Merge git://proxy01.pd.intel.com:9419/git/mesa/mesa into crestline
[mesa.git] / src / mesa / drivers / dri / i965 / brw_wm_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
33
34 #include "brw_context.h"
35 #include "brw_state.h"
36 #include "brw_defines.h"
37 #include "bufmgr.h"
38
39 /***********************************************************************
40 * WM unit - fragment programs and rasterization
41 */
42
43 static void invalidate_scratch_cb( struct intel_context *intel,
44 void *unused )
45 {
46 /* nothing */
47 }
48
49
50 static void upload_wm_unit(struct brw_context *brw )
51 {
52 struct intel_context *intel = &brw->intel;
53 struct brw_wm_unit_state wm;
54 GLuint max_threads;
55
56 if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
57 max_threads = 0;
58 else
59 max_threads = 31;
60
61
62 memset(&wm, 0, sizeof(wm));
63
64 /* CACHE_NEW_WM_PROG */
65 wm.thread0.grf_reg_count = ((brw->wm.prog_data->total_grf-1) & ~15) / 16;
66 wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6;
67 wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf;
68 wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length;
69 wm.thread3.const_urb_entry_read_length = brw->wm.prog_data->curb_read_length;
70
71 wm.wm5.max_threads = max_threads;
72
73 if (brw->wm.prog_data->total_scratch) {
74 GLuint per_thread = (brw->wm.prog_data->total_scratch + 1023) / 1024;
75 GLuint total = per_thread * (max_threads + 1);
76
77 /* Scratch space -- just have to make sure there is sufficient
78 * allocated for the active program and current number of threads.
79 */
80
81 if (!brw->wm.scratch_buffer) {
82 bmGenBuffers(intel, "wm scratch", 1, &brw->wm.scratch_buffer, 12);
83 bmBufferSetInvalidateCB(intel,
84 brw->wm.scratch_buffer,
85 invalidate_scratch_cb,
86 NULL,
87 GL_FALSE);
88 }
89
90 if (total > brw->wm.scratch_buffer_size) {
91 brw->wm.scratch_buffer_size = total;
92 bmBufferData(intel,
93 brw->wm.scratch_buffer,
94 brw->wm.scratch_buffer_size,
95 NULL,
96 0);
97 }
98
99 assert(per_thread <= 12 * 1024);
100 wm.thread2.per_thread_scratch_space = (per_thread / 1024) - 1;
101
102 /* XXX: could make this dynamic as this is so rarely active:
103 */
104 /* BRW_NEW_LOCK */
105 wm.thread2.scratch_space_base_pointer =
106 bmBufferOffset(intel, brw->wm.scratch_buffer) >> 10;
107 }
108
109 /* CACHE_NEW_SURFACE */
110 wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces;
111
112 /* BRW_NEW_CURBE_OFFSETS */
113 wm.thread3.const_urb_entry_read_offset = brw->curbe.wm_start * 2;
114
115 wm.thread3.urb_entry_read_offset = 0;
116 wm.thread1.depth_coef_urb_read_offset = 1;
117 wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
118
119 /* CACHE_NEW_SAMPLER */
120 wm.wm4.sampler_count = brw->wm.sampler_count;
121 wm.wm4.sampler_state_pointer = brw->wm.sampler_gs_offset >> 5;
122
123 /* BRW_NEW_FRAGMENT_PROGRAM */
124 {
125 const struct gl_fragment_program *fp = brw->fragment_program;
126
127 if (fp->Base.InputsRead & (1<<FRAG_ATTRIB_WPOS))
128 wm.wm5.program_uses_depth = 1; /* as far as we can tell */
129
130 if (fp->Base.OutputsWritten & (1<<FRAG_RESULT_DEPR))
131 wm.wm5.program_computes_depth = 1;
132
133 /* _NEW_COLOR */
134 if (fp->UsesKill ||
135 brw->attribs.Color->AlphaEnabled)
136 wm.wm5.program_uses_killpixel = 1;
137 }
138
139 wm.wm5.enable_16_pix = 1;
140 wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */
141 wm.wm5.legacy_line_rast = 0;
142 wm.wm5.legacy_global_depth_bias = 0;
143 wm.wm5.early_depth_test = 1; /* never need to disable */
144 wm.wm5.line_aa_region_width = 0;
145 wm.wm5.line_endcap_aa_region_width = 1;
146
147 /* _NEW_POLYGONSTIPPLE */
148 if (brw->attribs.Polygon->StippleFlag)
149 wm.wm5.polygon_stipple = 1;
150
151 /* _NEW_POLYGON */
152 if (brw->attribs.Polygon->OffsetFill) {
153 wm.wm5.depth_offset = 1;
154 /* Something wierd going on with legacy_global_depth_bias,
155 * offset_constant, scaling and MRD. This value passes glean
156 * but gives some odd results elsewere (eg. the
157 * quad-offset-units test).
158 */
159 wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2;
160
161 /* This is the only value that passes glean:
162 */
163 wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor;
164 }
165
166 /* _NEW_LINE */
167 if (brw->attribs.Line->StippleFlag) {
168 wm.wm5.line_stipple = 1;
169 }
170
171 if (INTEL_DEBUG & DEBUG_STATS || intel->stats_wm)
172 wm.wm4.stats_enable = 1;
173
174 brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm );
175 }
176
177 const struct brw_tracked_state brw_wm_unit = {
178 .dirty = {
179 .mesa = (_NEW_POLYGON |
180 _NEW_POLYGONSTIPPLE |
181 _NEW_LINE |
182 _NEW_COLOR),
183
184 .brw = (BRW_NEW_FRAGMENT_PROGRAM |
185 BRW_NEW_CURBE_OFFSETS |
186 BRW_NEW_LOCK),
187
188 .cache = (CACHE_NEW_SURFACE |
189 CACHE_NEW_WM_PROG |
190 CACHE_NEW_SAMPLER)
191 },
192 .update = upload_wm_unit
193 };
194