1 /*
2  * Copyright © 2019 Collabora, Ltd.
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  * Authors (Collabora):
24  *    Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
25  */
26 
27 #include "nir.h"
28 #include "nir_builder.h"
29 
30 /*
31  * Lowers SSBOs to globals, for hardware that lack native SSBO support. When
32  * lowering, *_ssbo_* instructions will become *_global_* instructions,
33  * augmented with load_ssbo_address.
34  *
35  * DOES NOT PERFORM BOUNDS CHECKING. DO NOT USE IN PRODUCTION ON UNTRUSTED
36  * CONTEXTS INCLUDING WEBGL 2.
37  */
38 
39 static nir_intrinsic_op
lower_ssbo_op(nir_intrinsic_op op)40 lower_ssbo_op(nir_intrinsic_op op)
41 {
42    switch (op) {
43    case nir_intrinsic_load_ssbo:
44       return nir_intrinsic_load_global;
45 
46    case nir_intrinsic_store_ssbo:
47       return nir_intrinsic_store_global;
48 
49    case nir_intrinsic_ssbo_atomic_add:
50       return nir_intrinsic_global_atomic_add;
51    case nir_intrinsic_ssbo_atomic_imin:
52       return nir_intrinsic_global_atomic_imin;
53    case nir_intrinsic_ssbo_atomic_umin:
54       return nir_intrinsic_global_atomic_umin;
55    case nir_intrinsic_ssbo_atomic_imax:
56       return nir_intrinsic_global_atomic_imax;
57    case nir_intrinsic_ssbo_atomic_umax:
58       return nir_intrinsic_global_atomic_umax;
59    case nir_intrinsic_ssbo_atomic_and:
60       return nir_intrinsic_global_atomic_and;
61    case nir_intrinsic_ssbo_atomic_or:
62       return nir_intrinsic_global_atomic_or;
63    case nir_intrinsic_ssbo_atomic_xor:
64       return nir_intrinsic_global_atomic_xor;
65    case nir_intrinsic_ssbo_atomic_exchange:
66       return nir_intrinsic_global_atomic_exchange;
67    case nir_intrinsic_ssbo_atomic_comp_swap:
68       return nir_intrinsic_global_atomic_comp_swap;
69 
70    case nir_intrinsic_ssbo_atomic_fadd:
71       return nir_intrinsic_global_atomic_fadd;
72    case nir_intrinsic_ssbo_atomic_fmin:
73       return nir_intrinsic_global_atomic_fmin;
74    case nir_intrinsic_ssbo_atomic_fmax:
75       return nir_intrinsic_global_atomic_fmax;
76    case nir_intrinsic_ssbo_atomic_fcomp_swap:
77       return nir_intrinsic_global_atomic_fcomp_swap;
78 
79    default:
80       unreachable("Invalid SSBO op");
81    }
82 }
83 
84 /* Like SSBO property sysvals, though SSBO index may be indirect. C.f.
85  * nir_load_system_value */
86 
87 static inline nir_ssa_def *
nir_load_ssbo_prop(nir_builder * b,nir_intrinsic_op op,nir_src * idx,unsigned bitsize)88 nir_load_ssbo_prop(nir_builder *b, nir_intrinsic_op op,
89       nir_src *idx, unsigned bitsize)
90 {
91    nir_intrinsic_instr *load = nir_intrinsic_instr_create(b->shader, op);
92    load->num_components = 1;
93    nir_src_copy(&load->src[0], idx);
94    nir_ssa_dest_init(&load->instr, &load->dest, 1, bitsize, NULL);
95    nir_builder_instr_insert(b, &load->instr);
96    return &load->dest.ssa;
97 }
98 
99 #define nir_ssbo_prop(b, prop, index, bitsize) \
100    nir_load_ssbo_prop(b, nir_intrinsic_##prop, index, bitsize)
101 
102 static nir_ssa_def *
lower_ssbo_instr(nir_builder * b,nir_intrinsic_instr * intr)103 lower_ssbo_instr(nir_builder *b, nir_intrinsic_instr *intr)
104 {
105    nir_intrinsic_op op = lower_ssbo_op(intr->intrinsic);
106    bool is_store = op == nir_intrinsic_store_global;
107    bool is_atomic = !is_store && op != nir_intrinsic_load_global;
108 
109    /* We have to calculate the address:
110     *
111     * &(SSBO[offset]) = &SSBO + offset
112     */
113 
114    nir_src index = intr->src[is_store ? 1 : 0];
115    nir_src *offset_src = nir_get_io_offset_src(intr);
116    nir_ssa_def *offset = nir_ssa_for_src(b, *offset_src, 1);
117 
118    nir_ssa_def *address =
119       nir_iadd(b,
120             nir_ssbo_prop(b, load_ssbo_address, &index, 64),
121             nir_u2u64(b, offset));
122 
123    /* Create the replacement intrinsic */
124 
125    nir_intrinsic_instr *global =
126       nir_intrinsic_instr_create(b->shader, op);
127 
128    global->num_components = intr->num_components;
129    global->src[is_store ? 1 : 0] = nir_src_for_ssa(address);
130 
131    if (!is_atomic) {
132       nir_intrinsic_set_align_mul(global, nir_intrinsic_align_mul(intr));
133       nir_intrinsic_set_align_offset(global, nir_intrinsic_align_offset(intr));
134    }
135 
136    if (is_store) {
137       nir_src_copy(&global->src[0], &intr->src[0]);
138       nir_intrinsic_set_write_mask(global, nir_intrinsic_write_mask(intr));
139    } else {
140       nir_ssa_dest_init(&global->instr, &global->dest,
141                         intr->dest.ssa.num_components,
142                         intr->dest.ssa.bit_size, NULL);
143 
144       if (is_atomic) {
145          nir_src_copy(&global->src[1], &intr->src[2]);
146          if (nir_intrinsic_infos[op].num_srcs > 2)
147             nir_src_copy(&global->src[2], &intr->src[3]);
148       }
149    }
150 
151    nir_builder_instr_insert(b, &global->instr);
152    return is_store ? NULL : &global->dest.ssa;
153 }
154 
155 static bool
should_lower_ssbo_instr(const nir_instr * instr)156 should_lower_ssbo_instr(const nir_instr *instr)
157 {
158    if (instr->type != nir_instr_type_intrinsic)
159       return false;
160 
161    const nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
162 
163    switch (intr->intrinsic) {
164    case nir_intrinsic_load_ssbo:
165    case nir_intrinsic_store_ssbo:
166    case nir_intrinsic_ssbo_atomic_add:
167    case nir_intrinsic_ssbo_atomic_imin:
168    case nir_intrinsic_ssbo_atomic_umin:
169    case nir_intrinsic_ssbo_atomic_imax:
170    case nir_intrinsic_ssbo_atomic_umax:
171    case nir_intrinsic_ssbo_atomic_and:
172    case nir_intrinsic_ssbo_atomic_or:
173    case nir_intrinsic_ssbo_atomic_xor:
174    case nir_intrinsic_ssbo_atomic_exchange:
175    case nir_intrinsic_ssbo_atomic_comp_swap:
176    case nir_intrinsic_ssbo_atomic_fadd:
177    case nir_intrinsic_ssbo_atomic_fmin:
178    case nir_intrinsic_ssbo_atomic_fmax:
179    case nir_intrinsic_ssbo_atomic_fcomp_swap:
180       return true;
181    default:
182       return false;
183    }
184 
185    return false;
186 }
187 
188 bool
nir_lower_ssbo(nir_shader * shader)189 nir_lower_ssbo(nir_shader *shader)
190 {
191    bool progress = false;
192 
193    nir_foreach_function(function, shader) {
194       nir_function_impl *impl = function->impl;
195       nir_builder b;
196       nir_builder_init(&b, impl);
197 
198       nir_foreach_block(block, impl) {
199          nir_foreach_instr_safe(instr, block) {
200             if (!should_lower_ssbo_instr(instr)) continue;
201             progress = true;
202             b.cursor = nir_before_instr(instr);
203 
204             nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
205             nir_ssa_def *replace = lower_ssbo_instr(&b, intr);
206 
207             if (replace)  {
208                nir_ssa_def_rewrite_uses(&intr->dest.ssa,
209                                      replace);
210             }
211 
212             nir_instr_remove(instr);
213          }
214       }
215    }
216 
217    return progress;
218 }
219