src/glsl/nir/nir_intrinsics.h

   1 /*
   2  * Copyright © 2014 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  * Authors:
  24  *    Connor Abbott (cwabbott0@gmail.com)
  25  *
  26  */
  27
  28 /**
  29  * This header file defines all the available intrinsics in one place. It
  30  * expands to a list of macros of the form:
  31  *
  32  * INTRINSIC(name, num_srcs, src_components, has_dest, dest_components,
  33  *              num_variables, num_indices, flags)
  34  *
  35  * Which should correspond one-to-one with the nir_intrinsic_info structure. It
  36  * is included in both ir.h to create the nir_intrinsic enum (with members of
  37  * the form nir_intrinsic_(name)) and and in opcodes.c to create
  38  * nir_intrinsic_infos, which is a const array of nir_intrinsic_info structures
  39  * for each intrinsic.
  40  */
  41
  42 #define ARR(...) { __VA_ARGS__ }
  43
  44
  45 INTRINSIC(load_var, 0, ARR(), true, 0, 1, 0, NIR_INTRINSIC_CAN_ELIMINATE)
  46 INTRINSIC(store_var, 1, ARR(0), false, 0, 1, 0, 0)
  47 INTRINSIC(copy_var, 0, ARR(), false, 0, 2, 0, 0)
  48
  49 /*
  50  * Interpolation of input.  The interp_var_at* intrinsics are similar to the
  51  * load_var intrinsic acting an a shader input except that they interpolate
  52  * the input differently.  The at_sample and at_offset intrinsics take an
  53  * aditional source that is a integer sample id or a vec2 position offset
  54  * respectively.
  55  */
  56
  57 INTRINSIC(interp_var_at_centroid, 0, ARR(0), true, 0, 1, 0,
  58           NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
  59 INTRINSIC(interp_var_at_sample, 1, ARR(1), true, 0, 1, 0,
  60           NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
  61 INTRINSIC(interp_var_at_offset, 1, ARR(2), true, 0, 1, 0,
  62           NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
  63
  64 /*
  65  * a barrier is an intrinsic with no inputs/outputs but which can't be moved
  66  * around/optimized in general
  67  */
  68 #define BARRIER(name) INTRINSIC(name, 0, ARR(), false, 0, 0, 0, 0)
  69
  70 BARRIER(barrier)
  71 BARRIER(discard)
  72
  73 /*
  74  * Memory barrier with semantics analogous to the memoryBarrier() GLSL
  75  * intrinsic.
  76  */
  77 BARRIER(memory_barrier)
  78
  79 /** A conditional discard, with a single boolean source. */
  80 INTRINSIC(discard_if, 1, ARR(1), false, 0, 0, 0, 0)
  81
  82 INTRINSIC(emit_vertex,   0, ARR(), false, 0, 0, 1, 0)
  83 INTRINSIC(end_primitive, 0, ARR(), false, 0, 0, 1, 0)
  84
  85 /*
  86  * Atomic counters
  87  *
  88  * The *_var variants take an atomic_uint nir_variable, while the other,
  89  * lowered, variants take a constant buffer index and register offset.
  90  */
  91
  92 #define ATOMIC(name, flags) \
  93    INTRINSIC(atomic_counter_##name##_var, 0, ARR(), true, 1, 1, 0, flags) \
  94    INTRINSIC(atomic_counter_##name, 1, ARR(1), true, 1, 0, 1, flags)
  95
  96 ATOMIC(inc, 0)
  97 ATOMIC(dec, 0)
  98 ATOMIC(read, NIR_INTRINSIC_CAN_ELIMINATE)
  99
 100 /*
 101  * Image load, store and atomic intrinsics.
 102  *
 103  * All image intrinsics take an image target passed as a nir_variable.  Image
 104  * variables contain a number of memory and layout qualifiers that influence
 105  * the semantics of the intrinsic.
 106  *
 107  * All image intrinsics take a four-coordinate vector and a sample index as
 108  * first two sources, determining the location within the image that will be
 109  * accessed by the intrinsic.  Components not applicable to the image target
 110  * in use are undefined.  Image store takes an additional four-component
 111  * argument with the value to be written, and image atomic operations take
 112  * either one or two additional scalar arguments with the same meaning as in
 113  * the ARB_shader_image_load_store specification.
 114  */
 115 INTRINSIC(image_load, 2, ARR(4, 1), true, 4, 1, 0,
 116           NIR_INTRINSIC_CAN_ELIMINATE)
 117 INTRINSIC(image_store, 3, ARR(4, 1, 4), false, 0, 1, 0, 0)
 118 INTRINSIC(image_atomic_add, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 119 INTRINSIC(image_atomic_min, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 120 INTRINSIC(image_atomic_max, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 121 INTRINSIC(image_atomic_and, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 122 INTRINSIC(image_atomic_or, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 123 INTRINSIC(image_atomic_xor, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 124 INTRINSIC(image_atomic_exchange, 3, ARR(4, 1, 1), true, 1, 1, 0, 0)
 125 INTRINSIC(image_atomic_comp_swap, 4, ARR(4, 1, 1, 1), true, 1, 1, 0, 0)
 126 INTRINSIC(image_size, 0, ARR(), true, 4, 1, 0,
 127           NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 128
 129 #define SYSTEM_VALUE(name, components) \
 130    INTRINSIC(load_##name, 0, ARR(), true, components, 0, 0, \
 131    NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 132
 133 SYSTEM_VALUE(front_face, 1)
 134 SYSTEM_VALUE(vertex_id, 1)
 135 SYSTEM_VALUE(vertex_id_zero_base, 1)
 136 SYSTEM_VALUE(base_vertex, 1)
 137 SYSTEM_VALUE(instance_id, 1)
 138 SYSTEM_VALUE(sample_id, 1)
 139 SYSTEM_VALUE(sample_pos, 2)
 140 SYSTEM_VALUE(sample_mask_in, 1)
 141 SYSTEM_VALUE(invocation_id, 1)
 142
 143 /*
 144  * The format of the indices depends on the type of the load.  For uniforms,
 145  * the first index is the base address and the second index is an offset that
 146  * should be added to the base address.  (This way you can determine in the
 147  * back-end which variable is being accessed even in an array.)  For inputs,
 148  * the one and only index corresponds to the attribute slot.  UBO loads
 149  * have two indices the first of which is the descriptor set and the second
 150  * is the base address to load from.
 151  *
 152  * UBO loads have a (possibly constant) source which is the UBO buffer index.
 153  * For each type of load, the _indirect variant has one additional source
 154  * (the second in the case of UBO's) that is the is an indirect to be added to
 155  * the constant address or base offset to compute the final offset.
 156  *
 157  * For vector backends, the address is in terms of one vec4, and so each array
 158  * element is +4 scalar components from the previous array element. For scalar
 159  * backends, the address is in terms of a single 4-byte float/int and arrays
 160  * elements begin immediately after the previous array element.
 161  */
 162
 163 #define LOAD(name, extra_srcs, indices, flags) \
 164    INTRINSIC(load_##name, extra_srcs, ARR(1), true, 0, 0, indices, flags) \
 165    INTRINSIC(load_##name##_indirect, extra_srcs + 1, ARR(1, 1), \
 166              true, 0, 0, indices, flags)
 167
 168 LOAD(uniform, 0, 2, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 169 LOAD(ubo, 1, 2, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 170 LOAD(input, 0, 1, NIR_INTRINSIC_CAN_ELIMINATE | NIR_INTRINSIC_CAN_REORDER)
 171 /* LOAD(ssbo, 1, 0) */
 172
 173 /*
 174  * Stores work the same way as loads, except now the first register input is
 175  * the value or array to store and the optional second input is the indirect
 176  * offset.
 177  */
 178
 179 #define STORE(name, num_indices, flags) \
 180    INTRINSIC(store_##name, 1, ARR(0), false, 0, 0, num_indices, flags) \
 181    INTRINSIC(store_##name##_indirect, 2, ARR(0, 1), false, 0, 0, \
 182              num_indices, flags) \
 183
 184 STORE(output, 1, 0)
 185 /* STORE(ssbo, 2, 0) */
 186
 187 LAST_INTRINSIC(store_output_indirect)