55a1acd563db98b1d1761f0c86119a630154024c
[mesa.git] / src / mesa / drivers / dri / i965 / gen7_wm_state.c
1 /*
2 * Copyright © 2011 Intel Corporation
3 *
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:
10 *
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
13 * Software.
14 *
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
21 * IN THE SOFTWARE.
22 */
23
24 #include <stdbool.h>
25 #include "brw_context.h"
26 #include "brw_state.h"
27 #include "brw_defines.h"
28 #include "brw_util.h"
29 #include "brw_wm.h"
30 #include "program/program.h"
31 #include "program/prog_parameter.h"
32 #include "program/prog_statevars.h"
33 #include "intel_batchbuffer.h"
34
35 static void
36 upload_wm_state(struct brw_context *brw)
37 {
38 struct gl_context *ctx = &brw->ctx;
39 /* BRW_NEW_FRAGMENT_PROGRAM */
40 const struct brw_fragment_program *fp =
41 brw_fragment_program_const(brw->fragment_program);
42 /* BRW_NEW_FS_PROG_DATA */
43 const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
44 bool writes_depth = prog_data->computed_depth_mode != BRW_PSCDEPTH_OFF;
45 uint32_t dw1, dw2;
46
47 /* _NEW_BUFFERS */
48 bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
49
50 dw1 = dw2 = 0;
51 dw1 |= GEN7_WM_STATISTICS_ENABLE;
52 dw1 |= GEN7_WM_LINE_AA_WIDTH_1_0;
53 dw1 |= GEN7_WM_LINE_END_CAP_AA_WIDTH_0_5;
54
55 /* _NEW_LINE */
56 if (ctx->Line.StippleFlag)
57 dw1 |= GEN7_WM_LINE_STIPPLE_ENABLE;
58
59 /* _NEW_POLYGON */
60 if (ctx->Polygon.StippleFlag)
61 dw1 |= GEN7_WM_POLYGON_STIPPLE_ENABLE;
62
63 if (fp->program.Base.InputsRead & VARYING_BIT_POS)
64 dw1 |= GEN7_WM_USES_SOURCE_DEPTH | GEN7_WM_USES_SOURCE_W;
65
66 dw1 |= prog_data->computed_depth_mode << GEN7_WM_COMPUTED_DEPTH_MODE_SHIFT;
67 dw1 |= prog_data->barycentric_interp_modes <<
68 GEN7_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT;
69
70 /* _NEW_COLOR, _NEW_MULTISAMPLE */
71 /* Enable if the pixel shader kernel generates and outputs oMask.
72 */
73 if (prog_data->uses_kill || ctx->Color.AlphaEnabled ||
74 ctx->Multisample.SampleAlphaToCoverage ||
75 prog_data->uses_omask) {
76 dw1 |= GEN7_WM_KILL_ENABLE;
77 }
78
79 /* _NEW_BUFFERS | _NEW_COLOR */
80 if (brw_color_buffer_write_enabled(brw) || writes_depth ||
81 dw1 & GEN7_WM_KILL_ENABLE) {
82 dw1 |= GEN7_WM_DISPATCH_ENABLE;
83 }
84 if (multisampled_fbo) {
85 /* _NEW_MULTISAMPLE */
86 if (ctx->Multisample.Enabled)
87 dw1 |= GEN7_WM_MSRAST_ON_PATTERN;
88 else
89 dw1 |= GEN7_WM_MSRAST_OFF_PIXEL;
90
91 if (_mesa_get_min_invocations_per_fragment(ctx, brw->fragment_program, false) > 1)
92 dw2 |= GEN7_WM_MSDISPMODE_PERSAMPLE;
93 else
94 dw2 |= GEN7_WM_MSDISPMODE_PERPIXEL;
95 } else {
96 dw1 |= GEN7_WM_MSRAST_OFF_PIXEL;
97 dw2 |= GEN7_WM_MSDISPMODE_PERSAMPLE;
98 }
99
100 if (fp->program.Base.SystemValuesRead & SYSTEM_BIT_SAMPLE_MASK_IN) {
101 dw1 |= GEN7_WM_USES_INPUT_COVERAGE_MASK;
102 }
103
104 BEGIN_BATCH(3);
105 OUT_BATCH(_3DSTATE_WM << 16 | (3 - 2));
106 OUT_BATCH(dw1);
107 OUT_BATCH(dw2);
108 ADVANCE_BATCH();
109 }
110
111 const struct brw_tracked_state gen7_wm_state = {
112 .dirty = {
113 .mesa = _NEW_BUFFERS |
114 _NEW_COLOR |
115 _NEW_LINE |
116 _NEW_MULTISAMPLE |
117 _NEW_POLYGON,
118 .brw = BRW_NEW_BATCH |
119 BRW_NEW_FRAGMENT_PROGRAM |
120 BRW_NEW_FS_PROG_DATA,
121 },
122 .emit = upload_wm_state,
123 };
124
125 static void
126 upload_ps_state(struct brw_context *brw)
127 {
128 struct gl_context *ctx = &brw->ctx;
129 uint32_t dw2, dw4, dw5, ksp0, ksp2;
130 const int max_threads_shift = brw->is_haswell ?
131 HSW_PS_MAX_THREADS_SHIFT : IVB_PS_MAX_THREADS_SHIFT;
132
133 /* BRW_NEW_FS_PROG_DATA */
134 const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
135
136 dw2 = dw4 = dw5 = ksp2 = 0;
137
138 const unsigned sampler_count =
139 DIV_ROUND_UP(CLAMP(brw->wm.base.sampler_count, 0, 16), 4);
140 dw2 |= SET_FIELD(sampler_count, GEN7_PS_SAMPLER_COUNT);
141
142 dw2 |= ((prog_data->base.binding_table.size_bytes / 4) <<
143 GEN7_PS_BINDING_TABLE_ENTRY_COUNT_SHIFT);
144
145 if (prog_data->base.use_alt_mode)
146 dw2 |= GEN7_PS_FLOATING_POINT_MODE_ALT;
147
148 /* Haswell requires the sample mask to be set in this packet as well as
149 * in 3DSTATE_SAMPLE_MASK; the values should match. */
150 /* _NEW_BUFFERS, _NEW_MULTISAMPLE */
151 if (brw->is_haswell)
152 dw4 |= SET_FIELD(gen6_determine_sample_mask(brw), HSW_PS_SAMPLE_MASK);
153
154 dw4 |= (brw->max_wm_threads - 1) << max_threads_shift;
155
156 if (prog_data->base.nr_params > 0)
157 dw4 |= GEN7_PS_PUSH_CONSTANT_ENABLE;
158
159 /* From the IVB PRM, volume 2 part 1, page 287:
160 * "This bit is inserted in the PS payload header and made available to
161 * the DataPort (either via the message header or via header bypass) to
162 * indicate that oMask data (one or two phases) is included in Render
163 * Target Write messages. If present, the oMask data is used to mask off
164 * samples."
165 */
166 if (prog_data->uses_omask)
167 dw4 |= GEN7_PS_OMASK_TO_RENDER_TARGET;
168
169 /* From the IVB PRM, volume 2 part 1, page 287:
170 * "If the PS kernel does not need the Position XY Offsets to
171 * compute a Position Value, then this field should be programmed
172 * to POSOFFSET_NONE."
173 * "SW Recommendation: If the PS kernel needs the Position Offsets
174 * to compute a Position XY value, this field should match Position
175 * ZW Interpolation Mode to ensure a consistent position.xyzw
176 * computation."
177 * We only require XY sample offsets. So, this recommendation doesn't
178 * look useful at the moment. We might need this in future.
179 */
180 if (prog_data->uses_pos_offset)
181 dw4 |= GEN7_PS_POSOFFSET_SAMPLE;
182 else
183 dw4 |= GEN7_PS_POSOFFSET_NONE;
184
185 /* BRW_NEW_FS_PROG_DATA | _NEW_COLOR
186 *
187 * The hardware wedges if you have this bit set but don't turn on any dual
188 * source blend factors.
189 */
190 if (prog_data->dual_src_blend &&
191 (ctx->Color.BlendEnabled & 1) &&
192 ctx->Color.Blend[0]._UsesDualSrc) {
193 dw4 |= GEN7_PS_DUAL_SOURCE_BLEND_ENABLE;
194 }
195
196 /* BRW_NEW_FS_PROG_DATA */
197 if (prog_data->num_varying_inputs != 0)
198 dw4 |= GEN7_PS_ATTRIBUTE_ENABLE;
199
200 /* In case of non 1x per sample shading, only one of SIMD8 and SIMD16
201 * should be enabled. We do 'SIMD16 only' dispatch if a SIMD16 shader
202 * is successfully compiled. In majority of the cases that bring us
203 * better performance than 'SIMD8 only' dispatch.
204 */
205 int min_inv_per_frag =
206 _mesa_get_min_invocations_per_fragment(ctx, brw->fragment_program, false);
207 assert(min_inv_per_frag >= 1);
208
209 if (prog_data->prog_offset_16 || prog_data->no_8) {
210 dw4 |= GEN7_PS_16_DISPATCH_ENABLE;
211 if (!prog_data->no_8 && min_inv_per_frag == 1) {
212 dw4 |= GEN7_PS_8_DISPATCH_ENABLE;
213 dw5 |= (prog_data->base.dispatch_grf_start_reg <<
214 GEN7_PS_DISPATCH_START_GRF_SHIFT_0);
215 dw5 |= (prog_data->dispatch_grf_start_reg_16 <<
216 GEN7_PS_DISPATCH_START_GRF_SHIFT_2);
217 ksp0 = brw->wm.base.prog_offset;
218 ksp2 = brw->wm.base.prog_offset + prog_data->prog_offset_16;
219 } else {
220 dw5 |= (prog_data->dispatch_grf_start_reg_16 <<
221 GEN7_PS_DISPATCH_START_GRF_SHIFT_0);
222 ksp0 = brw->wm.base.prog_offset + prog_data->prog_offset_16;
223 }
224 }
225 else {
226 dw4 |= GEN7_PS_8_DISPATCH_ENABLE;
227 dw5 |= (prog_data->base.dispatch_grf_start_reg <<
228 GEN7_PS_DISPATCH_START_GRF_SHIFT_0);
229 ksp0 = brw->wm.base.prog_offset;
230 }
231
232 dw4 |= brw->wm.fast_clear_op;
233
234 BEGIN_BATCH(8);
235 OUT_BATCH(_3DSTATE_PS << 16 | (8 - 2));
236 OUT_BATCH(ksp0);
237 OUT_BATCH(dw2);
238 if (prog_data->base.total_scratch) {
239 OUT_RELOC(brw->wm.base.scratch_bo,
240 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
241 ffs(prog_data->base.total_scratch) - 11);
242 } else {
243 OUT_BATCH(0);
244 }
245 OUT_BATCH(dw4);
246 OUT_BATCH(dw5);
247 OUT_BATCH(0); /* kernel 1 pointer */
248 OUT_BATCH(ksp2);
249 ADVANCE_BATCH();
250 }
251
252 const struct brw_tracked_state gen7_ps_state = {
253 .dirty = {
254 .mesa = _NEW_BUFFERS |
255 _NEW_COLOR |
256 _NEW_MULTISAMPLE,
257 .brw = BRW_NEW_BATCH |
258 BRW_NEW_FRAGMENT_PROGRAM |
259 BRW_NEW_FS_PROG_DATA,
260 },
261 .emit = upload_ps_state,
262 };