src/panfrost/bifrost/bi_special.c

   1 /*
   2  * Copyright (C) 2020 Collabora Ltd.
   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 FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  *
  23  * Authors (Collabora):
  24  *      Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
  25  */
  26
  27 #include "compiler.h"
  28
  29 /* Bifrost requires special functions to be lowered in various machine specific
  30  * ways. The routines in this file are used in codegen for this. */
  31
  32 /* New Bifrost has a FEXP2_FAST instruction but requires an auxiliary
  33  * parameter. */
  34
  35 static void
  36 bi_emit_fexp2_new(bi_context *ctx, nir_alu_instr *instr)
  37 {
  38         /* FMA_MSCALE T, X, 1.0, 0, 0x18 */
  39
  40         bi_instruction mscale = {
  41                 .type = BI_FMA,
  42                 .op = { .mscale = true },
  43                 .dest = bi_make_temp(ctx),
  44                 .dest_type = nir_type_float32,
  45                 .src = {
  46                         pan_src_index(&instr->src[0].src),
  47                         BIR_INDEX_CONSTANT | 0,
  48                         BIR_INDEX_ZERO,
  49                         BIR_INDEX_CONSTANT | 32,
  50                 },
  51                 .src_types = {
  52                         nir_type_float32,
  53                         nir_type_float32,
  54                         nir_type_float32,
  55                         nir_type_int32,
  56                 },
  57                 .constant = {
  58                         /* 0x3f80000000 = 1.0f as fp32
  59                          * 24 = shift to multiply by 2^24 */
  60                         .u64 = (0x3f800000) | (24ull << 32)
  61                 }
  62         };
  63
  64         /* F2I_RTE T, T */
  65
  66         bi_instruction f2i = {
  67                 .type = BI_CONVERT,
  68                 .dest = bi_make_temp(ctx),
  69                 .dest_type = nir_type_int32,
  70                 .src = { mscale.dest },
  71                 .src_types = { nir_type_float32 },
  72                 .roundmode = BIFROST_RTE
  73         };
  74
  75         /* FEXP2_FAST T, T, X */
  76
  77         bi_instruction fexp = {
  78                 .type = BI_SPECIAL,
  79                 .op = { .special = BI_SPECIAL_EXP2_LOW },
  80                 .dest = pan_dest_index(&instr->dest.dest),
  81                 .dest_type = nir_type_float32,
  82                 .src = { f2i.dest, mscale.src[0] },
  83                 .src_types = { nir_type_int32, nir_type_float32 },
  84         };
  85
  86         bi_emit(ctx, mscale);
  87         bi_emit(ctx, f2i);
  88         bi_emit(ctx, fexp);
  89 }
  90
  91 /* Even on new Bifrost, there are a bunch of reductions to do */
  92
  93 static void
  94 bi_emit_flog2_new(bi_context *ctx, nir_alu_instr *instr)
  95 {
  96         /* LOG_FREXPE X */
  97         bi_instruction frexpe = {
  98                 .type = BI_FREXP,
  99                 .op = { .frexp = BI_FREXPE_LOG },
 100                 .dest = bi_make_temp(ctx),
 101                 .dest_type = nir_type_int32,
 102                 .src = { pan_src_index(&instr->src[0].src) },
 103                 .src_types = { nir_type_float32 }
 104         };
 105
 106         /* I32_TO_F32 m */
 107         bi_instruction i2f = {
 108                 .type = BI_CONVERT,
 109                 .dest = bi_make_temp(ctx),
 110                 .dest_type = nir_type_float32,
 111                 .src = { frexpe.dest },
 112                 .src_types = { nir_type_int32 },
 113                 .roundmode = BIFROST_RTZ
 114         };
 115
 116         /* ADD_FREXPM (x-1), -1.0, X */
 117         bi_instruction x_minus_1 = {
 118                 .type = BI_REDUCE_FMA,
 119                 .op = { .reduce = BI_REDUCE_ADD_FREXPM },
 120                 .dest = bi_make_temp(ctx),
 121                 .dest_type = nir_type_float32,
 122                 .src = {
 123                         BIR_INDEX_CONSTANT,
 124                         pan_src_index(&instr->src[0].src),
 125                 },
 126                 .src_types = { nir_type_float32, nir_type_float32 },
 127                 .constant = {
 128                         .u64 = 0xBF800000 /* -1.0 */
 129                 }
 130         };
 131
 132         /* FLOG2_HELP log2(x)/(x-1), x */
 133         bi_instruction help = {
 134                 .type = BI_TABLE,
 135                 .op = { .table = BI_TABLE_LOG2_U_OVER_U_1_LOW },
 136                 .dest = bi_make_temp(ctx),
 137                 .dest_type = nir_type_float32,
 138                 .src = { pan_src_index(&instr->src[0].src) },
 139                 .src_types = { nir_type_float32 },
 140         };
 141
 142         /* FMA log2(x)/(x - 1), (x - 1), M */
 143         bi_instruction fma = {
 144                 .type = BI_FMA,
 145                 .dest = pan_dest_index(&instr->dest.dest),
 146                 .dest_type = nir_type_float32,
 147                 .src = {
 148                         help.dest,
 149                         x_minus_1.dest,
 150                         i2f.dest
 151                 },
 152                 .src_types = {
 153                         nir_type_float32,
 154                         nir_type_float32,
 155                         nir_type_float32
 156                 }
 157         };
 158
 159         bi_emit(ctx, frexpe);
 160         bi_emit(ctx, i2f);
 161         bi_emit(ctx, x_minus_1);
 162         bi_emit(ctx, help);
 163         bi_emit(ctx, fma);
 164 }
 165
 166 void
 167 bi_emit_fexp2(bi_context *ctx, nir_alu_instr *instr)
 168 {
 169         /* TODO: G71 */
 170         bi_emit_fexp2_new(ctx, instr);
 171 }
 172
 173 void
 174 bi_emit_flog2(bi_context *ctx, nir_alu_instr *instr)
 175 {
 176         /* TODO: G71 */
 177         bi_emit_flog2_new(ctx, instr);
 178 }