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 float32_atomic_add;
  46    bool float64;
  47    bool float64_atomic_add;
  48    bool fragment_shader_sample_interlock;
  49    bool fragment_shader_pixel_interlock;
  50    bool geometry_streams;
  51    bool image_ms_array;
  52    bool image_read_without_format;
  53    bool image_write_without_format;
  54    bool int8;
  55    bool int16;
  56    bool int64;
  57    bool int64_atomics;
  58    bool integer_functions2;
  59    bool kernel;
  60    bool kernel_image;
  61    bool literal_sampler;
  62    bool min_lod;
  63    bool multiview;
  64    bool physical_storage_buffer_address;
  65    bool post_depth_coverage;
  66    bool runtime_descriptor_array;
  67    bool float_controls;
  68    bool shader_clock;
  69    bool shader_viewport_index_layer;
  70    bool stencil_export;
  71    bool storage_8bit;
  72    bool storage_16bit;
  73    bool storage_image_ms;
  74    bool subgroup_arithmetic;
  75    bool subgroup_ballot;
  76    bool subgroup_basic;
  77    bool subgroup_quad;
  78    bool subgroup_shuffle;
  79    bool subgroup_vote;
  80    bool tessellation;
  81    bool transform_feedback;
  82    bool variable_pointers;
  83    bool vk_memory_model;
  84    bool vk_memory_model_device_scope;
  85    bool float16;
  86    bool amd_fragment_mask;
  87    bool amd_gcn_shader;
  88    bool amd_shader_ballot;
  89    bool amd_trinary_minmax;
  90    bool amd_image_read_write_lod;
  91    bool amd_shader_explicit_vertex_parameter;
  92    bool amd_image_gather_bias_lod;
  93 };
  94
  95 typedef struct shader_info {
  96    const char *name;
  97
  98    /* Descriptive name provided by the client; may be NULL */
  99    const char *label;
 100
 101    /** The shader stage, such as MESA_SHADER_VERTEX. */
 102    gl_shader_stage stage:8;
 103
 104    /** The shader stage in a non SSO linked program that follows this stage,
 105      * such as MESA_SHADER_FRAGMENT.
 106      */
 107    gl_shader_stage next_stage:8;
 108
 109    /* Number of textures used by this shader */
 110    uint8_t num_textures;
 111    /* Number of uniform buffers used by this shader */
 112    uint8_t num_ubos;
 113    /* Number of atomic buffers used by this shader */
 114    uint8_t num_abos;
 115    /* Number of shader storage buffers (max .driver_location + 1) used by this
 116     * shader.  In the case of nir_lower_atomics_to_ssbo being used, this will
 117     * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs
 118     * and atomic counters in nir_shader->info.
 119     */
 120    uint8_t num_ssbos;
 121    /* Number of images used by this shader */
 122    uint8_t num_images;
 123
 124    /* Which inputs are actually read */
 125    uint64_t inputs_read;
 126    /* Which outputs are actually written */
 127    uint64_t outputs_written;
 128    /* Which outputs are actually read */
 129    uint64_t outputs_read;
 130    /* Which system values are actually read */
 131    uint64_t system_values_read;
 132
 133    /* Which patch inputs are actually read */
 134    uint32_t patch_inputs_read;
 135    /* Which patch outputs are actually written */
 136    uint32_t patch_outputs_written;
 137    /* Which patch outputs are read */
 138    uint32_t patch_outputs_read;
 139
 140    /* Which inputs are read indirectly (subset of inputs_read) */
 141    uint64_t inputs_read_indirectly;
 142    /* Which outputs are read or written indirectly */
 143    uint64_t outputs_accessed_indirectly;
 144    /* Which patch inputs are read indirectly (subset of patch_inputs_read) */
 145    uint64_t patch_inputs_read_indirectly;
 146    /* Which patch outputs are read or written indirectly */
 147    uint64_t patch_outputs_accessed_indirectly;
 148
 149    /** Bitfield of which textures are used */
 150    uint32_t textures_used;
 151
 152    /** Bitfield of which textures are used by texelFetch() */
 153    uint32_t textures_used_by_txf;
 154
 155    /** Bitfield of which images are used */
 156    uint32_t images_used;
 157    /** Bitfield of which images are buffers. */
 158    uint32_t image_buffers;
 159    /** Bitfield of which images are MSAA. */
 160    uint32_t msaa_images;
 161
 162    /* SPV_KHR_float_controls: execution mode for floating point ops */
 163    uint16_t float_controls_execution_mode;
 164
 165    /* The size of the gl_ClipDistance[] array, if declared. */
 166    uint8_t clip_distance_array_size:4;
 167
 168    /* The size of the gl_CullDistance[] array, if declared. */
 169    uint8_t cull_distance_array_size:4;
 170
 171    /* Whether or not this shader ever uses textureGather() */
 172    bool uses_texture_gather:1;
 173
 174    /**
 175     * True if this shader uses the fddx/fddy opcodes.
 176     *
 177     * Note that this does not include the "fine" and "coarse" variants.
 178     */
 179    bool uses_fddx_fddy:1;
 180
 181    /**
 182     * True if this shader uses 64-bit ALU operations
 183     */
 184    bool uses_64bit:1;
 185
 186    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 187    bool first_ubo_is_default_ubo:1;
 188
 189    /* Whether or not separate shader objects were used */
 190    bool separate_shader:1;
 191
 192    /** Was this shader linked with any transform feedback varyings? */
 193    bool has_transform_feedback_varyings:1;
 194
 195    /* Whether flrp has been lowered. */
 196    bool flrp_lowered:1;
 197
 198    /* Whether nir_lower_io has been called to lower derefs.
 199     * nir_variables for inputs and outputs might not be present in the IR.
 200     */
 201    bool io_lowered:1;
 202
 203    /* Whether the shader writes memory, including transform feedback. */
 204    bool writes_memory:1;
 205
 206    /* Whether gl_Layer is viewport-relative */
 207    bool layer_viewport_relative:1;
 208
 209    union {
 210       struct {
 211          /* Which inputs are doubles */
 212          uint64_t double_inputs;
 213
 214          /* For AMD-specific driver-internal shaders. It replaces vertex
 215           * buffer loads with code generating VS inputs from scalar registers.
 216           *
 217           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 218           */
 219          uint8_t blit_sgprs_amd:4;
 220
 221          /* True if the shader writes position in window space coordinates pre-transform */
 222          bool window_space_position:1;
 223       } vs;
 224
 225       struct {
 226          /** The output primitive type (GL enum value) */
 227          uint16_t output_primitive;
 228
 229          /** The input primitive type (GL enum value) */
 230          uint16_t input_primitive;
 231
 232          /** The maximum number of vertices the geometry shader might write. */
 233          uint16_t vertices_out;
 234
 235          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 236          uint8_t invocations;
 237
 238          /** The number of vertices recieves per input primitive (max. 6) */
 239          uint8_t vertices_in:3;
 240
 241          /** Whether or not this shader uses EndPrimitive */
 242          bool uses_end_primitive:1;
 243
 244          /** The streams used in this shaders (max. 4) */
 245          uint8_t active_stream_mask:4;
 246       } gs;
 247
 248       struct {
 249          bool uses_discard:1;
 250          bool uses_demote:1;
 251
 252          /**
 253           * True if this fragment shader requires helper invocations.  This
 254           * can be caused by the use of ALU derivative ops, texture
 255           * instructions which do implicit derivatives, and the use of quad
 256           * subgroup operations.
 257           */
 258          bool needs_helper_invocations:1;
 259
 260          /**
 261           * Whether any inputs are declared with the "sample" qualifier.
 262           */
 263          bool uses_sample_qualifier:1;
 264
 265          /**
 266           * Whether early fragment tests are enabled as defined by
 267           * ARB_shader_image_load_store.
 268           */
 269          bool early_fragment_tests:1;
 270
 271          /**
 272           * Defined by INTEL_conservative_rasterization.
 273           */
 274          bool inner_coverage:1;
 275
 276          bool post_depth_coverage:1;
 277
 278          /**
 279           * \name ARB_fragment_coord_conventions
 280           * @{
 281           */
 282          bool pixel_center_integer:1;
 283          bool origin_upper_left:1;
 284          /*@}*/
 285
 286          bool pixel_interlock_ordered:1;
 287          bool pixel_interlock_unordered:1;
 288          bool sample_interlock_ordered:1;
 289          bool sample_interlock_unordered:1;
 290
 291          /**
 292           * Flags whether NIR's base types on the FS color outputs should be
 293           * ignored.
 294           *
 295           * GLSL requires that fragment shader output base types match the
 296           * render target's base types for the behavior to be defined.  From
 297           * the GL 4.6 spec:
 298           *
 299           *     "If the values written by the fragment shader do not match the
 300           *      format(s) of the corresponding color buffer(s), the result is
 301           *      undefined."
 302           *
 303           * However, for NIR shaders translated from TGSI, we don't have the
 304           * output types any more, so the driver will need to do whatever
 305           * fixups are necessary to handle effectively untyped data being
 306           * output from the FS.
 307           */
 308          bool untyped_color_outputs:1;
 309
 310          /** gl_FragDepth layout for ARB_conservative_depth. */
 311          enum gl_frag_depth_layout depth_layout:3;
 312
 313          /**
 314           * Interpolation qualifiers for drivers that lowers color inputs
 315           * to system values.
 316           */
 317          unsigned color0_interp:3; /* glsl_interp_mode */
 318          bool color0_sample:1;
 319          bool color0_centroid:1;
 320          unsigned color1_interp:3; /* glsl_interp_mode */
 321          bool color1_sample:1;
 322          bool color1_centroid:1;
 323       } fs;
 324
 325       struct {
 326          uint16_t local_size[3];
 327
 328          bool local_size_variable:1;
 329          uint8_t user_data_components_amd:3;
 330
 331          /*
 332           * Arrangement of invocations used to calculate derivatives in a compute
 333           * shader.  From NV_compute_shader_derivatives.
 334           */
 335          enum gl_derivative_group derivative_group:2;
 336
 337          /**
 338           * Size of shared variables accessed by the compute shader.
 339           */
 340          unsigned shared_size;
 341
 342          /**
 343           * pointer size is:
 344           *   AddressingModelLogical:    0    (default)
 345           *   AddressingModelPhysical32: 32
 346           *   AddressingModelPhysical64: 64
 347           */
 348          unsigned ptr_size;
 349       } cs;
 350
 351       /* Applies to both TCS and TES. */
 352       struct {
 353          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 354
 355          /** The number of vertices in the TCS output patch. */
 356          uint8_t tcs_vertices_out;
 357          enum gl_tess_spacing spacing:2;
 358
 359          /** Is the vertex order counterclockwise? */
 360          bool ccw:1;
 361          bool point_mode:1;
 362
 363          /* Bit mask of TCS per-vertex inputs (VS outputs) that are used
 364           * with a vertex index that is NOT the invocation id
 365           */
 366          uint64_t tcs_cross_invocation_inputs_read;
 367
 368          /* Bit mask of TCS per-vertex outputs that are used
 369           * with a vertex index that is NOT the invocation id
 370           */
 371          uint64_t tcs_cross_invocation_outputs_read;
 372       } tess;
 373    };
 374 } shader_info;
 375
 376 #ifdef __cplusplus
 377 }
 378 #endif
 379
 380 #endif /* SHADER_INFO_H */