1/*
2 * Copyright © 2016 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 "nir.h"
25#include "nir_builder.h"
26
27static nir_ssa_def *
28load_frag_coord(const nir_input_attachment_options *options, nir_builder *b)
29{
30   if (options->use_fragcoord_sysval)
31      return nir_load_frag_coord(b);
32
33   nir_variable *pos =
34      nir_find_variable_with_location(b->shader, nir_var_shader_in,
35                                      VARYING_SLOT_POS);
36   if (pos == NULL) {
37      pos = nir_variable_create(b->shader, nir_var_shader_in,
38                                glsl_vec4_type(), NULL);
39      pos->data.location = VARYING_SLOT_POS;
40   }
41   /**
42    * From Vulkan spec:
43    *   "The OriginLowerLeft execution mode must not be used; fragment entry
44    *    points must declare OriginUpperLeft."
45    *
46    * So at this point origin_upper_left should be true
47    */
48   assert(b->shader->info.fs.origin_upper_left == true);
49
50   return nir_load_var(b, pos);
51}
52
53static nir_ssa_def *
54load_layer_id(const nir_input_attachment_options *options, nir_builder *b)
55{
56   if (options->use_layer_id_sysval) {
57      if (options->use_view_id_for_layer)
58         return nir_load_view_index(b);
59      else
60         return nir_load_layer_id(b);
61   }
62
63   gl_varying_slot slot = options->use_view_id_for_layer ?
64      VARYING_SLOT_VIEW_INDEX : VARYING_SLOT_LAYER;
65   nir_variable *layer_id =
66      nir_find_variable_with_location(b->shader, nir_var_shader_in, slot);
67
68   if (layer_id == NULL) {
69      layer_id = nir_variable_create(b->shader, nir_var_shader_in,
70                                     glsl_int_type(), NULL);
71      layer_id->data.location = slot;
72      layer_id->data.interpolation = INTERP_MODE_FLAT;
73      layer_id->data.driver_location = b->shader->num_inputs++;
74   }
75
76   return nir_load_var(b, layer_id);
77}
78
79static bool
80try_lower_input_load(nir_function_impl *impl, nir_intrinsic_instr *load,
81                     const nir_input_attachment_options *options)
82{
83   nir_deref_instr *deref = nir_src_as_deref(load->src[0]);
84   assert(glsl_type_is_image(deref->type));
85
86   enum glsl_sampler_dim image_dim = glsl_get_sampler_dim(deref->type);
87   if (image_dim != GLSL_SAMPLER_DIM_SUBPASS &&
88       image_dim != GLSL_SAMPLER_DIM_SUBPASS_MS)
89      return false;
90
91   const bool multisampled = (image_dim == GLSL_SAMPLER_DIM_SUBPASS_MS);
92
93   nir_builder b;
94   nir_builder_init(&b, impl);
95   b.cursor = nir_instr_remove(&load->instr);
96
97   nir_ssa_def *frag_coord = load_frag_coord(options, &b);
98   frag_coord = nir_f2i32(&b, frag_coord);
99   nir_ssa_def *offset = nir_ssa_for_src(&b, load->src[1], 2);
100   nir_ssa_def *pos = nir_iadd(&b, frag_coord, offset);
101
102   nir_ssa_def *layer = load_layer_id(options, &b);
103   nir_ssa_def *coord =
104      nir_vec3(&b, nir_channel(&b, pos, 0), nir_channel(&b, pos, 1), layer);
105
106   nir_tex_instr *tex = nir_tex_instr_create(b.shader, 3 + multisampled);
107
108   tex->op = nir_texop_txf;
109   tex->sampler_dim = image_dim;
110
111   tex->dest_type =
112      nir_get_nir_type_for_glsl_base_type(glsl_get_sampler_result_type(deref->type));
113   tex->is_array = true;
114   tex->is_shadow = false;
115   tex->is_sparse = load->intrinsic == nir_intrinsic_image_deref_sparse_load;
116
117   tex->texture_index = 0;
118   tex->sampler_index = 0;
119
120   tex->src[0].src_type = nir_tex_src_texture_deref;
121   tex->src[0].src = nir_src_for_ssa(&deref->dest.ssa);
122
123   tex->src[1].src_type = nir_tex_src_coord;
124   tex->src[1].src = nir_src_for_ssa(coord);
125   tex->coord_components = 3;
126
127   tex->src[2].src_type = nir_tex_src_lod;
128   tex->src[2].src = nir_src_for_ssa(nir_imm_int(&b, 0));
129
130   if (image_dim == GLSL_SAMPLER_DIM_SUBPASS_MS) {
131      tex->op = nir_texop_txf_ms;
132      tex->src[3].src_type = nir_tex_src_ms_index;
133      tex->src[3].src = load->src[2];
134   }
135
136   tex->texture_non_uniform = nir_intrinsic_access(load) & ACCESS_NON_UNIFORM;
137
138   nir_ssa_dest_init(&tex->instr, &tex->dest, nir_tex_instr_dest_size(tex), 32, NULL);
139   nir_builder_instr_insert(&b, &tex->instr);
140
141   if (tex->is_sparse) {
142      unsigned load_result_size = load->dest.ssa.num_components - 1;
143      unsigned load_result_mask = BITFIELD_MASK(load_result_size);
144      nir_ssa_def *res = nir_channels(
145         &b, &tex->dest.ssa, load_result_mask | 0x10);
146
147      nir_ssa_def_rewrite_uses(&load->dest.ssa, res);
148   } else {
149      nir_ssa_def_rewrite_uses(&load->dest.ssa,
150                               &tex->dest.ssa);
151   }
152
153   return true;
154}
155
156static bool
157try_lower_input_texop(nir_function_impl *impl, nir_tex_instr *tex,
158                      const nir_input_attachment_options *options)
159{
160   nir_deref_instr *deref = nir_src_as_deref(tex->src[0].src);
161
162   if (glsl_get_sampler_dim(deref->type) != GLSL_SAMPLER_DIM_SUBPASS_MS)
163      return false;
164
165   nir_builder b;
166   nir_builder_init(&b, impl);
167   b.cursor = nir_before_instr(&tex->instr);
168
169   nir_ssa_def *frag_coord = load_frag_coord(options, &b);
170   frag_coord = nir_f2i32(&b, frag_coord);
171
172   nir_ssa_def *layer = load_layer_id(options, &b);
173   nir_ssa_def *coord = nir_vec3(&b, nir_channel(&b, frag_coord, 0),
174                                     nir_channel(&b, frag_coord, 1), layer);
175
176   tex->coord_components = 3;
177
178   nir_instr_rewrite_src(&tex->instr, &tex->src[1].src, nir_src_for_ssa(coord));
179
180   return true;
181}
182
183bool
184nir_lower_input_attachments(nir_shader *shader,
185                            const nir_input_attachment_options *options)
186{
187   assert(shader->info.stage == MESA_SHADER_FRAGMENT);
188   bool progress = false;
189
190   nir_foreach_function(function, shader) {
191      if (!function->impl)
192         continue;
193
194      nir_foreach_block(block, function->impl) {
195         nir_foreach_instr_safe(instr, block) {
196            switch (instr->type) {
197            case nir_instr_type_tex: {
198               nir_tex_instr *tex = nir_instr_as_tex(instr);
199
200               if (tex->op == nir_texop_fragment_mask_fetch_amd ||
201                   tex->op == nir_texop_fragment_fetch_amd) {
202                  progress |= try_lower_input_texop(function->impl, tex,
203                                                    options);
204               }
205               break;
206            }
207            case nir_instr_type_intrinsic: {
208               nir_intrinsic_instr *load = nir_instr_as_intrinsic(instr);
209
210               if (load->intrinsic == nir_intrinsic_image_deref_load ||
211                   load->intrinsic == nir_intrinsic_image_deref_sparse_load) {
212                  progress |= try_lower_input_load(function->impl, load,
213                                                   options);
214               }
215               break;
216            }
217            default:
218               break;
219            }
220         }
221      }
222   }
223
224   return progress;
225}
226