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