximage(brw, intelImage, intelObj->mt,
false /* invalidate */);
}
/* After we're done, we'd better agree that our layout is
* appropriate, or we'll end up hitting this function again on the
* next draw
*/
assert(intel_miptree_match_image(intelObj->mt, &intelImage->base.Base));
}
}
intelObj->validated_first_level = validate_first_level;
intelObj->validated_last_level = validate_last_level;
intelObj->_Format = intelObj->mt->format;
intelObj->needs_validate = false;
return true;
}
/* Recalculate all state from scratch. Perhaps not the most
* efficient, but this has gotten complex enough that we need
* something which is understandable and reliable.
*/
static bool
i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
{
struct gl_context *ctx = &intel->ctx;
struct i915_context *i915 = i915_context(ctx);
struct gl_texture_unit *tUnit = &ctx->Texture.Unit[unit];
struct gl_texture_object *tObj = tUnit->_Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_texture_image *firstImage;
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint *state = i915->state.Tex[unit], format, pitch;
GLint lodbias, aniso = 0;
GLubyte border[4];
GLfloat maxlod;
memset(state, 0, sizeof(state));
/*We need to refcount these. */
if (i915->state.tex_buffer[unit] != NULL) {
drm_intel_bo_unreference(i915->state.tex_buffer[unit]);
i915->state.tex_buffer[unit] = NULL;
}
if (!intel_finalize_mipmap_tree(intel, unit))
return false;
/* Get first image here, since intelObj->firstLevel will get set in
* the intel_finalize_mipmap_tree() call above.
*/
firstImage = tObj->Image[0][tObj->BaseLevel];
drm_intel_bo_reference(intelObj->mt->region->bo);
i915->state.tex_buffer[unit] = intelObj->mt->region->bo;
i915->state.tex_offset[unit] = intelObj->mt->offset;
format = translate_texture_format(firstImage->TexFormat,
tObj->DepthMode);
pitch = intelObj->mt->region->pitch * intelObj->mt->cpp;
state[I915_TEXREG_MS3] =
(((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) |
((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format);
if (intelObj->mt->region->tiling != I915_TILING_NONE) {
state[I915_TEXREG_MS3] |= MS3_TILED_SURFACE;
if (intelObj->mt->region->tiling == I915_TILING_Y)
state[I915_TEXREG_MS3] |= MS3_TILE_WALK;
}
/* We get one field with fraction bits for the maximum addressable
* (lowest resolution) LOD. Use it to cover both MAX_LEVEL and
* MAX_LOD.
*/
maxlod = MIN2(sampler->MaxLod, tObj->_MaxLevel - tObj->BaseLevel);
state[I915_TEXREG_MS4] =
((((pitch / 4) - 1) << MS4_PITCH_SHIFT) |
MS4_CUBE_FACE_ENA_MASK |
(U_FIXED(CLAMP(maxlod, 0.0, 11.0), 2) << MS4_MAX_LOD_SHIFT) |
((firstImage->Depth - 1) << MS4_VOLUME_DEPTH_SHIFT));
{
GLuint minFilt, mipFilt, magFilt;
switch (sampler->MinFilter) {
case GL_NEAREST:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_NONE;
break;
case GL_LINEAR:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_NONE;
break;
case GL_NEAREST_MIPMAP_NEAREST:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_NEAREST;
break;
case GL_LINEAR_MIPMAP_NEAREST:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_NEAREST;
break;
case GL_NEAREST_MIPMAP_LINEAR:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_LINEAR;
break;
case GL_LINEAR_MIPMAP_LINEAR:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_LINEAR;
break;
default:
return false;
}
if (sampler->MaxAnisotropy > 1.0) {
minFilt = FILTER_ANISOTROPIC;
magFilt = FILTER_ANISOTROPIC;
if (sampler->MaxAnisotropy > 2.0)
aniso = SS2_MAX_ANISO_4;
else
aniso = SS2_MAX_ANISO_2;
}
else {
switch (sampler->MagFilter) {
case GL_NEAREST:
magFilt = FILTER_NEAREST;
break;
case GL_LINEAR:
magFilt = FILTER_LINEAR;
break;
default:
return false;
}
}
lodbias = (int) ((tUnit->LodBias + sampler->LodBias) * 16.0);
if (lodbias < -256)
lodbias = -256;
if (lodbias > 255)
lodbias = 255;
state[I915_TEXREG_SS2] = ((lodbias << SS2_LOD_BIAS_SHIFT) &
SS2_LOD_BIAS_MASK);
/* YUV conversion:
*/
if (firstImage->TexFormat == MESA_FORMAT_YCBCR ||
firstImage->TexFormat == MESA_FORMAT_YCBCR_REV)
state[I915_TEXREG_SS2] |= SS2_COLORSPACE_CONVERSION;
/* Shadow:
*/
if (sampler->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB &&
tObj->Target != GL_TEXTURE_3D) {
if (tObj->Target == GL_TEXTURE_1D)
return false;
state[I915_TEXREG_SS2] |=
(SS2_SHADOW_ENABLE |
intel_translate_shadow_compare_func(sampler->CompareFunc));
minFilt = FILTER_4X4_FLAT;
magFilt = FILTER_4X4_FLAT;
}
state[I915_TEXREG_SS2] |= ((minFilt << SS2_MIN_FILTER_SHIFT) |
(mipFilt << SS2_MIP_FILTER_SHIFT) |
(magFilt << SS2_MAG_FILTER_SHIFT) |
aniso);
}
{
GLenum ws = sampler->WrapS;
GLenum wt = sampler->WrapT;
GLenum wr = sampler->WrapR;
float minlod;
/* We program 1D textures as 2D textures, so the 2D texcoord could
* result in sampling border values if we don't set the T wrap to
* repeat.
*/
if (tObj->Target == GL_TEXTURE_1D)
wt = GL_REPEAT;
/* 3D textures don't seem to respect the border color.
* Fallback if there's ever a danger that they might refer to
* it.
*
* Effectively this means fallback on 3D clamp or
* clamp_to_border.
*/
if (tObj->Target == GL_TEXTURE_3D &&
(sampler->MinFilter != GL_NEAREST ||
sampler->MagFilter != GL_NEAREST) &&
(ws == GL_CLAMP ||
wt == GL_CLAMP ||
wr == GL_CLAMP ||
ws == GL_CLAMP_TO_BORDER ||
wt == GL_CLAMP_TO_BORDER || wr == GL_CLAMP_TO_BORDER))
return false;
/* Only support TEXCOORDMODE_CLAMP_EDGE and TEXCOORDMODE_CUBE (not
* used) when using cube map texture coordinates
*/
if (tObj->Target == GL_TEXTURE_CUBE_MAP_ARB &&
(((ws != GL_CLAMP) && (ws != GL_CLAMP_TO_EDGE)) ||
((wt != GL_CLAMP) && (wt != GL_CLAMP_TO_EDGE))))
return false;
/*
* According to 3DSTATE_MAP_STATE at page of 104 in Bspec
* Vol3d 3D Instructions:
* [DevGDG and DevAlv]: Must be a power of 2 for cube maps.
* [DevLPT, DevCST and DevBLB]: If not a power of 2, cube maps
* must have all faces enabled.
*
* But, as I tested on pineview(DevBLB derived), the rendering is
* bad(you will find the color isn't samplered right in some
* fragments). After checking, it seems that the texture layout is
* wrong: making the width and height align of 4(although this
* doesn't make much sense) will fix this issue and also broke some
* others. Well, Bspec mentioned nothing about the layout alignment
* and layout for NPOT cube map. I guess the Bspec just assume it's
* a POT cube map.
*
* Thus, I guess we need do this for other platforms as well.
*/
if (tObj->Target == GL_TEXTURE_CUBE_MAP_ARB &&
!is_power_of_two(firstImage->Height))
return false;
state[I915_TEXREG_SS3] = ss3; /* SS3_NORMALIZED_COORDS */
state[I915_TEXREG_SS3] |=
((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) |
(translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) |
(translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT));
minlod = MIN2(sampler->MinLod, tObj->_MaxLevel - tObj->BaseLevel);
state[I915_TEXREG_SS3] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT);
state[I915_TEXREG_SS3] |= (U_FIXED(CLAMP(minlod, 0.0, 11.0), 4) <<
SS3_MIN_LOD_SHIFT);
}
/* convert border color from float to ubyte */
CLAMPED_FLOAT_TO_UBYTE(border[0], sampler->BorderColor.f[0]);
CLAMPED_FLOAT_TO_UBYTE(border[1], sampler->BorderColor.f[1]);
CLAMPED_FLOAT_TO_UBYTE(border[2], sampler->BorderColor.f[2]);
CLAMPED_FLOAT_TO_UBYTE(border[3], sampler->BorderColor.f[3]);
if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
/* GL specs that border color for depth textures is taken from the
* R channel, while the hardware uses A. Spam R into all the channels
* for safety.
*/
state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[0],
border[0],
border[0],
border[0]);
} else {
state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[3],
border[0],
border[1],
border[2]);
}
I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), true);
/* memcmp was already disabled, but definitely won't work as the
* region might now change and that wouldn't be detected:
*/
I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit));
#if 0
DBG(TEXTURE, "state[I915_TEXREG_SS2] = 0x%x\n", state[I915_TEXREG_SS2]);
DBG(TEXTURE, "state[I915_TEXREG_SS3] = 0x%x\n", state[I915_TEXREG_SS3]);
DBG(TEXTURE, "state[I915_TEXREG_SS4] = 0x%x\n", state[I915_TEXREG_SS4]);
DBG(TEXTURE, "state[I915_TEXREG_MS2] = 0x%x\n", state[I915_TEXREG_MS2]);
DBG(TEXTURE, "state[I915_TEXREG_MS3] = 0x%x\n", state[I915_TEXREG_MS3]);
DBG(TEXTURE, "state[I915_TEXREG_MS4] = 0x%x\n", state[I915_TEXREG_MS4]);
#endif
return true;
}
/**
* Sets the sampler state for a single unit.
*/
static void
gen7_update_sampler_state(struct brw_context *brw, int unit, int ss_index,
struct gen7_sampler_state *sampler,
uint32_t *sdc_offset)
{
struct gl_context *ctx = &brw->ctx;
struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
struct gl_texture_object *texObj = texUnit->_Current;
struct gl_sampler_object *gl_sampler = _mesa_get_samplerobj(ctx, unit);
bool using_nearest = false;
/* These don't use samplers at all. */
if (texObj->Target == GL_TEXTURE_BUFFER)
return;
switch (gl_sampler->MinFilter) {
case GL_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
using_nearest = true;
break;
case GL_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
break;
case GL_NEAREST_MIPMAP_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
break;
case GL_LINEAR_MIPMAP_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
break;
case GL_NEAREST_MIPMAP_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
break;
case GL_LINEAR_MIPMAP_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
break;
default:
break;
}
/* Set Anisotropy: */
if (gl_sampler->MaxAnisotropy > 1.0) {
sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC;
sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
sampler->ss0.aniso_algorithm = 1;
if (gl_sampler->MaxAnisotropy > 2.0) {
sampler->ss3.max_aniso = MIN2((gl_sampler->MaxAnisotropy - 2) / 2,
BRW_ANISORATIO_16);
}
}
else {
switch (gl_sampler->MagFilter) {
case GL_NEAREST:
sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
using_nearest = true;
break;
case GL_LINEAR:
sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
break;
default:
break;
}
}
sampler->ss3.r_wrap_mode = translate_wrap_mode(gl_sampler->WrapR,
using_nearest);
sampler->ss3.s_wrap_mode = translate_wrap_mode(gl_sampler->WrapS,
using_nearest);
sampler->ss3.t_wrap_mode = translate_wrap_mode(gl_sampler->WrapT,
using_nearest);
/* Cube-maps on 965 and later must use the same wrap mode for all 3
* coordinate dimensions. Futher, only CUBE and CLAMP are valid.
*/
if (texObj->Target == GL_TEXTURE_CUBE_MAP ||
texObj->Target == GL_TEXTURE_CUBE_MAP_ARRAY) {
if ((ctx->Texture.CubeMapSeamless || gl_sampler->CubeMapSeamless) &&
(gl_sampler->MinFilter != GL_NEAREST ||
gl_sampler->MagFilter != GL_NEAREST)) {
sampler->ss3.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss3.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
} else {
sampler->ss3.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
}
} else if (texObj->Target == GL_TEXTURE_1D) {
/* There's a bug in 1D texture sampling - it actually pays
* attention to the wrap_t value, though it should not.
* Override the wrap_t value here to GL_REPEAT to keep
* any nonexistent border pixels from floating in.
*/
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_WRAP;
}
/* Set shadow function: */
if (gl_sampler->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB) {
/* Shadowing is "enabled" by emitting a particular sampler
* message (sample_c). So need to recompile WM program when
* shadow comparison is enabled on each/any texture unit.
*/
sampler->ss1.shadow_function =
intel_translate_shadow_compare_func(gl_sampler->CompareFunc);
}
/* Set LOD bias: */
sampler->ss0.lod_bias = S_FIXED(CLAMP(texUnit->LodBias +
gl_sampler->LodBias, -16, 15), 8);
sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
sampler->ss0.base_level = U_FIXED(0, 1);
sampler->ss1.max_lod = U_FIXED(CLAMP(gl_sampler->MaxLod, 0, 13), 8);
sampler->ss1.min_lod = U_FIXED(CLAMP(gl_sampler->MinLod, 0, 13), 8);
/* The sampler can handle non-normalized texture rectangle coordinates
* natively
*/
if (texObj->Target == GL_TEXTURE_RECTANGLE) {
sampler->ss3.non_normalized_coord = 1;
}
upload_default_color(brw, gl_sampler, unit, sdc_offset);
sampler->ss2.default_color_pointer = *sdc_offset >> 5;
if (sampler->ss0.min_filter != BRW_MAPFILTER_NEAREST)
sampler->ss3.address_round |= BRW_ADDRESS_ROUNDING_ENABLE_U_MIN |
BRW_ADDRESS_ROUNDING_ENABLE_V_MIN |
BRW_ADDRESS_ROUNDING_ENABLE_R_MIN;
if (sampler->ss0.mag_filter != BRW_MAPFILTER_NEAREST)
sampler->ss3.address_round |= BRW_ADDRESS_ROUNDING_ENABLE_U_MAG |
BRW_ADDRESS_ROUNDING_ENABLE_V_MAG |
BRW_ADDRESS_ROUNDING_ENABLE_R_MAG;
}
static void
convert_sampler(struct st_context *st,
struct pipe_sampler_state *sampler,
GLuint texUnit)
{
struct gl_texture_object *texobj;
struct gl_context *ctx = st->ctx;
struct gl_sampler_object *msamp;
texobj = ctx->Texture.Unit[texUnit]._Current;
if (!texobj) {
texobj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
}
msamp = _mesa_get_samplerobj(ctx, texUnit);
memset(sampler, 0, sizeof(*sampler));
sampler->wrap_s = gl_wrap_xlate(msamp->WrapS);
sampler->wrap_t = gl_wrap_xlate(msamp->WrapT);
sampler->wrap_r = gl_wrap_xlate(msamp->WrapR);
sampler->min_img_filter = gl_filter_to_img_filter(msamp->MinFilter);
sampler->min_mip_filter = gl_filter_to_mip_filter(msamp->MinFilter);
sampler->mag_img_filter = gl_filter_to_img_filter(msamp->MagFilter);
if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB)
sampler->normalized_coords = 1;
sampler->lod_bias = ctx->Texture.Unit[texUnit].LodBias + msamp->LodBias;
sampler->min_lod = CLAMP(msamp->MinLod,
0.0f,
(GLfloat) texobj->MaxLevel - texobj->BaseLevel);
sampler->max_lod = MIN2((GLfloat) texobj->MaxLevel - texobj->BaseLevel,
msamp->MaxLod);
if (sampler->max_lod < sampler->min_lod) {
/* The GL spec doesn't seem to specify what to do in this case.
* Swap the values.
*/
float tmp = sampler->max_lod;
sampler->max_lod = sampler->min_lod;
sampler->min_lod = tmp;
assert(sampler->min_lod <= sampler->max_lod);
}
if (msamp->BorderColor.ui[0] ||
msamp->BorderColor.ui[1] ||
msamp->BorderColor.ui[2] ||
msamp->BorderColor.ui[3]) {
struct st_texture_object *stobj = st_texture_object(texobj);
struct gl_texture_image *teximg;
GLboolean is_integer = GL_FALSE;
union pipe_color_union border_color;
teximg = texobj->Image[0][texobj->BaseLevel];
if (teximg) {
is_integer = _mesa_is_enum_format_integer(teximg->InternalFormat);
}
if (st->apply_texture_swizzle_to_border_color && stobj->sampler_view) {
const unsigned char swz[4] =
{
stobj->sampler_view->swizzle_r,
stobj->sampler_view->swizzle_g,
stobj->sampler_view->swizzle_b,
stobj->sampler_view->swizzle_a,
};
st_translate_color(&msamp->BorderColor,
&border_color,
teximg ? teximg->_BaseFormat : GL_RGBA, is_integer);
util_format_apply_color_swizzle(&sampler->border_color,
&border_color, swz, is_integer);
} else {
st_translate_color(&msamp->BorderColor,
&sampler->border_color,
teximg ? teximg->_BaseFormat : GL_RGBA, is_integer);
}
}
sampler->max_anisotropy = (msamp->MaxAnisotropy == 1.0 ?
0 : (GLuint) msamp->MaxAnisotropy);
/* only care about ARB_shadow, not SGI shadow */
if (msamp->CompareMode == GL_COMPARE_R_TO_TEXTURE) {
sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE;
sampler->compare_func
= st_compare_func_to_pipe(msamp->CompareFunc);
}
sampler->seamless_cube_map =
ctx->Texture.CubeMapSeamless || msamp->CubeMapSeamless;
}
/**
* Sets the sampler state for a single unit.
*/
static void
gen7_update_sampler_state(struct brw_context *brw, int unit,
struct gen7_sampler_state *sampler)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
struct gl_texture_object *texObj = texUnit->_Current;
struct gl_sampler_object *gl_sampler = _mesa_get_samplerobj(ctx, unit);
bool using_nearest = false;
switch (gl_sampler->MinFilter) {
case GL_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
using_nearest = true;
break;
case GL_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
break;
case GL_NEAREST_MIPMAP_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
break;
case GL_LINEAR_MIPMAP_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
break;
case GL_NEAREST_MIPMAP_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
break;
case GL_LINEAR_MIPMAP_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
break;
default:
break;
}
/* Set Anisotropy: */
if (gl_sampler->MaxAnisotropy > 1.0) {
sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC;
sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
if (gl_sampler->MaxAnisotropy > 2.0) {
sampler->ss3.max_aniso = MIN2((gl_sampler->MaxAnisotropy - 2) / 2,
BRW_ANISORATIO_16);
}
}
else {
switch (gl_sampler->MagFilter) {
case GL_NEAREST:
sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
using_nearest = true;
break;
case GL_LINEAR:
sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
break;
default:
break;
}
}
sampler->ss3.r_wrap_mode = translate_wrap_mode(gl_sampler->WrapR,
using_nearest);
sampler->ss3.s_wrap_mode = translate_wrap_mode(gl_sampler->WrapS,
using_nearest);
sampler->ss3.t_wrap_mode = translate_wrap_mode(gl_sampler->WrapT,
using_nearest);
/* Cube-maps on 965 and later must use the same wrap mode for all 3
* coordinate dimensions. Futher, only CUBE and CLAMP are valid.
*/
if (texObj->Target == GL_TEXTURE_CUBE_MAP) {
if (ctx->Texture.CubeMapSeamless &&
(gl_sampler->MinFilter != GL_NEAREST ||
gl_sampler->MagFilter != GL_NEAREST)) {
sampler->ss3.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss3.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
} else {
sampler->ss3.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
}
} else if (texObj->Target == GL_TEXTURE_1D) {
/* There's a bug in 1D texture sampling - it actually pays
* attention to the wrap_t value, though it should not.
* Override the wrap_t value here to GL_REPEAT to keep
* any nonexistent border pixels from floating in.
*/
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_WRAP;
}
/* Set shadow function: */
if (gl_sampler->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB) {
/* Shadowing is "enabled" by emitting a particular sampler
* message (sample_c). So need to recompile WM program when
* shadow comparison is enabled on each/any texture unit.
*/
sampler->ss1.shadow_function =
intel_translate_shadow_compare_func(gl_sampler->CompareFunc);
}
/* Set LOD bias: */
sampler->ss0.lod_bias = S_FIXED(CLAMP(texUnit->LodBias +
gl_sampler->LodBias, -16, 15), 8);
sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
/* Set BaseMipLevel, MaxLOD, MinLOD:
*
* XXX: I don't think that using firstLevel, lastLevel works,
* because we always setup the surface state as if firstLevel ==
* level zero. Probably have to subtract firstLevel from each of
* these:
*/
sampler->ss0.base_level = U_FIXED(0, 1);
sampler->ss1.max_lod = U_FIXED(CLAMP(gl_sampler->MaxLod, 0, 13), 8);
sampler->ss1.min_lod = U_FIXED(CLAMP(gl_sampler->MinLod, 0, 13), 8);
upload_default_color(brw, gl_sampler, unit);
sampler->ss2.default_color_pointer = brw->wm.sdc_offset[unit] >> 5;
}
GLuint
intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
{
struct gl_context *ctx = &intel->ctx;
struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint face, i;
GLuint nr_faces = 0;
struct intel_texture_image *firstImage;
int width, height, depth;
/* TBOs require no validation -- they always just point to their BO. */
if (tObj->Target == GL_TEXTURE_BUFFER)
return true;
/* We know/require this is true by now:
*/
assert(intelObj->base._BaseComplete);
/* What levels must the tree include at a minimum?
*/
intel_update_max_level(intelObj, sampler);
firstImage = intel_texture_image(tObj->Image[0][tObj->BaseLevel]);
/* Check tree can hold all active levels. Check tree matches
* target, imageFormat, etc.
*
* For pre-gen4, we have to match first_level == tObj->BaseLevel,
* because we don't have the control that gen4 does to make min/mag
* determination happen at a nonzero (hardware) baselevel. Because
* of that, we just always relayout on baselevel change.
*/
if (intelObj->mt &&
(!intel_miptree_match_image(intelObj->mt, &firstImage->base.Base) ||
intelObj->mt->first_level != tObj->BaseLevel ||
intelObj->mt->last_level < intelObj->_MaxLevel)) {
intel_miptree_release(&intelObj->mt);
}
/* May need to create a new tree:
*/
if (!intelObj->mt) {
intel_miptree_get_dimensions_for_image(&firstImage->base.Base,
&width, &height, &depth);
intelObj->mt = intel_miptree_create(intel,
intelObj->base.Target,
firstImage->base.Base.TexFormat,
tObj->BaseLevel,
intelObj->_MaxLevel,
width,
height,
depth,
true,
0 /* num_samples */,
false /* msaa_is_interleaved */);
if (!intelObj->mt)
return false;
}
/* Pull in any images not in the object's tree:
*/
nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
for (face = 0; face < nr_faces; face++) {
for (i = tObj->BaseLevel; i <= intelObj->_MaxLevel; i++) {
struct intel_texture_image *intelImage =
intel_texture_image(intelObj->base.Image[face][i]);
/* skip too small size mipmap */
if (intelImage == NULL)
break;
if (intelObj->mt != intelImage->mt) {
intel_miptree_copy_teximage(intel, intelImage, intelObj->mt);
}
}
}
return true;
}
/* Recalculate all state from scratch. Perhaps not the most
* efficient, but this has gotten complex enough that we need
* something which is understandable and reliable.
*/
static bool
i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
{
struct gl_context *ctx = &intel->ctx;
struct i830_context *i830 = i830_context(ctx);
struct gl_texture_unit *tUnit = &ctx->Texture.Unit[unit];
struct gl_texture_object *tObj = tUnit->_Current;
struct intel_texture_object *intelObj = intel_texture_object(tObj);
struct gl_texture_image *firstImage;
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
GLuint *state = i830->state.Tex[unit], format, pitch;
GLint lodbias;
GLubyte border[4];
GLuint dst_x, dst_y;
memset(state, 0, sizeof(*state));
/*We need to refcount these. */
if (i830->state.tex_buffer[unit] != NULL) {
drm_intel_bo_unreference(i830->state.tex_buffer[unit]);
i830->state.tex_buffer[unit] = NULL;
}
if (!intel_finalize_mipmap_tree(intel, unit))
return false;
/* Get first image here, since intelObj->firstLevel will get set in
* the intel_finalize_mipmap_tree() call above.
*/
firstImage = tObj->Image[0][tObj->BaseLevel];
intel_miptree_get_image_offset(intelObj->mt, tObj->BaseLevel, 0,
&dst_x, &dst_y);
drm_intel_bo_reference(intelObj->mt->region->bo);
i830->state.tex_buffer[unit] = intelObj->mt->region->bo;
pitch = intelObj->mt->region->pitch;
/* XXX: This calculation is probably broken for tiled images with
* a non-page-aligned offset.
*/
i830->state.tex_offset[unit] = dst_x * intelObj->mt->cpp + dst_y * pitch;
format = translate_texture_format(firstImage->TexFormat);
state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 |
(LOAD_TEXTURE_MAP0 << unit) | 4);
state[I830_TEXREG_TM0S1] =
(((firstImage->Height - 1) << TM0S1_HEIGHT_SHIFT) |
((firstImage->Width - 1) << TM0S1_WIDTH_SHIFT) | format);
if (intelObj->mt->region->tiling != I915_TILING_NONE) {
state[I830_TEXREG_TM0S1] |= TM0S1_TILED_SURFACE;
if (intelObj->mt->region->tiling == I915_TILING_Y)
state[I830_TEXREG_TM0S1] |= TM0S1_TILE_WALK;
}
state[I830_TEXREG_TM0S2] =
((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK);
{
if (tObj->Target == GL_TEXTURE_CUBE_MAP)
state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit) |
CUBE_NEGX_ENABLE |
CUBE_POSX_ENABLE |
CUBE_NEGY_ENABLE |
CUBE_POSY_ENABLE |
CUBE_NEGZ_ENABLE | CUBE_POSZ_ENABLE);
else
state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit));
}
{
GLuint minFilt, mipFilt, magFilt;
float maxlod;
uint32_t minlod_fixed, maxlod_fixed;
switch (sampler->MinFilter) {
case GL_NEAREST:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_NONE;
break;
case GL_LINEAR:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_NONE;
break;
case GL_NEAREST_MIPMAP_NEAREST:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_NEAREST;
break;
case GL_LINEAR_MIPMAP_NEAREST:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_NEAREST;
break;
case GL_NEAREST_MIPMAP_LINEAR:
minFilt = FILTER_NEAREST;
mipFilt = MIPFILTER_LINEAR;
break;
case GL_LINEAR_MIPMAP_LINEAR:
minFilt = FILTER_LINEAR;
mipFilt = MIPFILTER_LINEAR;
break;
default:
return false;
}
if (sampler->MaxAnisotropy > 1.0) {
minFilt = FILTER_ANISOTROPIC;
magFilt = FILTER_ANISOTROPIC;
}
else {
switch (sampler->MagFilter) {
case GL_NEAREST:
magFilt = FILTER_NEAREST;
break;
case GL_LINEAR:
magFilt = FILTER_LINEAR;
break;
default:
return false;
}
}
lodbias = (int) ((tUnit->LodBias + sampler->LodBias) * 16.0);
if (lodbias < -64)
lodbias = -64;
if (lodbias > 63)
lodbias = 63;
state[I830_TEXREG_TM0S3] = ((lodbias << TM0S3_LOD_BIAS_SHIFT) &
TM0S3_LOD_BIAS_MASK);
#if 0
/* YUV conversion:
*/
if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR ||
firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV)
state[I830_TEXREG_TM0S3] |= SS2_COLORSPACE_CONVERSION;
#endif
/* We get one field with fraction bits for the maximum
* addressable (smallest resolution) LOD. Use it to cover both
* MAX_LEVEL and MAX_LOD.
*/
minlod_fixed = U_FIXED(CLAMP(sampler->MinLod, 0.0, 11), 4);
maxlod = MIN2(sampler->MaxLod, tObj->_MaxLevel - tObj->BaseLevel);
if (intel->intelScreen->deviceID == PCI_CHIP_I855_GM ||
intel->intelScreen->deviceID == PCI_CHIP_I865_G) {
maxlod_fixed = U_FIXED(CLAMP(maxlod, 0.0, 11.75), 2);
maxlod_fixed = MAX2(maxlod_fixed, (minlod_fixed + 3) >> 2);
state[I830_TEXREG_TM0S3] |= maxlod_fixed << TM0S3_MIN_MIP_SHIFT;
state[I830_TEXREG_TM0S2] |= TM0S2_LOD_PRECLAMP;
} else {
void
nv20_emit_tex_obj(struct gl_context *ctx, int emit)
{
const int i = emit - NOUVEAU_STATE_TEX_OBJ0;
struct nouveau_pushbuf *push = context_push(ctx);
const int bo_flags = NOUVEAU_BO_RD | NOUVEAU_BO_GART | NOUVEAU_BO_VRAM;
struct gl_texture_object *t;
struct nouveau_surface *s;
struct gl_texture_image *ti;
const struct gl_sampler_object *sa;
uint32_t tx_format, tx_filter, tx_wrap, tx_enable;
PUSH_RESET(push, BUFCTX_TEX(i));
if (!ctx->Texture.Unit[i]._ReallyEnabled) {
BEGIN_NV04(push, NV20_3D(TEX_ENABLE(i)), 1);
PUSH_DATA (push, 0);
context_dirty(ctx, TEX_SHADER);
return;
}
t = ctx->Texture.Unit[i]._Current;
s = &to_nouveau_texture(t)->surfaces[t->BaseLevel];
ti = t->Image[0][t->BaseLevel];
sa = _mesa_get_samplerobj(ctx, i);
if (!nouveau_texture_validate(ctx, t))
return;
/* Recompute the texturing registers. */
tx_format = ti->DepthLog2 << 28
| ti->HeightLog2 << 24
| ti->WidthLog2 << 20
| NV20_3D_TEX_FORMAT_DIMS_2D
| NV20_3D_TEX_FORMAT_NO_BORDER
| 1 << 16;
tx_wrap = nvgl_wrap_mode(sa->WrapR) << 16
| nvgl_wrap_mode(sa->WrapT) << 8
| nvgl_wrap_mode(sa->WrapS) << 0;
tx_filter = nvgl_filter_mode(sa->MagFilter) << 24
| nvgl_filter_mode(sa->MinFilter) << 16
| 2 << 12;
tx_enable = NV20_3D_TEX_ENABLE_ENABLE
| log2i(sa->MaxAnisotropy) << 4;
if (t->Target == GL_TEXTURE_RECTANGLE) {
BEGIN_NV04(push, NV20_3D(TEX_NPOT_PITCH(i)), 1);
PUSH_DATA (push, s->pitch << 16);
BEGIN_NV04(push, NV20_3D(TEX_NPOT_SIZE(i)), 1);
PUSH_DATA (push, s->width << 16 | s->height);
tx_format |= get_tex_format_rect(ti);
} else {
tx_format |= get_tex_format_pot(ti);
}
if (sa->MinFilter != GL_NEAREST &&
sa->MinFilter != GL_LINEAR) {
int lod_min = sa->MinLod;
int lod_max = MIN2(sa->MaxLod, t->_MaxLambda);
int lod_bias = sa->LodBias
+ ctx->Texture.Unit[i].LodBias;
lod_max = CLAMP(lod_max, 0, 15);
lod_min = CLAMP(lod_min, 0, 15);
lod_bias = CLAMP(lod_bias, 0, 15);
tx_format |= NV20_3D_TEX_FORMAT_MIPMAP;
tx_filter |= lod_bias << 8;
tx_enable |= lod_min << 26
| lod_max << 14;
}
/* Write it to the hardware. */
BEGIN_NV04(push, NV20_3D(TEX_FORMAT(i)), 1);
PUSH_MTHD (push, NV20_3D(TEX_FORMAT(i)), BUFCTX_TEX(i),
s->bo, tx_format, bo_flags | NOUVEAU_BO_OR,
NV20_3D_TEX_FORMAT_DMA0,
NV20_3D_TEX_FORMAT_DMA1);
BEGIN_NV04(push, NV20_3D(TEX_OFFSET(i)), 1);
PUSH_MTHDl(push, NV20_3D(TEX_OFFSET(i)), BUFCTX_TEX(i),
s->bo, s->offset, bo_flags);
BEGIN_NV04(push, NV20_3D(TEX_WRAP(i)), 1);
PUSH_DATA (push, tx_wrap);
BEGIN_NV04(push, NV20_3D(TEX_FILTER(i)), 1);
PUSH_DATA (push, tx_filter);
BEGIN_NV04(push, NV20_3D(TEX_ENABLE(i)), 1);
PUSH_DATA (push, tx_enable);
context_dirty(ctx, TEX_SHADER);
}
static void
convert_sampler(struct st_context *st,
struct pipe_sampler_state *sampler,
GLuint texUnit)
{
const struct gl_texture_object *texobj;
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *msamp;
GLenum texBaseFormat;
texobj = ctx->Texture.Unit[texUnit]._Current;
if (!texobj) {
texobj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
msamp = &texobj->Sampler;
} else {
msamp = _mesa_get_samplerobj(ctx, texUnit);
}
texBaseFormat = _mesa_texture_base_format(texobj);
memset(sampler, 0, sizeof(*sampler));
sampler->wrap_s = gl_wrap_xlate(msamp->WrapS);
sampler->wrap_t = gl_wrap_xlate(msamp->WrapT);
sampler->wrap_r = gl_wrap_xlate(msamp->WrapR);
sampler->min_img_filter = gl_filter_to_img_filter(msamp->MinFilter);
sampler->min_mip_filter = gl_filter_to_mip_filter(msamp->MinFilter);
sampler->mag_img_filter = gl_filter_to_img_filter(msamp->MagFilter);
if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB)
sampler->normalized_coords = 1;
sampler->lod_bias = ctx->Texture.Unit[texUnit].LodBias + msamp->LodBias;
sampler->min_lod = MAX2(msamp->MinLod, 0.0f);
sampler->max_lod = msamp->MaxLod;
if (sampler->max_lod < sampler->min_lod) {
/* The GL spec doesn't seem to specify what to do in this case.
* Swap the values.
*/
float tmp = sampler->max_lod;
sampler->max_lod = sampler->min_lod;
sampler->min_lod = tmp;
assert(sampler->min_lod <= sampler->max_lod);
}
/* For non-black borders... */
if (msamp->BorderColor.ui[0] ||
msamp->BorderColor.ui[1] ||
msamp->BorderColor.ui[2] ||
msamp->BorderColor.ui[3]) {
const struct st_texture_object *stobj = st_texture_object_const(texobj);
const GLboolean is_integer = texobj->_IsIntegerFormat;
const struct pipe_sampler_view *sv = NULL;
union pipe_color_union border_color;
GLuint i;
/* Just search for the first used view. We can do this because the
swizzle is per-texture, not per context. */
/* XXX: clean that up to not use the sampler view at all */
for (i = 0; i < stobj->num_sampler_views; ++i) {
if (stobj->sampler_views[i]) {
sv = stobj->sampler_views[i];
break;
}
}
if (st->apply_texture_swizzle_to_border_color && sv) {
const unsigned char swz[4] =
{
sv->swizzle_r,
sv->swizzle_g,
sv->swizzle_b,
sv->swizzle_a,
};
st_translate_color(&msamp->BorderColor,
&border_color,
texBaseFormat, is_integer);
util_format_apply_color_swizzle(&sampler->border_color,
&border_color, swz, is_integer);
} else {
st_translate_color(&msamp->BorderColor,
&sampler->border_color,
texBaseFormat, is_integer);
}
}
sampler->max_anisotropy = (msamp->MaxAnisotropy == 1.0 ?
0 : (GLuint) msamp->MaxAnisotropy);
/* If sampling a depth texture and using shadow comparison */
if ((texBaseFormat == GL_DEPTH_COMPONENT ||
texBaseFormat == GL_DEPTH_STENCIL) &&
msamp->CompareMode == GL_COMPARE_R_TO_TEXTURE) {
sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE;
sampler->compare_func = st_compare_func_to_pipe(msamp->CompareFunc);
}
sampler->seamless_cube_map =
ctx->Texture.CubeMapSeamless || msamp->CubeMapSeamless;
}
void
brw_populate_sampler_prog_key_data(struct gl_context *ctx,
const struct gl_program *prog,
struct brw_sampler_prog_key_data *key)
{
struct brw_context *brw = brw_context(ctx);
const struct gen_device_info *devinfo = &brw->screen->devinfo;
GLbitfield mask = prog->SamplersUsed;
while (mask) {
const int s = u_bit_scan(&mask);
key->swizzles[s] = SWIZZLE_NOOP;
key->scale_factors[s] = 0.0f;
int unit_id = prog->SamplerUnits[s];
const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id];
if (unit->_Current && unit->_Current->Target != GL_TEXTURE_BUFFER) {
const struct gl_texture_object *t = unit->_Current;
const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id);
const bool alpha_depth = t->DepthMode == GL_ALPHA &&
(img->_BaseFormat == GL_DEPTH_COMPONENT ||
img->_BaseFormat == GL_DEPTH_STENCIL);
/* Haswell handles texture swizzling as surface format overrides
* (except for GL_ALPHA); all other platforms need MOVs in the shader.
*/
if (alpha_depth || (devinfo->gen < 8 && !devinfo->is_haswell))
key->swizzles[s] = brw_get_texture_swizzle(ctx, t);
if (devinfo->gen < 8 &&
sampler->MinFilter != GL_NEAREST &&
sampler->MagFilter != GL_NEAREST) {
if (sampler->WrapS == GL_CLAMP)
key->gl_clamp_mask[0] |= 1 << s;
if (sampler->WrapT == GL_CLAMP)
key->gl_clamp_mask[1] |= 1 << s;
if (sampler->WrapR == GL_CLAMP)
key->gl_clamp_mask[2] |= 1 << s;
}
/* gather4 for RG32* is broken in multiple ways on Gen7. */
if (devinfo->gen == 7 && prog->info.uses_texture_gather) {
switch (img->InternalFormat) {
case GL_RG32I:
case GL_RG32UI: {
/* We have to override the format to R32G32_FLOAT_LD.
* This means that SCS_ALPHA and SCS_ONE will return 0x3f8
* (1.0) rather than integer 1. This needs shader hacks.
*
* On Ivybridge, we whack W (alpha) to ONE in our key's
* swizzle. On Haswell, we look at the original texture
* swizzle, and use XYZW with channels overridden to ONE,
* leaving normal texture swizzling to SCS.
*/
unsigned src_swizzle =
devinfo->is_haswell ? t->_Swizzle : key->swizzles[s];
for (int i = 0; i < 4; i++) {
unsigned src_comp = GET_SWZ(src_swizzle, i);
if (src_comp == SWIZZLE_ONE || src_comp == SWIZZLE_W) {
key->swizzles[i] &= ~(0x7 << (3 * i));
key->swizzles[i] |= SWIZZLE_ONE << (3 * i);
}
}
/* fallthrough */
}
case GL_RG32F:
/* The channel select for green doesn't work - we have to
* request blue. Haswell can use SCS for this, but Ivybridge
* needs a shader workaround.
*/
if (!devinfo->is_haswell)
key->gather_channel_quirk_mask |= 1 << s;
break;
}
}
/* Gen6's gather4 is broken for UINT/SINT; we treat them as
* UNORM/FLOAT instead and fix it in the shader.
*/
if (devinfo->gen == 6 && prog->info.uses_texture_gather) {
key->gen6_gather_wa[s] = gen6_gather_workaround(img->InternalFormat);
}
/* If this is a multisample sampler, and uses the CMS MSAA layout,
* then we need to emit slightly different code to first sample the
* MCS surface.
*/
struct intel_texture_object *intel_tex =
intel_texture_object((struct gl_texture_object *)t);
/* From gen9 onwards some single sampled buffers can also be
* compressed. These don't need ld2dms sampling along with mcs fetch.
*/
if (intel_tex->mt->aux_usage == ISL_AUX_USAGE_MCS) {
assert(devinfo->gen >= 7);
assert(intel_tex->mt->surf.samples > 1);
assert(intel_tex->mt->aux_buf);
assert(intel_tex->mt->surf.msaa_layout == ISL_MSAA_LAYOUT_ARRAY);
key->compressed_multisample_layout_mask |= 1 << s;
if (intel_tex->mt->surf.samples >= 16) {
assert(devinfo->gen >= 9);
key->msaa_16 |= 1 << s;
}
}
if (t->Target == GL_TEXTURE_EXTERNAL_OES && intel_tex->planar_format) {
/* Setup possible scaling factor. */
key->scale_factors[s] = intel_tex->planar_format->scaling_factor;
switch (intel_tex->planar_format->components) {
case __DRI_IMAGE_COMPONENTS_Y_UV:
key->y_uv_image_mask |= 1 << s;
break;
case __DRI_IMAGE_COMPONENTS_Y_U_V:
key->y_u_v_image_mask |= 1 << s;
break;
case __DRI_IMAGE_COMPONENTS_Y_XUXV:
key->yx_xuxv_image_mask |= 1 << s;
break;
case __DRI_IMAGE_COMPONENTS_Y_UXVX:
key->xy_uxvx_image_mask |= 1 << s;
break;
case __DRI_IMAGE_COMPONENTS_AYUV:
key->ayuv_image_mask |= 1 << s;
break;
case __DRI_IMAGE_COMPONENTS_XYUV:
key->xyuv_image_mask |= 1 << s;
break;
default:
break;
}
}
}
}
}
static GLboolean
update_single_texture(struct st_context *st,
struct pipe_sampler_view **sampler_view,
GLuint texUnit)
{
struct pipe_context *pipe = st->pipe;
struct gl_context *ctx = st->ctx;
const struct gl_sampler_object *samp;
struct gl_texture_object *texObj;
struct st_texture_object *stObj;
enum pipe_format st_view_format;
GLboolean retval;
samp = _mesa_get_samplerobj(ctx, texUnit);
texObj = ctx->Texture.Unit[texUnit]._Current;
if (!texObj) {
texObj = _mesa_get_fallback_texture(ctx, TEXTURE_2D_INDEX);
samp = &texObj->Sampler;
}
stObj = st_texture_object(texObj);
retval = st_finalize_texture(ctx, st->pipe, texObj);
if (!retval) {
/* out of mem */
return GL_FALSE;
}
/* Determine the format of the texture sampler view */
st_view_format = stObj->pt->format;
{
const struct st_texture_image *firstImage =
st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]);
const gl_format texFormat = firstImage->base.TexFormat;
enum pipe_format firstImageFormat =
st_mesa_format_to_pipe_format(texFormat);
if ((samp->sRGBDecode == GL_SKIP_DECODE_EXT) &&
(_mesa_get_format_color_encoding(texFormat) == GL_SRGB)) {
/* Don't do sRGB->RGB conversion. Interpret the texture data as
* linear values.
*/
const gl_format linearFormat =
_mesa_get_srgb_format_linear(texFormat);
firstImageFormat = st_mesa_format_to_pipe_format(linearFormat);
}
if (firstImageFormat != stObj->pt->format)
st_view_format = firstImageFormat;
}
/* if sampler view has changed dereference it */
if (stObj->sampler_view) {
if (check_sampler_swizzle(stObj->sampler_view,
stObj->base._Swizzle,
samp->DepthMode) ||
(st_view_format != stObj->sampler_view->format) ||
stObj->base.BaseLevel != stObj->sampler_view->u.tex.first_level) {
pipe_sampler_view_reference(&stObj->sampler_view, NULL);
}
}
*sampler_view = st_get_texture_sampler_view_from_stobj(stObj, pipe,
samp,
st_view_format);
return GL_TRUE;
}