static void
upload_clip_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t dw2 = 0;
/* _NEW_BUFFERS */
struct gl_framebuffer *fb = ctx->DrawBuffer;
/* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->barycentric_interp_modes &
BRW_WM_NONPERSPECTIVE_BARYCENTRIC_BITS) {
dw2 |= GEN6_CLIP_NON_PERSPECTIVE_BARYCENTRIC_ENABLE;
}
if (!ctx->Transform.DepthClamp)
dw2 |= GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
dw2 |=
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
dw2 |=
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
dw2 |= (ctx->Transform.ClipPlanesEnabled <<
GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT);
if (ctx->Viewport.X == 0 &&
ctx->Viewport.Y == 0 &&
ctx->Viewport.Width == fb->Width &&
ctx->Viewport.Height == fb->Height) {
dw2 |= GEN6_CLIP_GB_TEST;
}
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(GEN6_CLIP_STATISTICS_ENABLE);
OUT_BATCH(GEN6_CLIP_ENABLE |
GEN6_CLIP_API_OGL |
GEN6_CLIP_MODE_NORMAL |
GEN6_CLIP_XY_TEST |
dw2);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
GEN6_CLIP_FORCE_ZERO_RTAINDEX);
ADVANCE_BATCH();
}
/**
* SAMPLER_STATE. See gen7_update_sampler_state().
*/
static uint32_t
gen7_blorp_emit_sampler_state(struct brw_context *brw,
const brw_blorp_params *params)
{
uint32_t sampler_offset;
struct gen7_sampler_state *sampler = (struct gen7_sampler_state *)
brw_state_batch(brw, AUB_TRACE_SAMPLER_STATE,
sizeof(struct gen7_sampler_state),
32, &sampler_offset);
memset(sampler, 0, sizeof(*sampler));
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
sampler->ss3.r_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.s_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
sampler->ss3.t_wrap_mode = BRW_TEXCOORDMODE_CLAMP;
// sampler->ss0.min_mag_neq = 1;
/* Set LOD bias:
*/
sampler->ss0.lod_bias = 0;
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(0, 8);
sampler->ss1.min_lod = U_FIXED(0, 8);
sampler->ss3.non_normalized_coord = 1;
sampler->ss3.address_round |= BRW_ADDRESS_ROUNDING_ENABLE_U_MIN |
BRW_ADDRESS_ROUNDING_ENABLE_V_MIN |
BRW_ADDRESS_ROUNDING_ENABLE_R_MIN;
sampler->ss3.address_round |= BRW_ADDRESS_ROUNDING_ENABLE_U_MAG |
BRW_ADDRESS_ROUNDING_ENABLE_V_MAG |
BRW_ADDRESS_ROUNDING_ENABLE_R_MAG;
return sampler_offset;
}
static void
upload_clip_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t depth_clamp = 0;
uint32_t provoking, userclip;
uint32_t nonperspective_barycentric_enable_flag = 0;
/* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->barycentric_interp_modes &
BRW_WM_NONPERSPECTIVE_BARYCENTRIC_BITS) {
nonperspective_barycentric_enable_flag =
GEN6_CLIP_NON_PERSPECTIVE_BARYCENTRIC_ENABLE;
}
if (!ctx->Transform.DepthClamp)
depth_clamp = GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
provoking =
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
provoking =
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
userclip = ctx->Transform.ClipPlanesEnabled;
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(GEN6_CLIP_STATISTICS_ENABLE);
OUT_BATCH(GEN6_CLIP_ENABLE |
GEN6_CLIP_API_OGL |
GEN6_CLIP_MODE_NORMAL |
nonperspective_barycentric_enable_flag |
GEN6_CLIP_XY_TEST |
GEN6_CLIP_GB_TEST |
userclip << GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT |
depth_clamp |
provoking);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
GEN6_CLIP_FORCE_ZERO_RTAINDEX);
ADVANCE_BATCH();
}
static void
upload_sf(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
uint32_t dw1 = 0, dw2 = 0, dw3 = 0;
float point_size;
dw1 = GEN6_SF_STATISTICS_ENABLE;
if (brw->sf.viewport_transform_enable)
dw1 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
/* _NEW_LINE */
uint32_t line_width_u3_7 = brw_get_line_width(brw);
if (brw->gen >= 9 || brw->is_cherryview) {
dw1 |= line_width_u3_7 << GEN9_SF_LINE_WIDTH_SHIFT;
} else {
dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
}
if (ctx->Line.SmoothFlag) {
dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
/* _NEW_POINT - Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw3 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3);
/* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */
if (use_state_point_size(brw))
dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* _NEW_POINT | _NEW_MULTISAMPLE */
if ((ctx->Point.SmoothFlag || _mesa_is_multisample_enabled(ctx)) &&
!ctx->Point.PointSprite) {
dw3 |= GEN8_SF_SMOOTH_POINT_ENABLE;
}
dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw3 |= (2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw3 |= (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_SF << 16 | (4 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
ADVANCE_BATCH();
}
static void
upload_clip_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t depth_clamp = 0;
uint32_t provoking, userclip;
if (!ctx->Transform.DepthClamp)
depth_clamp = GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
provoking =
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
provoking =
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
userclip = (1 << brw_count_bits(ctx->Transform.ClipPlanesEnabled)) - 1;
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(GEN6_CLIP_STATISTICS_ENABLE);
OUT_BATCH(GEN6_CLIP_ENABLE |
GEN6_CLIP_API_OGL |
GEN6_CLIP_MODE_NORMAL |
GEN6_CLIP_XY_TEST |
userclip << GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT |
depth_clamp |
provoking);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
GEN6_CLIP_FORCE_ZERO_RTAINDEX);
ADVANCE_BATCH();
}
static void
upload_clip_state(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
/* BRW_NEW_META_IN_PROGRESS */
uint32_t dw1 = brw->meta_in_progress ? 0 : GEN6_CLIP_STATISTICS_ENABLE;
uint32_t dw2 = 0;
/* _NEW_BUFFERS */
struct gl_framebuffer *fb = ctx->DrawBuffer;
/* BRW_NEW_FS_PROG_DATA */
if (brw->wm.prog_data->barycentric_interp_modes &
BRW_WM_NONPERSPECTIVE_BARYCENTRIC_BITS) {
dw2 |= GEN6_CLIP_NON_PERSPECTIVE_BARYCENTRIC_ENABLE;
}
dw1 |= brw->vs.prog_data->base.cull_distance_mask;
if (brw->gen >= 7)
dw1 |= GEN7_CLIP_EARLY_CULL;
if (brw->gen == 7) {
/* _NEW_POLYGON */
if (ctx->Polygon._FrontBit == _mesa_is_user_fbo(fb))
dw1 |= GEN7_CLIP_WINDING_CCW;
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw1 |= GEN7_CLIP_CULLMODE_FRONT;
break;
case GL_BACK:
dw1 |= GEN7_CLIP_CULLMODE_BACK;
break;
case GL_FRONT_AND_BACK:
dw1 |= GEN7_CLIP_CULLMODE_BOTH;
break;
default:
unreachable("Should not get here: invalid CullFlag");
}
} else {
dw1 |= GEN7_CLIP_CULLMODE_NONE;
}
}
if (brw->gen < 8 && !ctx->Transform.DepthClamp)
dw2 |= GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
dw2 |=
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
dw2 |=
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
dw2 |= (ctx->Transform.ClipPlanesEnabled <<
GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT);
if (ctx->Transform.ClipDepthMode == GL_ZERO_TO_ONE)
dw2 |= GEN6_CLIP_API_D3D;
else
dw2 |= GEN6_CLIP_API_OGL;
dw2 |= GEN6_CLIP_GB_TEST;
/* We need to disable guardband clipping if the guardband (which we always
* program to the maximum screen-space bounding box of 8K x 8K) will be
* smaller than the viewport.
*
* Closely examining the clip determination formulas in the documentation
* reveals that objects will be discarded entirely if they're outside the
* (small) guardband, even if they're within the (large) viewport:
*
* TR = TR_GB || TR_VPXY || TR_VPZ || TR_UC || TR_NEGW
* TA = !TR && TA_GB && TA_VPZ && TA_NEGW
* MC = !(TA || TR)
*
* (TA is "Trivial Accept", TR is "Trivial Reject", MC is "Must Clip".)
*
* Disabling guardband clipping removes the TR_GB condition, which means
* they'll be considered MC ("Must Clip") unless they're rejected for
* some other reason.
*
* Note that there is no TA_VPXY condition. If there were, objects entirely
* inside a 16384x16384 viewport would be trivially accepted, breaking the
* "objects must have a screenspace bounding box not exceeding 8K in the X
* or Y direction" restriction. Instead, they're clipped.
*/
for (unsigned i = 0; i < ctx->Const.MaxViewports; i++) {
if (ctx->ViewportArray[i].Width > 8192 ||
ctx->ViewportArray[i].Height > 8192) {
dw2 &= ~GEN6_CLIP_GB_TEST;
break;
}
}
/* If the viewport dimensions are smaller than the drawable dimensions,
* we have to disable guardband clipping prior to Gen8. We always program
* the guardband to a fixed size, which is almost always larger than the
* viewport. Any geometry which intersects the viewport but lies within
* the guardband would bypass the 3D clipping stage, so it wouldn't be
* clipped to the viewport. Rendering would happen beyond the viewport,
* but still inside the drawable.
*
* Gen8+ introduces a viewport extents test which restricts rendering to
* the viewport, so we can ignore this restriction.
*/
if (brw->gen < 8) {
const float fb_width = (float)_mesa_geometric_width(fb);
const float fb_height = (float)_mesa_geometric_height(fb);
for (unsigned i = 0; i < ctx->Const.MaxViewports; i++) {
if (ctx->ViewportArray[i].X != 0 ||
ctx->ViewportArray[i].Y != 0 ||
ctx->ViewportArray[i].Width != fb_width ||
ctx->ViewportArray[i].Height != fb_height) {
dw2 &= ~GEN6_CLIP_GB_TEST;
break;
}
}
}
/* BRW_NEW_RASTERIZER_DISCARD */
if (ctx->RasterDiscard) {
dw2 |= GEN6_CLIP_MODE_REJECT_ALL;
if (brw->gen == 6) {
perf_debug("Rasterizer discard is currently implemented via the "
"clipper; having the GS not write primitives would "
"likely be faster.\n");
}
}
uint32_t enable;
if (brw->primitive == _3DPRIM_RECTLIST)
enable = 0;
else
enable = GEN6_CLIP_ENABLE;
if (!is_drawing_points(brw) && !is_drawing_lines(brw))
dw2 |= GEN6_CLIP_XY_TEST;
/* BRW_NEW_VUE_MAP_GEOM_OUT */
const int max_vp_index =
(brw->vue_map_geom_out.slots_valid & VARYING_BIT_VIEWPORT) != 0 ?
ctx->Const.MaxViewports : 1;
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(dw1);
OUT_BATCH(enable |
GEN6_CLIP_MODE_NORMAL |
dw2);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
(_mesa_geometric_layers(fb) > 0 ? 0 : GEN6_CLIP_FORCE_ZERO_RTAINDEX) |
((max_vp_index - 1) & GEN6_CLIP_MAX_VP_INDEX_MASK));
ADVANCE_BATCH();
}
static void
upload_sf_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t urb_entry_read_length;
/* BRW_NEW_FRAGMENT_PROGRAM */
uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
/* _NEW_LIGHT */
bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
uint32_t dw1, dw2, dw3, dw4, dw16, dw17;
int i;
/* _NEW_BUFFER */
bool render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
int attr = 0, input_index = 0;
int urb_entry_read_offset = 1;
float point_size;
uint16_t attr_overrides[FRAG_ATTRIB_MAX];
uint32_t point_sprite_origin;
/* CACHE_NEW_VS_PROG */
urb_entry_read_length = ((brw->vs.prog_data->vue_map.num_slots + 1) / 2 -
urb_entry_read_offset);
if (urb_entry_read_length == 0) {
/* Setting the URB entry read length to 0 causes undefined behavior, so
* if we have no URB data to read, set it to 1.
*/
urb_entry_read_length = 1;
}
dw1 =
GEN6_SF_SWIZZLE_ENABLE |
num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT |
urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
dw2 = GEN6_SF_STATISTICS_ENABLE |
GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
dw3 = 0;
dw4 = 0;
dw16 = 0;
dw17 = 0;
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw2 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
if (ctx->Polygon.OffsetLine)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
if (ctx->Polygon.OffsetPoint)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
switch (ctx->Polygon.FrontMode) {
case GL_FILL:
dw2 |= GEN6_SF_FRONT_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_FRONT_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_FRONT_POINT;
break;
default:
assert(0);
break;
}
switch (ctx->Polygon.BackMode) {
case GL_FILL:
dw2 |= GEN6_SF_BACK_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_BACK_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_BACK_POINT;
break;
default:
assert(0);
break;
}
/* _NEW_SCISSOR */
if (ctx->Scissor.Enabled)
dw3 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_POLYGON */
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw3 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw3 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw3 |= GEN6_SF_CULL_BOTH;
break;
default:
assert(0);
break;
}
} else {
dw3 |= GEN6_SF_CULL_NONE;
}
/* _NEW_LINE */
dw3 |= U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7) <<
GEN6_SF_LINE_WIDTH_SHIFT;
if (ctx->Line.SmoothFlag) {
dw3 |= GEN6_SF_LINE_AA_ENABLE;
dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
/* _NEW_PROGRAM | _NEW_POINT */
if (!(ctx->VertexProgram.PointSizeEnabled ||
ctx->Point._Attenuated))
dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw4 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
/*
* Window coordinates in an FBO are inverted, which means point
* sprite origin must be inverted, too.
*/
if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) {
point_sprite_origin = GEN6_SF_POINT_SPRITE_LOWERLEFT;
} else {
point_sprite_origin = GEN6_SF_POINT_SPRITE_UPPERLEFT;
}
dw1 |= point_sprite_origin;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw4 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw4 |=
(1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
/* Create the mapping from the FS inputs we produce to the VS outputs
* they source from.
*/
for (; attr < FRAG_ATTRIB_MAX; attr++) {
enum glsl_interp_qualifier interp_qualifier =
brw->fragment_program->InterpQualifier[attr];
bool is_gl_Color = attr == FRAG_ATTRIB_COL0 || attr == FRAG_ATTRIB_COL1;
if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
continue;
/* _NEW_POINT */
if (ctx->Point.PointSprite &&
(attr >= FRAG_ATTRIB_TEX0 && attr <= FRAG_ATTRIB_TEX7) &&
ctx->Point.CoordReplace[attr - FRAG_ATTRIB_TEX0]) {
dw16 |= (1 << input_index);
}
if (attr == FRAG_ATTRIB_PNTC)
dw16 |= (1 << input_index);
/* flat shading */
if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
(shade_model_flat && is_gl_Color &&
interp_qualifier == INTERP_QUALIFIER_NONE))
dw17 |= (1 << input_index);
/* The hardware can only do the overrides on 16 overrides at a
* time, and the other up to 16 have to be lined up so that the
* input index = the output index. We'll need to do some
* tweaking to make sure that's the case.
*/
assert(input_index < 16 || attr == input_index);
/* CACHE_NEW_VS_PROG | _NEW_LIGHT | _NEW_PROGRAM */
attr_overrides[input_index++] =
get_attr_override(&brw->vs.prog_data->vue_map,
urb_entry_read_offset, attr,
ctx->VertexProgram._TwoSideEnabled);
}
for (; input_index < FRAG_ATTRIB_MAX; input_index++)
attr_overrides[input_index] = 0;
BEGIN_BATCH(20);
OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH(dw4);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
for (i = 0; i < 8; i++) {
OUT_BATCH(attr_overrides[i * 2] | attr_overrides[i * 2 + 1] << 16);
}
OUT_BATCH(dw16); /* point sprite texcoord bitmask */
OUT_BATCH(dw17); /* constant interp bitmask */
OUT_BATCH(0); /* wrapshortest enables 0-7 */
OUT_BATCH(0); /* wrapshortest enables 8-15 */
ADVANCE_BATCH();
}
static void
upload_sf_state(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
uint32_t dw1, dw2, dw3;
float point_size;
/* _NEW_BUFFERS */
bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1;
dw1 = GEN6_SF_STATISTICS_ENABLE;
if (brw->sf.viewport_transform_enable)
dw1 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
/* _NEW_BUFFERS */
dw1 |= (brw_depthbuffer_format(brw) << GEN7_SF_DEPTH_BUFFER_SURFACE_FORMAT_SHIFT);
/* _NEW_POLYGON */
if (ctx->Polygon._FrontBit == render_to_fbo)
dw1 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
if (ctx->Polygon.OffsetLine)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
if (ctx->Polygon.OffsetPoint)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
switch (ctx->Polygon.FrontMode) {
case GL_FILL:
dw1 |= GEN6_SF_FRONT_SOLID;
break;
case GL_LINE:
dw1 |= GEN6_SF_FRONT_WIREFRAME;
break;
case GL_POINT:
dw1 |= GEN6_SF_FRONT_POINT;
break;
default:
unreachable("not reached");
}
switch (ctx->Polygon.BackMode) {
case GL_FILL:
dw1 |= GEN6_SF_BACK_SOLID;
break;
case GL_LINE:
dw1 |= GEN6_SF_BACK_WIREFRAME;
break;
case GL_POINT:
dw1 |= GEN6_SF_BACK_POINT;
break;
default:
unreachable("not reached");
}
dw2 = 0;
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw2 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw2 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw2 |= GEN6_SF_CULL_BOTH;
break;
default:
unreachable("not reached");
}
} else {
dw2 |= GEN6_SF_CULL_NONE;
}
/* _NEW_SCISSOR _NEW_POLYGON BRW_NEW_GEOMETRY_PROGRAM BRW_NEW_PRIMITIVE */
if (ctx->Scissor.EnableFlags ||
is_drawing_points(brw) || is_drawing_lines(brw))
dw2 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_LINE */
{
uint32_t line_width_u3_7 = brw_get_line_width(brw);
dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
}
if (ctx->Line.SmoothFlag) {
dw2 |= GEN6_SF_LINE_AA_ENABLE;
dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
if (ctx->Line.StippleFlag && brw->is_haswell) {
dw2 |= HSW_SF_LINE_STIPPLE_ENABLE;
}
/* _NEW_MULTISAMPLE */
if (multisampled_fbo && ctx->Multisample.Enabled)
dw2 |= GEN6_SF_MSRAST_ON_PATTERN;
/* FINISHME: Last Pixel Enable? Vertex Sub Pixel Precision Select?
*/
dw3 = GEN6_SF_LINE_AA_MODE_TRUE;
/* _NEW_PROGRAM | _NEW_POINT */
if (!(ctx->VertexProgram.PointSizeEnabled || ctx->Point._Attenuated))
dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw3 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3);
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw3 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw3 |= (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_SF << 16 | (7 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(ctx->Polygon.OffsetClamp); /* global depth offset clamp */
ADVANCE_BATCH();
}
static void
upload_clip_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t dw1 = brw->meta_in_progress ? 0 : GEN6_CLIP_STATISTICS_ENABLE;
uint32_t dw2 = 0;
/* _NEW_BUFFERS */
struct gl_framebuffer *fb = ctx->DrawBuffer;
/* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->barycentric_interp_modes &
BRW_WM_NONPERSPECTIVE_BARYCENTRIC_BITS) {
dw2 |= GEN6_CLIP_NON_PERSPECTIVE_BARYCENTRIC_ENABLE;
}
if (!ctx->Transform.DepthClamp)
dw2 |= GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
dw2 |=
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
dw2 |=
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
dw2 |= (ctx->Transform.ClipPlanesEnabled <<
GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT);
if (ctx->Viewport.X == 0 &&
ctx->Viewport.Y == 0 &&
ctx->Viewport.Width == fb->Width &&
ctx->Viewport.Height == fb->Height) {
dw2 |= GEN6_CLIP_GB_TEST;
}
/* BRW_NEW_RASTERIZER_DISCARD */
if (ctx->RasterDiscard) {
dw2 |= GEN6_CLIP_MODE_REJECT_ALL;
perf_debug("Rasterizer discard is currently implemented via the clipper; "
"having the GS not write primitives would likely be faster.");
}
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(dw1);
OUT_BATCH(GEN6_CLIP_ENABLE |
GEN6_CLIP_API_OGL |
GEN6_CLIP_MODE_NORMAL |
GEN6_CLIP_XY_TEST |
dw2);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
GEN6_CLIP_FORCE_ZERO_RTAINDEX);
ADVANCE_BATCH();
}
/* 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;
}
static void
upload_sf_state(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
/* CACHE_NEW_WM_PROG */
uint32_t num_outputs = brw->wm.prog_data->num_varying_inputs;
uint32_t dw1, dw2, dw3, dw4;
uint32_t point_sprite_enables;
uint32_t flat_enables;
int i;
/* _NEW_BUFFER */
bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
float point_size;
uint16_t attr_overrides[16];
uint32_t point_sprite_origin;
dw1 = GEN6_SF_SWIZZLE_ENABLE | num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT;
dw2 = GEN6_SF_STATISTICS_ENABLE |
GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
dw3 = 0;
dw4 = 0;
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw2 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
if (ctx->Polygon.OffsetLine)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
if (ctx->Polygon.OffsetPoint)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
switch (ctx->Polygon.FrontMode) {
case GL_FILL:
dw2 |= GEN6_SF_FRONT_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_FRONT_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_FRONT_POINT;
break;
default:
assert(0);
break;
}
switch (ctx->Polygon.BackMode) {
case GL_FILL:
dw2 |= GEN6_SF_BACK_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_BACK_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_BACK_POINT;
break;
default:
assert(0);
break;
}
/* _NEW_SCISSOR */
if (ctx->Scissor.Enabled)
dw3 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_POLYGON */
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw3 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw3 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw3 |= GEN6_SF_CULL_BOTH;
break;
default:
assert(0);
break;
}
} else {
dw3 |= GEN6_SF_CULL_NONE;
}
/* _NEW_LINE */
{
uint32_t line_width_u3_7 = U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7);
/* TODO: line width of 0 is not allowed when MSAA enabled */
if (line_width_u3_7 == 0)
line_width_u3_7 = 1;
dw3 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
}
if (ctx->Line.SmoothFlag) {
dw3 |= GEN6_SF_LINE_AA_ENABLE;
dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
/* _NEW_MULTISAMPLE */
if (multisampled_fbo && ctx->Multisample.Enabled)
dw3 |= GEN6_SF_MSRAST_ON_PATTERN;
/* _NEW_PROGRAM | _NEW_POINT */
if (!(ctx->VertexProgram.PointSizeEnabled ||
ctx->Point._Attenuated))
dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw4 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
/*
* Window coordinates in an FBO are inverted, which means point
* sprite origin must be inverted, too.
*/
if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) {
point_sprite_origin = GEN6_SF_POINT_SPRITE_LOWERLEFT;
} else {
point_sprite_origin = GEN6_SF_POINT_SPRITE_UPPERLEFT;
}
dw1 |= point_sprite_origin;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw4 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw4 |=
(1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
/* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM |
* CACHE_NEW_WM_PROG
*/
uint32_t urb_entry_read_length;
calculate_attr_overrides(brw, attr_overrides, &point_sprite_enables,
&flat_enables, &urb_entry_read_length);
dw1 |= (urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT);
BEGIN_BATCH(20);
OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH(dw4);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
for (i = 0; i < 8; i++) {
OUT_BATCH(attr_overrides[i * 2] | attr_overrides[i * 2 + 1] << 16);
}
OUT_BATCH(point_sprite_enables); /* dw16 */
OUT_BATCH(flat_enables);
OUT_BATCH(0); /* wrapshortest enables 0-7 */
OUT_BATCH(0); /* wrapshortest enables 8-15 */
ADVANCE_BATCH();
}
static void
upload_clip_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t depth_clamp = 0;
uint32_t provoking, userclip;
uint32_t dw1 = GEN6_CLIP_STATISTICS_ENABLE;
uint32_t nonperspective_barycentric_enable_flag = 0;
/* _NEW_BUFFERS */
bool render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
/* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->barycentric_interp_modes &
(1 << BRW_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC)) {
nonperspective_barycentric_enable_flag =
GEN6_CLIP_NON_PERSPECTIVE_BARYCENTRIC_ENABLE;
}
dw1 |= GEN7_CLIP_EARLY_CULL;
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw1 |= GEN7_CLIP_WINDING_CCW;
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw1 |= GEN7_CLIP_CULLMODE_FRONT;
break;
case GL_BACK:
dw1 |= GEN7_CLIP_CULLMODE_BACK;
break;
case GL_FRONT_AND_BACK:
dw1 |= GEN7_CLIP_CULLMODE_BOTH;
break;
default:
assert(!"Should not get here: invalid CullFlag");
break;
}
} else {
dw1 |= GEN7_CLIP_CULLMODE_NONE;
}
/* _NEW_TRANSFORM */
if (!ctx->Transform.DepthClamp)
depth_clamp = GEN6_CLIP_Z_TEST;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION) {
provoking =
(0 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(0 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
} else {
provoking =
(2 << GEN6_CLIP_TRI_PROVOKE_SHIFT) |
(2 << GEN6_CLIP_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_CLIP_LINE_PROVOKE_SHIFT);
}
/* _NEW_TRANSFORM */
userclip = ctx->Transform.ClipPlanesEnabled;
BEGIN_BATCH(4);
OUT_BATCH(_3DSTATE_CLIP << 16 | (4 - 2));
OUT_BATCH(dw1);
OUT_BATCH(GEN6_CLIP_ENABLE |
GEN6_CLIP_API_OGL |
GEN6_CLIP_MODE_NORMAL |
nonperspective_barycentric_enable_flag |
GEN6_CLIP_XY_TEST |
userclip << GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT |
depth_clamp |
provoking);
OUT_BATCH(U_FIXED(0.125, 3) << GEN6_CLIP_MIN_POINT_WIDTH_SHIFT |
U_FIXED(255.875, 3) << GEN6_CLIP_MAX_POINT_WIDTH_SHIFT |
GEN6_CLIP_FORCE_ZERO_RTAINDEX);
ADVANCE_BATCH();
}
/* 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 {
#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 {
maxlod_fixed = U_FIXED(CLAMP(maxlod, 0.0, 11), 0);
maxlod_fixed = MAX2(maxlod_fixed, (minlod_fixed + 15) >> 4);
state[I830_TEXREG_TM0S3] |= maxlod_fixed << TM0S3_MIN_MIP_SHIFT_830;
}
state[I830_TEXREG_TM0S3] |= minlod_fixed << TM0S3_MAX_MIP_SHIFT;
state[I830_TEXREG_TM0S3] |= ((minFilt << TM0S3_MIN_FILTER_SHIFT) |
(mipFilt << TM0S3_MIP_FILTER_SHIFT) |
(magFilt << TM0S3_MAG_FILTER_SHIFT));
}
{
GLenum ws = sampler->WrapS;
GLenum wt = sampler->WrapT;
/* 3D textures not available on i830
/**
* 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;
}
static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_sampler,
unsigned sdc_gs_offset,
struct brw_sampler_state *sampler)
{
memset(sampler, 0, sizeof(*sampler));
switch (pipe_sampler->min_mip_filter) {
case PIPE_TEX_FILTER_NEAREST:
sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
break;
case PIPE_TEX_FILTER_LINEAR:
sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
break;
case PIPE_TEX_FILTER_ANISO:
sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC;
break;
default:
break;
}
switch (pipe_sampler->min_mip_filter) {
case PIPE_TEX_MIPFILTER_NEAREST:
sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
break;
case PIPE_TEX_MIPFILTER_LINEAR:
sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
break;
case PIPE_TEX_MIPFILTER_NONE:
sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
break;
default:
break;
}
/* Set Anisotropy:
*/
switch (pipe_sampler->mag_img_filter) {
case PIPE_TEX_FILTER_NEAREST:
sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
break;
case PIPE_TEX_FILTER_LINEAR:
sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
break;
case PIPE_TEX_FILTER_ANISO:
sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
break;
default:
break;
}
if (pipe_sampler->max_anisotropy > 2.0) {
sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2,
BRW_ANISORATIO_16);
}
sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s);
sampler->ss1.r_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_r);
sampler->ss1.t_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_t);
/* Fulsim complains if I don't do this. Hardware doesn't mind:
*/
#if 0
if (texObj->Target == GL_TEXTURE_CUBE_MAP_ARB) {
sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
}
#endif
/* Set shadow function:
*/
if (pipe_sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
/* 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->ss0.shadow_function = intel_translate_shadow_compare_func(pipe_sampler->compare_func);
}
/* Set LOD bias:
*/
sampler->ss0.lod_bias = S_FIXED(CLAMP(pipe_sampler->lod_bias, -16, 15), 6);
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(MIN2(MAX2(pipe_sampler->max_lod, 0), 13), 6);
sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(pipe_sampler->min_lod, 0), 13), 6);
sampler->ss2.default_color_pointer = sdc_gs_offset >> 5;
}
/**
* 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
upload_sf_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
uint32_t dw1, dw2, dw3;
float point_size;
/* _NEW_BUFFERS */
bool render_to_fbo = _mesa_is_user_fbo(brw->intel.ctx.DrawBuffer);
bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
dw1 = GEN6_SF_STATISTICS_ENABLE |
GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
/* _NEW_BUFFERS */
dw1 |= (brw_depthbuffer_format(brw) << GEN7_SF_DEPTH_BUFFER_SURFACE_FORMAT_SHIFT);
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw1 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
if (ctx->Polygon.OffsetLine)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
if (ctx->Polygon.OffsetPoint)
dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
switch (ctx->Polygon.FrontMode) {
case GL_FILL:
dw1 |= GEN6_SF_FRONT_SOLID;
break;
case GL_LINE:
dw1 |= GEN6_SF_FRONT_WIREFRAME;
break;
case GL_POINT:
dw1 |= GEN6_SF_FRONT_POINT;
break;
default:
assert(0);
break;
}
switch (ctx->Polygon.BackMode) {
case GL_FILL:
dw1 |= GEN6_SF_BACK_SOLID;
break;
case GL_LINE:
dw1 |= GEN6_SF_BACK_WIREFRAME;
break;
case GL_POINT:
dw1 |= GEN6_SF_BACK_POINT;
break;
default:
assert(0);
break;
}
dw2 = 0;
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw2 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw2 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw2 |= GEN6_SF_CULL_BOTH;
break;
default:
assert(0);
break;
}
} else {
dw2 |= GEN6_SF_CULL_NONE;
}
/* _NEW_SCISSOR */
if (ctx->Scissor.Enabled)
dw2 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_LINE */
{
uint32_t line_width_u3_7 = U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7);
/* TODO: line width of 0 is not allowed when MSAA enabled */
if (line_width_u3_7 == 0)
line_width_u3_7 = 1;
dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
}
if (ctx->Line.SmoothFlag) {
dw2 |= GEN6_SF_LINE_AA_ENABLE;
dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
if (ctx->Line.StippleFlag && intel->is_haswell) {
dw2 |= HSW_SF_LINE_STIPPLE_ENABLE;
}
/* _NEW_MULTISAMPLE */
if (multisampled_fbo && ctx->Multisample.Enabled)
dw2 |= GEN6_SF_MSRAST_ON_PATTERN;
/* FINISHME: Last Pixel Enable? Vertex Sub Pixel Precision Select?
*/
dw3 = GEN6_SF_LINE_AA_MODE_TRUE;
/* _NEW_PROGRAM | _NEW_POINT */
if (!(ctx->VertexProgram.PointSizeEnabled || ctx->Point._Attenuated))
dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw3 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw3 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw3 |= (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
BEGIN_BATCH(7);
OUT_BATCH(_3DSTATE_SF << 16 | (7 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
ADVANCE_BATCH();
}
static void
upload_sf_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
/* BRW_NEW_FRAGMENT_PROGRAM */
uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
/* _NEW_LIGHT */
bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
uint32_t dw1, dw2, dw3, dw4, dw16, dw17;
int i;
/* _NEW_BUFFER */
bool render_to_fbo = _mesa_is_user_fbo(brw->intel.ctx.DrawBuffer);
bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
int attr = 0, input_index = 0;
int urb_entry_read_offset = 1;
float point_size;
uint16_t attr_overrides[FRAG_ATTRIB_MAX];
uint32_t point_sprite_origin;
dw1 = GEN6_SF_SWIZZLE_ENABLE | num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT;
dw2 = GEN6_SF_STATISTICS_ENABLE |
GEN6_SF_VIEWPORT_TRANSFORM_ENABLE;
dw3 = 0;
dw4 = 0;
dw16 = 0;
dw17 = 0;
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw2 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
if (ctx->Polygon.OffsetLine)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME;
if (ctx->Polygon.OffsetPoint)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT;
switch (ctx->Polygon.FrontMode) {
case GL_FILL:
dw2 |= GEN6_SF_FRONT_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_FRONT_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_FRONT_POINT;
break;
default:
assert(0);
break;
}
switch (ctx->Polygon.BackMode) {
case GL_FILL:
dw2 |= GEN6_SF_BACK_SOLID;
break;
case GL_LINE:
dw2 |= GEN6_SF_BACK_WIREFRAME;
break;
case GL_POINT:
dw2 |= GEN6_SF_BACK_POINT;
break;
default:
assert(0);
break;
}
/* _NEW_SCISSOR */
if (ctx->Scissor.Enabled)
dw3 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_POLYGON */
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw3 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw3 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw3 |= GEN6_SF_CULL_BOTH;
break;
default:
assert(0);
break;
}
} else {
dw3 |= GEN6_SF_CULL_NONE;
}
/* _NEW_LINE */
{
uint32_t line_width_u3_7 = U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7);
/* TODO: line width of 0 is not allowed when MSAA enabled */
if (line_width_u3_7 == 0)
line_width_u3_7 = 1;
dw3 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT;
}
if (ctx->Line.SmoothFlag) {
dw3 |= GEN6_SF_LINE_AA_ENABLE;
dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
/* _NEW_MULTISAMPLE */
if (multisampled_fbo && ctx->Multisample.Enabled)
dw3 |= GEN6_SF_MSRAST_ON_PATTERN;
/* _NEW_PROGRAM | _NEW_POINT */
if (!(ctx->VertexProgram.PointSizeEnabled ||
ctx->Point._Attenuated))
dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH;
/* Clamp to ARB_point_parameters user limits */
point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
/* Clamp to the hardware limits and convert to fixed point */
dw4 |= U_FIXED(CLAMP(point_size, 0.125, 255.875), 3);
/*
* Window coordinates in an FBO are inverted, which means point
* sprite origin must be inverted, too.
*/
if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) {
point_sprite_origin = GEN6_SF_POINT_SPRITE_LOWERLEFT;
} else {
point_sprite_origin = GEN6_SF_POINT_SPRITE_UPPERLEFT;
}
dw1 |= point_sprite_origin;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw4 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw4 |=
(1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
/* Create the mapping from the FS inputs we produce to the VS outputs
* they source from.
*/
uint32_t max_source_attr = 0;
for (; attr < FRAG_ATTRIB_MAX; attr++) {
enum glsl_interp_qualifier interp_qualifier =
brw->fragment_program->InterpQualifier[attr];
bool is_gl_Color = attr == FRAG_ATTRIB_COL0 || attr == FRAG_ATTRIB_COL1;
if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
continue;
/* _NEW_POINT */
if (ctx->Point.PointSprite &&
(attr >= FRAG_ATTRIB_TEX0 && attr <= FRAG_ATTRIB_TEX7) &&
ctx->Point.CoordReplace[attr - FRAG_ATTRIB_TEX0]) {
dw16 |= (1 << input_index);
}
if (attr == FRAG_ATTRIB_PNTC)
dw16 |= (1 << input_index);
/* flat shading */
if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
(shade_model_flat && is_gl_Color &&
interp_qualifier == INTERP_QUALIFIER_NONE))
dw17 |= (1 << input_index);
/* The hardware can only do the overrides on 16 overrides at a
* time, and the other up to 16 have to be lined up so that the
* input index = the output index. We'll need to do some
* tweaking to make sure that's the case.
*/
assert(input_index < 16 || attr == input_index);
/* CACHE_NEW_VS_PROG | _NEW_LIGHT | _NEW_PROGRAM */
attr_overrides[input_index++] =
get_attr_override(&brw->vs.prog_data->vue_map,
urb_entry_read_offset, attr,
ctx->VertexProgram._TwoSideEnabled,
&max_source_attr);
}
for (; input_index < FRAG_ATTRIB_MAX; input_index++)
attr_overrides[input_index] = 0;
/* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
* 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
*
* "This field should be set to the minimum length required to read the
* maximum source attribute. The maximum source attribute is indicated
* by the maximum value of the enabled Attribute # Source Attribute if
* Attribute Swizzle Enable is set, Number of Output Attributes-1 if
* enable is not set.
* read_length = ceiling((max_source_attr + 1) / 2)
*
* [errata] Corruption/Hang possible if length programmed larger than
* recommended"
*/
uint32_t urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
dw1 |= urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
BEGIN_BATCH(20);
OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH(dw4);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
for (i = 0; i < 8; i++) {
OUT_BATCH(attr_overrides[i * 2] | attr_overrides[i * 2 + 1] << 16);
}
OUT_BATCH(dw16); /* point sprite texcoord bitmask */
OUT_BATCH(dw17); /* constant interp bitmask */
OUT_BATCH(0); /* wrapshortest enables 0-7 */
OUT_BATCH(0); /* wrapshortest enables 8-15 */
ADVANCE_BATCH();
}
/* 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 GLboolean
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;
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 GL_FALSE;
/* Get first image here, since intelObj->firstLevel will get set in
* the intel_finalize_mipmap_tree() call above.
*/
firstImage = tObj->Image[0][intelObj->firstLevel];
drm_intel_bo_reference(intelObj->mt->region->buffer);
i915->state.tex_buffer[unit] = intelObj->mt->region->buffer;
i915->state.tex_offset[unit] = 0; /* Always the origin of the miptree */
format = translate_texture_format(firstImage->TexFormat,
firstImage->InternalFormat,
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(tObj->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 (tObj->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 GL_FALSE;
}
if (tObj->MaxAnisotropy > 1.0) {
minFilt = FILTER_ANISOTROPIC;
magFilt = FILTER_ANISOTROPIC;
if (tObj->MaxAnisotropy > 2.0)
aniso = SS2_MAX_ANISO_4;
else
aniso = SS2_MAX_ANISO_2;
}
else {
switch (tObj->MagFilter) {
case GL_NEAREST:
magFilt = FILTER_NEAREST;
break;
case GL_LINEAR:
magFilt = FILTER_LINEAR;
break;
default:
return GL_FALSE;
}
}
lodbias = (int) ((tUnit->LodBias + tObj->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 (tObj->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB &&
tObj->Target != GL_TEXTURE_3D) {
if (tObj->Target == GL_TEXTURE_1D)
return GL_FALSE;
state[I915_TEXREG_SS2] |=
(SS2_SHADOW_ENABLE |
intel_translate_shadow_compare_func(tObj->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 = tObj->WrapS;
GLenum wt = tObj->WrapT;
GLenum wr = tObj->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 &&
(tObj->MinFilter != GL_NEAREST ||
tObj->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 GL_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 GL_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(tObj->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], tObj->BorderColor.f[0]);
CLAMPED_FLOAT_TO_UBYTE(border[1], tObj->BorderColor.f[1]);
CLAMPED_FLOAT_TO_UBYTE(border[2], tObj->BorderColor.f[2]);
CLAMPED_FLOAT_TO_UBYTE(border[3], tObj->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), GL_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 GL_TRUE;
}
static void
upload_sf_state(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
GLcontext *ctx = &intel->ctx;
/* CACHE_NEW_VS_PROG */
uint32_t num_inputs = brw_count_bits(brw->vs.prog_data->outputs_written);
uint32_t num_outputs = brw_count_bits(brw->fragment_program->Base.InputsRead);
uint32_t dw1, dw2, dw3, dw4, dw16;
int i;
/* _NEW_BUFFER */
GLboolean render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
int attr = 0;
dw1 =
num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT |
(num_inputs + 1) / 2 << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
1 << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
dw2 = GEN6_SF_VIEWPORT_TRANSFORM_ENABLE |
GEN6_SF_STATISTICS_ENABLE;
dw3 = 0;
dw4 = 0;
dw16 = 0;
/* _NEW_POLYGON */
if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo)
dw2 |= GEN6_SF_WINDING_CCW;
if (ctx->Polygon.OffsetFill)
dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID;
/* _NEW_SCISSOR */
if (ctx->Scissor.Enabled)
dw3 |= GEN6_SF_SCISSOR_ENABLE;
/* _NEW_POLYGON */
if (ctx->Polygon.CullFlag) {
switch (ctx->Polygon.CullFaceMode) {
case GL_FRONT:
dw3 |= GEN6_SF_CULL_FRONT;
break;
case GL_BACK:
dw3 |= GEN6_SF_CULL_BACK;
break;
case GL_FRONT_AND_BACK:
dw3 |= GEN6_SF_CULL_BOTH;
break;
default:
assert(0);
break;
}
} else {
dw3 |= GEN6_SF_CULL_NONE;
}
/* _NEW_LINE */
dw3 |= U_FIXED(CLAMP(ctx->Line.Width, 0.0, 7.99), 7) <<
GEN6_SF_LINE_WIDTH_SHIFT;
if (ctx->Line.SmoothFlag) {
dw3 |= GEN6_SF_LINE_AA_ENABLE;
dw3 |= GEN6_SF_LINE_AA_MODE_TRUE;
dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0;
}
/* _NEW_POINT */
if (ctx->Point._Attenuated)
dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH;
dw4 |= U_FIXED(CLAMP(ctx->Point.Size, 0.125, 225.875), 3) <<
GEN6_SF_POINT_WIDTH_SHIFT;
if (ctx->Point.SpriteOrigin == GL_LOWER_LEFT)
dw1 |= GEN6_SF_POINT_SPRITE_LOWERLEFT;
/* _NEW_LIGHT */
if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) {
dw4 |=
(2 << GEN6_SF_TRI_PROVOKE_SHIFT) |
(2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) |
(1 << GEN6_SF_LINE_PROVOKE_SHIFT);
} else {
dw4 |=
(1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
}
if (ctx->Point.PointSprite) {
for (i = 0; i < 8; i++) {
if (ctx->Point.CoordReplace[i])
dw16 |= (1 << i);
}
}
BEGIN_BATCH(20);
OUT_BATCH(CMD_3D_SF_STATE << 16 | (20 - 2));
OUT_BATCH(dw1);
OUT_BATCH(dw2);
OUT_BATCH(dw3);
OUT_BATCH(dw4);
OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */
OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */
OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
for (i = 0; i < 8; i++) {
uint32_t attr_overrides = 0;
for (; attr < 64; attr++) {
if (brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)) {
attr_overrides |= get_attr_override(brw, attr);
attr++;
break;
}
}
for (; attr < 64; attr++) {
if (brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)) {
attr_overrides |= get_attr_override(brw, attr) << 16;
attr++;
break;
}
}
OUT_BATCH(attr_overrides);
}
OUT_BATCH(dw16); /* point sprite texcoord bitmask */
OUT_BATCH(0); /* constant interp bitmask */
OUT_BATCH(0); /* wrapshortest enables 0-7 */
OUT_BATCH(0); /* wrapshortest enables 8-15 */
ADVANCE_BATCH();
intel_batchbuffer_emit_mi_flush(intel->batch);
}