2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics 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 <keithw@vmware.com>
32 #include "main/glheader.h"
33 #include "main/macros.h"
34 #include "main/enums.h"
35 #include "program/program.h"
37 #include "intel_batchbuffer.h"
39 #include "brw_defines.h"
40 #include "brw_context.h"
46 /* This is performed against the original triangles, so no indirection
50 static void compute_tri_direction( struct brw_clip_compile
*c
)
52 struct brw_codegen
*p
= &c
->func
;
53 struct brw_reg e
= c
->reg
.tmp0
;
54 struct brw_reg f
= c
->reg
.tmp1
;
55 GLuint hpos_offset
= brw_varying_to_offset(&c
->vue_map
, VARYING_SLOT_POS
);
56 struct brw_reg v0
= byte_offset(c
->reg
.vertex
[0], hpos_offset
);
57 struct brw_reg v1
= byte_offset(c
->reg
.vertex
[1], hpos_offset
);
58 struct brw_reg v2
= byte_offset(c
->reg
.vertex
[2], hpos_offset
);
61 struct brw_reg v0n
= get_tmp(c
);
62 struct brw_reg v1n
= get_tmp(c
);
63 struct brw_reg v2n
= get_tmp(c
);
66 * NOTE: We can't modify the original vertex coordinates,
67 * as it may impact further operations.
68 * So, we have to keep normalized coordinates in temp registers.
71 * Try to optimize unnecessary MOV's.
77 brw_clip_project_position(c
, v0n
);
78 brw_clip_project_position(c
, v1n
);
79 brw_clip_project_position(c
, v2n
);
81 /* Calculate the vectors of two edges of the triangle:
83 brw_ADD(p
, e
, v0n
, negate(v2n
));
84 brw_ADD(p
, f
, v1n
, negate(v2n
));
86 /* Take their crossproduct:
88 brw_set_default_access_mode(p
, BRW_ALIGN_16
);
89 brw_MUL(p
, vec4(brw_null_reg()), brw_swizzle(e
, 1,2,0,3), brw_swizzle(f
,2,0,1,3));
90 brw_MAC(p
, vec4(e
), negate(brw_swizzle(e
, 2,0,1,3)), brw_swizzle(f
,1,2,0,3));
91 brw_set_default_access_mode(p
, BRW_ALIGN_1
);
93 brw_MUL(p
, c
->reg
.dir
, c
->reg
.dir
, vec4(e
));
97 static void cull_direction( struct brw_clip_compile
*c
)
99 struct brw_codegen
*p
= &c
->func
;
102 assert (!(c
->key
.fill_ccw
== CLIP_CULL
&&
103 c
->key
.fill_cw
== CLIP_CULL
));
105 if (c
->key
.fill_ccw
== CLIP_CULL
)
106 conditional
= BRW_CONDITIONAL_GE
;
108 conditional
= BRW_CONDITIONAL_L
;
111 vec1(brw_null_reg()),
113 get_element(c
->reg
.dir
, 2),
116 brw_IF(p
, BRW_EXECUTE_1
);
118 brw_clip_kill_thread(c
);
125 static void copy_bfc( struct brw_clip_compile
*c
)
127 struct brw_codegen
*p
= &c
->func
;
130 /* Do we have any colors to copy?
132 if (!(brw_clip_have_varying(c
, VARYING_SLOT_COL0
) &&
133 brw_clip_have_varying(c
, VARYING_SLOT_BFC0
)) &&
134 !(brw_clip_have_varying(c
, VARYING_SLOT_COL1
) &&
135 brw_clip_have_varying(c
, VARYING_SLOT_BFC1
)))
138 /* In some weird degenerate cases we can end up testing the
139 * direction twice, once for culling and once for bfc copying. Oh
140 * well, that's what you get for setting weird GL state.
142 if (c
->key
.copy_bfc_ccw
)
143 conditional
= BRW_CONDITIONAL_GE
;
145 conditional
= BRW_CONDITIONAL_L
;
148 vec1(brw_null_reg()),
150 get_element(c
->reg
.dir
, 2),
153 brw_IF(p
, BRW_EXECUTE_1
);
157 for (i
= 0; i
< 3; i
++) {
158 if (brw_clip_have_varying(c
, VARYING_SLOT_COL0
) &&
159 brw_clip_have_varying(c
, VARYING_SLOT_BFC0
))
161 byte_offset(c
->reg
.vertex
[i
],
162 brw_varying_to_offset(&c
->vue_map
,
164 byte_offset(c
->reg
.vertex
[i
],
165 brw_varying_to_offset(&c
->vue_map
,
166 VARYING_SLOT_BFC0
)));
168 if (brw_clip_have_varying(c
, VARYING_SLOT_COL1
) &&
169 brw_clip_have_varying(c
, VARYING_SLOT_BFC1
))
171 byte_offset(c
->reg
.vertex
[i
],
172 brw_varying_to_offset(&c
->vue_map
,
174 byte_offset(c
->reg
.vertex
[i
],
175 brw_varying_to_offset(&c
->vue_map
,
176 VARYING_SLOT_BFC1
)));
186 GLfloat iz = 1.0 / dir.z;
187 GLfloat ac = dir.x * iz;
188 GLfloat bc = dir.y * iz;
189 offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE;
190 offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor;
193 static void compute_offset( struct brw_clip_compile
*c
)
195 struct brw_codegen
*p
= &c
->func
;
196 struct brw_reg off
= c
->reg
.offset
;
197 struct brw_reg dir
= c
->reg
.dir
;
199 brw_math_invert(p
, get_element(off
, 2), get_element(dir
, 2));
200 brw_MUL(p
, vec2(off
), vec2(dir
), get_element(off
, 2));
203 vec1(brw_null_reg()),
205 brw_abs(get_element(off
, 0)),
206 brw_abs(get_element(off
, 1)));
208 brw_SEL(p
, vec1(off
),
209 brw_abs(get_element(off
, 0)), brw_abs(get_element(off
, 1)));
210 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
212 brw_MUL(p
, vec1(off
), vec1(off
), brw_imm_f(c
->key
.offset_factor
));
213 brw_ADD(p
, vec1(off
), vec1(off
), brw_imm_f(c
->key
.offset_units
));
217 static void merge_edgeflags( struct brw_clip_compile
*c
)
219 struct brw_codegen
*p
= &c
->func
;
220 struct brw_reg tmp0
= get_element_ud(c
->reg
.tmp0
, 0);
222 brw_AND(p
, tmp0
, get_element_ud(c
->reg
.R0
, 2), brw_imm_ud(PRIM_MASK
));
224 vec1(brw_null_reg()),
227 brw_imm_ud(_3DPRIM_POLYGON
));
229 /* Get away with using reg.vertex because we know that this is not
230 * a _3DPRIM_TRISTRIP_REVERSE:
232 brw_IF(p
, BRW_EXECUTE_1
);
234 brw_AND(p
, vec1(brw_null_reg()), get_element_ud(c
->reg
.R0
, 2), brw_imm_ud(1<<8));
235 brw_inst_set_cond_modifier(p
->devinfo
, brw_last_inst
, BRW_CONDITIONAL_EQ
);
236 brw_MOV(p
, byte_offset(c
->reg
.vertex
[0],
237 brw_varying_to_offset(&c
->vue_map
,
240 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
242 brw_AND(p
, vec1(brw_null_reg()), get_element_ud(c
->reg
.R0
, 2), brw_imm_ud(1<<9));
243 brw_inst_set_cond_modifier(p
->devinfo
, brw_last_inst
, BRW_CONDITIONAL_EQ
);
244 brw_MOV(p
, byte_offset(c
->reg
.vertex
[2],
245 brw_varying_to_offset(&c
->vue_map
,
248 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
255 static void apply_one_offset( struct brw_clip_compile
*c
,
256 struct brw_indirect vert
)
258 struct brw_codegen
*p
= &c
->func
;
259 GLuint ndc_offset
= brw_varying_to_offset(&c
->vue_map
,
260 BRW_VARYING_SLOT_NDC
);
261 struct brw_reg z
= deref_1f(vert
, ndc_offset
+
262 2 * type_sz(BRW_REGISTER_TYPE_F
));
264 brw_ADD(p
, z
, z
, vec1(c
->reg
.offset
));
269 /***********************************************************************
270 * Output clipped polygon as an unfilled primitive:
272 static void emit_lines(struct brw_clip_compile
*c
,
275 struct brw_codegen
*p
= &c
->func
;
276 struct brw_indirect v0
= brw_indirect(0, 0);
277 struct brw_indirect v1
= brw_indirect(1, 0);
278 struct brw_indirect v0ptr
= brw_indirect(2, 0);
279 struct brw_indirect v1ptr
= brw_indirect(3, 0);
281 /* Need a separate loop for offset:
284 brw_MOV(p
, c
->reg
.loopcount
, c
->reg
.nr_verts
);
285 brw_MOV(p
, get_addr_reg(v0ptr
), brw_address(c
->reg
.inlist
));
287 brw_DO(p
, BRW_EXECUTE_1
);
289 brw_MOV(p
, get_addr_reg(v0
), deref_1uw(v0ptr
, 0));
290 brw_ADD(p
, get_addr_reg(v0ptr
), get_addr_reg(v0ptr
), brw_imm_uw(2));
292 apply_one_offset(c
, v0
);
294 brw_ADD(p
, c
->reg
.loopcount
, c
->reg
.loopcount
, brw_imm_d(-1));
295 brw_inst_set_cond_modifier(p
->devinfo
, brw_last_inst
, BRW_CONDITIONAL_G
);
298 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
301 /* v1ptr = &inlist[nr_verts]
304 brw_MOV(p
, c
->reg
.loopcount
, c
->reg
.nr_verts
);
305 brw_MOV(p
, get_addr_reg(v0ptr
), brw_address(c
->reg
.inlist
));
306 brw_ADD(p
, get_addr_reg(v1ptr
), get_addr_reg(v0ptr
), retype(c
->reg
.nr_verts
, BRW_REGISTER_TYPE_UW
));
307 brw_ADD(p
, get_addr_reg(v1ptr
), get_addr_reg(v1ptr
), retype(c
->reg
.nr_verts
, BRW_REGISTER_TYPE_UW
));
308 brw_MOV(p
, deref_1uw(v1ptr
, 0), deref_1uw(v0ptr
, 0));
310 brw_DO(p
, BRW_EXECUTE_1
);
312 brw_MOV(p
, get_addr_reg(v0
), deref_1uw(v0ptr
, 0));
313 brw_MOV(p
, get_addr_reg(v1
), deref_1uw(v0ptr
, 2));
314 brw_ADD(p
, get_addr_reg(v0ptr
), get_addr_reg(v0ptr
), brw_imm_uw(2));
316 /* draw edge if edgeflag != 0 */
318 vec1(brw_null_reg()), BRW_CONDITIONAL_NZ
,
319 deref_1f(v0
, brw_varying_to_offset(&c
->vue_map
,
322 brw_IF(p
, BRW_EXECUTE_1
);
324 brw_clip_emit_vue(c
, v0
, BRW_URB_WRITE_ALLOCATE_COMPLETE
,
325 (_3DPRIM_LINESTRIP
<< URB_WRITE_PRIM_TYPE_SHIFT
)
326 | URB_WRITE_PRIM_START
);
327 brw_clip_emit_vue(c
, v1
, BRW_URB_WRITE_ALLOCATE_COMPLETE
,
328 (_3DPRIM_LINESTRIP
<< URB_WRITE_PRIM_TYPE_SHIFT
)
329 | URB_WRITE_PRIM_END
);
333 brw_ADD(p
, c
->reg
.loopcount
, c
->reg
.loopcount
, brw_imm_d(-1));
334 brw_inst_set_cond_modifier(p
->devinfo
, brw_last_inst
, BRW_CONDITIONAL_NZ
);
337 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
342 static void emit_points(struct brw_clip_compile
*c
,
345 struct brw_codegen
*p
= &c
->func
;
347 struct brw_indirect v0
= brw_indirect(0, 0);
348 struct brw_indirect v0ptr
= brw_indirect(2, 0);
350 brw_MOV(p
, c
->reg
.loopcount
, c
->reg
.nr_verts
);
351 brw_MOV(p
, get_addr_reg(v0ptr
), brw_address(c
->reg
.inlist
));
353 brw_DO(p
, BRW_EXECUTE_1
);
355 brw_MOV(p
, get_addr_reg(v0
), deref_1uw(v0ptr
, 0));
356 brw_ADD(p
, get_addr_reg(v0ptr
), get_addr_reg(v0ptr
), brw_imm_uw(2));
358 /* draw if edgeflag != 0
361 vec1(brw_null_reg()), BRW_CONDITIONAL_NZ
,
362 deref_1f(v0
, brw_varying_to_offset(&c
->vue_map
,
365 brw_IF(p
, BRW_EXECUTE_1
);
368 apply_one_offset(c
, v0
);
370 brw_clip_emit_vue(c
, v0
, BRW_URB_WRITE_ALLOCATE_COMPLETE
,
371 (_3DPRIM_POINTLIST
<< URB_WRITE_PRIM_TYPE_SHIFT
)
372 | URB_WRITE_PRIM_START
| URB_WRITE_PRIM_END
);
376 brw_ADD(p
, c
->reg
.loopcount
, c
->reg
.loopcount
, brw_imm_d(-1));
377 brw_inst_set_cond_modifier(p
->devinfo
, brw_last_inst
, BRW_CONDITIONAL_NZ
);
380 brw_inst_set_pred_control(p
->devinfo
, brw_last_inst
, BRW_PREDICATE_NORMAL
);
389 static void emit_primitives( struct brw_clip_compile
*c
,
395 brw_clip_tri_emit_polygon(c
);
399 emit_lines(c
, do_offset
);
403 emit_points(c
, do_offset
);
407 unreachable("not reached");
413 static void emit_unfilled_primitives( struct brw_clip_compile
*c
)
415 struct brw_codegen
*p
= &c
->func
;
417 /* Direction culling has already been done.
419 if (c
->key
.fill_ccw
!= c
->key
.fill_cw
&&
420 c
->key
.fill_ccw
!= CLIP_CULL
&&
421 c
->key
.fill_cw
!= CLIP_CULL
)
424 vec1(brw_null_reg()),
426 get_element(c
->reg
.dir
, 2),
429 brw_IF(p
, BRW_EXECUTE_1
);
431 emit_primitives(c
, c
->key
.fill_ccw
, c
->key
.offset_ccw
);
435 emit_primitives(c
, c
->key
.fill_cw
, c
->key
.offset_cw
);
439 else if (c
->key
.fill_cw
!= CLIP_CULL
) {
440 emit_primitives(c
, c
->key
.fill_cw
, c
->key
.offset_cw
);
442 else if (c
->key
.fill_ccw
!= CLIP_CULL
) {
443 emit_primitives(c
, c
->key
.fill_ccw
, c
->key
.offset_ccw
);
450 static void check_nr_verts( struct brw_clip_compile
*c
)
452 struct brw_codegen
*p
= &c
->func
;
454 brw_CMP(p
, vec1(brw_null_reg()), BRW_CONDITIONAL_L
, c
->reg
.nr_verts
, brw_imm_d(3));
455 brw_IF(p
, BRW_EXECUTE_1
);
457 brw_clip_kill_thread(c
);
463 void brw_emit_unfilled_clip( struct brw_clip_compile
*c
)
465 struct brw_codegen
*p
= &c
->func
;
467 c
->need_direction
= ((c
->key
.offset_ccw
|| c
->key
.offset_cw
) ||
468 (c
->key
.fill_ccw
!= c
->key
.fill_cw
) ||
469 c
->key
.fill_ccw
== CLIP_CULL
||
470 c
->key
.fill_cw
== CLIP_CULL
||
471 c
->key
.copy_bfc_cw
||
472 c
->key
.copy_bfc_ccw
);
474 brw_clip_tri_alloc_regs(c
, 3 + c
->key
.nr_userclip
+ 6);
475 brw_clip_tri_init_vertices(c
);
476 brw_clip_init_ff_sync(c
);
478 assert(brw_clip_have_varying(c
, VARYING_SLOT_EDGE
));
480 if (c
->key
.fill_ccw
== CLIP_CULL
&&
481 c
->key
.fill_cw
== CLIP_CULL
) {
482 brw_clip_kill_thread(c
);
488 /* Need to use the inlist indirection here:
490 if (c
->need_direction
)
491 compute_tri_direction(c
);
493 if (c
->key
.fill_ccw
== CLIP_CULL
||
494 c
->key
.fill_cw
== CLIP_CULL
)
497 if (c
->key
.offset_ccw
||
501 if (c
->key
.copy_bfc_ccw
||
505 /* Need to do this whether we clip or not:
507 if (c
->has_flat_shading
)
508 brw_clip_tri_flat_shade(c
);
510 brw_clip_init_clipmask(c
);
511 brw_CMP(p
, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ
, c
->reg
.planemask
, brw_imm_ud(0));
512 brw_IF(p
, BRW_EXECUTE_1
);
514 brw_clip_init_planes(c
);
520 emit_unfilled_primitives(c
);
521 brw_clip_kill_thread(c
);