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 integer_functions2;
  57    bool kernel;
  58    bool min_lod;
  59    bool multiview;
  60    bool physical_storage_buffer_address;
  61    bool post_depth_coverage;
  62    bool runtime_descriptor_array;
  63    bool float_controls;
  64    bool shader_clock;
  65    bool shader_viewport_index_layer;
  66    bool stencil_export;
  67    bool storage_8bit;
  68    bool storage_16bit;
  69    bool storage_image_ms;
  70    bool subgroup_arithmetic;
  71    bool subgroup_ballot;
  72    bool subgroup_basic;
  73    bool subgroup_quad;
  74    bool subgroup_shuffle;
  75    bool subgroup_vote;
  76    bool tessellation;
  77    bool transform_feedback;
  78    bool variable_pointers;
  79    bool vk_memory_model;
  80    bool vk_memory_model_device_scope;
  81    bool float16;
  82    bool amd_fragment_mask;
  83    bool amd_gcn_shader;
  84    bool amd_shader_ballot;
  85    bool amd_trinary_minmax;
  86    bool amd_image_read_write_lod;
  87    bool amd_shader_explicit_vertex_parameter;
  88 };
  89
  90 typedef struct shader_info {
  91    const char *name;
  92
  93    /* Descriptive name provided by the client; may be NULL */
  94    const char *label;
  95
  96    /** The shader stage, such as MESA_SHADER_VERTEX. */
  97    gl_shader_stage stage:8;
  98
  99    /** The shader stage in a non SSO linked program that follows this stage,
 100      * such as MESA_SHADER_FRAGMENT.
 101      */
 102    gl_shader_stage next_stage:8;
 103
 104    /* Number of textures used by this shader */
 105    uint8_t num_textures;
 106    /* Number of uniform buffers used by this shader */
 107    uint8_t num_ubos;
 108    /* Number of atomic buffers used by this shader */
 109    uint8_t num_abos;
 110    /* Number of shader storage buffers (max .driver_location + 1) used by this
 111     * shader.  In the case of nir_lower_atomics_to_ssbo being used, this will
 112     * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs
 113     * and atomic counters in nir_shader->info.
 114     */
 115    uint8_t num_ssbos;
 116    /* Number of images used by this shader */
 117    uint8_t num_images;
 118    /* Index of the last MSAA image. */
 119    int8_t last_msaa_image;
 120
 121    /* Which inputs are actually read */
 122    uint64_t inputs_read;
 123    /* Which outputs are actually written */
 124    uint64_t outputs_written;
 125    /* Which outputs are actually read */
 126    uint64_t outputs_read;
 127    /* Which system values are actually read */
 128    uint64_t system_values_read;
 129
 130    /* Which patch inputs are actually read */
 131    uint32_t patch_inputs_read;
 132    /* Which patch outputs are actually written */
 133    uint32_t patch_outputs_written;
 134    /* Which patch outputs are read */
 135    uint32_t patch_outputs_read;
 136
 137    /* Which inputs are read indirectly (subset of inputs_read) */
 138    uint64_t inputs_read_indirectly;
 139    /* Which outputs are read or written indirectly */
 140    uint64_t outputs_accessed_indirectly;
 141    /* Which patch inputs are read indirectly (subset of patch_inputs_read) */
 142    uint64_t patch_inputs_read_indirectly;
 143    /* Which patch outputs are read or written indirectly */
 144    uint64_t patch_outputs_accessed_indirectly;
 145
 146    /** Bitfield of which textures are used */
 147    uint32_t textures_used;
 148
 149    /** Bitfield of which textures are used by texelFetch() */
 150    uint32_t textures_used_by_txf;
 151
 152    /** Bitfield of which images are used */
 153    uint32_t images_used;
 154
 155    /* SPV_KHR_float_controls: execution mode for floating point ops */
 156    uint16_t float_controls_execution_mode;
 157
 158    /* The size of the gl_ClipDistance[] array, if declared. */
 159    uint8_t clip_distance_array_size:4;
 160
 161    /* The size of the gl_CullDistance[] array, if declared. */
 162    uint8_t cull_distance_array_size:4;
 163
 164    /* Whether or not this shader ever uses textureGather() */
 165    bool uses_texture_gather:1;
 166
 167    /**
 168     * True if this shader uses the fddx/fddy opcodes.
 169     *
 170     * Note that this does not include the "fine" and "coarse" variants.
 171     */
 172    bool uses_fddx_fddy:1;
 173
 174    /**
 175     * True if this shader uses 64-bit ALU operations
 176     */
 177    bool uses_64bit:1;
 178
 179    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 180    bool first_ubo_is_default_ubo:1;
 181
 182    /* Whether or not separate shader objects were used */
 183    bool separate_shader:1;
 184
 185    /** Was this shader linked with any transform feedback varyings? */
 186    bool has_transform_feedback_varyings:1;
 187
 188    /* Whether flrp has been lowered. */
 189    bool flrp_lowered:1;
 190
 191    /* Whether the shader writes memory, including transform feedback. */
 192    bool writes_memory:1;
 193
 194    union {
 195       struct {
 196          /* Which inputs are doubles */
 197          uint64_t double_inputs;
 198
 199          /* For AMD-specific driver-internal shaders. It replaces vertex
 200           * buffer loads with code generating VS inputs from scalar registers.
 201           *
 202           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 203           */
 204          uint8_t blit_sgprs_amd:4;
 205
 206          /* True if the shader writes position in window space coordinates pre-transform */
 207          bool window_space_position:1;
 208       } vs;
 209
 210       struct {
 211          /** The output primitive type (GL enum value) */
 212          uint16_t output_primitive;
 213
 214          /** The input primitive type (GL enum value) */
 215          uint16_t input_primitive;
 216
 217          /** The maximum number of vertices the geometry shader might write. */
 218          uint16_t vertices_out;
 219
 220          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 221          uint8_t invocations;
 222
 223          /** The number of vertices recieves per input primitive (max. 6) */
 224          uint8_t vertices_in:3;
 225
 226          /** Whether or not this shader uses EndPrimitive */
 227          bool uses_end_primitive:1;
 228
 229          /** Whether or not this shader uses non-zero streams */
 230          bool uses_streams:1;
 231       } gs;
 232
 233       struct {
 234          bool uses_discard:1;
 235          bool uses_demote:1;
 236
 237          /**
 238           * True if this fragment shader requires helper invocations.  This
 239           * can be caused by the use of ALU derivative ops, texture
 240           * instructions which do implicit derivatives, and the use of quad
 241           * subgroup operations.
 242           */
 243          bool needs_helper_invocations:1;
 244
 245          /**
 246           * Whether any inputs are declared with the "sample" qualifier.
 247           */
 248          bool uses_sample_qualifier:1;
 249
 250          /**
 251           * Whether early fragment tests are enabled as defined by
 252           * ARB_shader_image_load_store.
 253           */
 254          bool early_fragment_tests:1;
 255
 256          /**
 257           * Defined by INTEL_conservative_rasterization.
 258           */
 259          bool inner_coverage:1;
 260
 261          bool post_depth_coverage:1;
 262
 263          /**
 264           * \name ARB_fragment_coord_conventions
 265           * @{
 266           */
 267          bool pixel_center_integer:1;
 268          bool origin_upper_left:1;
 269          /*@}*/
 270
 271          bool pixel_interlock_ordered:1;
 272          bool pixel_interlock_unordered:1;
 273          bool sample_interlock_ordered:1;
 274          bool sample_interlock_unordered:1;
 275
 276          /**
 277           * Flags whether NIR's base types on the FS color outputs should be
 278           * ignored.
 279           *
 280           * GLSL requires that fragment shader output base types match the
 281           * render target's base types for the behavior to be defined.  From
 282           * the GL 4.6 spec:
 283           *
 284           *     "If the values written by the fragment shader do not match the
 285           *      format(s) of the corresponding color buffer(s), the result is
 286           *      undefined."
 287           *
 288           * However, for NIR shaders translated from TGSI, we don't have the
 289           * output types any more, so the driver will need to do whatever
 290           * fixups are necessary to handle effectively untyped data being
 291           * output from the FS.
 292           */
 293          bool untyped_color_outputs:1;
 294
 295          /** gl_FragDepth layout for ARB_conservative_depth. */
 296          enum gl_frag_depth_layout depth_layout:3;
 297       } fs;
 298
 299       struct {
 300          uint16_t local_size[3];
 301
 302          bool local_size_variable:1;
 303          uint8_t user_data_components_amd:3;
 304
 305          /*
 306           * Arrangement of invocations used to calculate derivatives in a compute
 307           * shader.  From NV_compute_shader_derivatives.
 308           */
 309          enum gl_derivative_group derivative_group:2;
 310
 311          /**
 312           * Size of shared variables accessed by the compute shader.
 313           */
 314          unsigned shared_size;
 315
 316          /**
 317           * pointer size is:
 318           *   AddressingModelLogical:    0    (default)
 319           *   AddressingModelPhysical32: 32
 320           *   AddressingModelPhysical64: 64
 321           */
 322          unsigned ptr_size;
 323       } cs;
 324
 325       /* Applies to both TCS and TES. */
 326       struct {
 327          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 328
 329          /** The number of vertices in the TCS output patch. */
 330          uint8_t tcs_vertices_out;
 331          enum gl_tess_spacing spacing:2;
 332
 333          /* Bit mask of TCS per-vertex inputs (VS outputs) that are used
 334           * with a vertex index that is NOT the invocation id
 335           */
 336          uint64_t tcs_cross_invocation_inputs_read;
 337
 338          /* Bit mask of TCS per-vertex outputs that are used
 339           * with a vertex index that is NOT the invocation id
 340           */
 341          uint64_t tcs_cross_invocation_outputs_read;
 342
 343          /** Is the vertex order counterclockwise? */
 344          bool ccw:1;
 345          bool point_mode:1;
 346       } tess;
 347    };
 348 } shader_info;
 349
 350 #ifdef __cplusplus
 351 }
 352 #endif
 353
 354 #endif /* SHADER_INFO_H */