2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
6 Permission is hereby granted, free of charge, to any person obtaining
7 a copy of this software and associated documentation files (the
8 "Software"), to deal in the Software without restriction, including
9 without limitation the rights to use, copy, modify, merge, publish,
10 distribute, sublicense, and/or sell copies of the Software, and to
11 permit persons to whom the Software is furnished to do so, subject to
12 the following conditions:
14 The above copyright notice and this permission notice (including the
15 next paragraph) shall be included in all copies or substantial
16 portions of the Software.
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **********************************************************************/
29 * Keith Whitwell <keith@tungstengraphics.com>
33 #include "main/glheader.h"
34 #include "main/macros.h"
35 #include "main/enums.h"
36 #include "program/program.h"
38 #include "intel_batchbuffer.h"
40 #include "brw_defines.h"
41 #include "brw_context.h"
48 struct brw_reg
get_tmp( struct brw_clip_compile
*c
)
50 struct brw_reg tmp
= brw_vec4_grf(c
->last_tmp
, 0);
52 if (++c
->last_tmp
> c
->prog_data
.total_grf
)
53 c
->prog_data
.total_grf
= c
->last_tmp
;
58 static void release_tmp( struct brw_clip_compile
*c
, struct brw_reg tmp
)
60 if (tmp
.nr
== c
->last_tmp
-1)
65 static struct brw_reg
make_plane_ud(GLuint x
, GLuint y
, GLuint z
, GLuint w
)
67 return brw_imm_ud((w
<<24) | (z
<<16) | (y
<<8) | x
);
71 void brw_clip_init_planes( struct brw_clip_compile
*c
)
73 struct brw_compile
*p
= &c
->func
;
75 if (!c
->key
.nr_userclip
) {
76 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 0), make_plane_ud( 0, 0, 0xff, 1));
77 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 1), make_plane_ud( 0, 0, 1, 1));
78 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 2), make_plane_ud( 0, 0xff, 0, 1));
79 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 3), make_plane_ud( 0, 1, 0, 1));
80 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 4), make_plane_ud(0xff, 0, 0, 1));
81 brw_MOV(p
, get_element_ud(c
->reg
.fixed_planes
, 5), make_plane_ud( 1, 0, 0, 1));
89 /* Project 'pos' to screen space (or back again), overwrite with results:
91 void brw_clip_project_position(struct brw_clip_compile
*c
, struct brw_reg pos
)
93 struct brw_compile
*p
= &c
->func
;
97 brw_math_invert(p
, get_element(pos
, W
), get_element(pos
, W
));
99 /* value.xyz *= value.rhw
101 brw_set_access_mode(p
, BRW_ALIGN_16
);
102 brw_MUL(p
, brw_writemask(pos
, WRITEMASK_XYZ
), pos
, brw_swizzle1(pos
, W
));
103 brw_set_access_mode(p
, BRW_ALIGN_1
);
107 static void brw_clip_project_vertex( struct brw_clip_compile
*c
,
108 struct brw_indirect vert_addr
)
110 struct brw_compile
*p
= &c
->func
;
111 struct brw_reg tmp
= get_tmp(c
);
112 GLuint hpos_offset
= brw_varying_to_offset(&c
->vue_map
, VARYING_SLOT_POS
);
113 GLuint ndc_offset
= brw_varying_to_offset(&c
->vue_map
,
114 BRW_VARYING_SLOT_NDC
);
116 /* Fixup position. Extract from the original vertex and re-project
119 brw_MOV(p
, tmp
, deref_4f(vert_addr
, hpos_offset
));
120 brw_clip_project_position(c
, tmp
);
121 brw_MOV(p
, deref_4f(vert_addr
, ndc_offset
), tmp
);
129 /* Interpolate between two vertices and put the result into a0.0.
130 * Increment a0.0 accordingly.
132 void brw_clip_interp_vertex( struct brw_clip_compile
*c
,
133 struct brw_indirect dest_ptr
,
134 struct brw_indirect v0_ptr
, /* from */
135 struct brw_indirect v1_ptr
, /* to */
139 struct brw_compile
*p
= &c
->func
;
140 struct brw_reg tmp
= get_tmp(c
);
143 /* Just copy the vertex header:
146 * After CLIP stage, only first 256 bits of the VUE are read
147 * back on Ironlake, so needn't change it
149 brw_copy_indirect_to_indirect(p
, dest_ptr
, v0_ptr
, 1);
151 /* Iterate over each attribute (could be done in pairs?)
153 for (slot
= 0; slot
< c
->vue_map
.num_slots
; slot
++) {
154 int varying
= c
->vue_map
.slot_to_varying
[slot
];
155 GLuint delta
= brw_vue_slot_to_offset(slot
);
157 if (varying
== VARYING_SLOT_EDGE
) {
159 brw_MOV(p
, deref_4f(dest_ptr
, delta
), brw_imm_f(1));
161 brw_MOV(p
, deref_4f(dest_ptr
, delta
), deref_4f(v0_ptr
, delta
));
162 } else if (varying
== VARYING_SLOT_PSIZ
||
163 varying
== VARYING_SLOT_CLIP_DIST0
||
164 varying
== VARYING_SLOT_CLIP_DIST1
) {
165 /* PSIZ doesn't need interpolation because it isn't used by the
166 * fragment shader. CLIP_DIST0 and CLIP_DIST1 don't need
167 * intepolation because on pre-GEN6, these are just placeholder VUE
168 * slots that don't perform any action.
170 } else if (varying
< VARYING_SLOT_MAX
) {
171 /* This is a true vertex result (and not a special value for the VUE
172 * header), so interpolate:
174 * New = attr0 + t*attr1 - t*attr0
177 vec4(brw_null_reg()),
178 deref_4f(v1_ptr
, delta
),
183 negate(deref_4f(v0_ptr
, delta
)),
187 deref_4f(dest_ptr
, delta
),
188 deref_4f(v0_ptr
, delta
),
193 if (c
->vue_map
.num_slots
% 2) {
194 GLuint delta
= brw_vue_slot_to_offset(c
->vue_map
.num_slots
);
196 brw_MOV(p
, deref_4f(dest_ptr
, delta
), brw_imm_f(0));
201 /* Recreate the projected (NDC) coordinate in the new vertex
204 brw_clip_project_vertex(c
, dest_ptr
);
207 void brw_clip_emit_vue(struct brw_clip_compile
*c
,
208 struct brw_indirect vert
,
213 struct brw_compile
*p
= &c
->func
;
217 assert(!(allocate
&& eot
));
219 /* Copy the vertex from vertn into m1..mN+1:
221 brw_copy_from_indirect(p
, brw_message_reg(1), vert
, c
->nr_regs
);
223 /* Overwrite PrimType and PrimStart in the message header, for
224 * each vertex in turn:
226 brw_MOV(p
, get_element_ud(c
->reg
.R0
, 2), brw_imm_ud(header
));
229 /* Send each vertex as a seperate write to the urb. This
230 * is different to the concept in brw_sf_emit.c, where
231 * subsequent writes are used to build up a single urb
232 * entry. Each of these writes instantiates a seperate
233 * urb entry - (I think... what about 'allocate'?)
236 allocate
? c
->reg
.R0
: retype(brw_null_reg(), BRW_REGISTER_TYPE_UD
),
241 c
->nr_regs
+ 1, /* msg length */
242 allocate
? 1 : 0, /* response_length */
244 1, /* writes_complete */
246 BRW_URB_SWIZZLE_NONE
);
251 void brw_clip_kill_thread(struct brw_clip_compile
*c
)
253 struct brw_compile
*p
= &c
->func
;
256 /* Send an empty message to kill the thread and release any
257 * allocated urb entry:
260 retype(brw_null_reg(), BRW_REGISTER_TYPE_UD
),
266 0, /* response len */
268 1, /* writes complete */
270 BRW_URB_SWIZZLE_NONE
);
276 struct brw_reg
brw_clip_plane0_address( struct brw_clip_compile
*c
)
278 return brw_address(c
->reg
.fixed_planes
);
282 struct brw_reg
brw_clip_plane_stride( struct brw_clip_compile
*c
)
284 if (c
->key
.nr_userclip
) {
285 return brw_imm_uw(16);
288 return brw_imm_uw(4);
293 /* If flatshading, distribute color from provoking vertex prior to
296 void brw_clip_copy_colors( struct brw_clip_compile
*c
,
297 GLuint to
, GLuint from
)
299 struct brw_compile
*p
= &c
->func
;
301 if (brw_clip_have_varying(c
, VARYING_SLOT_COL0
))
303 byte_offset(c
->reg
.vertex
[to
],
304 brw_varying_to_offset(&c
->vue_map
,
306 byte_offset(c
->reg
.vertex
[from
],
307 brw_varying_to_offset(&c
->vue_map
,
308 VARYING_SLOT_COL0
)));
310 if (brw_clip_have_varying(c
, VARYING_SLOT_COL1
))
312 byte_offset(c
->reg
.vertex
[to
],
313 brw_varying_to_offset(&c
->vue_map
,
315 byte_offset(c
->reg
.vertex
[from
],
316 brw_varying_to_offset(&c
->vue_map
,
317 VARYING_SLOT_COL1
)));
319 if (brw_clip_have_varying(c
, VARYING_SLOT_BFC0
))
321 byte_offset(c
->reg
.vertex
[to
],
322 brw_varying_to_offset(&c
->vue_map
,
324 byte_offset(c
->reg
.vertex
[from
],
325 brw_varying_to_offset(&c
->vue_map
,
326 VARYING_SLOT_BFC0
)));
328 if (brw_clip_have_varying(c
, VARYING_SLOT_BFC1
))
330 byte_offset(c
->reg
.vertex
[to
],
331 brw_varying_to_offset(&c
->vue_map
,
333 byte_offset(c
->reg
.vertex
[from
],
334 brw_varying_to_offset(&c
->vue_map
,
335 VARYING_SLOT_BFC1
)));
340 void brw_clip_init_clipmask( struct brw_clip_compile
*c
)
342 struct brw_compile
*p
= &c
->func
;
343 struct brw_reg incoming
= get_element_ud(c
->reg
.R0
, 2);
345 /* Shift so that lowest outcode bit is rightmost:
347 brw_SHR(p
, c
->reg
.planemask
, incoming
, brw_imm_ud(26));
349 if (c
->key
.nr_userclip
) {
350 struct brw_reg tmp
= retype(vec1(get_tmp(c
)), BRW_REGISTER_TYPE_UD
);
352 /* Rearrange userclip outcodes so that they come directly after
353 * the fixed plane bits.
355 brw_AND(p
, tmp
, incoming
, brw_imm_ud(0x3f<<14));
356 brw_SHR(p
, tmp
, tmp
, brw_imm_ud(8));
357 brw_OR(p
, c
->reg
.planemask
, c
->reg
.planemask
, tmp
);
363 void brw_clip_ff_sync(struct brw_clip_compile
*c
)
365 struct intel_context
*intel
= &c
->func
.brw
->intel
;
367 if (intel
->needs_ff_sync
) {
368 struct brw_compile
*p
= &c
->func
;
370 brw_set_conditionalmod(p
, BRW_CONDITIONAL_Z
);
371 brw_AND(p
, brw_null_reg(), c
->reg
.ff_sync
, brw_imm_ud(0x1));
372 brw_IF(p
, BRW_EXECUTE_1
);
374 brw_OR(p
, c
->reg
.ff_sync
, c
->reg
.ff_sync
, brw_imm_ud(0x1));
380 1, /* response length */
384 brw_set_predicate_control(p
, BRW_PREDICATE_NONE
);
388 void brw_clip_init_ff_sync(struct brw_clip_compile
*c
)
390 struct intel_context
*intel
= &c
->func
.brw
->intel
;
392 if (intel
->needs_ff_sync
) {
393 struct brw_compile
*p
= &c
->func
;
395 brw_MOV(p
, c
->reg
.ff_sync
, brw_imm_ud(0));