2 * Copyright (C) 2018-2019 Alyssa Rosenzweig <alyssa@rosenzweig.io>
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:
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
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
27 #include "compiler/nir/nir.h"
28 #include "util/u_dynarray.h"
29 #include "util/register_allocate.h"
31 /* To be shoved inside panfrost_screen for the Gallium driver, or somewhere
32 * else for Vulkan/standalone. The single compiler "screen" to be shared across
33 * all shader compiles, used to store complex initialization (for instance,
34 * related to register allocation) */
36 struct midgard_screen
{
37 /* Precomputed register allocation sets for varying numbers of work
38 * registers. The zeroeth entry corresponds to 8 work registers. The
39 * eighth entry corresponds to 16 work registers. NULL if this set has
40 * not been allocated yet. */
42 struct ra_regs
*regs
[9];
44 /* Work register classes corresponds to the above register sets */
45 unsigned reg_classes
[9][4];
48 /* Define the general compiler entry point */
50 #define MAX_SYSVAL_COUNT 32
52 /* Allow 2D of sysval IDs, while allowing nonparametric sysvals to equal
53 * their class for equal comparison */
55 #define PAN_SYSVAL(type, no) (((no) << 16) | PAN_SYSVAL_##type)
56 #define PAN_SYSVAL_TYPE(sysval) ((sysval) & 0xffff)
57 #define PAN_SYSVAL_ID(sysval) ((sysval) >> 16)
59 /* Define some common types. We start at one for easy indexing of hash
60 * tables internal to the compiler */
63 PAN_SYSVAL_VIEWPORT_SCALE
= 1,
64 PAN_SYSVAL_VIEWPORT_OFFSET
= 2,
65 PAN_SYSVAL_TEXTURE_SIZE
= 3,
68 #define PAN_TXS_SYSVAL_ID(texidx, dim, is_array) \
69 ((texidx) | ((dim) << 7) | ((is_array) ? (1 << 9) : 0))
71 #define PAN_SYSVAL_ID_TO_TXS_TEX_IDX(id) ((id) & 0x7f)
72 #define PAN_SYSVAL_ID_TO_TXS_DIM(id) (((id) >> 7) & 0x3)
73 #define PAN_SYSVAL_ID_TO_TXS_IS_ARRAY(id) !!((id) & (1 << 9))
76 int work_register_count
;
83 /* Prepended before uniforms, mapping to SYSVAL_ names for the
86 unsigned sysval_count
;
87 unsigned sysvals
[MAX_SYSVAL_COUNT
];
89 unsigned varyings
[32];
91 /* Boolean properties of the program */
92 bool writes_point_size
;
96 struct util_dynarray compiled
;
98 /* For a blend shader using a constant color -- patch point. If
99 * negative, there's no constant. */
101 int blend_patch_offset
;
103 /* The number of bytes to allocate per-thread for Thread Local Storage
104 * (register spilling), or zero if no spilling is used */
107 /* IN: For a fragment shader with a lowered alpha test, the ref value */
112 midgard_compile_shader_nir(struct midgard_screen
*screen
, nir_shader
*nir
, midgard_program
*program
, bool is_blend
);
114 /* NIR options are shared between the standalone compiler and the online
115 * compiler. Defining it here is the simplest, though maybe not the Right
118 static const nir_shader_compiler_options midgard_nir_options
= {
122 .lower_flrp32
= true,
123 .lower_flrp64
= true,
124 .lower_ffract
= true,
130 .lower_find_lsb
= true,
132 .lower_wpos_pntc
= true,
134 /* TODO: We have native ops to help here, which we'll want to look into
138 .vertex_id_zero_based
= true,
139 .lower_extract_byte
= true,
140 .lower_extract_word
= true,
141 .lower_rotate
= true,
143 .lower_doubles_options
= nir_lower_dmod
,
145 .vectorize_io
= true,