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_gcn_shader;
  83    bool amd_shader_ballot;
  84    bool amd_trinary_minmax;
  85    bool amd_image_read_write_lod;
  86 };
  87
  88 typedef struct shader_info {
  89    const char *name;
  90
  91    /* Descriptive name provided by the client; may be NULL */
  92    const char *label;
  93
  94    /** The shader stage, such as MESA_SHADER_VERTEX. */
  95    gl_shader_stage stage:8;
  96
  97    /** The shader stage in a non SSO linked program that follows this stage,
  98      * such as MESA_SHADER_FRAGMENT.
  99      */
 100    gl_shader_stage next_stage:8;
 101
 102    /* Number of textures used by this shader */
 103    uint8_t num_textures;
 104    /* Number of uniform buffers used by this shader */
 105    uint8_t num_ubos;
 106    /* Number of atomic buffers used by this shader */
 107    uint8_t num_abos;
 108    /* Number of shader storage buffers (max .driver_location + 1) used by this
 109     * shader.  In the case of nir_lower_atomics_to_ssbo being used, this will
 110     * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs
 111     * and atomic counters in nir_shader->info.
 112     */
 113    uint8_t num_ssbos;
 114    /* Number of images used by this shader */
 115    uint8_t num_images;
 116    /* Index of the last MSAA image. */
 117    int8_t last_msaa_image;
 118
 119    /* Which inputs are actually read */
 120    uint64_t inputs_read;
 121    /* Which outputs are actually written */
 122    uint64_t outputs_written;
 123    /* Which outputs are actually read */
 124    uint64_t outputs_read;
 125    /* Which system values are actually read */
 126    uint64_t system_values_read;
 127
 128    /* Which patch inputs are actually read */
 129    uint32_t patch_inputs_read;
 130    /* Which patch outputs are actually written */
 131    uint32_t patch_outputs_written;
 132    /* Which patch outputs are read */
 133    uint32_t patch_outputs_read;
 134
 135    /** Bitfield of which textures are used */
 136    uint32_t textures_used;
 137
 138    /** Bitfield of which textures are used by texelFetch() */
 139    uint32_t textures_used_by_txf;
 140
 141    /* SPV_KHR_float_controls: execution mode for floating point ops */
 142    uint16_t float_controls_execution_mode;
 143
 144    /* The size of the gl_ClipDistance[] array, if declared. */
 145    uint8_t clip_distance_array_size:4;
 146
 147    /* The size of the gl_CullDistance[] array, if declared. */
 148    uint8_t cull_distance_array_size:4;
 149
 150    /* Whether or not this shader ever uses textureGather() */
 151    bool uses_texture_gather:1;
 152
 153    /**
 154     * True if this shader uses the fddx/fddy opcodes.
 155     *
 156     * Note that this does not include the "fine" and "coarse" variants.
 157     */
 158    bool uses_fddx_fddy:1;
 159
 160    /**
 161     * True if this shader uses 64-bit ALU operations
 162     */
 163    bool uses_64bit:1;
 164
 165    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 166    bool first_ubo_is_default_ubo:1;
 167
 168    /* Whether or not separate shader objects were used */
 169    bool separate_shader:1;
 170
 171    /** Was this shader linked with any transform feedback varyings? */
 172    bool has_transform_feedback_varyings:1;
 173
 174    /* Whether flrp has been lowered. */
 175    bool flrp_lowered:1;
 176
 177    union {
 178       struct {
 179          /* Which inputs are doubles */
 180          uint64_t double_inputs;
 181
 182          /* For AMD-specific driver-internal shaders. It replaces vertex
 183           * buffer loads with code generating VS inputs from scalar registers.
 184           *
 185           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 186           */
 187          uint8_t blit_sgprs_amd:4;
 188
 189          /* True if the shader writes position in window space coordinates pre-transform */
 190          bool window_space_position:1;
 191       } vs;
 192
 193       struct {
 194          /** The output primitive type (GL enum value) */
 195          uint16_t output_primitive;
 196
 197          /** The input primitive type (GL enum value) */
 198          uint16_t input_primitive;
 199
 200          /** The maximum number of vertices the geometry shader might write. */
 201          uint16_t vertices_out;
 202
 203          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 204          uint8_t invocations;
 205
 206          /** The number of vertices recieves per input primitive (max. 6) */
 207          uint8_t vertices_in:3;
 208
 209          /** Whether or not this shader uses EndPrimitive */
 210          bool uses_end_primitive:1;
 211
 212          /** Whether or not this shader uses non-zero streams */
 213          bool uses_streams:1;
 214       } gs;
 215
 216       struct {
 217          bool uses_discard:1;
 218
 219          /**
 220           * True if this fragment shader requires helper invocations.  This
 221           * can be caused by the use of ALU derivative ops, texture
 222           * instructions which do implicit derivatives, and the use of quad
 223           * subgroup operations.
 224           */
 225          bool needs_helper_invocations:1;
 226
 227          /**
 228           * Whether any inputs are declared with the "sample" qualifier.
 229           */
 230          bool uses_sample_qualifier:1;
 231
 232          /**
 233           * Whether early fragment tests are enabled as defined by
 234           * ARB_shader_image_load_store.
 235           */
 236          bool early_fragment_tests:1;
 237
 238          /**
 239           * Defined by INTEL_conservative_rasterization.
 240           */
 241          bool inner_coverage:1;
 242
 243          bool post_depth_coverage:1;
 244
 245          /**
 246           * \name ARB_fragment_coord_conventions
 247           * @{
 248           */
 249          bool pixel_center_integer:1;
 250          bool origin_upper_left:1;
 251          /*@}*/
 252
 253          bool pixel_interlock_ordered:1;
 254          bool pixel_interlock_unordered:1;
 255          bool sample_interlock_ordered:1;
 256          bool sample_interlock_unordered:1;
 257
 258          /**
 259           * Flags whether NIR's base types on the FS color outputs should be
 260           * ignored.
 261           *
 262           * GLSL requires that fragment shader output base types match the
 263           * render target's base types for the behavior to be defined.  From
 264           * the GL 4.6 spec:
 265           *
 266           *     "If the values written by the fragment shader do not match the
 267           *      format(s) of the corresponding color buffer(s), the result is
 268           *      undefined."
 269           *
 270           * However, for NIR shaders translated from TGSI, we don't have the
 271           * output types any more, so the driver will need to do whatever
 272           * fixups are necessary to handle effectively untyped data being
 273           * output from the FS.
 274           */
 275          bool untyped_color_outputs:1;
 276
 277          /** gl_FragDepth layout for ARB_conservative_depth. */
 278          enum gl_frag_depth_layout depth_layout:3;
 279       } fs;
 280
 281       struct {
 282          uint16_t local_size[3];
 283
 284          bool local_size_variable:1;
 285          uint8_t user_data_components_amd:3;
 286
 287          /*
 288           * Arrangement of invocations used to calculate derivatives in a compute
 289           * shader.  From NV_compute_shader_derivatives.
 290           */
 291          enum gl_derivative_group derivative_group:2;
 292
 293          /**
 294           * Size of shared variables accessed by the compute shader.
 295           */
 296          unsigned shared_size;
 297
 298          /**
 299           * pointer size is:
 300           *   AddressingModelLogical:    0    (default)
 301           *   AddressingModelPhysical32: 32
 302           *   AddressingModelPhysical64: 64
 303           */
 304          unsigned ptr_size;
 305       } cs;
 306
 307       /* Applies to both TCS and TES. */
 308       struct {
 309          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 310
 311          /** The number of vertices in the TCS output patch. */
 312          uint8_t tcs_vertices_out;
 313          enum gl_tess_spacing spacing:2;
 314
 315          /** Is the vertex order counterclockwise? */
 316          bool ccw:1;
 317          bool point_mode:1;
 318       } tess;
 319    };
 320 } shader_info;
 321
 322 #ifdef __cplusplus
 323 }
 324 #endif
 325
 326 #endif /* SHADER_INFO_H */