src/compiler/nir/nir_lower_int_to_float.c

   1 /*
   2  * Copyright © 2018 Intel Corporation
   3  * Copyright © 2019 Vasily Khoruzhick <anarsoul@gmail.com>
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a
   6  * copy of this software and associated documentation files (the "Software"),
   7  * to deal in the Software without restriction, including without limitation
   8  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   9  * and/or sell copies of the Software, and to permit persons to whom the
  10  * Software is furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice (including the next
  13  * paragraph) shall be included in all copies or substantial portions of the
  14  * Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  21  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  22  * IN THE SOFTWARE.
  23  */
  24
  25 #include "nir.h"
  26 #include "nir_builder.h"
  27
  28 static bool
  29 assert_ssa_def_is_not_int(nir_ssa_def *def, void *arg)
  30 {
  31    MAYBE_UNUSED BITSET_WORD *int_types = arg;
  32    assert(!BITSET_TEST(int_types, def->index));
  33    return true;
  34 }
  35
  36 static bool
  37 rewrite_bool_ssa_def_to_float(nir_ssa_def *def, void *_progress)
  38 {
  39    bool *progress = _progress;
  40    if (def->bit_size == 1) {
  41       def->bit_size = 32;
  42       *progress = true;
  43    }
  44    return true;
  45 }
  46
  47 static bool
  48 lower_alu_instr(nir_builder *b, nir_alu_instr *alu)
  49 {
  50    const nir_op_info *op_info = &nir_op_infos[alu->op];
  51
  52    b->cursor = nir_before_instr(&alu->instr);
  53
  54    /* Replacement SSA value */
  55    nir_ssa_def *rep = NULL;
  56    switch (alu->op) {
  57    case nir_op_vec2:
  58    case nir_op_vec3:
  59    case nir_op_vec4:
  60       /* These we expect to have integers or booleans but the opcode doesn't change */
  61       break;
  62
  63    case nir_op_b2f32: alu->op = nir_op_mov; break;
  64    case nir_op_b2i32: alu->op = nir_op_mov; break;
  65    case nir_op_i2f32: alu->op = nir_op_mov; break;
  66    case nir_op_f2i32: alu->op = nir_op_mov; break;
  67    case nir_op_f2i1:
  68    case nir_op_f2b1:
  69    case nir_op_i2b1:
  70       rep = nir_sne(b, nir_ssa_for_alu_src(b, alu, 0),
  71                        nir_imm_float(b, 0));
  72       break;
  73
  74    case nir_op_flt: alu->op = nir_op_slt; break;
  75    case nir_op_fge: alu->op = nir_op_sge; break;
  76    case nir_op_feq: alu->op = nir_op_seq; break;
  77    case nir_op_fne: alu->op = nir_op_sne; break;
  78    case nir_op_ilt: alu->op = nir_op_slt; break;
  79    case nir_op_ige: alu->op = nir_op_sge; break;
  80    case nir_op_ieq: alu->op = nir_op_seq; break;
  81    case nir_op_ine: alu->op = nir_op_sne; break;
  82    case nir_op_ult: alu->op = nir_op_slt; break;
  83    case nir_op_uge: alu->op = nir_op_sge; break;
  84    case nir_op_ineg: alu->op = nir_op_fneg; break;
  85
  86    case nir_op_ball_fequal2:  alu->op = nir_op_fall_equal2; break;
  87    case nir_op_ball_fequal3:  alu->op = nir_op_fall_equal3; break;
  88    case nir_op_ball_fequal4:  alu->op = nir_op_fall_equal4; break;
  89    case nir_op_bany_fnequal2: alu->op = nir_op_fany_nequal2; break;
  90    case nir_op_bany_fnequal3: alu->op = nir_op_fany_nequal3; break;
  91    case nir_op_bany_fnequal4: alu->op = nir_op_fany_nequal4; break;
  92    case nir_op_ball_iequal2:  alu->op = nir_op_fall_equal2; break;
  93    case nir_op_ball_iequal3:  alu->op = nir_op_fall_equal3; break;
  94    case nir_op_ball_iequal4:  alu->op = nir_op_fall_equal4; break;
  95    case nir_op_bany_inequal2: alu->op = nir_op_fany_nequal2; break;
  96    case nir_op_bany_inequal3: alu->op = nir_op_fany_nequal3; break;
  97    case nir_op_bany_inequal4: alu->op = nir_op_fany_nequal4; break;
  98
  99    case nir_op_bcsel: alu->op = nir_op_fcsel; break;
 100
 101    case nir_op_iadd: alu->op = nir_op_fadd; break;
 102    case nir_op_isub: alu->op = nir_op_fsub; break;
 103    case nir_op_imul: alu->op = nir_op_fmul; break;
 104    case nir_op_iand: alu->op = nir_op_fmul; break;
 105    case nir_op_ixor: alu->op = nir_op_sne; break;
 106    case nir_op_ior: alu->op = nir_op_fmax; break;
 107
 108    case nir_op_idiv:
 109       rep = nir_ftrunc(b, nir_fdiv(b,
 110                                    nir_ssa_for_alu_src(b, alu, 0),
 111                                    nir_ssa_for_alu_src(b, alu, 1)));
 112       break;
 113
 114    case nir_op_inot:
 115       rep = nir_seq(b, nir_ssa_for_alu_src(b, alu, 0),
 116                        nir_imm_float(b, 0));
 117       break;
 118
 119    default:
 120       assert(alu->dest.dest.ssa.bit_size > 1);
 121       for (unsigned i = 0; i < op_info->num_inputs; i++)
 122          assert(alu->src[i].src.ssa->bit_size > 1);
 123       return false;
 124    }
 125
 126    if (rep) {
 127       /* We've emitted a replacement instruction */
 128       nir_ssa_def_rewrite_uses(&alu->dest.dest.ssa, nir_src_for_ssa(rep));
 129       nir_instr_remove(&alu->instr);
 130    } else {
 131       if (alu->dest.dest.ssa.bit_size == 1)
 132          alu->dest.dest.ssa.bit_size = 32;
 133    }
 134
 135    return true;
 136 }
 137
 138 static bool
 139 nir_lower_int_to_float_impl(nir_function_impl *impl)
 140 {
 141    bool progress = false;
 142    BITSET_WORD *float_types = NULL, *int_types = NULL;
 143
 144    nir_builder b;
 145    nir_builder_init(&b, impl);
 146
 147    nir_index_ssa_defs(impl);
 148    float_types = calloc(BITSET_WORDS(impl->ssa_alloc),
 149                         sizeof(BITSET_WORD));
 150    int_types = calloc(BITSET_WORDS(impl->ssa_alloc),
 151                       sizeof(BITSET_WORD));
 152    nir_gather_ssa_types(impl, float_types, int_types);
 153
 154    nir_foreach_block(block, impl) {
 155       nir_foreach_instr_safe(instr, block) {
 156          switch (instr->type) {
 157          case nir_instr_type_alu:
 158             progress |= lower_alu_instr(&b, nir_instr_as_alu(instr));
 159             break;
 160
 161          case nir_instr_type_load_const: {
 162             nir_load_const_instr *load = nir_instr_as_load_const(instr);
 163             if (load->def.bit_size == 1) {
 164                for (unsigned i = 0; i < load->def.num_components; i++)
 165                   load->value[i].f32 = load->value[i].b ? 1.0 : 0.0;
 166                load->def.bit_size = 32;
 167             } else if (BITSET_TEST(int_types, load->def.index)) {
 168                for (unsigned i = 0; i < load->def.num_components; i++)
 169                   load->value[i].f32 = load->value[i].i32;
 170             }
 171             break;
 172          }
 173
 174          case nir_instr_type_intrinsic:
 175          case nir_instr_type_ssa_undef:
 176          case nir_instr_type_phi:
 177          case nir_instr_type_tex:
 178             nir_foreach_ssa_def(instr, rewrite_bool_ssa_def_to_float,
 179                                 &progress);
 180             break;
 181
 182          default:
 183             nir_foreach_ssa_def(instr, assert_ssa_def_is_not_int, (void *)int_types);
 184             break;
 185          }
 186       }
 187    }
 188
 189    if (progress) {
 190       nir_metadata_preserve(impl, nir_metadata_block_index |
 191                                   nir_metadata_dominance);
 192    }
 193
 194    free(float_types);
 195    free(int_types);
 196
 197    return progress;
 198 }
 199
 200 bool
 201 nir_lower_int_to_float(nir_shader *shader)
 202 {
 203    bool progress = false;
 204
 205    nir_foreach_function(function, shader) {
 206       if (function->impl && nir_lower_int_to_float_impl(function->impl))
 207          progress = true;
 208    }
 209
 210    return progress;
 211 }