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