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