2 * Copyright © 2009 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
24 * Eric Anholt <eric@anholt.net>
28 #include "brw_context.h"
29 #include "brw_state.h"
30 #include "brw_defines.h"
32 #include "main/macros.h"
33 #include "main/fbobject.h"
34 #include "intel_batchbuffer.h"
37 * Determine the appropriate attribute override value to store into the
38 * 3DSTATE_SF structure for a given fragment shader attribute. The attribute
39 * override value contains two pieces of information: the location of the
40 * attribute in the VUE (relative to urb_entry_read_offset, see below), and a
41 * flag indicating whether to "swizzle" the attribute based on the direction
42 * the triangle is facing.
44 * If an attribute is "swizzled", then the given VUE location is used for
45 * front-facing triangles, and the VUE location that immediately follows is
46 * used for back-facing triangles. We use this to implement the mapping from
47 * gl_FrontColor/gl_BackColor to gl_Color.
49 * urb_entry_read_offset is the offset into the VUE at which the SF unit is
50 * being instructed to begin reading attribute data. It can be set to a
51 * nonzero value to prevent the SF unit from wasting time reading elements of
52 * the VUE that are not needed by the fragment shader. It is measured in
56 get_attr_override(struct brw_vue_map
*vue_map
, int urb_entry_read_offset
,
57 int fs_attr
, bool two_side_color
, uint32_t *max_source_attr
)
59 int vs_attr
= _mesa_frag_attrib_to_vert_result(fs_attr
);
60 if (vs_attr
< 0 || vs_attr
== VERT_RESULT_HPOS
) {
61 /* These attributes will be overwritten by the fragment shader's
62 * interpolation code (see emit_interp() in brw_wm_fp.c), so just let
63 * them reference the first available attribute.
68 /* Find the VUE slot for this attribute. */
69 int slot
= vue_map
->vert_result_to_slot
[vs_attr
];
71 /* If there was only a back color written but not front, use back
72 * as the color instead of undefined
74 if (slot
== -1 && vs_attr
== VERT_RESULT_COL0
)
75 slot
= vue_map
->vert_result_to_slot
[VERT_RESULT_BFC0
];
76 if (slot
== -1 && vs_attr
== VERT_RESULT_COL1
)
77 slot
= vue_map
->vert_result_to_slot
[VERT_RESULT_BFC1
];
80 /* This attribute does not exist in the VUE--that means that the vertex
81 * shader did not write to it. Behavior is undefined in this case, so
82 * just reference the first available attribute.
87 /* Compute the location of the attribute relative to urb_entry_read_offset.
88 * Each increment of urb_entry_read_offset represents a 256-bit value, so
89 * it counts for two 128-bit VUE slots.
91 int source_attr
= slot
- 2 * urb_entry_read_offset
;
92 assert(source_attr
>= 0 && source_attr
< 32);
94 /* If we are doing two-sided color, and the VUE slot following this one
95 * represents a back-facing color, then we need to instruct the SF unit to
96 * do back-facing swizzling.
98 bool swizzling
= two_side_color
&&
99 ((vue_map
->slot_to_vert_result
[slot
] == VERT_RESULT_COL0
&&
100 vue_map
->slot_to_vert_result
[slot
+1] == VERT_RESULT_BFC0
) ||
101 (vue_map
->slot_to_vert_result
[slot
] == VERT_RESULT_COL1
&&
102 vue_map
->slot_to_vert_result
[slot
+1] == VERT_RESULT_BFC1
));
104 /* Update max_source_attr. If swizzling, the SF will read this slot + 1. */
105 if (*max_source_attr
< source_attr
+ swizzling
)
106 *max_source_attr
= source_attr
+ swizzling
;
110 (ATTRIBUTE_SWIZZLE_INPUTATTR_FACING
<< ATTRIBUTE_SWIZZLE_SHIFT
);
117 upload_sf_state(struct brw_context
*brw
)
119 struct intel_context
*intel
= &brw
->intel
;
120 struct gl_context
*ctx
= &intel
->ctx
;
121 /* BRW_NEW_FRAGMENT_PROGRAM */
122 uint32_t num_outputs
= _mesa_bitcount_64(brw
->fragment_program
->Base
.InputsRead
);
124 bool shade_model_flat
= ctx
->Light
.ShadeModel
== GL_FLAT
;
125 uint32_t dw1
, dw2
, dw3
, dw4
, dw16
, dw17
;
128 bool render_to_fbo
= _mesa_is_user_fbo(brw
->intel
.ctx
.DrawBuffer
);
129 bool multisampled_fbo
= ctx
->DrawBuffer
->Visual
.samples
> 1;
131 int attr
= 0, input_index
= 0;
132 int urb_entry_read_offset
= 1;
134 uint16_t attr_overrides
[FRAG_ATTRIB_MAX
];
135 uint32_t point_sprite_origin
;
137 dw1
= GEN6_SF_SWIZZLE_ENABLE
| num_outputs
<< GEN6_SF_NUM_OUTPUTS_SHIFT
;
139 dw2
= GEN6_SF_STATISTICS_ENABLE
|
140 GEN6_SF_VIEWPORT_TRANSFORM_ENABLE
;
148 if ((ctx
->Polygon
.FrontFace
== GL_CCW
) ^ render_to_fbo
)
149 dw2
|= GEN6_SF_WINDING_CCW
;
151 if (ctx
->Polygon
.OffsetFill
)
152 dw2
|= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID
;
154 if (ctx
->Polygon
.OffsetLine
)
155 dw2
|= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME
;
157 if (ctx
->Polygon
.OffsetPoint
)
158 dw2
|= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT
;
160 switch (ctx
->Polygon
.FrontMode
) {
162 dw2
|= GEN6_SF_FRONT_SOLID
;
166 dw2
|= GEN6_SF_FRONT_WIREFRAME
;
170 dw2
|= GEN6_SF_FRONT_POINT
;
178 switch (ctx
->Polygon
.BackMode
) {
180 dw2
|= GEN6_SF_BACK_SOLID
;
184 dw2
|= GEN6_SF_BACK_WIREFRAME
;
188 dw2
|= GEN6_SF_BACK_POINT
;
197 if (ctx
->Scissor
.Enabled
)
198 dw3
|= GEN6_SF_SCISSOR_ENABLE
;
201 if (ctx
->Polygon
.CullFlag
) {
202 switch (ctx
->Polygon
.CullFaceMode
) {
204 dw3
|= GEN6_SF_CULL_FRONT
;
207 dw3
|= GEN6_SF_CULL_BACK
;
209 case GL_FRONT_AND_BACK
:
210 dw3
|= GEN6_SF_CULL_BOTH
;
217 dw3
|= GEN6_SF_CULL_NONE
;
222 uint32_t line_width_u3_7
= U_FIXED(CLAMP(ctx
->Line
.Width
, 0.0, 7.99), 7);
223 /* TODO: line width of 0 is not allowed when MSAA enabled */
224 if (line_width_u3_7
== 0)
226 dw3
|= line_width_u3_7
<< GEN6_SF_LINE_WIDTH_SHIFT
;
228 if (ctx
->Line
.SmoothFlag
) {
229 dw3
|= GEN6_SF_LINE_AA_ENABLE
;
230 dw3
|= GEN6_SF_LINE_AA_MODE_TRUE
;
231 dw3
|= GEN6_SF_LINE_END_CAP_WIDTH_1_0
;
233 /* _NEW_MULTISAMPLE */
234 if (multisampled_fbo
&& ctx
->Multisample
.Enabled
)
235 dw3
|= GEN6_SF_MSRAST_ON_PATTERN
;
237 /* _NEW_PROGRAM | _NEW_POINT */
238 if (!(ctx
->VertexProgram
.PointSizeEnabled
||
239 ctx
->Point
._Attenuated
))
240 dw4
|= GEN6_SF_USE_STATE_POINT_WIDTH
;
242 /* Clamp to ARB_point_parameters user limits */
243 point_size
= CLAMP(ctx
->Point
.Size
, ctx
->Point
.MinSize
, ctx
->Point
.MaxSize
);
245 /* Clamp to the hardware limits and convert to fixed point */
246 dw4
|= U_FIXED(CLAMP(point_size
, 0.125, 255.875), 3);
249 * Window coordinates in an FBO are inverted, which means point
250 * sprite origin must be inverted, too.
252 if ((ctx
->Point
.SpriteOrigin
== GL_LOWER_LEFT
) != render_to_fbo
) {
253 point_sprite_origin
= GEN6_SF_POINT_SPRITE_LOWERLEFT
;
255 point_sprite_origin
= GEN6_SF_POINT_SPRITE_UPPERLEFT
;
257 dw1
|= point_sprite_origin
;
260 if (ctx
->Light
.ProvokingVertex
!= GL_FIRST_VERTEX_CONVENTION
) {
262 (2 << GEN6_SF_TRI_PROVOKE_SHIFT
) |
263 (2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT
) |
264 (1 << GEN6_SF_LINE_PROVOKE_SHIFT
);
267 (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT
);
270 /* Create the mapping from the FS inputs we produce to the VS outputs
273 uint32_t max_source_attr
= 0;
274 for (; attr
< FRAG_ATTRIB_MAX
; attr
++) {
275 enum glsl_interp_qualifier interp_qualifier
=
276 brw
->fragment_program
->InterpQualifier
[attr
];
277 bool is_gl_Color
= attr
== FRAG_ATTRIB_COL0
|| attr
== FRAG_ATTRIB_COL1
;
279 if (!(brw
->fragment_program
->Base
.InputsRead
& BITFIELD64_BIT(attr
)))
283 if (ctx
->Point
.PointSprite
&&
284 (attr
>= FRAG_ATTRIB_TEX0
&& attr
<= FRAG_ATTRIB_TEX7
) &&
285 ctx
->Point
.CoordReplace
[attr
- FRAG_ATTRIB_TEX0
]) {
286 dw16
|= (1 << input_index
);
289 if (attr
== FRAG_ATTRIB_PNTC
)
290 dw16
|= (1 << input_index
);
293 if (interp_qualifier
== INTERP_QUALIFIER_FLAT
||
294 (shade_model_flat
&& is_gl_Color
&&
295 interp_qualifier
== INTERP_QUALIFIER_NONE
))
296 dw17
|= (1 << input_index
);
298 /* The hardware can only do the overrides on 16 overrides at a
299 * time, and the other up to 16 have to be lined up so that the
300 * input index = the output index. We'll need to do some
301 * tweaking to make sure that's the case.
303 assert(input_index
< 16 || attr
== input_index
);
305 /* CACHE_NEW_VS_PROG | _NEW_LIGHT | _NEW_PROGRAM */
306 attr_overrides
[input_index
++] =
307 get_attr_override(&brw
->vs
.prog_data
->vue_map
,
308 urb_entry_read_offset
, attr
,
309 ctx
->VertexProgram
._TwoSideEnabled
,
313 for (; input_index
< FRAG_ATTRIB_MAX
; input_index
++)
314 attr_overrides
[input_index
] = 0;
316 /* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
317 * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
319 * "This field should be set to the minimum length required to read the
320 * maximum source attribute. The maximum source attribute is indicated
321 * by the maximum value of the enabled Attribute # Source Attribute if
322 * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
324 * read_length = ceiling((max_source_attr + 1) / 2)
326 * [errata] Corruption/Hang possible if length programmed larger than
329 uint32_t urb_entry_read_length
= ALIGN(max_source_attr
+ 1, 2) / 2;
330 dw1
|= urb_entry_read_length
<< GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT
|
331 urb_entry_read_offset
<< GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT
;
334 OUT_BATCH(_3DSTATE_SF
<< 16 | (20 - 2));
339 OUT_BATCH_F(ctx
->Polygon
.OffsetUnits
* 2); /* constant. copied from gen4 */
340 OUT_BATCH_F(ctx
->Polygon
.OffsetFactor
); /* scale */
341 OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */
342 for (i
= 0; i
< 8; i
++) {
343 OUT_BATCH(attr_overrides
[i
* 2] | attr_overrides
[i
* 2 + 1] << 16);
345 OUT_BATCH(dw16
); /* point sprite texcoord bitmask */
346 OUT_BATCH(dw17
); /* constant interp bitmask */
347 OUT_BATCH(0); /* wrapshortest enables 0-7 */
348 OUT_BATCH(0); /* wrapshortest enables 8-15 */
352 const struct brw_tracked_state gen6_sf_state
= {
354 .mesa
= (_NEW_LIGHT
|
362 .brw
= (BRW_NEW_CONTEXT
|
363 BRW_NEW_FRAGMENT_PROGRAM
),
364 .cache
= CACHE_NEW_VS_PROG
366 .emit
= upload_sf_state
,