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 kernel;
  57    bool min_lod;
  58    bool multiview;
  59    bool physical_storage_buffer_address;
  60    bool post_depth_coverage;
  61    bool runtime_descriptor_array;
  62    bool float_controls;
  63    bool shader_clock;
  64    bool shader_viewport_index_layer;
  65    bool stencil_export;
  66    bool storage_8bit;
  67    bool storage_16bit;
  68    bool storage_image_ms;
  69    bool subgroup_arithmetic;
  70    bool subgroup_ballot;
  71    bool subgroup_basic;
  72    bool subgroup_quad;
  73    bool subgroup_shuffle;
  74    bool subgroup_vote;
  75    bool tessellation;
  76    bool transform_feedback;
  77    bool variable_pointers;
  78    bool vk_memory_model;
  79    bool vk_memory_model_device_scope;
  80    bool float16;
  81    bool amd_gcn_shader;
  82    bool amd_shader_ballot;
  83    bool amd_trinary_minmax;
  84    bool amd_image_read_write_lod;
  85 };
  86
  87 typedef struct shader_info {
  88    const char *name;
  89
  90    /* Descriptive name provided by the client; may be NULL */
  91    const char *label;
  92
  93    /** The shader stage, such as MESA_SHADER_VERTEX. */
  94    gl_shader_stage stage:8;
  95
  96    /** The shader stage in a non SSO linked program that follows this stage,
  97      * such as MESA_SHADER_FRAGMENT.
  98      */
  99    gl_shader_stage next_stage:8;
 100
 101    /* Number of textures used by this shader */
 102    uint8_t num_textures;
 103    /* Number of uniform buffers used by this shader */
 104    uint8_t num_ubos;
 105    /* Number of atomic buffers used by this shader */
 106    uint8_t num_abos;
 107    /* Number of shader storage buffers used by this shader */
 108    uint8_t num_ssbos;
 109    /* Number of images used by this shader */
 110    uint8_t num_images;
 111    /* Index of the last MSAA image. */
 112    int8_t last_msaa_image;
 113
 114    /* Which inputs are actually read */
 115    uint64_t inputs_read;
 116    /* Which outputs are actually written */
 117    uint64_t outputs_written;
 118    /* Which outputs are actually read */
 119    uint64_t outputs_read;
 120    /* Which system values are actually read */
 121    uint64_t system_values_read;
 122
 123    /* Which patch inputs are actually read */
 124    uint32_t patch_inputs_read;
 125    /* Which patch outputs are actually written */
 126    uint32_t patch_outputs_written;
 127    /* Which patch outputs are read */
 128    uint32_t patch_outputs_read;
 129
 130    /** Bitfield of which textures are used */
 131    uint32_t textures_used;
 132
 133    /** Bitfield of which textures are used by texelFetch() */
 134    uint32_t textures_used_by_txf;
 135
 136    /* SPV_KHR_float_controls: execution mode for floating point ops */
 137    uint16_t float_controls_execution_mode;
 138
 139    /* The size of the gl_ClipDistance[] array, if declared. */
 140    uint8_t clip_distance_array_size:4;
 141
 142    /* The size of the gl_CullDistance[] array, if declared. */
 143    uint8_t cull_distance_array_size:4;
 144
 145    /* Whether or not this shader ever uses textureGather() */
 146    bool uses_texture_gather:1;
 147
 148    /**
 149     * True if this shader uses the fddx/fddy opcodes.
 150     *
 151     * Note that this does not include the "fine" and "coarse" variants.
 152     */
 153    bool uses_fddx_fddy:1;
 154
 155    /**
 156     * True if this shader uses 64-bit ALU operations
 157     */
 158    bool uses_64bit:1;
 159
 160    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 161    bool first_ubo_is_default_ubo:1;
 162
 163    /* Whether or not separate shader objects were used */
 164    bool separate_shader:1;
 165
 166    /** Was this shader linked with any transform feedback varyings? */
 167    bool has_transform_feedback_varyings:1;
 168
 169    /* Whether flrp has been lowered. */
 170    bool flrp_lowered:1;
 171
 172    union {
 173       struct {
 174          /* Which inputs are doubles */
 175          uint64_t double_inputs;
 176
 177          /* For AMD-specific driver-internal shaders. It replaces vertex
 178           * buffer loads with code generating VS inputs from scalar registers.
 179           *
 180           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 181           */
 182          uint8_t blit_sgprs_amd:4;
 183
 184          /* True if the shader writes position in window space coordinates pre-transform */
 185          bool window_space_position:1;
 186       } vs;
 187
 188       struct {
 189          /** The output primitive type (GL enum value) */
 190          uint16_t output_primitive;
 191
 192          /** The input primitive type (GL enum value) */
 193          uint16_t input_primitive;
 194
 195          /** The maximum number of vertices the geometry shader might write. */
 196          uint16_t vertices_out;
 197
 198          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 199          uint8_t invocations;
 200
 201          /** The number of vertices recieves per input primitive (max. 6) */
 202          uint8_t vertices_in:3;
 203
 204          /** Whether or not this shader uses EndPrimitive */
 205          bool uses_end_primitive:1;
 206
 207          /** Whether or not this shader uses non-zero streams */
 208          bool uses_streams:1;
 209       } gs;
 210
 211       struct {
 212          bool uses_discard:1;
 213
 214          /**
 215           * True if this fragment shader requires helper invocations.  This
 216           * can be caused by the use of ALU derivative ops, texture
 217           * instructions which do implicit derivatives, and the use of quad
 218           * subgroup operations.
 219           */
 220          bool needs_helper_invocations:1;
 221
 222          /**
 223           * Whether any inputs are declared with the "sample" qualifier.
 224           */
 225          bool uses_sample_qualifier:1;
 226
 227          /**
 228           * Whether early fragment tests are enabled as defined by
 229           * ARB_shader_image_load_store.
 230           */
 231          bool early_fragment_tests:1;
 232
 233          /**
 234           * Defined by INTEL_conservative_rasterization.
 235           */
 236          bool inner_coverage:1;
 237
 238          bool post_depth_coverage:1;
 239
 240          /**
 241           * \name ARB_fragment_coord_conventions
 242           * @{
 243           */
 244          bool pixel_center_integer:1;
 245          bool origin_upper_left:1;
 246          /*@}*/
 247
 248          bool pixel_interlock_ordered:1;
 249          bool pixel_interlock_unordered:1;
 250          bool sample_interlock_ordered:1;
 251          bool sample_interlock_unordered:1;
 252
 253          /**
 254           * Flags whether NIR's base types on the FS color outputs should be
 255           * ignored.
 256           *
 257           * GLSL requires that fragment shader output base types match the
 258           * render target's base types for the behavior to be defined.  From
 259           * the GL 4.6 spec:
 260           *
 261           *     "If the values written by the fragment shader do not match the
 262           *      format(s) of the corresponding color buffer(s), the result is
 263           *      undefined."
 264           *
 265           * However, for NIR shaders translated from TGSI, we don't have the
 266           * output types any more, so the driver will need to do whatever
 267           * fixups are necessary to handle effectively untyped data being
 268           * output from the FS.
 269           */
 270          bool untyped_color_outputs:1;
 271
 272          /** gl_FragDepth layout for ARB_conservative_depth. */
 273          enum gl_frag_depth_layout depth_layout:3;
 274       } fs;
 275
 276       struct {
 277          uint16_t local_size[3];
 278
 279          bool local_size_variable:1;
 280          uint8_t user_data_components_amd:3;
 281
 282          /*
 283           * Arrangement of invocations used to calculate derivatives in a compute
 284           * shader.  From NV_compute_shader_derivatives.
 285           */
 286          enum gl_derivative_group derivative_group:2;
 287
 288          /**
 289           * Size of shared variables accessed by the compute shader.
 290           */
 291          unsigned shared_size;
 292
 293          /**
 294           * pointer size is:
 295           *   AddressingModelLogical:    0    (default)
 296           *   AddressingModelPhysical32: 32
 297           *   AddressingModelPhysical64: 64
 298           */
 299          unsigned ptr_size;
 300       } cs;
 301
 302       /* Applies to both TCS and TES. */
 303       struct {
 304          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 305
 306          /** The number of vertices in the TCS output patch. */
 307          uint8_t tcs_vertices_out;
 308          enum gl_tess_spacing spacing:2;
 309
 310          /** Is the vertex order counterclockwise? */
 311          bool ccw:1;
 312          bool point_mode:1;
 313       } tess;
 314    };
 315 } shader_info;
 316
 317 #ifdef __cplusplus
 318 }
 319 #endif
 320
 321 #endif /* SHADER_INFO_H */