i965: Move must_use/has_separate_stencil fields to brw_context.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_clip_state.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 "brw_context.h"
33 #include "brw_state.h"
34 #include "brw_defines.h"
35
36 static void
37 upload_clip_vp(struct brw_context *brw)
38 {
39 struct intel_context *intel = &brw->intel;
40 struct gl_context *ctx = &intel->ctx;
41 struct brw_clipper_viewport *vp;
42
43 vp = brw_state_batch(brw, AUB_TRACE_CLIP_VP_STATE,
44 sizeof(*vp), 32, &brw->clip.vp_offset);
45
46 const float maximum_post_clamp_delta = 4096;
47 float gbx = maximum_post_clamp_delta / (float) ctx->Viewport.Width;
48 float gby = maximum_post_clamp_delta / (float) ctx->Viewport.Height;
49
50 vp->xmin = -gbx;
51 vp->xmax = gbx;
52 vp->ymin = -gby;
53 vp->ymax = gby;
54 }
55
56 static void
57 brw_upload_clip_unit(struct brw_context *brw)
58 {
59 struct intel_context *intel = &brw->intel;
60 struct gl_context *ctx = &intel->ctx;
61 struct brw_clip_unit_state *clip;
62
63 /* _NEW_BUFFERS */
64 struct gl_framebuffer *fb = ctx->DrawBuffer;
65
66 upload_clip_vp(brw);
67
68 clip = brw_state_batch(brw, AUB_TRACE_CLIP_STATE,
69 sizeof(*clip), 32, &brw->clip.state_offset);
70 memset(clip, 0, sizeof(*clip));
71
72 /* BRW_NEW_PROGRAM_CACHE | CACHE_NEW_CLIP_PROG */
73 clip->thread0.grf_reg_count = (ALIGN(brw->clip.prog_data->total_grf, 16) /
74 16 - 1);
75 clip->thread0.kernel_start_pointer =
76 brw_program_reloc(brw,
77 brw->clip.state_offset +
78 offsetof(struct brw_clip_unit_state, thread0),
79 brw->clip.prog_offset +
80 (clip->thread0.grf_reg_count << 1)) >> 6;
81
82 clip->thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
83 clip->thread1.single_program_flow = 1;
84
85 clip->thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length;
86 clip->thread3.const_urb_entry_read_length =
87 brw->clip.prog_data->curb_read_length;
88
89 /* BRW_NEW_CURBE_OFFSETS */
90 clip->thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2;
91 clip->thread3.dispatch_grf_start_reg = 1;
92 clip->thread3.urb_entry_read_offset = 0;
93
94 /* BRW_NEW_URB_FENCE */
95 clip->thread4.nr_urb_entries = brw->urb.nr_clip_entries;
96 clip->thread4.urb_entry_allocation_size = brw->urb.vsize - 1;
97 /* If we have enough clip URB entries to run two threads, do so.
98 */
99 if (brw->urb.nr_clip_entries >= 10) {
100 /* Half of the URB entries go to each thread, and it has to be an
101 * even number.
102 */
103 assert(brw->urb.nr_clip_entries % 2 == 0);
104
105 /* Although up to 16 concurrent Clip threads are allowed on Ironlake,
106 * only 2 threads can output VUEs at a time.
107 */
108 if (intel->gen == 5)
109 clip->thread4.max_threads = 16 - 1;
110 else
111 clip->thread4.max_threads = 2 - 1;
112 } else {
113 assert(brw->urb.nr_clip_entries >= 5);
114 clip->thread4.max_threads = 1 - 1;
115 }
116
117 if (unlikely(INTEL_DEBUG & DEBUG_STATS))
118 clip->thread4.stats_enable = 1;
119
120 clip->clip5.userclip_enable_flags = 0x7f;
121 clip->clip5.userclip_must_clip = 1;
122
123 /* enable guardband clipping if we can */
124 if (ctx->Viewport.X == 0 &&
125 ctx->Viewport.Y == 0 &&
126 ctx->Viewport.Width == fb->Width &&
127 ctx->Viewport.Height == fb->Height)
128 {
129 clip->clip5.guard_band_enable = 1;
130 clip->clip6.clipper_viewport_state_ptr =
131 (brw->batch.bo->offset + brw->clip.vp_offset) >> 5;
132
133 /* emit clip viewport relocation */
134 drm_intel_bo_emit_reloc(brw->batch.bo,
135 (brw->clip.state_offset +
136 offsetof(struct brw_clip_unit_state, clip6)),
137 brw->batch.bo, brw->clip.vp_offset,
138 I915_GEM_DOMAIN_INSTRUCTION, 0);
139 }
140
141 /* _NEW_TRANSFORM */
142 if (!ctx->Transform.DepthClamp)
143 clip->clip5.viewport_z_clip_enable = 1;
144 clip->clip5.viewport_xy_clip_enable = 1;
145 clip->clip5.vertex_position_space = BRW_CLIP_NDCSPACE;
146 clip->clip5.api_mode = BRW_CLIP_API_OGL;
147 clip->clip5.clip_mode = brw->clip.prog_data->clip_mode;
148
149 if (intel->is_g4x)
150 clip->clip5.negative_w_clip_test = 1;
151
152 clip->viewport_xmin = -1;
153 clip->viewport_xmax = 1;
154 clip->viewport_ymin = -1;
155 clip->viewport_ymax = 1;
156
157 brw->state.dirty.cache |= CACHE_NEW_CLIP_UNIT;
158 }
159
160 const struct brw_tracked_state brw_clip_unit = {
161 .dirty = {
162 .mesa = _NEW_TRANSFORM | _NEW_BUFFERS | _NEW_VIEWPORT,
163 .brw = (BRW_NEW_BATCH |
164 BRW_NEW_PROGRAM_CACHE |
165 BRW_NEW_CURBE_OFFSETS |
166 BRW_NEW_URB_FENCE),
167 .cache = CACHE_NEW_CLIP_PROG
168 },
169 .emit = brw_upload_clip_unit,
170 };