2 * Copyright © 2012 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
25 #include "program/program.h"
26 #include "brw_state.h"
27 #include "brw_defines.h"
29 #include "intel_batchbuffer.h"
32 gen8_upload_ps_extra(struct brw_context
*brw
,
33 const struct gl_fragment_program
*fp
,
34 const struct brw_wm_prog_data
*prog_data
,
35 bool multisampled_fbo
)
37 struct gl_context
*ctx
= &brw
->ctx
;
40 dw1
|= GEN8_PSX_PIXEL_SHADER_VALID
;
41 dw1
|= prog_data
->computed_depth_mode
<< GEN8_PSX_COMPUTED_DEPTH_MODE_SHIFT
;
43 if (prog_data
->uses_kill
)
44 dw1
|= GEN8_PSX_KILL_ENABLE
;
46 if (prog_data
->num_varying_inputs
!= 0)
47 dw1
|= GEN8_PSX_ATTRIBUTE_ENABLE
;
49 if (fp
->Base
.InputsRead
& VARYING_BIT_POS
)
50 dw1
|= GEN8_PSX_USES_SOURCE_DEPTH
| GEN8_PSX_USES_SOURCE_W
;
52 if (multisampled_fbo
&&
53 _mesa_get_min_invocations_per_fragment(ctx
, fp
, false) > 1)
54 dw1
|= GEN8_PSX_SHADER_IS_PER_SAMPLE
;
56 if (fp
->Base
.SystemValuesRead
& SYSTEM_BIT_SAMPLE_MASK_IN
) {
58 dw1
|= BRW_PSICMS_INNER
<< GEN9_PSX_SHADER_NORMAL_COVERAGE_MASK_SHIFT
;
60 dw1
|= GEN8_PSX_SHADER_USES_INPUT_COVERAGE_MASK
;
63 if (prog_data
->uses_omask
)
64 dw1
|= GEN8_PSX_OMASK_TO_RENDER_TARGET
;
66 if (brw
->gen
>= 9 && prog_data
->pulls_bary
)
67 dw1
|= GEN9_PSX_SHADER_PULLS_BARY
;
69 /* The stricter cross-primitive coherency guarantees that the hardware
70 * gives us with the "Accesses UAV" bit set for at least one shader stage
71 * and the "UAV coherency required" bit set on the 3DPRIMITIVE command are
72 * redundant within the current image, atomic counter and SSBO GL APIs,
73 * which all have very loose ordering and coherency requirements and
74 * generally rely on the application to insert explicit barriers when a
75 * shader invocation is expected to see the memory writes performed by the
76 * invocations of some previous primitive. Regardless of the value of "UAV
77 * coherency required", the "Accesses UAV" bits will implicitly cause an in
78 * most cases useless DC flush when the lowermost stage with the bit set
81 * It would be nice to disable it, but in some cases we can't because on
82 * Gen8+ it also has an influence on rasterization via the PS UAV-only
83 * signal (which could be set independently from the coherency mechanism in
84 * the 3DSTATE_WM command on Gen7), and because in some cases it will
85 * determine whether the hardware skips execution of the fragment shader or
86 * not via the ThreadDispatchEnable signal. However if we know that
87 * GEN8_PS_BLEND_HAS_WRITEABLE_RT is going to be set and
88 * GEN8_PSX_PIXEL_SHADER_NO_RT_WRITE is not set it shouldn't make any
89 * difference so we may just disable it here.
91 * BRW_NEW_FS_PROG_DATA | BRW_NEW_FRAGMENT_PROGRAM | _NEW_BUFFERS | _NEW_COLOR
93 if ((_mesa_active_fragment_shader_has_atomic_ops(&brw
->ctx
) ||
94 prog_data
->base
.nr_image_params
) &&
95 !brw_color_buffer_write_enabled(brw
))
96 dw1
|= GEN8_PSX_SHADER_HAS_UAV
;
99 OUT_BATCH(_3DSTATE_PS_EXTRA
<< 16 | (2 - 2));
105 upload_ps_extra(struct brw_context
*brw
)
107 /* BRW_NEW_FRAGMENT_PROGRAM */
108 const struct brw_fragment_program
*fp
=
109 brw_fragment_program_const(brw
->fragment_program
);
110 /* BRW_NEW_FS_PROG_DATA */
111 const struct brw_wm_prog_data
*prog_data
= brw
->wm
.prog_data
;
112 /* BRW_NEW_NUM_SAMPLES */
113 const bool multisampled_fbo
= brw
->num_samples
> 1;
115 gen8_upload_ps_extra(brw
, &fp
->program
, prog_data
, multisampled_fbo
);
118 const struct brw_tracked_state gen8_ps_extra
= {
120 .mesa
= _NEW_BUFFERS
| _NEW_COLOR
,
121 .brw
= BRW_NEW_CONTEXT
|
122 BRW_NEW_FRAGMENT_PROGRAM
|
123 BRW_NEW_FS_PROG_DATA
|
126 .emit
= upload_ps_extra
,
130 upload_wm_state(struct brw_context
*brw
)
132 struct gl_context
*ctx
= &brw
->ctx
;
135 dw1
|= GEN7_WM_STATISTICS_ENABLE
;
136 dw1
|= GEN7_WM_LINE_AA_WIDTH_1_0
;
137 dw1
|= GEN7_WM_LINE_END_CAP_AA_WIDTH_0_5
;
138 dw1
|= GEN7_WM_POINT_RASTRULE_UPPER_RIGHT
;
141 if (ctx
->Line
.StippleFlag
)
142 dw1
|= GEN7_WM_LINE_STIPPLE_ENABLE
;
145 if (ctx
->Polygon
.StippleFlag
)
146 dw1
|= GEN7_WM_POLYGON_STIPPLE_ENABLE
;
148 /* BRW_NEW_FS_PROG_DATA */
149 dw1
|= brw
->wm
.prog_data
->barycentric_interp_modes
<<
150 GEN7_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT
;
152 /* BRW_NEW_FS_PROG_DATA */
153 if (brw
->wm
.prog_data
->early_fragment_tests
)
154 dw1
|= GEN7_WM_EARLY_DS_CONTROL_PREPS
;
155 else if (brw
->wm
.prog_data
->base
.nr_image_params
)
156 dw1
|= GEN7_WM_EARLY_DS_CONTROL_PSEXEC
;
159 OUT_BATCH(_3DSTATE_WM
<< 16 | (2 - 2));
164 const struct brw_tracked_state gen8_wm_state
= {
168 .brw
= BRW_NEW_CONTEXT
|
169 BRW_NEW_FS_PROG_DATA
,
171 .emit
= upload_wm_state
,
175 gen8_upload_ps_state(struct brw_context
*brw
,
176 const struct gl_fragment_program
*fp
,
177 const struct brw_stage_state
*stage_state
,
178 const struct brw_wm_prog_data
*prog_data
,
179 uint32_t fast_clear_op
)
181 struct gl_context
*ctx
= &brw
->ctx
;
182 uint32_t dw3
= 0, dw6
= 0, dw7
= 0, ksp0
, ksp2
= 0;
184 /* Initialize the execution mask with VMask. Otherwise, derivatives are
185 * incorrect for subspans where some of the pixels are unlit. We believe
186 * the bit just didn't take effect in previous generations.
188 dw3
|= GEN7_PS_VECTOR_MASK_ENABLE
;
190 const unsigned sampler_count
=
191 DIV_ROUND_UP(CLAMP(stage_state
->sampler_count
, 0, 16), 4);
192 dw3
|= SET_FIELD(sampler_count
, GEN7_PS_SAMPLER_COUNT
);
194 /* BRW_NEW_FS_PROG_DATA */
196 ((prog_data
->base
.binding_table
.size_bytes
/ 4) <<
197 GEN7_PS_BINDING_TABLE_ENTRY_COUNT_SHIFT
);
199 if (prog_data
->base
.use_alt_mode
)
200 dw3
|= GEN7_PS_FLOATING_POINT_MODE_ALT
;
202 /* 3DSTATE_PS expects the number of threads per PSD, which is always 64;
203 * it implicitly scales for different GT levels (which have some # of PSDs).
205 * In Gen8 the format is U8-2 whereas in Gen9 it is U8-1.
208 dw6
|= (64 - 1) << HSW_PS_MAX_THREADS_SHIFT
;
210 dw6
|= (64 - 2) << HSW_PS_MAX_THREADS_SHIFT
;
212 if (prog_data
->base
.nr_params
> 0)
213 dw6
|= GEN7_PS_PUSH_CONSTANT_ENABLE
;
215 /* From the documentation for this packet:
216 * "If the PS kernel does not need the Position XY Offsets to
217 * compute a Position Value, then this field should be programmed
218 * to POSOFFSET_NONE."
220 * "SW Recommendation: If the PS kernel needs the Position Offsets
221 * to compute a Position XY value, this field should match Position
222 * ZW Interpolation Mode to ensure a consistent position.xyzw
225 * We only require XY sample offsets. So, this recommendation doesn't
226 * look useful at the moment. We might need this in future.
228 if (prog_data
->uses_pos_offset
)
229 dw6
|= GEN7_PS_POSOFFSET_SAMPLE
;
231 dw6
|= GEN7_PS_POSOFFSET_NONE
;
233 dw6
|= fast_clear_op
;
236 * In case of non 1x per sample shading, only one of SIMD8 and SIMD16
237 * should be enabled. We do 'SIMD16 only' dispatch if a SIMD16 shader
238 * is successfully compiled. In majority of the cases that bring us
239 * better performance than 'SIMD8 only' dispatch.
241 int min_invocations_per_fragment
=
242 _mesa_get_min_invocations_per_fragment(ctx
, fp
, false);
243 assert(min_invocations_per_fragment
>= 1);
245 if (prog_data
->prog_offset_16
|| prog_data
->no_8
) {
246 dw6
|= GEN7_PS_16_DISPATCH_ENABLE
;
247 if (!prog_data
->no_8
&& min_invocations_per_fragment
== 1) {
248 dw6
|= GEN7_PS_8_DISPATCH_ENABLE
;
249 dw7
|= (prog_data
->base
.dispatch_grf_start_reg
<<
250 GEN7_PS_DISPATCH_START_GRF_SHIFT_0
);
251 dw7
|= (prog_data
->dispatch_grf_start_reg_16
<<
252 GEN7_PS_DISPATCH_START_GRF_SHIFT_2
);
253 ksp0
= stage_state
->prog_offset
;
254 ksp2
= stage_state
->prog_offset
+ prog_data
->prog_offset_16
;
256 dw7
|= (prog_data
->dispatch_grf_start_reg_16
<<
257 GEN7_PS_DISPATCH_START_GRF_SHIFT_0
);
259 ksp0
= stage_state
->prog_offset
+ prog_data
->prog_offset_16
;
262 dw6
|= GEN7_PS_8_DISPATCH_ENABLE
;
263 dw7
|= (prog_data
->base
.dispatch_grf_start_reg
<<
264 GEN7_PS_DISPATCH_START_GRF_SHIFT_0
);
265 ksp0
= stage_state
->prog_offset
;
269 OUT_BATCH(_3DSTATE_PS
<< 16 | (12 - 2));
273 if (prog_data
->base
.total_scratch
) {
274 OUT_RELOC64(stage_state
->scratch_bo
,
275 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
276 ffs(prog_data
->base
.total_scratch
) - 11);
283 OUT_BATCH(0); /* kernel 1 pointer */
291 upload_ps_state(struct brw_context
*brw
)
293 /* BRW_NEW_FS_PROG_DATA */
294 const struct brw_wm_prog_data
*prog_data
= brw
->wm
.prog_data
;
295 gen8_upload_ps_state(brw
, brw
->fragment_program
, &brw
->wm
.base
, prog_data
,
296 brw
->wm
.fast_clear_op
);
299 const struct brw_tracked_state gen8_ps_state
= {
301 .mesa
= _NEW_MULTISAMPLE
,
302 .brw
= BRW_NEW_BATCH
|
303 BRW_NEW_FRAGMENT_PROGRAM
|
304 BRW_NEW_FS_PROG_DATA
,
306 .emit
= upload_ps_state
,