i965: Move intel_context::is_<platform> flags to brw_context.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_lower_texture_gradients.cpp
1 /*
2 * Copyright © 2012 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 /**
25 * \file brw_lower_texture_gradients.cpp
26 */
27
28 #include "glsl/ir.h"
29 #include "glsl/ir_builder.h"
30 #include "program/prog_instruction.h"
31 #include "brw_context.h"
32
33 using namespace ir_builder;
34
35 class lower_texture_grad_visitor : public ir_hierarchical_visitor {
36 public:
37 lower_texture_grad_visitor(bool has_sample_d_c)
38 : has_sample_d_c(has_sample_d_c)
39 {
40 progress = false;
41 }
42
43 ir_visitor_status visit_leave(ir_texture *ir);
44
45
46 bool progress;
47 bool has_sample_d_c;
48
49 private:
50 void emit(ir_variable *, ir_rvalue *);
51 };
52
53 /**
54 * Emit a variable declaration and an assignment to initialize it.
55 */
56 void
57 lower_texture_grad_visitor::emit(ir_variable *var, ir_rvalue *value)
58 {
59 base_ir->insert_before(var);
60 base_ir->insert_before(assign(var, value));
61 }
62
63 static const glsl_type *
64 txs_type(const glsl_type *type)
65 {
66 unsigned dims;
67 switch (type->sampler_dimensionality) {
68 case GLSL_SAMPLER_DIM_1D:
69 dims = 1;
70 break;
71 case GLSL_SAMPLER_DIM_2D:
72 case GLSL_SAMPLER_DIM_RECT:
73 case GLSL_SAMPLER_DIM_CUBE:
74 dims = 2;
75 break;
76 case GLSL_SAMPLER_DIM_3D:
77 dims = 3;
78 break;
79 default:
80 assert(!"Should not get here: invalid sampler dimensionality");
81 dims = 2;
82 }
83
84 if (type->sampler_array)
85 dims++;
86
87 return glsl_type::get_instance(GLSL_TYPE_INT, dims, 1);
88 }
89
90 ir_visitor_status
91 lower_texture_grad_visitor::visit_leave(ir_texture *ir)
92 {
93 /* Only lower textureGrad with shadow samplers */
94 if (ir->op != ir_txd || !ir->shadow_comparitor)
95 return visit_continue;
96
97 /* Lower textureGrad() with samplerCubeShadow even if we have the sample_d_c
98 * message. GLSL provides gradients for the 'r' coordinate. Unfortunately:
99 *
100 * From the Ivybridge PRM, Volume 4, Part 1, sample_d message description:
101 * "The r coordinate contains the faceid, and the r gradients are ignored
102 * by hardware."
103 *
104 * We likely need to do a similar treatment for samplerCube and
105 * samplerCubeArray, but we have insufficient testing for that at the moment.
106 */
107 bool need_lowering = !has_sample_d_c ||
108 ir->sampler->type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE;
109
110 if (!need_lowering)
111 return visit_continue;
112
113 void *mem_ctx = ralloc_parent(ir);
114
115 const glsl_type *grad_type = ir->lod_info.grad.dPdx->type;
116
117 /* Use textureSize() to get the width and height of LOD 0; swizzle away
118 * the depth/number of array slices.
119 */
120 ir_texture *txs = new(mem_ctx) ir_texture(ir_txs);
121 txs->set_sampler(ir->sampler->clone(mem_ctx, NULL),
122 txs_type(ir->sampler->type));
123 txs->lod_info.lod = new(mem_ctx) ir_constant(0);
124 ir_variable *size =
125 new(mem_ctx) ir_variable(grad_type, "size", ir_var_temporary);
126 if (ir->sampler->type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE) {
127 base_ir->insert_before(size);
128 base_ir->insert_before(assign(size, expr(ir_unop_i2f, txs), WRITEMASK_XY));
129 base_ir->insert_before(assign(size, new(mem_ctx) ir_constant(1.0f), WRITEMASK_Z));
130 } else {
131 emit(size, expr(ir_unop_i2f,
132 swizzle_for_size(txs, grad_type->vector_elements)));
133 }
134
135 /* Scale the gradients by width and height. Effectively, the incoming
136 * gradients are s'(x,y), t'(x,y), and r'(x,y) from equation 3.19 in the
137 * GL 3.0 spec; we want u'(x,y), which is w_t * s'(x,y).
138 */
139 ir_variable *dPdx =
140 new(mem_ctx) ir_variable(grad_type, "dPdx", ir_var_temporary);
141 emit(dPdx, mul(size, ir->lod_info.grad.dPdx));
142
143 ir_variable *dPdy =
144 new(mem_ctx) ir_variable(grad_type, "dPdy", ir_var_temporary);
145 emit(dPdy, mul(size, ir->lod_info.grad.dPdy));
146
147 /* Calculate rho from equation 3.20 of the GL 3.0 specification. */
148 ir_rvalue *rho;
149 if (dPdx->type->is_scalar()) {
150 rho = expr(ir_binop_max, expr(ir_unop_abs, dPdx),
151 expr(ir_unop_abs, dPdy));
152 } else {
153 rho = expr(ir_binop_max, expr(ir_unop_sqrt, dot(dPdx, dPdx)),
154 expr(ir_unop_sqrt, dot(dPdy, dPdy)));
155 }
156
157 /* lambda_base = log2(rho). We're ignoring GL state biases for now. */
158 ir->op = ir_txl;
159 ir->lod_info.lod = expr(ir_unop_log2, rho);
160
161 progress = true;
162 return visit_continue;
163 }
164
165 extern "C" {
166
167 bool
168 brw_lower_texture_gradients(struct brw_context *brw,
169 struct exec_list *instructions)
170 {
171 struct intel_context *intel = &brw->intel;
172 bool has_sample_d_c = intel->gen >= 8 || brw->is_haswell;
173 lower_texture_grad_visitor v(has_sample_d_c);
174
175 visit_list_elements(&v, instructions);
176
177 return v.progress;
178 }
179
180 }