src/compiler/shader_info.h

   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
  25 #ifndef SHADER_INFO_H
  26 #define SHADER_INFO_H
  27
  28 #include "shader_enums.h"
  29 #include <stdint.h>
  30
  31 #ifdef __cplusplus
  32 extern "C" {
  33 #endif
  34
  35 struct spirv_supported_capabilities {
  36    bool address;
  37    bool atomic_storage;
  38    bool demote_to_helper_invocation;
  39    bool derivative_group;
  40    bool descriptor_array_dynamic_indexing;
  41    bool descriptor_array_non_uniform_indexing;
  42    bool descriptor_indexing;
  43    bool device_group;
  44    bool draw_parameters;
  45    bool float64;
  46    bool fragment_shader_sample_interlock;
  47    bool fragment_shader_pixel_interlock;
  48    bool geometry_streams;
  49    bool image_ms_array;
  50    bool image_read_without_format;
  51    bool image_write_without_format;
  52    bool int8;
  53    bool int16;
  54    bool int64;
  55    bool int64_atomics;
  56    bool kernel;
  57    bool min_lod;
  58    bool multiview;
  59    bool physical_storage_buffer_address;
  60    bool post_depth_coverage;
  61    bool runtime_descriptor_array;
  62    bool float_controls;
  63    bool shader_viewport_index_layer;
  64    bool stencil_export;
  65    bool storage_8bit;
  66    bool storage_16bit;
  67    bool storage_image_ms;
  68    bool subgroup_arithmetic;
  69    bool subgroup_ballot;
  70    bool subgroup_basic;
  71    bool subgroup_quad;
  72    bool subgroup_shuffle;
  73    bool subgroup_vote;
  74    bool tessellation;
  75    bool transform_feedback;
  76    bool variable_pointers;
  77    bool float16;
  78    bool amd_gcn_shader;
  79    bool amd_shader_ballot;
  80    bool amd_trinary_minmax;
  81 };
  82
  83 typedef struct shader_info {
  84    const char *name;
  85
  86    /* Descriptive name provided by the client; may be NULL */
  87    const char *label;
  88
  89    /** The shader stage, such as MESA_SHADER_VERTEX. */
  90    gl_shader_stage stage;
  91
  92    /** The shader stage in a non SSO linked program that follows this stage,
  93      * such as MESA_SHADER_FRAGMENT.
  94      */
  95    gl_shader_stage next_stage;
  96
  97    /* Number of textures used by this shader */
  98    unsigned num_textures;
  99    /* Number of uniform buffers used by this shader */
 100    unsigned num_ubos;
 101    /* Number of atomic buffers used by this shader */
 102    unsigned num_abos;
 103    /* Number of shader storage buffers used by this shader */
 104    unsigned num_ssbos;
 105    /* Number of images used by this shader */
 106    unsigned num_images;
 107
 108    /* Which inputs are actually read */
 109    uint64_t inputs_read;
 110    /* Which outputs are actually written */
 111    uint64_t outputs_written;
 112    /* Which outputs are actually read */
 113    uint64_t outputs_read;
 114    /* Which system values are actually read */
 115    uint64_t system_values_read;
 116
 117    /* Which patch inputs are actually read */
 118    uint32_t patch_inputs_read;
 119    /* Which patch outputs are actually written */
 120    uint32_t patch_outputs_written;
 121    /* Which patch outputs are read */
 122    uint32_t patch_outputs_read;
 123
 124    /* Whether or not this shader ever uses textureGather() */
 125    bool uses_texture_gather;
 126
 127    /** Bitfield of which textures are used */
 128    uint32_t textures_used;
 129
 130    /** Bitfield of which textures are used by texelFetch() */
 131    uint32_t textures_used_by_txf;
 132
 133    /**
 134     * True if this shader uses the fddx/fddy opcodes.
 135     *
 136     * Note that this does not include the "fine" and "coarse" variants.
 137     */
 138    bool uses_fddx_fddy;
 139
 140    /**
 141     * True if this shader uses 64-bit ALU operations
 142     */
 143    bool uses_64bit;
 144
 145    /* The size of the gl_ClipDistance[] array, if declared. */
 146    unsigned clip_distance_array_size;
 147
 148    /* The size of the gl_CullDistance[] array, if declared. */
 149    unsigned cull_distance_array_size;
 150
 151    /* Whether or not separate shader objects were used */
 152    bool separate_shader;
 153
 154    /** Was this shader linked with any transform feedback varyings? */
 155    bool has_transform_feedback_varyings;
 156
 157    union {
 158       struct {
 159          /* Which inputs are doubles */
 160          uint64_t double_inputs;
 161
 162          /* For AMD-specific driver-internal shaders. It replaces vertex
 163           * buffer loads with code generating VS inputs from scalar registers.
 164           *
 165           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 166           */
 167          unsigned blit_sgprs_amd;
 168
 169          /* True if the shader writes position in window space coordinates pre-transform */
 170          bool window_space_position;
 171       } vs;
 172
 173       struct {
 174          /** The number of vertices recieves per input primitive */
 175          unsigned vertices_in;
 176
 177          /** The output primitive type (GL enum value) */
 178          unsigned output_primitive;
 179
 180          /** The input primitive type (GL enum value) */
 181          unsigned input_primitive;
 182
 183          /** The maximum number of vertices the geometry shader might write. */
 184          unsigned vertices_out;
 185
 186          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 187          unsigned invocations;
 188
 189          /** Whether or not this shader uses EndPrimitive */
 190          bool uses_end_primitive;
 191
 192          /** Whether or not this shader uses non-zero streams */
 193          bool uses_streams;
 194       } gs;
 195
 196       struct {
 197          bool uses_discard;
 198
 199          /**
 200           * True if this fragment shader requires helper invocations.  This
 201           * can be caused by the use of ALU derivative ops, texture
 202           * instructions which do implicit derivatives, and the use of quad
 203           * subgroup operations.
 204           */
 205          bool needs_helper_invocations;
 206
 207          /**
 208           * Whether any inputs are declared with the "sample" qualifier.
 209           */
 210          bool uses_sample_qualifier;
 211
 212          /**
 213           * Whether early fragment tests are enabled as defined by
 214           * ARB_shader_image_load_store.
 215           */
 216          bool early_fragment_tests;
 217
 218          /**
 219           * Defined by INTEL_conservative_rasterization.
 220           */
 221          bool inner_coverage;
 222
 223          bool post_depth_coverage;
 224
 225          /**
 226           * \name ARB_fragment_coord_conventions
 227           * @{
 228           */
 229          bool pixel_center_integer;
 230          bool origin_upper_left;
 231          /*@}*/
 232
 233          bool pixel_interlock_ordered;
 234          bool pixel_interlock_unordered;
 235          bool sample_interlock_ordered;
 236          bool sample_interlock_unordered;
 237
 238          /**
 239           * Flags whether NIR's base types on the FS color outputs should be
 240           * ignored.
 241           *
 242           * GLSL requires that fragment shader output base types match the
 243           * render target's base types for the behavior to be defined.  From
 244           * the GL 4.6 spec:
 245           *
 246           *     "If the values written by the fragment shader do not match the
 247           *      format(s) of the corresponding color buffer(s), the result is
 248           *      undefined."
 249           *
 250           * However, for NIR shaders translated from TGSI, we don't have the
 251           * output types any more, so the driver will need to do whatever
 252           * fixups are necessary to handle effectively untyped data being
 253           * output from the FS.
 254           */
 255          bool untyped_color_outputs;
 256
 257          /** gl_FragDepth layout for ARB_conservative_depth. */
 258          enum gl_frag_depth_layout depth_layout;
 259       } fs;
 260
 261       struct {
 262          unsigned local_size[3];
 263
 264          bool local_size_variable;
 265          char user_data_components_amd;
 266
 267          /**
 268           * Size of shared variables accessed by the compute shader.
 269           */
 270          unsigned shared_size;
 271
 272
 273          /**
 274           * pointer size is:
 275           *   AddressingModelLogical:    0    (default)
 276           *   AddressingModelPhysical32: 32
 277           *   AddressingModelPhysical64: 64
 278           */
 279          unsigned ptr_size;
 280
 281          /*
 282           * Arrangement of invocations used to calculate derivatives in a compute
 283           * shader.  From NV_compute_shader_derivatives.
 284           */
 285          enum gl_derivative_group derivative_group;
 286       } cs;
 287
 288       /* Applies to both TCS and TES. */
 289       struct {
 290          /** The number of vertices in the TCS output patch. */
 291          unsigned tcs_vertices_out;
 292
 293          uint32_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 294          enum gl_tess_spacing spacing;
 295          /** Is the vertex order counterclockwise? */
 296          bool ccw;
 297          bool point_mode;
 298       } tess;
 299    };
 300 } shader_info;
 301
 302 #ifdef __cplusplus
 303 }
 304 #endif
 305
 306 #endif /* SHADER_INFO_H */