src/panfrost/midgard/midgard_compile.h

   1 /*
   2  * Copyright (C) 2018-2019 Alyssa Rosenzweig <alyssa@rosenzweig.io>
   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 #ifndef __MIDGARD_H_
  25 #define __MIDGARD_H_
  26
  27 #include "compiler/nir/nir.h"
  28 #include "util/u_dynarray.h"
  29 #include "panfrost-job.h"
  30
  31 /* Define the general compiler entry point */
  32
  33 #define MAX_SYSVAL_COUNT 32
  34
  35 /* Allow 2D of sysval IDs, while allowing nonparametric sysvals to equal
  36  * their class for equal comparison */
  37
  38 #define PAN_SYSVAL(type, no) (((no) << 16) | PAN_SYSVAL_##type)
  39 #define PAN_SYSVAL_TYPE(sysval) ((sysval) & 0xffff)
  40 #define PAN_SYSVAL_ID(sysval) ((sysval) >> 16)
  41
  42 /* Define some common types. We start at one for easy indexing of hash
  43  * tables internal to the compiler */
  44
  45 enum {
  46         PAN_SYSVAL_VIEWPORT_SCALE = 1,
  47         PAN_SYSVAL_VIEWPORT_OFFSET = 2,
  48         PAN_SYSVAL_TEXTURE_SIZE = 3,
  49         PAN_SYSVAL_SSBO = 4,
  50         PAN_SYSVAL_NUM_WORK_GROUPS = 5,
  51         PAN_SYSVAL_SAMPLER = 7,
  52 } pan_sysval;
  53
  54 #define PAN_TXS_SYSVAL_ID(texidx, dim, is_array)          \
  55         ((texidx) | ((dim) << 7) | ((is_array) ? (1 << 9) : 0))
  56
  57 #define PAN_SYSVAL_ID_TO_TXS_TEX_IDX(id)        ((id) & 0x7f)
  58 #define PAN_SYSVAL_ID_TO_TXS_DIM(id)            (((id) >> 7) & 0x3)
  59 #define PAN_SYSVAL_ID_TO_TXS_IS_ARRAY(id)       !!((id) & (1 << 9))
  60
  61 /* Special attribute slots for vertex builtins. Sort of arbitrary but let's be
  62  * consistent with the blob so we can compare traces easier. */
  63
  64 enum {
  65         PAN_VERTEX_ID   = 16,
  66         PAN_INSTANCE_ID = 17,
  67         PAN_MAX_ATTRIBUTE
  68 } pan_special_attributes;
  69
  70 typedef struct {
  71         int work_register_count;
  72         int uniform_count;
  73         int uniform_cutoff;
  74
  75         /* Prepended before uniforms, mapping to SYSVAL_ names for the
  76          * sysval */
  77
  78         unsigned sysval_count;
  79         unsigned sysvals[MAX_SYSVAL_COUNT];
  80
  81         unsigned varyings[32];
  82         enum mali_format varying_type[32];
  83
  84         /* Boolean properties of the program */
  85         bool writes_point_size;
  86
  87         int first_tag;
  88
  89         struct util_dynarray compiled;
  90
  91         /* For a blend shader using a constant color -- patch point. If
  92          * negative, there's no constant. */
  93
  94         int blend_patch_offset;
  95
  96         /* The number of bytes to allocate per-thread for Thread Local Storage
  97          * (register spilling), or zero if no spilling is used */
  98         unsigned tls_size;
  99
 100         /* IN: For a fragment shader with a lowered alpha test, the ref value */
 101         float alpha_ref;
 102 } midgard_program;
 103
 104 int
 105 midgard_compile_shader_nir(nir_shader *nir, midgard_program *program, bool is_blend, unsigned blend_rt, unsigned gpu_id, bool shaderdb);
 106
 107 /* NIR options are shared between the standalone compiler and the online
 108  * compiler. Defining it here is the simplest, though maybe not the Right
 109  * solution. */
 110
 111 static const nir_shader_compiler_options midgard_nir_options = {
 112         .lower_ffma = true,
 113         .lower_sub = true,
 114         .lower_scmp = true,
 115         .lower_flrp32 = true,
 116         .lower_flrp64 = true,
 117         .lower_ffract = true,
 118         .lower_fmod = true,
 119         .lower_fdiv = true,
 120         .lower_idiv = true,
 121         .lower_isign = true,
 122         .lower_fpow = true,
 123         .lower_find_lsb = true,
 124         .lower_fdph = true,
 125
 126         .lower_wpos_pntc = true,
 127
 128         /* TODO: We have native ops to help here, which we'll want to look into
 129          * eventually */
 130         .lower_fsign = true,
 131
 132         .lower_extract_byte = true,
 133         .lower_extract_word = true,
 134         .lower_rotate = true,
 135
 136         .lower_pack_half_2x16 = true,
 137         .lower_pack_half_2x16_split = true,
 138         .lower_pack_unorm_2x16 = true,
 139         .lower_pack_snorm_2x16 = true,
 140         .lower_pack_unorm_4x8 = true,
 141         .lower_pack_snorm_4x8 = true,
 142         .lower_unpack_half_2x16 = true,
 143         .lower_unpack_half_2x16_split = true,
 144         .lower_unpack_unorm_2x16 = true,
 145         .lower_unpack_snorm_2x16 = true,
 146         .lower_unpack_unorm_4x8 = true,
 147         .lower_unpack_snorm_4x8 = true,
 148
 149         .lower_doubles_options = nir_lower_dmod,
 150
 151         .lower_bitfield_extract_to_shifts = true,
 152         .vectorize_io = true,
 153         .use_interpolated_input_intrinsics = true
 154 };
 155
 156 #endif