1b8e80941Smrg/* 2b8e80941Smrg * Copyright © 2018 Valve Corporation 3b8e80941Smrg * Copyright © 2017 Red Hat 4b8e80941Smrg * 5b8e80941Smrg * Permission is hereby granted, free of charge, to any person obtaining a 6b8e80941Smrg * copy of this software and associated documentation files (the "Software"), 7b8e80941Smrg * to deal in the Software without restriction, including without limitation 8b8e80941Smrg * the rights to use, copy, modify, merge, publish, distribute, sublicense, 9b8e80941Smrg * and/or sell copies of the Software, and to permit persons to whom the 10b8e80941Smrg * Software is furnished to do so, subject to the following conditions: 11b8e80941Smrg * 12b8e80941Smrg * The above copyright notice and this permission notice (including the next 13b8e80941Smrg * paragraph) shall be included in all copies or substantial portions of the 14b8e80941Smrg * Software. 15b8e80941Smrg * 16b8e80941Smrg * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17b8e80941Smrg * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18b8e80941Smrg * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19b8e80941Smrg * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20b8e80941Smrg * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 21b8e80941Smrg * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 22b8e80941Smrg * IN THE SOFTWARE. 23b8e80941Smrg * 24b8e80941Smrg */ 25b8e80941Smrg 26b8e80941Smrg#include "vtn_private.h" 27b8e80941Smrg#include "GLSL.ext.AMD.h" 28b8e80941Smrg 29b8e80941Smrgbool 30b8e80941Smrgvtn_handle_amd_gcn_shader_instruction(struct vtn_builder *b, SpvOp ext_opcode, 31b8e80941Smrg const uint32_t *w, unsigned count) 32b8e80941Smrg{ 33b8e80941Smrg const struct glsl_type *dest_type = 34b8e80941Smrg vtn_value(b, w[1], vtn_value_type_type)->type->type; 35b8e80941Smrg struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); 36b8e80941Smrg val->ssa = vtn_create_ssa_value(b, dest_type); 37b8e80941Smrg 38b8e80941Smrg switch ((enum GcnShaderAMD)ext_opcode) { 39b8e80941Smrg case CubeFaceIndexAMD: 40b8e80941Smrg val->ssa->def = nir_cube_face_index(&b->nb, vtn_ssa_value(b, w[5])->def); 41b8e80941Smrg break; 42b8e80941Smrg case CubeFaceCoordAMD: 43b8e80941Smrg val->ssa->def = nir_cube_face_coord(&b->nb, vtn_ssa_value(b, w[5])->def); 44b8e80941Smrg break; 45b8e80941Smrg case TimeAMD: { 46b8e80941Smrg nir_intrinsic_instr *intrin = nir_intrinsic_instr_create(b->nb.shader, 47b8e80941Smrg nir_intrinsic_shader_clock); 48b8e80941Smrg nir_ssa_dest_init(&intrin->instr, &intrin->dest, 2, 32, NULL); 49b8e80941Smrg nir_builder_instr_insert(&b->nb, &intrin->instr); 50b8e80941Smrg val->ssa->def = nir_pack_64_2x32(&b->nb, &intrin->dest.ssa); 51b8e80941Smrg break; 52b8e80941Smrg } 53b8e80941Smrg default: 54b8e80941Smrg unreachable("Invalid opcode"); 55b8e80941Smrg } 56b8e80941Smrg return true; 57b8e80941Smrg} 58b8e80941Smrg 59b8e80941Smrgbool 60b8e80941Smrgvtn_handle_amd_shader_trinary_minmax_instruction(struct vtn_builder *b, SpvOp ext_opcode, 61b8e80941Smrg const uint32_t *w, unsigned count) 62b8e80941Smrg{ 63b8e80941Smrg struct nir_builder *nb = &b->nb; 64b8e80941Smrg const struct glsl_type *dest_type = 65b8e80941Smrg vtn_value(b, w[1], vtn_value_type_type)->type->type; 66b8e80941Smrg struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa); 67b8e80941Smrg val->ssa = vtn_create_ssa_value(b, dest_type); 68b8e80941Smrg 69b8e80941Smrg unsigned num_inputs = count - 5; 70b8e80941Smrg assert(num_inputs == 3); 71b8e80941Smrg nir_ssa_def *src[3] = { NULL, }; 72b8e80941Smrg for (unsigned i = 0; i < num_inputs; i++) 73b8e80941Smrg src[i] = vtn_ssa_value(b, w[i + 5])->def; 74b8e80941Smrg 75b8e80941Smrg switch ((enum ShaderTrinaryMinMaxAMD)ext_opcode) { 76b8e80941Smrg case FMin3AMD: 77b8e80941Smrg val->ssa->def = nir_fmin3(nb, src[0], src[1], src[2]); 78b8e80941Smrg break; 79b8e80941Smrg case UMin3AMD: 80b8e80941Smrg val->ssa->def = nir_umin3(nb, src[0], src[1], src[2]); 81b8e80941Smrg break; 82b8e80941Smrg case SMin3AMD: 83b8e80941Smrg val->ssa->def = nir_imin3(nb, src[0], src[1], src[2]); 84b8e80941Smrg break; 85b8e80941Smrg case FMax3AMD: 86b8e80941Smrg val->ssa->def = nir_fmax3(nb, src[0], src[1], src[2]); 87b8e80941Smrg break; 88b8e80941Smrg case UMax3AMD: 89b8e80941Smrg val->ssa->def = nir_umax3(nb, src[0], src[1], src[2]); 90b8e80941Smrg break; 91b8e80941Smrg case SMax3AMD: 92b8e80941Smrg val->ssa->def = nir_imax3(nb, src[0], src[1], src[2]); 93b8e80941Smrg break; 94b8e80941Smrg case FMid3AMD: 95b8e80941Smrg val->ssa->def = nir_fmed3(nb, src[0], src[1], src[2]); 96b8e80941Smrg break; 97b8e80941Smrg case UMid3AMD: 98b8e80941Smrg val->ssa->def = nir_umed3(nb, src[0], src[1], src[2]); 99b8e80941Smrg break; 100b8e80941Smrg case SMid3AMD: 101b8e80941Smrg val->ssa->def = nir_imed3(nb, src[0], src[1], src[2]); 102b8e80941Smrg break; 103b8e80941Smrg default: 104b8e80941Smrg unreachable("unknown opcode\n"); 105b8e80941Smrg break; 106b8e80941Smrg } 107b8e80941Smrg 108b8e80941Smrg return true; 109b8e80941Smrg} 110