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