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
 120    /* Which inputs are actually read */
 121    uint64_t inputs_read;
 122    /* Which outputs are actually written */
 123    uint64_t outputs_written;
 124    /* Which outputs are actually read */
 125    uint64_t outputs_read;
 126    /* Which system values are actually read */
 127    uint64_t system_values_read;
 128
 129    /* Which patch inputs are actually read */
 130    uint32_t patch_inputs_read;
 131    /* Which patch outputs are actually written */
 132    uint32_t patch_outputs_written;
 133    /* Which patch outputs are read */
 134    uint32_t patch_outputs_read;
 135
 136    /* Which inputs are read indirectly (subset of inputs_read) */
 137    uint64_t inputs_read_indirectly;
 138    /* Which outputs are read or written indirectly */
 139    uint64_t outputs_accessed_indirectly;
 140    /* Which patch inputs are read indirectly (subset of patch_inputs_read) */
 141    uint64_t patch_inputs_read_indirectly;
 142    /* Which patch outputs are read or written indirectly */
 143    uint64_t patch_outputs_accessed_indirectly;
 144
 145    /** Bitfield of which textures are used */
 146    uint32_t textures_used;
 147
 148    /** Bitfield of which textures are used by texelFetch() */
 149    uint32_t textures_used_by_txf;
 150
 151    /** Bitfield of which images are used */
 152    uint32_t images_used;
 153    /** Bitfield of which images are buffers. */
 154    uint32_t image_buffers;
 155    /** Bitfield of which images are MSAA. */
 156    uint32_t msaa_images;
 157
 158    /* SPV_KHR_float_controls: execution mode for floating point ops */
 159    uint16_t float_controls_execution_mode;
 160
 161    /* The size of the gl_ClipDistance[] array, if declared. */
 162    uint8_t clip_distance_array_size:4;
 163
 164    /* The size of the gl_CullDistance[] array, if declared. */
 165    uint8_t cull_distance_array_size:4;
 166
 167    /* Whether or not this shader ever uses textureGather() */
 168    bool uses_texture_gather:1;
 169
 170    /**
 171     * True if this shader uses the fddx/fddy opcodes.
 172     *
 173     * Note that this does not include the "fine" and "coarse" variants.
 174     */
 175    bool uses_fddx_fddy:1;
 176
 177    /**
 178     * True if this shader uses 64-bit ALU operations
 179     */
 180    bool uses_64bit:1;
 181
 182    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 183    bool first_ubo_is_default_ubo:1;
 184
 185    /* Whether or not separate shader objects were used */
 186    bool separate_shader:1;
 187
 188    /** Was this shader linked with any transform feedback varyings? */
 189    bool has_transform_feedback_varyings:1;
 190
 191    /* Whether flrp has been lowered. */
 192    bool flrp_lowered:1;
 193
 194    /* Whether the shader writes memory, including transform feedback. */
 195    bool writes_memory:1;
 196
 197    /* Whether gl_Layer is viewport-relative */
 198    bool layer_viewport_relative:1;
 199
 200    union {
 201       struct {
 202          /* Which inputs are doubles */
 203          uint64_t double_inputs;
 204
 205          /* For AMD-specific driver-internal shaders. It replaces vertex
 206           * buffer loads with code generating VS inputs from scalar registers.
 207           *
 208           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 209           */
 210          uint8_t blit_sgprs_amd:4;
 211
 212          /* True if the shader writes position in window space coordinates pre-transform */
 213          bool window_space_position:1;
 214       } vs;
 215
 216       struct {
 217          /** The output primitive type (GL enum value) */
 218          uint16_t output_primitive;
 219
 220          /** The input primitive type (GL enum value) */
 221          uint16_t input_primitive;
 222
 223          /** The maximum number of vertices the geometry shader might write. */
 224          uint16_t vertices_out;
 225
 226          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 227          uint8_t invocations;
 228
 229          /** The number of vertices recieves per input primitive (max. 6) */
 230          uint8_t vertices_in:3;
 231
 232          /** Whether or not this shader uses EndPrimitive */
 233          bool uses_end_primitive:1;
 234
 235          /** Whether or not this shader uses non-zero streams */
 236          bool uses_streams:1;
 237       } gs;
 238
 239       struct {
 240          bool uses_discard:1;
 241          bool uses_demote:1;
 242
 243          /**
 244           * True if this fragment shader requires helper invocations.  This
 245           * can be caused by the use of ALU derivative ops, texture
 246           * instructions which do implicit derivatives, and the use of quad
 247           * subgroup operations.
 248           */
 249          bool needs_helper_invocations:1;
 250
 251          /**
 252           * Whether any inputs are declared with the "sample" qualifier.
 253           */
 254          bool uses_sample_qualifier:1;
 255
 256          /**
 257           * Whether early fragment tests are enabled as defined by
 258           * ARB_shader_image_load_store.
 259           */
 260          bool early_fragment_tests:1;
 261
 262          /**
 263           * Defined by INTEL_conservative_rasterization.
 264           */
 265          bool inner_coverage:1;
 266
 267          bool post_depth_coverage:1;
 268
 269          /**
 270           * \name ARB_fragment_coord_conventions
 271           * @{
 272           */
 273          bool pixel_center_integer:1;
 274          bool origin_upper_left:1;
 275          /*@}*/
 276
 277          bool pixel_interlock_ordered:1;
 278          bool pixel_interlock_unordered:1;
 279          bool sample_interlock_ordered:1;
 280          bool sample_interlock_unordered:1;
 281
 282          /**
 283           * Flags whether NIR's base types on the FS color outputs should be
 284           * ignored.
 285           *
 286           * GLSL requires that fragment shader output base types match the
 287           * render target's base types for the behavior to be defined.  From
 288           * the GL 4.6 spec:
 289           *
 290           *     "If the values written by the fragment shader do not match the
 291           *      format(s) of the corresponding color buffer(s), the result is
 292           *      undefined."
 293           *
 294           * However, for NIR shaders translated from TGSI, we don't have the
 295           * output types any more, so the driver will need to do whatever
 296           * fixups are necessary to handle effectively untyped data being
 297           * output from the FS.
 298           */
 299          bool untyped_color_outputs:1;
 300
 301          /** gl_FragDepth layout for ARB_conservative_depth. */
 302          enum gl_frag_depth_layout depth_layout:3;
 303       } fs;
 304
 305       struct {
 306          uint16_t local_size[3];
 307
 308          bool local_size_variable:1;
 309          uint8_t user_data_components_amd:3;
 310
 311          /*
 312           * Arrangement of invocations used to calculate derivatives in a compute
 313           * shader.  From NV_compute_shader_derivatives.
 314           */
 315          enum gl_derivative_group derivative_group:2;
 316
 317          /**
 318           * Size of shared variables accessed by the compute shader.
 319           */
 320          unsigned shared_size;
 321
 322          /**
 323           * pointer size is:
 324           *   AddressingModelLogical:    0    (default)
 325           *   AddressingModelPhysical32: 32
 326           *   AddressingModelPhysical64: 64
 327           */
 328          unsigned ptr_size;
 329       } cs;
 330
 331       /* Applies to both TCS and TES. */
 332       struct {
 333          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 334
 335          /** The number of vertices in the TCS output patch. */
 336          uint8_t tcs_vertices_out;
 337          enum gl_tess_spacing spacing:2;
 338
 339          /** Is the vertex order counterclockwise? */
 340          bool ccw:1;
 341          bool point_mode:1;
 342
 343          /* Bit mask of TCS per-vertex inputs (VS outputs) that are used
 344           * with a vertex index that is NOT the invocation id
 345           */
 346          uint64_t tcs_cross_invocation_inputs_read;
 347
 348          /* Bit mask of TCS per-vertex outputs that are used
 349           * with a vertex index that is NOT the invocation id
 350           */
 351          uint64_t tcs_cross_invocation_outputs_read;
 352       } tess;
 353    };
 354 } shader_info;
 355
 356 #ifdef __cplusplus
 357 }
 358 #endif
 359
 360 #endif /* SHADER_INFO_H */