1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
|
/*
Copyright (C) Intel Corp. 2006. All Rights Reserved.
Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
develop this 3D driver.
Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:
The above copyright notice and this permission notice (including the
next paragraph) shall be included in all copies or substantial
portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
**********************************************************************/
/*
* Authors:
* Keith Whitwell <keith@tungstengraphics.com>
*/
#include "brw_context.h"
#include "brw_state.h"
#include "brw_defines.h"
#include "dri_bufmgr.h"
#include "brw_wm.h"
/***********************************************************************
* WM unit - fragment programs and rasterization
*/
struct brw_wm_unit_key {
unsigned int total_grf, total_scratch;
unsigned int urb_entry_read_length;
unsigned int curb_entry_read_length;
unsigned int dispatch_grf_start_reg;
unsigned int curbe_offset;
unsigned int urb_size;
unsigned int max_threads;
unsigned int nr_surfaces, sampler_count;
GLboolean uses_depth, computes_depth, uses_kill, is_glsl;
GLboolean polygon_stipple, stats_wm;
};
static void
wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
{
const struct gl_fragment_program *fp = brw->fragment_program;
struct intel_context *intel = &brw->intel;
memset(key, 0, sizeof(*key));
if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
key->max_threads = 1;
else
key->max_threads = 32;
/* CACHE_NEW_WM_PROG */
key->total_grf = brw->wm.prog_data->total_grf;
key->urb_entry_read_length = brw->wm.prog_data->urb_read_length;
key->curb_entry_read_length = brw->wm.prog_data->curb_read_length;
key->dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf;
key->total_scratch = ALIGN(brw->wm.prog_data->total_scratch, 1024);
/* BRW_NEW_URB_FENCE */
key->urb_size = brw->urb.vsize;
/* BRW_NEW_CURBE_OFFSETS */
key->curbe_offset = brw->curbe.wm_start;
/* CACHE_NEW_SURFACE */
key->nr_surfaces = brw->wm.nr_surfaces;
/* CACHE_NEW_SAMPLER */
key->sampler_count = brw->wm.sampler_count;
/* _NEW_POLYGONSTIPPLE */
key->polygon_stipple = brw->attribs.Polygon->StippleFlag;
/* BRW_NEW_FRAGMENT_PROGRAM */
key->uses_depth = (fp->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS)) != 0;
/* as far as we can tell */
key->computes_depth =
(fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) != 0;
/* _NEW_COLOR */
key->uses_kill = fp->UsesKill || brw->attribs.Color->AlphaEnabled;
key->is_glsl = brw_wm_is_glsl(fp);
/* XXX: This needs a flag to indicate when it changes. */
key->stats_wm = intel->stats_wm;
}
static dri_bo *
wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
dri_bo **reloc_bufs)
{
struct brw_wm_unit_state wm;
dri_bo *bo;
memset(&wm, 0, sizeof(wm));
wm.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
wm.thread0.kernel_start_pointer = brw->wm.prog_bo->offset >> 6; /* reloc */
wm.thread1.depth_coef_urb_read_offset = 1;
wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
wm.thread1.binding_table_entry_count = key->nr_surfaces;
if (key->total_scratch != 0) {
wm.thread2.scratch_space_base_pointer =
brw->wm.scratch_buffer->offset >> 10; /* reloc */
wm.thread2.per_thread_scratch_space = key->total_scratch / 1024 - 1;
} else {
wm.thread2.scratch_space_base_pointer = 0;
wm.thread2.per_thread_scratch_space = 0;
}
wm.thread3.dispatch_grf_start_reg = key->dispatch_grf_start_reg;
wm.thread3.urb_entry_read_length = key->urb_entry_read_length;
wm.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
wm.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
wm.thread3.urb_entry_read_offset = 0;
wm.wm4.sampler_count = (key->sampler_count + 1) / 4;
if (brw->wm.sampler_bo != NULL) {
/* reloc */
wm.wm4.sampler_state_pointer = brw->wm.sampler_bo->offset >> 5;
} else {
wm.wm4.sampler_state_pointer = 0;
}
wm.wm5.program_uses_depth = key->uses_depth;
wm.wm5.program_computes_depth = key->computes_depth;
wm.wm5.program_uses_killpixel = key->uses_kill;
if (key->is_glsl)
wm.wm5.enable_8_pix = 1;
else
wm.wm5.enable_16_pix = 1;
wm.wm5.max_threads = key->max_threads - 1;
wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */
wm.wm5.legacy_line_rast = 0;
wm.wm5.legacy_global_depth_bias = 0;
wm.wm5.early_depth_test = 1; /* never need to disable */
wm.wm5.line_aa_region_width = 0;
wm.wm5.line_endcap_aa_region_width = 1;
wm.wm5.polygon_stipple = key->polygon_stipple;
/* _NEW_POLYGON */
if (brw->attribs.Polygon->OffsetFill) {
wm.wm5.depth_offset = 1;
/* Something wierd going on with legacy_global_depth_bias,
* offset_constant, scaling and MRD. This value passes glean
* but gives some odd results elsewere (eg. the
* quad-offset-units test).
*/
wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2;
/* This is the only value that passes glean:
*/
wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor;
}
/* _NEW_LINE */
if (brw->attribs.Line->StippleFlag) {
wm.wm5.line_stipple = 1;
}
if (INTEL_DEBUG & DEBUG_STATS || key->stats_wm)
wm.wm4.stats_enable = 1;
bo = brw_upload_cache(&brw->cache, BRW_WM_UNIT,
key, sizeof(*key),
reloc_bufs, 3,
&wm, sizeof(wm),
NULL, NULL);
/* Emit WM program relocation */
dri_emit_reloc(bo,
DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
wm.thread0.grf_reg_count << 1,
offsetof(struct brw_wm_unit_state, thread0),
brw->wm.prog_bo);
/* Emit scratch space relocation */
if (key->total_scratch != 0) {
dri_emit_reloc(bo,
DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
wm.thread2.per_thread_scratch_space,
offsetof(struct brw_wm_unit_state, thread2),
brw->wm.scratch_buffer);
}
/* Emit sampler state relocation */
if (key->sampler_count != 0) {
dri_emit_reloc(bo,
DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
wm.wm4.stats_enable | (wm.wm4.sampler_count << 2),
offsetof(struct brw_wm_unit_state, wm4),
brw->wm.sampler_bo);
}
return bo;
}
static void upload_wm_unit( struct brw_context *brw )
{
struct intel_context *intel = &brw->intel;
struct brw_wm_unit_key key;
dri_bo *reloc_bufs[3];
wm_unit_populate_key(brw, &key);
/* Allocate the necessary scratch space if we haven't already. Don't
* bother reducing the allocation later, since we use scratch so
* rarely.
*/
assert(key.total_scratch <= 12 * 1024);
if (key.total_scratch) {
GLuint total = key.total_scratch * key.max_threads;
if (brw->wm.scratch_buffer && total > brw->wm.scratch_buffer->size) {
dri_bo_unreference(brw->wm.scratch_buffer);
brw->wm.scratch_buffer = NULL;
}
if (brw->wm.scratch_buffer == NULL) {
brw->wm.scratch_buffer = dri_bo_alloc(intel->bufmgr,
"wm scratch",
total,
4096, DRM_BO_FLAG_MEM_TT);
}
}
reloc_bufs[0] = brw->wm.prog_bo;
reloc_bufs[1] = brw->wm.scratch_buffer;
reloc_bufs[2] = brw->wm.sampler_bo;
dri_bo_unreference(brw->wm.state_bo);
brw->wm.state_bo = brw_search_cache(&brw->cache, BRW_WM_UNIT,
&key, sizeof(key),
reloc_bufs, 3,
NULL);
if (brw->wm.state_bo == NULL) {
brw->wm.state_bo = wm_unit_create_from_key(brw, &key, reloc_bufs);
}
}
const struct brw_tracked_state brw_wm_unit = {
.dirty = {
.mesa = (_NEW_POLYGON |
_NEW_POLYGONSTIPPLE |
_NEW_LINE |
_NEW_COLOR),
.brw = (BRW_NEW_FRAGMENT_PROGRAM |
BRW_NEW_CURBE_OFFSETS |
BRW_NEW_LOCK),
.cache = (CACHE_NEW_SURFACE |
CACHE_NEW_WM_PROG |
CACHE_NEW_SAMPLER)
},
.update = upload_wm_unit,
};
|
