1/* 2 * Copyright © 2021 Valve 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 * Timur Kristóf 25 * 26 */ 27 28#include "nir.h" 29#include "nir_builder.h" 30 31typedef struct 32{ 33 struct hash_table *range_ht; 34 const nir_opt_offsets_options *options; 35} opt_offsets_state; 36 37static nir_ssa_scalar 38try_extract_const_addition(nir_builder *b, nir_ssa_scalar val, opt_offsets_state *state, unsigned *out_const, uint32_t max) 39{ 40 val = nir_ssa_scalar_chase_movs(val); 41 42 if (!nir_ssa_scalar_is_alu(val)) 43 return val; 44 45 nir_alu_instr *alu = nir_instr_as_alu(val.def->parent_instr); 46 if (alu->op != nir_op_iadd || 47 !alu->src[0].src.is_ssa || 48 !alu->src[1].src.is_ssa || 49 alu->src[0].negate || alu->src[0].abs || 50 alu->src[1].negate || alu->src[1].abs) 51 return val; 52 53 nir_ssa_scalar src[2] = { 54 {alu->src[0].src.ssa, alu->src[0].swizzle[val.comp]}, 55 {alu->src[1].src.ssa, alu->src[1].swizzle[val.comp]}, 56 }; 57 58 /* Make sure that we aren't taking out an addition that could trigger 59 * unsigned wrapping in a way that would change the semantics of the load. 60 * Ignored for ints-as-floats (lower_bitops is a proxy for that), where 61 * unsigned wrapping doesn't make sense. 62 */ 63 if (!alu->no_unsigned_wrap && !b->shader->options->lower_bitops) { 64 if (!state->range_ht) { 65 /* Cache for nir_unsigned_upper_bound */ 66 state->range_ht = _mesa_pointer_hash_table_create(NULL); 67 } 68 69 /* Check if there can really be an unsigned wrap. */ 70 uint32_t ub0 = nir_unsigned_upper_bound(b->shader, state->range_ht, src[0], NULL); 71 uint32_t ub1 = nir_unsigned_upper_bound(b->shader, state->range_ht, src[1], NULL); 72 73 if ((UINT32_MAX - ub0) < ub1) 74 return val; 75 76 /* We proved that unsigned wrap won't be possible, so we can set the flag too. */ 77 alu->no_unsigned_wrap = true; 78 } 79 80 for (unsigned i = 0; i < 2; ++i) { 81 src[i] = nir_ssa_scalar_chase_movs(src[i]); 82 if (nir_ssa_scalar_is_const(src[i])) { 83 uint32_t offset = nir_ssa_scalar_as_uint(src[i]); 84 if (offset + *out_const <= max) { 85 *out_const += offset; 86 return try_extract_const_addition(b, src[1 - i], state, out_const, max); 87 } 88 } 89 } 90 91 uint32_t orig_offset = *out_const; 92 src[0] = try_extract_const_addition(b, src[0], state, out_const, max); 93 src[1] = try_extract_const_addition(b, src[1], state, out_const, max); 94 if (*out_const == orig_offset) 95 return val; 96 97 b->cursor = nir_before_instr(&alu->instr); 98 nir_ssa_def *r = 99 nir_iadd(b, nir_channel(b, src[0].def, src[0].comp), 100 nir_channel(b, src[1].def, src[1].comp)); 101 return nir_get_ssa_scalar(r, 0); 102} 103 104static bool 105try_fold_load_store(nir_builder *b, 106 nir_intrinsic_instr *intrin, 107 opt_offsets_state *state, 108 unsigned offset_src_idx, 109 uint32_t max) 110{ 111 /* Assume that BASE is the constant offset of a load/store. 112 * Try to constant-fold additions to the offset source 113 * into the actual const offset of the instruction. 114 */ 115 116 unsigned off_const = nir_intrinsic_base(intrin); 117 nir_src *off_src = &intrin->src[offset_src_idx]; 118 nir_ssa_def *replace_src = NULL; 119 120 if (!off_src->is_ssa || off_src->ssa->bit_size != 32) 121 return false; 122 123 if (!nir_src_is_const(*off_src)) { 124 uint32_t add_offset = 0; 125 nir_ssa_scalar val = {.def = off_src->ssa, .comp = 0}; 126 val = try_extract_const_addition(b, val, state, &add_offset, max); 127 if (add_offset == 0) 128 return false; 129 off_const += add_offset; 130 b->cursor = nir_before_instr(&intrin->instr); 131 replace_src = nir_channel(b, val.def, val.comp); 132 } else if (nir_src_as_uint(*off_src) && off_const + nir_src_as_uint(*off_src) <= max) { 133 off_const += nir_src_as_uint(*off_src); 134 b->cursor = nir_before_instr(&intrin->instr); 135 replace_src = nir_imm_zero(b, off_src->ssa->num_components, off_src->ssa->bit_size); 136 } 137 138 if (!replace_src) 139 return false; 140 141 nir_instr_rewrite_src(&intrin->instr, &intrin->src[offset_src_idx], nir_src_for_ssa(replace_src)); 142 nir_intrinsic_set_base(intrin, off_const); 143 return true; 144} 145 146static bool 147try_fold_shared2(nir_builder *b, 148 nir_intrinsic_instr *intrin, 149 opt_offsets_state *state, 150 unsigned offset_src_idx) 151{ 152 unsigned comp_size = (intrin->intrinsic == nir_intrinsic_load_shared2_amd ? 153 intrin->dest.ssa.bit_size : intrin->src[0].ssa->bit_size) / 8; 154 unsigned stride = (nir_intrinsic_st64(intrin) ? 64 : 1) * comp_size; 155 unsigned offset0 = nir_intrinsic_offset0(intrin) * stride; 156 unsigned offset1 = nir_intrinsic_offset1(intrin) * stride; 157 nir_src *off_src = &intrin->src[offset_src_idx]; 158 159 if (!nir_src_is_const(*off_src)) 160 return false; 161 162 unsigned const_offset = nir_src_as_uint(*off_src); 163 offset0 += const_offset; 164 offset1 += const_offset; 165 bool st64 = offset0 % (64 * comp_size) == 0 && offset1 % (64 * comp_size) == 0; 166 stride = (st64 ? 64 : 1) * comp_size; 167 if (const_offset % stride || offset0 > 255 * stride || offset1 > 255 * stride) 168 return false; 169 170 b->cursor = nir_before_instr(&intrin->instr); 171 nir_instr_rewrite_src(&intrin->instr, off_src, nir_src_for_ssa(nir_imm_zero(b, 1, 32))); 172 nir_intrinsic_set_offset0(intrin, offset0 / stride); 173 nir_intrinsic_set_offset1(intrin, offset1 / stride); 174 nir_intrinsic_set_st64(intrin, st64); 175 176 return true; 177} 178 179static bool 180process_instr(nir_builder *b, nir_instr *instr, void *s) 181{ 182 if (instr->type != nir_instr_type_intrinsic) 183 return false; 184 185 opt_offsets_state *state = (opt_offsets_state *) s; 186 nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); 187 188 switch (intrin->intrinsic) { 189 case nir_intrinsic_load_uniform: 190 return try_fold_load_store(b, intrin, state, 0, state->options->uniform_max); 191 case nir_intrinsic_load_ubo_vec4: 192 return try_fold_load_store(b, intrin, state, 1, state->options->ubo_vec4_max); 193 case nir_intrinsic_load_shared: 194 case nir_intrinsic_load_shared_ir3: 195 return try_fold_load_store(b, intrin, state, 0, state->options->shared_max); 196 case nir_intrinsic_store_shared: 197 case nir_intrinsic_store_shared_ir3: 198 return try_fold_load_store(b, intrin, state, 1, state->options->shared_max); 199 case nir_intrinsic_load_shared2_amd: 200 return try_fold_shared2(b, intrin, state, 0); 201 case nir_intrinsic_store_shared2_amd: 202 return try_fold_shared2(b, intrin, state, 1); 203 case nir_intrinsic_load_buffer_amd: 204 return try_fold_load_store(b, intrin, state, 1, state->options->buffer_max); 205 case nir_intrinsic_store_buffer_amd: 206 return try_fold_load_store(b, intrin, state, 2, state->options->buffer_max); 207 default: 208 return false; 209 } 210 211 unreachable("Can't reach here."); 212} 213 214bool 215nir_opt_offsets(nir_shader *shader, const nir_opt_offsets_options *options) 216{ 217 opt_offsets_state state; 218 state.range_ht = NULL; 219 state.options = options; 220 221 bool p = nir_shader_instructions_pass(shader, process_instr, 222 nir_metadata_block_index | 223 nir_metadata_dominance, 224 &state); 225 226 if (state.range_ht) 227 _mesa_hash_table_destroy(state.range_ht, NULL); 228 229 230 return p; 231} 232