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    bool amd_image_gather_bias_lod;
  89 };
  90
  91 typedef struct shader_info {
  92    const char *name;
  93
  94    /* Descriptive name provided by the client; may be NULL */
  95    const char *label;
  96
  97    /** The shader stage, such as MESA_SHADER_VERTEX. */
  98    gl_shader_stage stage:8;
  99
 100    /** The shader stage in a non SSO linked program that follows this stage,
 101      * such as MESA_SHADER_FRAGMENT.
 102      */
 103    gl_shader_stage next_stage:8;
 104
 105    /* Number of textures used by this shader */
 106    uint8_t num_textures;
 107    /* Number of uniform buffers used by this shader */
 108    uint8_t num_ubos;
 109    /* Number of atomic buffers used by this shader */
 110    uint8_t num_abos;
 111    /* Number of shader storage buffers (max .driver_location + 1) used by this
 112     * shader.  In the case of nir_lower_atomics_to_ssbo being used, this will
 113     * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs
 114     * and atomic counters in nir_shader->info.
 115     */
 116    uint8_t num_ssbos;
 117    /* Number of images used by this shader */
 118    uint8_t num_images;
 119    /* Index of the last MSAA image. */
 120    int8_t last_msaa_image;
 121
 122    /* Which inputs are actually read */
 123    uint64_t inputs_read;
 124    /* Which outputs are actually written */
 125    uint64_t outputs_written;
 126    /* Which outputs are actually read */
 127    uint64_t outputs_read;
 128    /* Which system values are actually read */
 129    uint64_t system_values_read;
 130
 131    /* Which patch inputs are actually read */
 132    uint32_t patch_inputs_read;
 133    /* Which patch outputs are actually written */
 134    uint32_t patch_outputs_written;
 135    /* Which patch outputs are read */
 136    uint32_t patch_outputs_read;
 137
 138    /* Which inputs are read indirectly (subset of inputs_read) */
 139    uint64_t inputs_read_indirectly;
 140    /* Which outputs are read or written indirectly */
 141    uint64_t outputs_accessed_indirectly;
 142    /* Which patch inputs are read indirectly (subset of patch_inputs_read) */
 143    uint64_t patch_inputs_read_indirectly;
 144    /* Which patch outputs are read or written indirectly */
 145    uint64_t patch_outputs_accessed_indirectly;
 146
 147    /** Bitfield of which textures are used */
 148    uint32_t textures_used;
 149
 150    /** Bitfield of which textures are used by texelFetch() */
 151    uint32_t textures_used_by_txf;
 152
 153    /** Bitfield of which images are used */
 154    uint32_t images_used;
 155
 156    /* SPV_KHR_float_controls: execution mode for floating point ops */
 157    uint16_t float_controls_execution_mode;
 158
 159    /* The size of the gl_ClipDistance[] array, if declared. */
 160    uint8_t clip_distance_array_size:4;
 161
 162    /* The size of the gl_CullDistance[] array, if declared. */
 163    uint8_t cull_distance_array_size:4;
 164
 165    /* Whether or not this shader ever uses textureGather() */
 166    bool uses_texture_gather:1;
 167
 168    /**
 169     * True if this shader uses the fddx/fddy opcodes.
 170     *
 171     * Note that this does not include the "fine" and "coarse" variants.
 172     */
 173    bool uses_fddx_fddy:1;
 174
 175    /**
 176     * True if this shader uses 64-bit ALU operations
 177     */
 178    bool uses_64bit:1;
 179
 180    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 181    bool first_ubo_is_default_ubo:1;
 182
 183    /* Whether or not separate shader objects were used */
 184    bool separate_shader:1;
 185
 186    /** Was this shader linked with any transform feedback varyings? */
 187    bool has_transform_feedback_varyings:1;
 188
 189    /* Whether flrp has been lowered. */
 190    bool flrp_lowered:1;
 191
 192    /* Whether the shader writes memory, including transform feedback. */
 193    bool writes_memory:1;
 194
 195    /* Whether gl_Layer is viewport-relative */
 196    bool layer_viewport_relative:1;
 197
 198    union {
 199       struct {
 200          /* Which inputs are doubles */
 201          uint64_t double_inputs;
 202
 203          /* For AMD-specific driver-internal shaders. It replaces vertex
 204           * buffer loads with code generating VS inputs from scalar registers.
 205           *
 206           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 207           */
 208          uint8_t blit_sgprs_amd:4;
 209
 210          /* True if the shader writes position in window space coordinates pre-transform */
 211          bool window_space_position:1;
 212       } vs;
 213
 214       struct {
 215          /** The output primitive type (GL enum value) */
 216          uint16_t output_primitive;
 217
 218          /** The input primitive type (GL enum value) */
 219          uint16_t input_primitive;
 220
 221          /** The maximum number of vertices the geometry shader might write. */
 222          uint16_t vertices_out;
 223
 224          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 225          uint8_t invocations;
 226
 227          /** The number of vertices recieves per input primitive (max. 6) */
 228          uint8_t vertices_in:3;
 229
 230          /** Whether or not this shader uses EndPrimitive */
 231          bool uses_end_primitive:1;
 232
 233          /** Whether or not this shader uses non-zero streams */
 234          bool uses_streams:1;
 235       } gs;
 236
 237       struct {
 238          bool uses_discard:1;
 239          bool uses_demote:1;
 240
 241          /**
 242           * True if this fragment shader requires helper invocations.  This
 243           * can be caused by the use of ALU derivative ops, texture
 244           * instructions which do implicit derivatives, and the use of quad
 245           * subgroup operations.
 246           */
 247          bool needs_helper_invocations:1;
 248
 249          /**
 250           * Whether any inputs are declared with the "sample" qualifier.
 251           */
 252          bool uses_sample_qualifier:1;
 253
 254          /**
 255           * Whether early fragment tests are enabled as defined by
 256           * ARB_shader_image_load_store.
 257           */
 258          bool early_fragment_tests:1;
 259
 260          /**
 261           * Defined by INTEL_conservative_rasterization.
 262           */
 263          bool inner_coverage:1;
 264
 265          bool post_depth_coverage:1;
 266
 267          /**
 268           * \name ARB_fragment_coord_conventions
 269           * @{
 270           */
 271          bool pixel_center_integer:1;
 272          bool origin_upper_left:1;
 273          /*@}*/
 274
 275          bool pixel_interlock_ordered:1;
 276          bool pixel_interlock_unordered:1;
 277          bool sample_interlock_ordered:1;
 278          bool sample_interlock_unordered:1;
 279
 280          /**
 281           * Flags whether NIR's base types on the FS color outputs should be
 282           * ignored.
 283           *
 284           * GLSL requires that fragment shader output base types match the
 285           * render target's base types for the behavior to be defined.  From
 286           * the GL 4.6 spec:
 287           *
 288           *     "If the values written by the fragment shader do not match the
 289           *      format(s) of the corresponding color buffer(s), the result is
 290           *      undefined."
 291           *
 292           * However, for NIR shaders translated from TGSI, we don't have the
 293           * output types any more, so the driver will need to do whatever
 294           * fixups are necessary to handle effectively untyped data being
 295           * output from the FS.
 296           */
 297          bool untyped_color_outputs:1;
 298
 299          /** gl_FragDepth layout for ARB_conservative_depth. */
 300          enum gl_frag_depth_layout depth_layout:3;
 301       } fs;
 302
 303       struct {
 304          uint16_t local_size[3];
 305
 306          bool local_size_variable:1;
 307          uint8_t user_data_components_amd:3;
 308
 309          /*
 310           * Arrangement of invocations used to calculate derivatives in a compute
 311           * shader.  From NV_compute_shader_derivatives.
 312           */
 313          enum gl_derivative_group derivative_group:2;
 314
 315          /**
 316           * Size of shared variables accessed by the compute shader.
 317           */
 318          unsigned shared_size;
 319
 320          /**
 321           * pointer size is:
 322           *   AddressingModelLogical:    0    (default)
 323           *   AddressingModelPhysical32: 32
 324           *   AddressingModelPhysical64: 64
 325           */
 326          unsigned ptr_size;
 327       } cs;
 328
 329       /* Applies to both TCS and TES. */
 330       struct {
 331          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 332
 333          /** The number of vertices in the TCS output patch. */
 334          uint8_t tcs_vertices_out;
 335          enum gl_tess_spacing spacing:2;
 336
 337          /** Is the vertex order counterclockwise? */
 338          bool ccw:1;
 339          bool point_mode:1;
 340
 341          /* Bit mask of TCS per-vertex inputs (VS outputs) that are used
 342           * with a vertex index that is NOT the invocation id
 343           */
 344          uint64_t tcs_cross_invocation_inputs_read;
 345
 346          /* Bit mask of TCS per-vertex outputs that are used
 347           * with a vertex index that is NOT the invocation id
 348           */
 349          uint64_t tcs_cross_invocation_outputs_read;
 350       } tess;
 351    };
 352 } shader_info;
 353
 354 #ifdef __cplusplus
 355 }
 356 #endif
 357
 358 #endif /* SHADER_INFO_H */