i965/fs: Don't propagate saturate modifiers into partial writes.
[mesa.git] / src / mesa / drivers / dri / i965 / gen7_gs_state.c
1 /*
2 * Copyright © 2013 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
21 * DEALINGS IN THE SOFTWARE.
22 */
23
24 #include "brw_context.h"
25 #include "brw_state.h"
26 #include "brw_defines.h"
27 #include "intel_batchbuffer.h"
28
29
30 static void
31 gen7_upload_gs_push_constants(struct brw_context *brw)
32 {
33 /* BRW_NEW_GEOMETRY_PROGRAM */
34 const struct brw_geometry_program *vp =
35 (struct brw_geometry_program *) brw->geometry_program;
36 if (!vp)
37 return;
38
39 /* CACHE_NEW_GS_PROG */
40 const struct brw_vec4_prog_data *prog_data = &brw->gs.prog_data->base;
41 struct brw_stage_state *stage_state = &brw->gs.base;
42
43 gen6_upload_vec4_push_constants(brw, &vp->program.Base, prog_data,
44 stage_state, AUB_TRACE_VS_CONSTANTS);
45 }
46
47 const struct brw_tracked_state gen7_gs_push_constants = {
48 .dirty = {
49 .mesa = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
50 .brw = (BRW_NEW_BATCH |
51 BRW_NEW_GEOMETRY_PROGRAM),
52 .cache = CACHE_NEW_GS_PROG,
53 },
54 .emit = gen7_upload_gs_push_constants,
55 };
56
57
58 static void
59 upload_gs_state(struct brw_context *brw)
60 {
61 const struct brw_stage_state *stage_state = &brw->gs.base;
62 const int max_threads_shift = brw->is_haswell ?
63 HSW_GS_MAX_THREADS_SHIFT : GEN6_GS_MAX_THREADS_SHIFT;
64 /* BRW_NEW_GEOMETRY_PROGRAM */
65 bool active = brw->geometry_program;
66 /* CACHE_NEW_GS_PROG */
67 const struct brw_vec4_prog_data *prog_data = &brw->gs.prog_data->base;
68
69 /* CACHE_NEW_SAMPLER */
70 BEGIN_BATCH(2);
71 OUT_BATCH(_3DSTATE_SAMPLER_STATE_POINTERS_GS << 16 | (2 - 2));
72 OUT_BATCH(stage_state->sampler_offset);
73 ADVANCE_BATCH();
74
75 gen7_upload_constant_state(brw, stage_state, active, _3DSTATE_CONSTANT_GS);
76
77 /**
78 * From Graphics BSpec: 3D-Media-GPGPU Engine > 3D Pipeline Stages >
79 * Geometry > Geometry Shader > State:
80 *
81 * "Note: Because of corruption in IVB:GT2, software needs to flush the
82 * whole fixed function pipeline when the GS enable changes value in
83 * the 3DSTATE_GS."
84 *
85 * The hardware architects have clarified that in this context "flush the
86 * whole fixed function pipeline" means to emit a PIPE_CONTROL with the "CS
87 * Stall" bit set.
88 */
89 if (!brw->is_haswell && brw->gt == 2 && brw->gs.enabled != active)
90 gen7_emit_cs_stall_flush(brw);
91
92 if (active) {
93 BEGIN_BATCH(7);
94 OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
95 OUT_BATCH(stage_state->prog_offset);
96 OUT_BATCH(((ALIGN(stage_state->sampler_count, 4)/4) <<
97 GEN6_GS_SAMPLER_COUNT_SHIFT) |
98 ((brw->gs.prog_data->base.base.binding_table.size_bytes / 4) <<
99 GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
100
101 if (brw->gs.prog_data->base.total_scratch) {
102 OUT_RELOC(stage_state->scratch_bo,
103 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
104 ffs(brw->gs.prog_data->base.total_scratch) - 11);
105 } else {
106 OUT_BATCH(0);
107 }
108
109 uint32_t dw4 =
110 ((brw->gs.prog_data->output_vertex_size_hwords * 2 - 1) <<
111 GEN7_GS_OUTPUT_VERTEX_SIZE_SHIFT) |
112 (brw->gs.prog_data->output_topology <<
113 GEN7_GS_OUTPUT_TOPOLOGY_SHIFT) |
114 (prog_data->urb_read_length <<
115 GEN6_GS_URB_READ_LENGTH_SHIFT) |
116 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT) |
117 (prog_data->dispatch_grf_start_reg <<
118 GEN6_GS_DISPATCH_START_GRF_SHIFT);
119
120 /* Note: the meaning of the GEN7_GS_REORDER_TRAILING bit changes between
121 * Ivy Bridge and Haswell.
122 *
123 * On Ivy Bridge, setting this bit causes the vertices of a triangle
124 * strip to be delivered to the geometry shader in an order that does
125 * not strictly follow the OpenGL spec, but preserves triangle
126 * orientation. For example, if the vertices are (1, 2, 3, 4, 5), then
127 * the geometry shader sees triangles:
128 *
129 * (1, 2, 3), (2, 4, 3), (3, 4, 5)
130 *
131 * (Clearing the bit is even worse, because it fails to preserve
132 * orientation).
133 *
134 * Triangle strips with adjacency always ordered in a way that preserves
135 * triangle orientation but does not strictly follow the OpenGL spec,
136 * regardless of the setting of this bit.
137 *
138 * On Haswell, both triangle strips and triangle strips with adjacency
139 * are always ordered in a way that preserves triangle orientation.
140 * Setting this bit causes the ordering to strictly follow the OpenGL
141 * spec.
142 *
143 * So in either case we want to set the bit. Unfortunately on Ivy
144 * Bridge this will get the order close to correct but not perfect.
145 */
146 uint32_t dw5 =
147 ((brw->max_gs_threads - 1) << max_threads_shift) |
148 (brw->gs.prog_data->control_data_header_size_hwords <<
149 GEN7_GS_CONTROL_DATA_HEADER_SIZE_SHIFT) |
150 ((brw->gs.prog_data->invocations - 1) <<
151 GEN7_GS_INSTANCE_CONTROL_SHIFT) |
152 (brw->gs.prog_data->dual_instanced_dispatch ?
153 GEN7_GS_DISPATCH_MODE_DUAL_INSTANCE :
154 GEN7_GS_DISPATCH_MODE_DUAL_OBJECT) |
155 GEN6_GS_STATISTICS_ENABLE |
156 (brw->gs.prog_data->include_primitive_id ?
157 GEN7_GS_INCLUDE_PRIMITIVE_ID : 0) |
158 GEN7_GS_REORDER_TRAILING |
159 GEN7_GS_ENABLE;
160 uint32_t dw6 = 0;
161
162 if (brw->is_haswell) {
163 dw6 |= brw->gs.prog_data->control_data_format <<
164 HSW_GS_CONTROL_DATA_FORMAT_SHIFT;
165 } else {
166 dw5 |= brw->gs.prog_data->control_data_format <<
167 IVB_GS_CONTROL_DATA_FORMAT_SHIFT;
168 }
169
170 OUT_BATCH(dw4);
171 OUT_BATCH(dw5);
172 OUT_BATCH(dw6);
173 ADVANCE_BATCH();
174 } else {
175 BEGIN_BATCH(7);
176 OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
177 OUT_BATCH(0); /* prog_bo */
178 OUT_BATCH((0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
179 (0 << GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
180 OUT_BATCH(0); /* scratch space base offset */
181 OUT_BATCH((1 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
182 (0 << GEN6_GS_URB_READ_LENGTH_SHIFT) |
183 GEN7_GS_INCLUDE_VERTEX_HANDLES |
184 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT));
185 OUT_BATCH((0 << GEN6_GS_MAX_THREADS_SHIFT) |
186 GEN6_GS_STATISTICS_ENABLE);
187 OUT_BATCH(0);
188 ADVANCE_BATCH();
189 }
190 brw->gs.enabled = active;
191 }
192
193 const struct brw_tracked_state gen7_gs_state = {
194 .dirty = {
195 .mesa = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
196 .brw = (BRW_NEW_CONTEXT |
197 BRW_NEW_GEOMETRY_PROGRAM |
198 BRW_NEW_GS_BINDING_TABLE |
199 BRW_NEW_BATCH |
200 BRW_NEW_PUSH_CONSTANT_ALLOCATION),
201 .cache = CACHE_NEW_GS_PROG | CACHE_NEW_SAMPLER
202 },
203 .emit = upload_gs_state,
204 };