2 * Copyright © 2009 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
24 * Eric Anholt <eric@anholt.net>
28 #include "brw_context.h"
29 #include "brw_state.h"
30 #include "brw_defines.h"
33 #include "program/program.h"
34 #include "program/prog_parameter.h"
35 #include "program/prog_statevars.h"
36 #include "main/framebuffer.h"
37 #include "intel_batchbuffer.h"
40 gen6_upload_wm_push_constants(struct brw_context
*brw
)
42 struct brw_stage_state
*stage_state
= &brw
->wm
.base
;
43 /* BRW_NEW_FRAGMENT_PROGRAM */
44 const struct brw_fragment_program
*fp
=
45 brw_fragment_program_const(brw
->fragment_program
);
46 /* BRW_NEW_FS_PROG_DATA */
47 const struct brw_wm_prog_data
*prog_data
= brw
->wm
.prog_data
;
49 gen6_upload_push_constants(brw
, &fp
->program
.Base
, &prog_data
->base
,
50 stage_state
, AUB_TRACE_WM_CONSTANTS
);
53 gen7_upload_constant_state(brw
, &brw
->wm
.base
, true,
54 _3DSTATE_CONSTANT_PS
);
58 const struct brw_tracked_state gen6_wm_push_constants
= {
60 .mesa
= _NEW_PROGRAM_CONSTANTS
,
61 .brw
= BRW_NEW_BATCH
|
63 BRW_NEW_FRAGMENT_PROGRAM
|
64 BRW_NEW_FS_PROG_DATA
|
65 BRW_NEW_PUSH_CONSTANT_ALLOCATION
,
67 .emit
= gen6_upload_wm_push_constants
,
71 gen6_upload_wm_state(struct brw_context
*brw
,
72 const struct brw_wm_prog_data
*prog_data
,
73 const struct brw_stage_state
*stage_state
,
74 bool multisampled_fbo
,
75 bool dual_source_blend_enable
, bool kill_enable
,
76 bool color_buffer_write_enable
, bool msaa_enabled
,
77 bool line_stipple_enable
, bool polygon_stipple_enable
,
78 bool statistic_enable
)
80 uint32_t dw2
, dw4
, dw5
, dw6
, ksp0
, ksp2
;
82 /* We can't fold this into gen6_upload_wm_push_constants(), because
83 * according to the SNB PRM, vol 2 part 1 section 7.2.2
84 * (3DSTATE_CONSTANT_PS [DevSNB]):
86 * "[DevSNB]: This packet must be followed by WM_STATE."
88 if (prog_data
->base
.nr_params
== 0) {
89 /* Disable the push constant buffers. */
91 OUT_BATCH(_3DSTATE_CONSTANT_PS
<< 16 | (5 - 2));
99 OUT_BATCH(_3DSTATE_CONSTANT_PS
<< 16 |
100 GEN6_CONSTANT_BUFFER_0_ENABLE
|
102 /* Pointer to the WM constant buffer. Covered by the set of
103 * state flags from gen6_upload_wm_push_constants.
105 OUT_BATCH(stage_state
->push_const_offset
+
106 stage_state
->push_const_size
- 1);
113 dw2
= dw4
= dw5
= dw6
= ksp2
= 0;
115 if (statistic_enable
)
116 dw4
|= GEN6_WM_STATISTICS_ENABLE
;
118 dw5
|= GEN6_WM_LINE_AA_WIDTH_1_0
;
119 dw5
|= GEN6_WM_LINE_END_CAP_AA_WIDTH_0_5
;
121 if (prog_data
->base
.use_alt_mode
)
122 dw2
|= GEN6_WM_FLOATING_POINT_MODE_ALT
;
124 dw2
|= (ALIGN(stage_state
->sampler_count
, 4) / 4) <<
125 GEN6_WM_SAMPLER_COUNT_SHIFT
;
127 dw2
|= ((prog_data
->base
.binding_table
.size_bytes
/ 4) <<
128 GEN6_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT
);
130 dw5
|= (brw
->max_wm_threads
- 1) << GEN6_WM_MAX_THREADS_SHIFT
;
132 if (prog_data
->dispatch_8
)
133 dw5
|= GEN6_WM_8_DISPATCH_ENABLE
;
135 if (prog_data
->dispatch_16
)
136 dw5
|= GEN6_WM_16_DISPATCH_ENABLE
;
138 dw4
|= prog_data
->base
.dispatch_grf_start_reg
<<
139 GEN6_WM_DISPATCH_START_GRF_SHIFT_0
;
140 dw4
|= prog_data
->dispatch_grf_start_reg_2
<<
141 GEN6_WM_DISPATCH_START_GRF_SHIFT_2
;
143 ksp0
= stage_state
->prog_offset
;
144 ksp2
= stage_state
->prog_offset
+ prog_data
->prog_offset_2
;
146 if (dual_source_blend_enable
)
147 dw5
|= GEN6_WM_DUAL_SOURCE_BLEND_ENABLE
;
149 if (line_stipple_enable
)
150 dw5
|= GEN6_WM_LINE_STIPPLE_ENABLE
;
152 if (polygon_stipple_enable
)
153 dw5
|= GEN6_WM_POLYGON_STIPPLE_ENABLE
;
155 if (prog_data
->uses_src_depth
)
156 dw5
|= GEN6_WM_USES_SOURCE_DEPTH
;
157 if (prog_data
->uses_src_w
)
158 dw5
|= GEN6_WM_USES_SOURCE_W
;
159 if (prog_data
->computed_depth_mode
!= BRW_PSCDEPTH_OFF
)
160 dw5
|= GEN6_WM_COMPUTED_DEPTH
;
161 dw6
|= prog_data
->barycentric_interp_modes
<<
162 GEN6_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT
;
165 dw5
|= GEN6_WM_KILL_ENABLE
;
167 if (color_buffer_write_enable
||
168 dw5
& (GEN6_WM_KILL_ENABLE
| GEN6_WM_COMPUTED_DEPTH
))
169 dw5
|= GEN6_WM_DISPATCH_ENABLE
;
171 /* From the SNB PRM, volume 2 part 1, page 278:
172 * "This bit is inserted in the PS payload header and made available to
173 * the DataPort (either via the message header or via header bypass) to
174 * indicate that oMask data (one or two phases) is included in Render
175 * Target Write messages. If present, the oMask data is used to mask off
178 if (prog_data
->uses_omask
)
179 dw5
|= GEN6_WM_OMASK_TO_RENDER_TARGET
;
181 dw6
|= prog_data
->num_varying_inputs
<<
182 GEN6_WM_NUM_SF_OUTPUTS_SHIFT
;
183 if (multisampled_fbo
) {
185 dw6
|= GEN6_WM_MSRAST_ON_PATTERN
;
187 dw6
|= GEN6_WM_MSRAST_OFF_PIXEL
;
189 if (prog_data
->persample_dispatch
)
190 dw6
|= GEN6_WM_MSDISPMODE_PERSAMPLE
;
192 dw6
|= GEN6_WM_MSDISPMODE_PERPIXEL
;
195 dw6
|= GEN6_WM_MSRAST_OFF_PIXEL
;
196 dw6
|= GEN6_WM_MSDISPMODE_PERSAMPLE
;
199 /* From the SNB PRM, volume 2 part 1, page 281:
200 * "If the PS kernel does not need the Position XY Offsets
201 * to compute a Position XY value, then this field should be
202 * programmed to POSOFFSET_NONE."
204 * "SW Recommendation: If the PS kernel needs the Position Offsets
205 * to compute a Position XY value, this field should match Position
206 * ZW Interpolation Mode to ensure a consistent position.xyzw
208 * We only require XY sample offsets. So, this recommendation doesn't
209 * look useful at the moment. We might need this in future.
211 if (prog_data
->uses_pos_offset
)
212 dw6
|= GEN6_WM_POSOFFSET_SAMPLE
;
214 dw6
|= GEN6_WM_POSOFFSET_NONE
;
217 OUT_BATCH(_3DSTATE_WM
<< 16 | (9 - 2));
220 if (prog_data
->base
.total_scratch
) {
221 OUT_RELOC(stage_state
->scratch_bo
,
222 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
223 ffs(prog_data
->base
.total_scratch
) - 11);
230 OUT_BATCH(0); /* kernel 1 pointer */
236 upload_wm_state(struct brw_context
*brw
)
238 struct gl_context
*ctx
= &brw
->ctx
;
239 /* BRW_NEW_FS_PROG_DATA */
240 const struct brw_wm_prog_data
*prog_data
= brw
->wm
.prog_data
;
243 const bool multisampled_fbo
= _mesa_geometric_samples(ctx
->DrawBuffer
) > 1;
245 /* BRW_NEW_FS_PROG_DATA | _NEW_COLOR */
246 const bool dual_src_blend_enable
= prog_data
->dual_src_blend
&&
247 (ctx
->Color
.BlendEnabled
& 1) &&
248 ctx
->Color
.Blend
[0]._UsesDualSrc
;
250 /* _NEW_COLOR, _NEW_MULTISAMPLE */
251 const bool kill_enable
= prog_data
->uses_kill
|| ctx
->Color
.AlphaEnabled
||
252 ctx
->Multisample
.SampleAlphaToCoverage
||
253 prog_data
->uses_omask
;
255 /* Rendering against the gl-context is always taken into account. */
256 const bool statistic_enable
= true;
258 /* _NEW_LINE | _NEW_POLYGON | _NEW_BUFFERS | _NEW_COLOR |
261 gen6_upload_wm_state(brw
, prog_data
, &brw
->wm
.base
,
263 dual_src_blend_enable
, kill_enable
,
264 brw_color_buffer_write_enabled(brw
),
265 ctx
->Multisample
.Enabled
,
266 ctx
->Line
.StippleFlag
, ctx
->Polygon
.StippleFlag
,
270 const struct brw_tracked_state gen6_wm_state
= {
272 .mesa
= _NEW_BUFFERS
|
277 _NEW_PROGRAM_CONSTANTS
,
278 .brw
= BRW_NEW_BATCH
|
280 BRW_NEW_FS_PROG_DATA
|
281 BRW_NEW_PUSH_CONSTANT_ALLOCATION
,
283 .emit
= upload_wm_state
,