Merge commit 'origin/gallium-0.1'
[mesa.git] / src / mesa / drivers / dri / i965 / brw_clip_line.c
1 /*
2 Copyright (C) Intel Corp. 2006. All Rights Reserved.
3 Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
4 develop this 3D driver.
5
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:
13
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.
17
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.
25
26 **********************************************************************/
27 /*
28 * Authors:
29 * Keith Whitwell <keith@tungstengraphics.com>
30 */
31
32 #include "main/glheader.h"
33 #include "main/macros.h"
34 #include "main/enums.h"
35 #include "shader/program.h"
36
37 #include "intel_batchbuffer.h"
38
39 #include "brw_defines.h"
40 #include "brw_context.h"
41 #include "brw_eu.h"
42 #include "brw_util.h"
43 #include "brw_clip.h"
44
45
46
47 static void brw_clip_line_alloc_regs( struct brw_clip_compile *c )
48 {
49 GLuint i = 0,j;
50
51 /* Register usage is static, precompute here:
52 */
53 c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;
54
55 if (c->key.nr_userclip) {
56 c->reg.fixed_planes = brw_vec4_grf(i, 0);
57 i += (6 + c->key.nr_userclip + 1) / 2;
58
59 c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2;
60 }
61 else
62 c->prog_data.curb_read_length = 0;
63
64
65 /* Payload vertices plus space for more generated vertices:
66 */
67 for (j = 0; j < 4; j++) {
68 c->reg.vertex[j] = brw_vec4_grf(i, 0);
69 i += c->nr_regs;
70 }
71
72 c->reg.t = brw_vec1_grf(i, 0);
73 c->reg.t0 = brw_vec1_grf(i, 1);
74 c->reg.t1 = brw_vec1_grf(i, 2);
75 c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD);
76 c->reg.plane_equation = brw_vec4_grf(i, 4);
77 i++;
78
79 c->reg.dp0 = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */
80 c->reg.dp1 = brw_vec1_grf(i, 4);
81 i++;
82
83 if (!c->key.nr_userclip) {
84 c->reg.fixed_planes = brw_vec8_grf(i, 0);
85 i++;
86 }
87
88
89 c->first_tmp = i;
90 c->last_tmp = i;
91
92 c->prog_data.urb_read_length = c->nr_regs; /* ? */
93 c->prog_data.total_grf = i;
94 }
95
96
97
98 /* Line clipping, more or less following the following algorithm:
99 *
100 * for (p=0;p<MAX_PLANES;p++) {
101 * if (clipmask & (1 << p)) {
102 * GLfloat dp0 = DOTPROD( vtx0, plane[p] );
103 * GLfloat dp1 = DOTPROD( vtx1, plane[p] );
104 *
105 * if (IS_NEGATIVE(dp1)) {
106 * GLfloat t = dp1 / (dp1 - dp0);
107 * if (t > t1) t1 = t;
108 * } else {
109 * GLfloat t = dp0 / (dp0 - dp1);
110 * if (t > t0) t0 = t;
111 * }
112 *
113 * if (t0 + t1 >= 1.0)
114 * return;
115 * }
116 * }
117 *
118 * interp( ctx, newvtx0, vtx0, vtx1, t0 );
119 * interp( ctx, newvtx1, vtx1, vtx0, t1 );
120 *
121 */
122 static void clip_and_emit_line( struct brw_clip_compile *c )
123 {
124 struct brw_compile *p = &c->func;
125 struct brw_indirect vtx0 = brw_indirect(0, 0);
126 struct brw_indirect vtx1 = brw_indirect(1, 0);
127 struct brw_indirect newvtx0 = brw_indirect(2, 0);
128 struct brw_indirect newvtx1 = brw_indirect(3, 0);
129 struct brw_indirect plane_ptr = brw_indirect(4, 0);
130 struct brw_instruction *plane_loop;
131 struct brw_instruction *plane_active;
132 struct brw_instruction *is_negative;
133 struct brw_instruction *is_neg2;
134 struct brw_instruction *not_culled;
135 struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD);
136
137 brw_MOV(p, get_addr_reg(vtx0), brw_address(c->reg.vertex[0]));
138 brw_MOV(p, get_addr_reg(vtx1), brw_address(c->reg.vertex[1]));
139 brw_MOV(p, get_addr_reg(newvtx0), brw_address(c->reg.vertex[2]));
140 brw_MOV(p, get_addr_reg(newvtx1), brw_address(c->reg.vertex[3]));
141 brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c));
142
143 /* Note: init t0, t1 together:
144 */
145 brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0));
146
147 brw_clip_init_planes(c);
148 brw_clip_init_clipmask(c);
149
150 /* -ve rhw workaround */
151 if (!BRW_IS_G4X(p->brw)) {
152 brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
153 brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
154 brw_imm_ud(1<<20));
155 brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f));
156 }
157
158 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
159
160 plane_loop = brw_DO(p, BRW_EXECUTE_1);
161 {
162 /* if (planemask & 1)
163 */
164 brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
165 brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1));
166
167 plane_active = brw_IF(p, BRW_EXECUTE_1);
168 {
169 if (c->key.nr_userclip)
170 brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0));
171 else
172 brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0));
173
174 /* dp = DP4(vtx->position, plane)
175 */
176 brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
177
178 /* if (IS_NEGATIVE(dp1))
179 */
180 brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
181 brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
182 is_negative = brw_IF(p, BRW_EXECUTE_1);
183 {
184 /*
185 * Both can be negative on GM965/G965 due to RHW workaround
186 * if so, this object should be rejected.
187 */
188 if (!BRW_IS_G4X(p->brw)) {
189 brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_LE, c->reg.dp0, brw_imm_f(0.0));
190 is_neg2 = brw_IF(p, BRW_EXECUTE_1);
191 {
192 brw_clip_kill_thread(c);
193 }
194 brw_ENDIF(p, is_neg2);
195 }
196
197 brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0));
198 brw_math_invert(p, c->reg.t, c->reg.t);
199 brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1);
200
201 brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 );
202 brw_MOV(p, c->reg.t1, c->reg.t);
203 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
204 }
205 is_negative = brw_ELSE(p, is_negative);
206 {
207 /* Coming back in. We know that both cannot be negative
208 * because the line would have been culled in that case.
209 */
210
211 /* If both are positive, do nothing */
212 /* Only on GM965/G965 */
213 if (!BRW_IS_G4X(p->brw)) {
214 brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0));
215 is_neg2 = brw_IF(p, BRW_EXECUTE_1);
216 }
217
218 {
219 brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1));
220 brw_math_invert(p, c->reg.t, c->reg.t);
221 brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0);
222
223 brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 );
224 brw_MOV(p, c->reg.t0, c->reg.t);
225 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
226 }
227
228 if (!BRW_IS_G4X(p->brw)) {
229 brw_ENDIF(p, is_neg2);
230 }
231 }
232 brw_ENDIF(p, is_negative);
233 }
234 brw_ENDIF(p, plane_active);
235
236 /* plane_ptr++;
237 */
238 brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c));
239
240 /* while (planemask>>=1) != 0
241 */
242 brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
243 brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1));
244 }
245 brw_WHILE(p, plane_loop);
246
247 brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1);
248 brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0));
249 not_culled = brw_IF(p, BRW_EXECUTE_1);
250 {
251 brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, GL_FALSE);
252 brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, GL_FALSE);
253
254 brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START);
255 brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END);
256 }
257 brw_ENDIF(p, not_culled);
258 brw_clip_kill_thread(c);
259 }
260
261
262
263 void brw_emit_line_clip( struct brw_clip_compile *c )
264 {
265 brw_clip_line_alloc_regs(c);
266
267 if (c->key.do_flat_shading)
268 brw_clip_copy_colors(c, 0, 1);
269
270 clip_and_emit_line(c);
271 }