1 /*
2  * Copyright © 2019 Google, Inc.
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 FROM,
20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21  * SOFTWARE.
22  */
23 
24 #include "nir.h"
25 #include "nir_builder.h"
26 
27 /* Lowing for fragment shader load_output.
28  *
29  * This pass supports the blend_equation_advanced, where a fragment
30  * shader loads the output (fragcolor) to read the current framebuffer.
31  * It does this by lowering the output read to a txf_ms_fb instruction.
32  * This instruction works similarly to a normal txf_ms except without
33  * taking a texture source argument.  (The driver backend is expected
34  * to wire this up to a free texture slot which is configured to read
35  * from the framebuffer.)
36  *
37  * This should be run after lower_wpos_ytransform, because the tex
38  * coordinates should be the physical fragcoord, not the logical
39  * y-flipped coord.
40  *
41  * Note that this pass explicitly does *not* add a sampler uniform
42  * (as txf_ms_fb does not reference a texture).  The driver backend
43  * is going to want nif->info.num_textures to include the count of
44  * number of textures *not* including the one it inserts to sample
45  * from the framebuffer, so it more easily knows where to insert the
46  * hidden texture to read from the fb.
47  */
48 
49 static void
lower_fb_read(nir_builder * b,nir_intrinsic_instr * intr)50 lower_fb_read(nir_builder *b, nir_intrinsic_instr *intr)
51 {
52    b->cursor = nir_before_instr(&intr->instr);
53 
54    nir_ssa_def *fragcoord = nir_load_frag_coord(b);
55    nir_ssa_def *sampid = nir_load_sample_id(b);
56 
57    fragcoord = nir_f2i32(b, fragcoord);
58 
59    nir_tex_instr *tex = nir_tex_instr_create(b->shader, 2);
60    tex->op = nir_texop_txf_ms_fb;
61    tex->sampler_dim = GLSL_SAMPLER_DIM_2D;
62    tex->coord_components = 2;
63    tex->dest_type = nir_type_float;
64    tex->src[0].src_type = nir_tex_src_coord;
65    tex->src[0].src = nir_src_for_ssa(nir_channels(b, fragcoord, 0x3));
66    tex->src[1].src_type = nir_tex_src_ms_index;
67    tex->src[1].src = nir_src_for_ssa(sampid);
68 
69    nir_ssa_dest_init(&tex->instr, &tex->dest, 4, 32, NULL);
70    nir_builder_instr_insert(b, &tex->instr);
71 
72    nir_ssa_def_rewrite_uses(&intr->dest.ssa, nir_src_for_ssa(&tex->dest.ssa));
73 }
74 
75 bool
nir_lower_fb_read(nir_shader * shader)76 nir_lower_fb_read(nir_shader *shader)
77 {
78    bool progress = false;
79 
80    assert(shader->info.stage == MESA_SHADER_FRAGMENT);
81 
82    nir_foreach_function(function, shader) {
83       nir_function_impl *impl = function->impl;
84 
85       if (!impl)
86          continue;
87 
88       nir_foreach_block(block, impl) {
89          nir_foreach_instr_safe(instr, block) {
90             if (instr->type != nir_instr_type_intrinsic)
91                continue;
92 
93             nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
94             if (intr->intrinsic != nir_intrinsic_load_output)
95                continue;
96 
97             /* TODO KHR_blend_equation_advanced is limited to non-MRT
98              * scenarios.. but possible there are other extensions
99              * where this pass would be useful that do support MRT?
100              *
101              * I guess for now I'll leave that as an exercise for the
102              * reader.
103              */
104             if (nir_intrinsic_base(intr) != 0 ||
105                 nir_src_as_uint(intr->src[0]) != 0)
106                continue;
107 
108             nir_builder b;
109             nir_builder_init(&b, impl);
110             lower_fb_read(&b, intr);
111             progress = true;
112          }
113       }
114 
115       nir_metadata_preserve(impl, nir_metadata_block_index |
116                                   nir_metadata_dominance);
117 
118    }
119 
120    return progress;
121 }
122