// Copyright (c) Epic Games Tools // Licensed under the MIT license (https://opensource.org/license/mit/) #include "lib_rdi_make/rdi_make.c" internal RDIM_DataModel rdim_data_model_from_os_arch(OperatingSystem os, RDI_Arch arch) { RDIM_DataModel data_model = RDIM_DataModel_Null; #define Case(os_name, arch_name, model_name) if(os == OperatingSystem_##os_name && arch == Arch_##arch_name) { data_model = RDIM_DataModel_##model_name; } Case(Windows, x86, LLP64); Case(Windows, x64, LLP64); Case(Linux, x86, ILP32); Case(Linux, x64, LLP64); Case(Mac, x64, LP64); #undef Case return data_model; } internal RDIM_TopLevelInfo rdim_make_top_level_info(String8 image_name, Arch arch, U64 exe_hash, RDIM_BinarySectionList sections) { // convert arch RDI_Arch arch_rdi = RDI_Arch_NULL; switch(arch) { default:{}break; case Arch_x64:{arch_rdi = RDI_Arch_X64;}break; case Arch_x86:{arch_rdi = RDI_Arch_X86;}break; } // find max VOFF U64 exe_voff_max = 0; for EachNode(sect_n, RDIM_BinarySectionNode, sections.first) { exe_voff_max = Max(exe_voff_max, sect_n->v.voff_opl); } // fill out top level info RDIM_TopLevelInfo top_level_info = {0}; top_level_info.arch = arch_rdi; top_level_info.exe_hash = exe_hash; top_level_info.voff_max = exe_voff_max; top_level_info.producer_name = str8_lit(BUILD_TITLE_STRING_LITERAL); return top_level_info; } internal RDIM_BakeResults rdim_bake(Arena *arena, RDIM_BakeParams *params) { ////////////////////////////////////////////////////////////// //- rjf: set up shared state // if(lane_idx() == 0) { rdim_shared = push_array(arena, RDIM_Shared, 1); } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake vmaps // ProfScope("bake vmaps") { Temp scratch = scratch_begin(&arena, 1); #pragma pack(push, 1) typedef struct VMapRecord VMapRecord; struct VMapRecord { union { struct { U32 negative_size; U64 voff; }; U8 digits[12]; } key; U32 idx; }; #pragma pack(pop) //////////////////////////// //- rjf: gather unsorted scope vmap records // VMapRecord *scope_vmap_records = 0; U64 scope_vmap_records_count = 0; ProfScope("gather unsorted scope vmap records") { //- rjf: calculate per-lane-chunk counts U64 *lane_chunk_range_counts = 0; if(lane_idx() == 0) { lane_chunk_range_counts = push_array(scratch.arena, U64, params->scopes.chunk_count * lane_count()); } lane_sync_u64(&lane_chunk_range_counts, 0); { U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { U64 slot_idx = lane_idx()*params->scopes.chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { lane_chunk_range_counts[slot_idx] += n->v[n_idx].voff_ranges.count; } chunk_idx += 1; } } lane_sync(); //- rjf: calculate per-lane-chunk offsets U64 *lane_chunk_range_offs = 0; U64 total_range_count = 0; if(lane_idx() == 0) { lane_chunk_range_offs = push_array(scratch.arena, U64, params->scopes.chunk_count * lane_count()); U64 off = 0; U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->scopes.chunk_count + chunk_idx; lane_chunk_range_offs[slot_idx] = off; off += lane_chunk_range_counts[slot_idx]; } chunk_idx += 1; } total_range_count = off; } lane_sync_u64(&lane_chunk_range_offs, 0); lane_sync_u64(&total_range_count, 0); //- rjf: allocate records if(lane_idx() == 0) { scope_vmap_records_count = total_range_count; scope_vmap_records = push_array_no_zero(scratch.arena, VMapRecord, scope_vmap_records_count); } lane_sync_u64(&scope_vmap_records, 0); lane_sync_u64(&scope_vmap_records_count, 0); //- rjf: fill records { U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { U64 slot_idx = lane_idx()*params->scopes.chunk_count + chunk_idx; U64 off = lane_chunk_range_offs[slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDI_U32 scope_idx = (RDI_U32)rdim_idx_from_scope(&n->v[n_idx]); // TODO(rjf): @u64_to_u32 for EachNode(rng_n, RDIM_Rng1U64Node, n->v[n_idx].voff_ranges.first) { scope_vmap_records[off].key.voff = rng_n->v.min; scope_vmap_records[off].key.negative_size = -(RDI_U32)(rng_n->v.max - rng_n->v.min); scope_vmap_records[off].idx = scope_idx; off += 1; } } chunk_idx += 1; } } } lane_sync(); //////////////////////////// //- rjf: gather unsorted global vmap records // VMapRecord *global_vmap_records = 0; U64 global_vmap_records_count = 0; ProfScope("gather unsorted global vmap records") { //- rjf: calculate per-lane-chunk counts U64 *lane_chunk_range_counts = 0; if(lane_idx() == 0) { lane_chunk_range_counts = push_array(scratch.arena, U64, params->global_variables.chunk_count * lane_count()); } lane_sync_u64(&lane_chunk_range_counts, 0); { U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->global_variables.first) { U64 slot_idx = lane_idx()*params->global_variables.chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); lane_chunk_range_counts[slot_idx] += dim_1u64(range); chunk_idx += 1; } } lane_sync(); //- rjf: calculate per-lane-chunk offsets U64 *lane_chunk_range_offs = 0; U64 total_range_count = 0; if(lane_idx() == 0) { lane_chunk_range_offs = push_array(scratch.arena, U64, params->global_variables.chunk_count * lane_count()); U64 off = 0; U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->global_variables.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->global_variables.chunk_count + chunk_idx; lane_chunk_range_offs[slot_idx] = off; off += lane_chunk_range_counts[slot_idx]; } chunk_idx += 1; } total_range_count = off; } lane_sync_u64(&lane_chunk_range_offs, 0); lane_sync_u64(&total_range_count, 0); //- rjf: allocate records if(lane_idx() == 0) { global_vmap_records_count = total_range_count; global_vmap_records = push_array_no_zero(scratch.arena, VMapRecord, global_vmap_records_count); } lane_sync_u64(&global_vmap_records, 0); lane_sync_u64(&global_vmap_records_count, 0); //- rjf: fill records { U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->global_variables.first) { U64 slot_idx = lane_idx()*params->global_variables.chunk_count + chunk_idx; U64 off = lane_chunk_range_offs[slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDI_U32 global_idx = (RDI_U32)rdim_idx_from_symbol(&n->v[n_idx]); // TODO(rjf): @u64_to_u32 RDI_U32 global_size = (RDI_U32)(n->v[n_idx].type ? n->v[n_idx].type->byte_size : 1); RDI_U64 global_voff = n->v[n_idx].offset; global_vmap_records[off].key.voff = global_voff; global_vmap_records[off].key.negative_size = -global_size; global_vmap_records[off].idx = global_idx; off += 1; } chunk_idx += 1; } } } lane_sync(); //////////////////////////// //- rjf: gather unsorted unit vmap records // VMapRecord *unit_vmap_records = 0; U64 unit_vmap_records_count = 0; ProfScope("gather unsorted unit vmap records") { //- rjf: calculate per-lane-chunk counts U64 *lane_chunk_range_counts = 0; if(lane_idx() == 0) { lane_chunk_range_counts = push_array(scratch.arena, U64, params->units.chunk_count * lane_count()); } lane_sync_u64(&lane_chunk_range_counts, 0); { U64 chunk_idx = 0; for EachNode(n, RDIM_UnitChunkNode, params->units.first) { U64 slot_idx = lane_idx()*params->units.chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { lane_chunk_range_counts[slot_idx] += n->v[n_idx].voff_ranges.total_count; } chunk_idx += 1; } } lane_sync(); //- rjf: calculate per-lane-chunk offsets U64 *lane_chunk_range_offs = 0; U64 total_range_count = 0; if(lane_idx() == 0) { lane_chunk_range_offs = push_array(scratch.arena, U64, params->units.chunk_count * lane_count()); U64 off = 0; U64 chunk_idx = 0; for EachNode(n, RDIM_UnitChunkNode, params->units.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->units.chunk_count + chunk_idx; lane_chunk_range_offs[slot_idx] = off; off += lane_chunk_range_counts[slot_idx]; } chunk_idx += 1; } total_range_count = off; } lane_sync_u64(&lane_chunk_range_offs, 0); lane_sync_u64(&total_range_count, 0); //- rjf: allocate records if(lane_idx() == 0) { unit_vmap_records_count = total_range_count; unit_vmap_records = push_array_no_zero(scratch.arena, VMapRecord, unit_vmap_records_count); } lane_sync_u64(&unit_vmap_records, 0); lane_sync_u64(&unit_vmap_records_count, 0); //- rjf: fill records { U64 chunk_idx = 0; for EachNode(n, RDIM_UnitChunkNode, params->units.first) { U64 slot_idx = lane_idx()*params->units.chunk_count + chunk_idx; U64 off = lane_chunk_range_offs[slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDI_U32 unit_idx = (RDI_U32)rdim_idx_from_unit(&n->v[n_idx]); // TODO(rjf): @u64_to_u32 for EachNode(rng_n, RDIM_Rng1U64ChunkNode, n->v[n_idx].voff_ranges.first) { for EachIndex(rng_n_idx, rng_n->count) { unit_vmap_records[off].key.voff = rng_n->v[rng_n_idx].min; unit_vmap_records[off].key.negative_size = -(RDI_U32)(rng_n->v[rng_n_idx].max - rng_n->v[rng_n_idx].min); unit_vmap_records[off].idx = unit_idx; off += 1; } } } chunk_idx += 1; } } } lane_sync(); //////////////////////////// //- rjf: sort & bake all vmaps // struct { String8 name; VMapRecord *records; U64 records_count; RDI_VMapEntry **vmap_out; U32 *vmap_count_out; } vmap_tasks[] = { {str8_lit_comp("scopes"), scope_vmap_records, scope_vmap_records_count, &rdim_shared->baked_scope_vmap.vmap.vmap, &rdim_shared->baked_scope_vmap.vmap.count}, {str8_lit_comp("globals"), global_vmap_records, global_vmap_records_count, &rdim_shared->baked_global_vmap.vmap.vmap, &rdim_shared->baked_global_vmap.vmap.count}, {str8_lit_comp("units"), unit_vmap_records, unit_vmap_records_count, &rdim_shared->baked_unit_vmap.vmap.vmap, &rdim_shared->baked_unit_vmap.vmap.count}, }; ProfScope("sort & bake all vmaps") { for EachElement(vmap_task_idx, vmap_tasks) ProfScope("sort & bake vmap for %.*s", str8_varg(vmap_tasks[vmap_task_idx].name)) { VMapRecord *records = vmap_tasks[vmap_task_idx].records; U64 records_count = vmap_tasks[vmap_task_idx].records_count; //////////////////////// //- rjf: sort // ProfScope("sort") { //- rjf: set up constants U64 bytes_per_digit = 1; U64 num_possible_values_per_digit = 1<<(bytes_per_digit*8); U64 digits_count = sizeof(((VMapRecord *)0)->key)/bytes_per_digit; //- rjf: set up swap buffer / lane counters VMapRecord *records__swap = 0; U32 **lane_digit_counts = 0; U32 **lane_digit_offs = 0; if(lane_idx() == 0) { records__swap = push_array_no_zero(scratch.arena, VMapRecord, records_count); lane_digit_counts = push_array_no_zero(scratch.arena, U32 *, lane_count()); lane_digit_offs = push_array_no_zero(scratch.arena, U32 *, lane_count()); } lane_sync_u64(&records__swap, 0); lane_sync_u64(&lane_digit_counts, 0); lane_sync_u64(&lane_digit_offs, 0); lane_digit_counts[lane_idx()] = push_array_no_zero(scratch.arena, U32, num_possible_values_per_digit); lane_digit_offs[lane_idx()] = push_array_no_zero(scratch.arena, U32, num_possible_values_per_digit); //- rjf: do all sort passes { VMapRecord *src = records; VMapRecord *dst = records__swap; for EachIndex(digit_idx, digits_count) { // rjf: count digit value occurrences per-lane { U32 *digit_counts = lane_digit_counts[lane_idx()]; MemoryZero(digit_counts, sizeof(digit_counts[0])*num_possible_values_per_digit); Rng1U64 range = lane_range(records_count); for EachInRange(idx, range) { VMapRecord *rec = &src[idx]; U16 digit_value = (U16)rec->key.digits[digit_idx]; digit_counts[digit_value] += 1; } } lane_sync(); // rjf: compute thread * digit value *relative* offset table { Rng1U64 range = lane_range(num_possible_values_per_digit); for EachInRange(value_idx, range) { U64 layout_off = 0; for EachIndex(lane_idx, lane_count()) { lane_digit_offs[lane_idx][value_idx] = layout_off; layout_off += lane_digit_counts[lane_idx][value_idx]; } } } lane_sync(); // rjf: convert relative offsets -> absolute offsets if(lane_idx() == 0) { U64 last_off = 0; U64 num_of_nonzero_digit = 0; for EachIndex(value_idx, num_possible_values_per_digit) { for EachIndex(lane_idx, lane_count()) { lane_digit_offs[lane_idx][value_idx] += last_off; } last_off = lane_digit_offs[lane_count()-1][value_idx] + lane_digit_counts[lane_count()-1][value_idx]; } // NOTE(rjf): required that: (last_off == element_count) } lane_sync(); // rjf: move { U32 *lane_digit_offsets = lane_digit_offs[lane_idx()]; Rng1U64 range = lane_range(records_count); for EachInRange(idx, range) { VMapRecord *src_rec = &src[idx]; U16 digit_value = (U16)src_rec->key.digits[digit_idx]; U64 dst_off = lane_digit_offsets[digit_value]; lane_digit_offsets[digit_value] += 1; MemoryCopyStruct(&dst[dst_off], src_rec); } } lane_sync(); // rjf: swap source with destination for next pass Swap(VMapRecord *, src, dst); } } } lane_sync(); //////////////////////// //- rjf: bake // RDI_VMapEntry *vmap = 0; RDI_U64 vmap_count = 0; { //- rjf: allocate vmap RDI_U64 vmap_count__cap = records_count*2 + 1; if(lane_idx() == 0) { vmap = push_array(arena, RDI_VMapEntry, vmap_count__cap); } lane_sync_u64(&vmap, 0); //- rjf: bake if(lane_idx() == 0) { typedef struct RangeNode RangeNode; struct RangeNode { RangeNode *next; Rng1U64 voff_range; U64 idx; }; RDI_VMapEntry *vmap_ptr = vmap; RDI_VMapEntry *vmap_opl = vmap + vmap_count__cap; RangeNode *top_range = 0; RangeNode *free_range = 0; U64 last_recorded_voff = 0; for(U64 record_idx = 0; record_idx <= records_count; record_idx += 1) { // rjf: get next voff range and index Rng1U64 voff_range = r1u64(max_U64, max_U64); U64 idx = 0; if(record_idx < records_count) { VMapRecord *record = &records[record_idx]; voff_range = r1u64(record->key.voff, record->key.voff + -record->key.negative_size); idx = (U64)record->idx; } // rjf: pop nodes we've advanced past { for(RangeNode *n = top_range, *next = 0; n != 0; n = next) { next = n->next; if(n->voff_range.max <= voff_range.min) { SLLStackPop(top_range); SLLStackPush(free_range, n); if(n->voff_range.max != last_recorded_voff) { vmap_ptr += 1; } vmap_ptr->voff = n->voff_range.max; vmap_ptr->idx = next ? next->idx : 0; last_recorded_voff = vmap_ptr->voff; } else { break; } } } // rjf: push this node if(record_idx < records_count) { RangeNode *r = free_range; if(r) { SLLStackPop(free_range); } else { r = push_array(scratch.arena, RangeNode, 1); } SLLStackPush(top_range, r); r->voff_range = voff_range; r->idx = idx; if(voff_range.min != last_recorded_voff || (vmap_ptr->idx != idx && vmap_ptr->idx != 0)) { vmap_ptr += 1; } vmap_ptr->voff = voff_range.min; vmap_ptr->idx = idx; last_recorded_voff = voff_range.min; } } if(last_recorded_voff != 0) { vmap_ptr += 1; } vmap_count = (vmap_ptr - vmap); } lane_sync_u64(&vmap_count, 0); } lane_sync(); //////////////////////// //- rjf: store // if(lane_idx() == 0) { vmap_tasks[vmap_task_idx].vmap_out[0] = vmap; vmap_tasks[vmap_task_idx].vmap_count_out[0] = vmap_count; } } } scratch_end(scratch); } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage build interned path tree // if(lane_idx() == 0) ProfScope("build interned path tree") { //- rjf: set up tree RDIM_BakePathTree *tree = rdim_push_array(arena, RDIM_BakePathTree, 1); rdim_bake_path_tree_insert(arena, tree, rdim_str8_lit("")); //- rjf: bake unit file paths RDIM_ProfScope("bake unit file paths") { for(RDIM_UnitChunkNode *n = params->units.first; n != 0; n = n->next) { for(RDI_U64 idx = 0; idx < n->count; idx += 1) { rdim_bake_path_tree_insert(arena, tree, n->v[idx].source_file); rdim_bake_path_tree_insert(arena, tree, n->v[idx].object_file); rdim_bake_path_tree_insert(arena, tree, n->v[idx].archive_file); rdim_bake_path_tree_insert(arena, tree, n->v[idx].build_path); } } } //- rjf: bake source file paths RDIM_ProfScope("bake source file paths") { for(RDIM_SrcFileChunkNode *n = params->src_files.first; n != 0; n = n->next) { for(RDI_U64 idx = 0; idx < n->count; idx += 1) { RDIM_BakePathNode *node = rdim_bake_path_tree_insert(arena, tree, n->v[idx].path); node->src_file = &n->v[idx]; } } } rdim_shared->path_tree = tree; } lane_sync(); RDIM_BakePathTree *path_tree = rdim_shared->path_tree; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage gather all unsorted, joined, line table info; & sort // ProfScope("gather all unsorted, joined, line table info; & sort") { //- rjf: set up outputs ProfScope("set up outputs") { // rjf: calculate header info if(lane_idx() == 0) { rdim_shared->line_tables_count = params->line_tables.total_count; rdim_shared->src_line_tables = push_array(arena, RDIM_LineTable *, rdim_shared->line_tables_count); ProfScope("flatten chunk list") { U64 joined_idx = 0; for(RDIM_LineTableChunkNode *n = params->line_tables.first; n != 0; n = n->next) { for EachIndex(idx, n->count) { rdim_shared->src_line_tables[joined_idx] = &n->v[idx]; joined_idx += 1; } } } rdim_shared->baked_line_tables.line_tables_count = params->line_tables.total_count + 1; rdim_shared->baked_line_tables.line_table_voffs_count = params->line_tables.total_line_count + 2*params->line_tables.total_seq_count; rdim_shared->baked_line_tables.line_table_lines_count = params->line_tables.total_line_count + params->line_tables.total_seq_count; rdim_shared->baked_line_tables.line_table_columns_count= 1; rdim_shared->line_table_block_take_counter = 0; } lane_sync(); // rjf: allocate outputs ProfScope("allocate outputs") { if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->unsorted_joined_line_tables = push_array(arena, RDIM_UnsortedJoinedLineTable, rdim_shared->line_tables_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->sorted_line_table_keys = push_array(arena, RDIM_SortKey *, rdim_shared->line_tables_count); } if(lane_idx() == lane_from_task_idx(2)) { rdim_shared->baked_line_tables.line_tables = push_array(arena, RDI_LineTable, rdim_shared->baked_line_tables.line_tables_count); ProfScope("lay out line tables") { U64 voffs_base_idx = 0; U64 lines_base_idx = 0; U64 cols_base_idx = 0; for EachIndex(idx, rdim_shared->line_tables_count) { U64 final_idx = idx+1; // NOTE(rjf): +1, to reserve [0] for nil RDIM_LineTable *src = rdim_shared->src_line_tables[idx]; RDI_LineTable *dst = &rdim_shared->baked_line_tables.line_tables[final_idx]; dst->voffs_base_idx = voffs_base_idx; // TODO(rjf): @u64_to_u32 dst->lines_base_idx = lines_base_idx; // TODO(rjf): @u64_to_u32 dst->cols_base_idx = cols_base_idx; // TODO(rjf): @u64_to_u32 dst->lines_count = src->line_count + src->seq_count; // TODO(rjf): @u64_to_u32 voffs_base_idx += src->line_count + 2*src->seq_count; lines_base_idx += src->line_count + 1*src->seq_count; } } } if(lane_idx() == lane_from_task_idx(3)) { rdim_shared->baked_line_tables.line_table_voffs = push_array(arena, RDI_U64, rdim_shared->baked_line_tables.line_table_voffs_count); } if(lane_idx() == lane_from_task_idx(4)) { rdim_shared->baked_line_tables.line_table_lines = push_array(arena, RDI_Line, rdim_shared->baked_line_tables.line_table_lines_count); } if(lane_idx() == lane_from_task_idx(5)) { rdim_shared->baked_line_tables.line_table_columns = push_array(arena, RDI_Column, rdim_shared->baked_line_tables.line_table_columns_count); } } } lane_sync(); //- rjf: wide bake ProfScope("wide bake") { U64 line_table_block_size = 4096; U64 line_table_block_count = (rdim_shared->line_tables_count + line_table_block_size - 1) / line_table_block_size; for(;;) { U64 line_table_block_num = ins_atomic_u64_inc_eval(&rdim_shared->line_table_block_take_counter); if(0 == line_table_block_num || line_table_block_count < line_table_block_num) { break; } U64 line_table_block_idx = line_table_block_num-1; Rng1U64 line_table_range = r1u64(line_table_block_idx*line_table_block_size, (line_table_block_idx+1)*line_table_block_size); line_table_range.max = Min(rdim_shared->line_tables_count, line_table_range.max); for EachInRange(line_table_idx, line_table_range) { RDIM_LineTable *src = rdim_shared->src_line_tables[line_table_idx]; RDIM_UnsortedJoinedLineTable *dst = &rdim_shared->unsorted_joined_line_tables[line_table_idx]; //- rjf: gather dst->line_count = src->line_count; dst->seq_count = src->seq_count; dst->key_count = dst->line_count + dst->seq_count; dst->line_keys = rdim_push_array_no_zero(arena, RDIM_SortKey, dst->key_count); dst->line_recs = rdim_push_array_no_zero(arena, RDIM_LineRec, dst->line_count); { RDIM_SortKey *key_ptr = dst->line_keys; RDIM_LineRec *rec_ptr = dst->line_recs; for(RDIM_LineSequenceNode *seq_n = src->first_seq; seq_n != 0; seq_n = seq_n->next) { RDIM_LineSequence *seq = &seq_n->v; for(RDI_U64 line_idx = 0; line_idx < seq->line_count; line_idx += 1) { key_ptr->key = seq->voffs[line_idx]; key_ptr->val = rec_ptr; key_ptr += 1; rec_ptr->file_id = (RDI_U32)rdim_idx_from_src_file(seq->src_file); // TODO(rjf): @u64_to_u32 rec_ptr->line_num = seq->line_nums[line_idx]; if(seq->col_nums != 0) { rec_ptr->col_first = seq->col_nums[line_idx*2]; rec_ptr->col_opl = seq->col_nums[line_idx*2 + 1]; } rec_ptr += 1; } key_ptr->key = seq->voffs[seq->line_count]; key_ptr->val = 0; key_ptr += 1; } } //- rjf: sort rdim_shared->sorted_line_table_keys[line_table_idx] = rdim_sort_key_array(arena, rdim_shared->unsorted_joined_line_tables[line_table_idx].line_keys, rdim_shared->unsorted_joined_line_tables[line_table_idx].key_count); //- rjf: fill RDIM_SortKey *sorted_line_keys = rdim_shared->sorted_line_table_keys[line_table_idx]; U64 sorted_line_keys_count = rdim_shared->unsorted_joined_line_tables[line_table_idx].key_count; RDI_LineTable *dst_line_table = &rdim_shared->baked_line_tables.line_tables[line_table_idx+1]; U64 *arranged_voffs = rdim_shared->baked_line_tables.line_table_voffs + dst_line_table->voffs_base_idx; RDI_Line *arranged_lines = rdim_shared->baked_line_tables.line_table_lines + dst_line_table->lines_base_idx; RDI_Column *arranged_cols = rdim_shared->baked_line_tables.line_table_columns + dst_line_table->cols_base_idx; { for EachIndex(idx, sorted_line_keys_count) { arranged_voffs[idx] = sorted_line_keys[idx].key; } arranged_voffs[sorted_line_keys_count] = ~0ull; for EachIndex(idx, sorted_line_keys_count) { RDIM_LineRec *rec = (RDIM_LineRec*)sorted_line_keys[idx].val; if(rec != 0) { arranged_lines[idx].file_idx = rec->file_id; arranged_lines[idx].line_num = rec->line_num; } else { arranged_lines[idx].file_idx = 0; arranged_lines[idx].line_num = 0; } } } } } } } lane_sync(); RDI_U64 line_tables_count = rdim_shared->line_tables_count; RDIM_LineTable **src_line_tables = rdim_shared->src_line_tables; RDIM_UnsortedJoinedLineTable *unsorted_joined_line_tables = rdim_shared->unsorted_joined_line_tables; RDIM_SortKey **sorted_line_table_keys = rdim_shared->sorted_line_table_keys; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage build string map // ProfScope("build string map") { Temp scratch = scratch_begin(&arena, 1); //- rjf: set up per-lane outputs RDIM_BakeStringMapTopology *top = 0; RDIM_BakeStringMapLoose **lane_maps__loose = 0; RDIM_BakeStringMapLoose *map__loose = 0; if(lane_idx() == 0) ProfScope("set up per-lane outputs") { top = push_array(scratch.arena, RDIM_BakeStringMapTopology, 1); top->slots_count = (64 + params->procedures.total_count*1 + params->global_variables.total_count*1 + params->thread_variables.total_count*1 + params->types.total_count/2); lane_maps__loose = push_array(scratch.arena, RDIM_BakeStringMapLoose *, lane_count()); map__loose = rdim_bake_string_map_loose_make(scratch.arena, top); } lane_sync_u64(&top, 0); lane_sync_u64(&lane_maps__loose, 0); lane_sync_u64(&map__loose, 0); //- rjf: set up this lane's map ProfScope("set up this lane's map") { lane_maps__loose[lane_idx()] = rdim_bake_string_map_loose_make(scratch.arena, top); } RDIM_BakeStringMapLoose *lane_map = lane_maps__loose[lane_idx()]; //- rjf: push all strings into this lane's map ProfScope("push all strings into this lane's map") { // rjf: push small top-level strings if(lane_idx() == 0) ProfScope("push small top-level strings") { rdim_bake_string_map_loose_insert(arena, top, lane_map, 1, params->top_level_info.exe_name); rdim_bake_string_map_loose_insert(arena, top, lane_map, 1, params->top_level_info.producer_name); for(RDIM_BinarySectionNode *n = params->binary_sections.first; n != 0; n = n->next) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 1, n->v.name); } for(RDIM_BakePathNode *n = path_tree->first; n != 0; n = n->next_order) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 1, n->name); } } // rjf: push strings from source files ProfScope("src files") { for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_String8 normalized_path = rdim_lower_from_str8(arena, n->v[n_idx].path); rdim_bake_string_map_loose_insert(arena, top, lane_map, 1, normalized_path); } } } // rjf: push strings from units ProfScope("units") { for EachNode(n, RDIM_UnitChunkNode, params->units.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].unit_name); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].compiler_name); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].source_file); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].object_file); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].archive_file); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].build_path); } } } // rjf: push strings from types ProfScope("types") { for EachNode(n, RDIM_TypeChunkNode, params->types.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].name); } } } // rjf: push strings from udts ProfScope("udts") { for EachNode(n, RDIM_UDTChunkNode, params->udts.first) { Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { for EachNode(mem, RDIM_UDTMember, n->v[idx].first_member) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, mem->name); } for EachNode(enum_val, RDIM_UDTEnumVal, n->v[idx].first_enum_val) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, enum_val->name); } } } } // rjf: push strings from symbols RDIM_SymbolChunkList *symbol_lists[] = { ¶ms->global_variables, ¶ms->thread_variables, ¶ms->procedures, ¶ms->constants, }; ProfScope("symbols") { for EachElement(list_idx, symbol_lists) { for EachNode(n, RDIM_SymbolChunkNode, symbol_lists[list_idx]->first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].name); rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].link_name); } } } } //- rjf: push strings from inline sites ProfScope("inline sites") { for EachNode(n, RDIM_InlineSiteChunkNode, params->inline_sites.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, n->v[n_idx].name); } } } //- rjf: push strings from scopes ProfScope("scopes") { for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { for EachNode(local, RDIM_Local, n->v[n_idx].first_local) { rdim_bake_string_map_loose_insert(arena, top, lane_map, 4, local->name); } } } } } lane_sync(); //- rjf: join ProfScope("join") { Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { for EachIndex(src_lane_idx, lane_count()) { RDIM_BakeStringMapLoose *src_map = lane_maps__loose[src_lane_idx]; RDIM_BakeStringMapLoose *dst_map = map__loose; if(dst_map->slots[slot_idx] == 0 && src_map->slots[slot_idx] != 0) { dst_map->slots[slot_idx] = src_map->slots[slot_idx]; } else if(dst_map->slots[slot_idx] != 0 && src_map->slots[slot_idx] != 0) { rdim_bake_string_chunk_list_concat_in_place(dst_map->slots[slot_idx], src_map->slots[slot_idx]); } } } } lane_sync(); //- rjf: sort ProfScope("sort") { RDIM_BakeStringMapLoose *map = map__loose; Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { if(map->slots[slot_idx] != 0) { *map->slots[slot_idx] = rdim_bake_string_chunk_list_sorted_from_unsorted(arena, map->slots[slot_idx]); } } } lane_sync(); //- rjf: tighten string table ProfScope("tighten string table") { RDIM_BakeStringMapLoose *map = map__loose; if(lane_idx() == 0) ProfScope("calc base indices, set up tight map") { RDIM_BakeStringMapBaseIndices bake_string_map_base_indices = rdim_bake_string_map_base_indices_from_map_loose(arena, top, map); rdim_shared->bake_strings.slots_count = top->slots_count; rdim_shared->bake_strings.slots = rdim_push_array(arena, RDIM_BakeStringChunkList, rdim_shared->bake_strings.slots_count); rdim_shared->bake_strings.slots_base_idxs = bake_string_map_base_indices.slots_base_idxs; rdim_shared->bake_strings.total_count = rdim_shared->bake_strings.slots_base_idxs[rdim_shared->bake_strings.slots_count]; } lane_sync(); ProfScope("fill tight map") { Rng1U64 slot_range = lane_range(rdim_shared->bake_strings.slots_count); for EachInRange(idx, slot_range) { if(map->slots[idx] != 0) { rdim_memcpy_struct(&rdim_shared->bake_strings.slots[idx], map->slots[idx]); } } } } scratch_end(scratch); } lane_sync(); RDIM_BakeStringMapTight *bake_strings = &rdim_shared->bake_strings; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage build name maps // B32 name_maps_need_build[RDI_NameMapKind_COUNT] = {0}; { name_maps_need_build[RDI_NameMapKind_GlobalVariables] = !!(params->subset_flags & RDIM_SubsetFlag_GlobalVariableNameMap); name_maps_need_build[RDI_NameMapKind_ThreadVariables] = !!(params->subset_flags & RDIM_SubsetFlag_ThreadVariableNameMap); name_maps_need_build[RDI_NameMapKind_Constants] = !!(params->subset_flags & RDIM_SubsetFlag_ConstantNameMap); name_maps_need_build[RDI_NameMapKind_Procedures] = !!(params->subset_flags & RDIM_SubsetFlag_ProcedureNameMap); name_maps_need_build[RDI_NameMapKind_Types] = !!(params->subset_flags & RDIM_SubsetFlag_TypeNameMap); name_maps_need_build[RDI_NameMapKind_LinkNameProcedures] = !!(params->subset_flags & RDIM_SubsetFlag_LinkNameProcedureNameMap); name_maps_need_build[RDI_NameMapKind_NormalSourcePaths] = !!(params->subset_flags & RDIM_SubsetFlag_NormalSourcePathNameMap); } RDIM_BakeNameMapTopology *bake_name_maps_tops = 0; RDIM_BakeNameMap **bake_name_maps = 0; ProfScope("build name maps") { Temp scratch = scratch_begin(&arena, 1); //- rjf: set up RDIM_BakeNameMap ***lane_maps = 0; if(lane_idx() == 0) { bake_name_maps_tops = push_array(arena, RDIM_BakeNameMapTopology, RDI_NameMapKind_COUNT); bake_name_maps = push_array(arena, RDIM_BakeNameMap *, RDI_NameMapKind_COUNT); lane_maps = push_array(scratch.arena, RDIM_BakeNameMap **, RDI_NameMapKind_COUNT); for EachNonZeroEnumVal(RDI_NameMapKind, k) { U64 slot_count = 0; switch((RDI_NameMapKindEnum)k) { case RDI_NameMapKind_NULL: case RDI_NameMapKind_COUNT: {}break; #define Case(name, total_count) case RDI_NameMapKind_##name:{slot_count = ((total_count) + (total_count)/4);}break Case(GlobalVariables, params->global_variables.total_count); Case(ThreadVariables, params->thread_variables.total_count); Case(Constants, params->constants.total_count); Case(Procedures, params->procedures.total_count); Case(LinkNameProcedures, params->procedures.total_count); Case(Types, params->types.total_count); Case(NormalSourcePaths, params->src_files.total_count); #undef Case } lane_maps[k] = push_array(arena, RDIM_BakeNameMap *, lane_count()); bake_name_maps_tops[k].slots_count = slot_count; } } lane_sync_u64(&bake_name_maps_tops, 0); lane_sync_u64(&bake_name_maps, 0); lane_sync_u64(&lane_maps, 0); //- rjf: wide build for EachNonZeroEnumVal(RDI_NameMapKind, k) ProfScope("name map build %.*s", str8_varg(rdi_string_from_name_map_kind(k))) { if(!name_maps_need_build[k]) { continue; } RDIM_BakeNameMapTopology *top = &bake_name_maps_tops[k]; lane_maps[k][lane_idx()] = rdim_bake_name_map_make(scratch.arena, top); RDIM_BakeNameMap *map = lane_maps[k][lane_idx()]; B32 link_names = 0; RDIM_SymbolChunkList *symbols = 0; switch((RDI_NameMapKindEnum)k) { case RDI_NameMapKind_NULL: case RDI_NameMapKind_COUNT: {}break; case RDI_NameMapKind_GlobalVariables: {symbols = ¶ms->global_variables;}goto symbol_name_map_build; case RDI_NameMapKind_ThreadVariables: {symbols = ¶ms->thread_variables;}goto symbol_name_map_build; case RDI_NameMapKind_Constants: {symbols = ¶ms->constants;}goto symbol_name_map_build; case RDI_NameMapKind_Procedures: {symbols = ¶ms->procedures;}goto symbol_name_map_build; case RDI_NameMapKind_LinkNameProcedures:{symbols = ¶ms->procedures; link_names = 1;}goto symbol_name_map_build; symbol_name_map_build:; { for EachNode(n, RDIM_SymbolChunkNode, symbols->first) { Rng1U64 n_range = lane_range(n->count); for EachInRange(n_idx, n_range) { RDIM_Symbol *symbol = &n->v[n_idx]; rdim_bake_name_map_insert(scratch.arena, top, map, 4, link_names ? symbol->link_name : symbol->name, rdim_idx_from_symbol(symbol)); } } }break; case RDI_NameMapKind_Types: { RDIM_TypeChunkList *types = ¶ms->types; for EachNode(n, RDIM_TypeChunkNode, types->first) { Rng1U64 n_range = lane_range(n->count); for EachInRange(n_idx, n_range) { RDIM_Type *type = &n->v[n_idx]; rdim_bake_name_map_insert(scratch.arena, top, map, 4, type->name, rdim_idx_from_type(type)); } } }break; case RDI_NameMapKind_NormalSourcePaths: { RDIM_SrcFileChunkList *src_files = ¶ms->src_files; for EachNode(n, RDIM_SrcFileChunkNode, src_files->first) { Rng1U64 n_range = lane_range(n->count); for EachInRange(n_idx, n_range) { RDIM_SrcFile *src_file = &n->v[n_idx]; RDIM_String8 normalized_path = rdim_lower_from_str8(arena, src_file->path); rdim_bake_name_map_insert(scratch.arena, top, map, 4, normalized_path, rdim_idx_from_src_file(src_file)); } } }break; } } lane_sync(); //- rjf: join & sort if(lane_idx() == 0) { for EachNonZeroEnumVal(RDI_NameMapKind, k) { if(!name_maps_need_build[k]) { continue; } bake_name_maps[k] = rdim_bake_name_map_make(arena, &bake_name_maps_tops[k]); } } lane_sync(); for EachNonZeroEnumVal(RDI_NameMapKind, k) ProfScope("name map join & sort %.*s", str8_varg(rdi_string_from_name_map_kind(k))) { if(!name_maps_need_build[k]) { continue; } RDIM_BakeNameMapTopology *top = &bake_name_maps_tops[k]; RDIM_BakeNameMap *map = bake_name_maps[k]; //- rjf: join ProfScope("join") { Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { for EachIndex(src_lane_idx, lane_count()) { RDIM_BakeNameMap *src_map = lane_maps[k][src_lane_idx]; RDIM_BakeNameMap *dst_map = map; if(dst_map->slots[slot_idx] == 0 && src_map->slots[slot_idx] != 0) { dst_map->slots[slot_idx] = src_map->slots[slot_idx]; } else if(dst_map->slots[slot_idx] != 0 && src_map->slots[slot_idx] != 0) { rdim_bake_name_chunk_list_concat_in_place(dst_map->slots[slot_idx], src_map->slots[slot_idx]); } } } } //- rjf: sort ProfScope("sort") { Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { if(map->slots[slot_idx] != 0) { RDIM_BakeNameChunkList *scratch_og_unsorted = map->slots[slot_idx]; map->slots[slot_idx] = push_array(arena, RDIM_BakeNameChunkList, 1); *map->slots[slot_idx] = rdim_bake_name_chunk_list_sorted_from_unsorted(arena, scratch_og_unsorted); } } } } lane_sync(); scratch_end(scratch); } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage build index runs // B32 need_index_runs = (!!(params->subset_flags & RDIM_SubsetFlag_NameMaps) || !!(params->subset_flags & RDIM_SubsetFlag_Types)); if(need_index_runs) ProfScope("build index runs") { Temp scratch = scratch_begin(&arena, 1); //- rjf: set up per-lane outputs RDIM_BakeIdxRunMapTopology *top = 0; RDIM_BakeIdxRunMapLoose **lane_maps__loose = 0; RDIM_BakeIdxRunMapLoose *map__loose = 0; if(lane_idx() == 0) ProfScope("set up per-lane outputs") { top = push_array(scratch.arena, RDIM_BakeIdxRunMapTopology, 1); top->slots_count = 64 + ((params->procedures.total_count + params->global_variables.total_count + params->thread_variables.total_count + params->udts.total_count) * 3) / 4; lane_maps__loose = push_array(scratch.arena, RDIM_BakeIdxRunMapLoose *, lane_count()); map__loose = rdim_bake_idx_run_map_loose_make(scratch.arena, top); } lane_sync_u64(&top, 0); lane_sync_u64(&lane_maps__loose, 0); lane_sync_u64(&map__loose, 0); //- rjf: set up this lane's map ProfScope("set up this lane's map") { lane_maps__loose[lane_idx()] = rdim_bake_idx_run_map_loose_make(scratch.arena, top); } RDIM_BakeIdxRunMapLoose *lane_map = lane_maps__loose[lane_idx()]; //- rjf: wide fill of all index runs ProfScope("fill all lane index run maps") { //- rjf: bake runs of function-type parameter lists ProfScope("bake runs of function-type parameter lists") { for EachNode(n, RDIM_TypeChunkNode, params->types.first) { Rng1U64 range = lane_range(n->count); ProfScope("[%I64u, %I64u)", range.min, range.max) for EachInRange(n_idx, range) { RDIM_Type *type = &n->v[n_idx]; if(type->count == 0) { continue; } if(type->kind == RDI_TypeKind_Function || type->kind == RDI_TypeKind_Method) { RDI_U32 param_idx_run_count = type->count; RDI_U32 *param_idx_run = rdim_push_array_no_zero(arena, RDI_U32, param_idx_run_count); for(RDI_U32 idx = 0; idx < param_idx_run_count; idx += 1) { param_idx_run[idx] = (RDI_U32)rdim_idx_from_type(type->param_types[idx]); // TODO(rjf): @u64_to_u32 } rdim_bake_idx_run_map_loose_insert(scratch.arena, top, lane_map, 4, param_idx_run, param_idx_run_count); } } } } //- rjf: bake runs of name map match lists for EachNonZeroEnumVal(RDI_NameMapKind, k) ProfScope("bake runs of name map match lists (%.*s)", str8_varg(rdi_string_from_name_map_kind(k))) { if(!name_maps_need_build[k]) { continue; } RDIM_BakeNameMapTopology *name_map_top = &bake_name_maps_tops[k]; RDIM_BakeNameMap *name_map = bake_name_maps[k]; Rng1U64 slot_idx_range = lane_range(name_map_top->slots_count); for EachInRange(slot_idx, slot_idx_range) { RDIM_BakeNameChunkList *slot = name_map->slots[slot_idx]; if(slot != 0) { typedef struct IdxRunNode IdxRunNode; struct IdxRunNode { IdxRunNode *next; RDI_U64 idx; }; IdxRunNode *first_idx_run_node = 0; IdxRunNode *last_idx_run_node = 0; U64 active_idx_count = 0; String8 active_string = {0}; RDIM_BakeNameChunkNode *n = slot->first; U64 n_idx = 0; for(;;) { // rjf: advance chunk if(n != 0 && n_idx >= n->count) { n = n->next; n_idx = 0; } // rjf: grab next element String8 string = {0}; U64 idx = 0; if(n != 0) { string = n->v[n_idx].string; idx = n->v[n_idx].idx; } // rjf: next element hash doesn't match the active? -> push index run, clear active list, start new list if(!str8_match(string, active_string, 0)) { if(active_idx_count > 1) { RDI_U64 idxs_count = active_idx_count; RDI_U32 *idxs = rdim_push_array(arena, RDI_U32, idxs_count); { U64 write_idx = 0; for EachNode(idx_run_n, IdxRunNode, first_idx_run_node) { idxs[write_idx] = (RDI_U32)idx_run_n->idx; // TODO(rjf): @u64_to_u32 write_idx += 1; } } rdim_bake_idx_run_map_loose_insert(scratch.arena, top, lane_map, 4, idxs, idxs_count); } active_string = string; first_idx_run_node = 0; last_idx_run_node = 0; active_idx_count = 0; } // rjf: new element matches the active list -> push if(active_string.size != 0 && str8_match(string, active_string, 0)) { IdxRunNode *idx_run_n = push_array(scratch.arena, IdxRunNode, 1); idx_run_n->idx = idx; SLLQueuePush(first_idx_run_node, last_idx_run_node, idx_run_n); active_idx_count += 1; } // rjf: advance index n_idx += 1; // rjf: end on zero node if(n == 0) { break; } } } } } } lane_sync(); //- rjf: join ProfScope("join") { Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { for EachIndex(src_lane_idx, lane_count()) { RDIM_BakeIdxRunMapLoose *src_map = lane_maps__loose[src_lane_idx]; RDIM_BakeIdxRunMapLoose *dst_map = map__loose; dst_map->slots_idx_counts[slot_idx] += src_map->slots_idx_counts[slot_idx]; if(dst_map->slots[slot_idx] == 0 && src_map->slots[slot_idx] != 0) { dst_map->slots[slot_idx] = src_map->slots[slot_idx]; } else if(dst_map->slots[slot_idx] != 0 && src_map->slots[slot_idx] != 0) { rdim_bake_idx_run_chunk_list_concat_in_place(dst_map->slots[slot_idx], src_map->slots[slot_idx]); } } } } lane_sync(); //- rjf: sort ProfScope("sort") { RDIM_BakeIdxRunMapLoose *map = map__loose; Rng1U64 slot_range = lane_range(top->slots_count); for EachInRange(slot_idx, slot_range) { if(map->slots[slot_idx] != 0) { *map->slots[slot_idx] = rdim_bake_idx_run_chunk_list_sorted_from_unsorted(arena, map->slots[slot_idx]); map->slots_idx_counts[slot_idx] = 0; for EachNode(n, RDIM_BakeIdxRunChunkNode, map->slots[slot_idx]->first) { for EachIndex(idx, n->count) { map->slots_idx_counts[slot_idx] += n->v[idx].count; } } } } } lane_sync(); //- rjf: tighten idx run table ProfScope("tighten idx run table") { RDIM_BakeIdxRunMapLoose *map = map__loose; if(lane_idx() == 0) ProfScope("calc base indices, set up tight map") { rdim_shared->bake_idx_runs.slots_count = top->slots_count; rdim_shared->bake_idx_runs.slots = rdim_push_array(arena, RDIM_BakeIdxRunChunkList, rdim_shared->bake_idx_runs.slots_count); rdim_shared->bake_idx_runs.slots_base_idxs = rdim_push_array(arena, RDI_U64, rdim_shared->bake_idx_runs.slots_count+1); RDI_U64 encoding_idx_off = 0; for(RDI_U64 slot_idx = 0; slot_idx < top->slots_count; slot_idx += 1) { rdim_shared->bake_idx_runs.slots_base_idxs[slot_idx] = encoding_idx_off; if(map->slots[slot_idx] != 0) { encoding_idx_off += map->slots_idx_counts[slot_idx]; } } rdim_shared->bake_idx_runs.slots_base_idxs[top->slots_count] = encoding_idx_off; } lane_sync(); ProfScope("fill tight map") { Rng1U64 slot_range = lane_range(rdim_shared->bake_idx_runs.slots_count); for EachInRange(idx, slot_range) { if(map->slots[idx] != 0) { rdim_memcpy_struct(&rdim_shared->bake_idx_runs.slots[idx], map->slots[idx]); } } } } lane_sync(); scratch_end(scratch); } lane_sync(); RDIM_BakeIdxRunMap *bake_idx_runs = &rdim_shared->bake_idx_runs; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake strings // ProfScope("bake strings") { // rjf: set up if(lane_idx() == 0) ProfScope("set up; lay out strings") { rdim_shared->baked_strings.string_offs_count = bake_strings->total_count + 1; rdim_shared->baked_strings.string_offs = rdim_push_array(arena, RDI_U32, rdim_shared->baked_strings.string_offs_count); RDI_U64 off_cursor = 0; for EachIndex(slot_idx, bake_strings->slots_count) { for EachNode(n, RDIM_BakeStringChunkNode, bake_strings->slots[slot_idx].first) { for EachIndex(n_idx, n->count) { RDIM_BakeString *src = &n->v[n_idx]; U64 dst_idx = bake_strings->slots_base_idxs[slot_idx] + n->base_idx + n_idx + 1; rdim_shared->baked_strings.string_offs[dst_idx] = off_cursor; off_cursor += src->string.size; } } } rdim_shared->baked_strings.string_data_size = off_cursor; rdim_shared->baked_strings.string_data = rdim_push_array(arena, RDI_U8, rdim_shared->baked_strings.string_data_size); } lane_sync(); // rjf: wide fill string data ProfScope("wide fill") { Rng1U64 slot_idx_range = lane_range(bake_strings->slots_count); for EachInRange(slot_idx, slot_idx_range) { for EachNode(n, RDIM_BakeStringChunkNode, bake_strings->slots[slot_idx].first) { for EachIndex(n_idx, n->count) { RDIM_BakeString *src = &n->v[n_idx]; U64 dst_idx = bake_strings->slots_base_idxs[slot_idx] + n->base_idx + n_idx + 1; U64 dst_off = rdim_shared->baked_strings.string_offs[dst_idx]; rdim_memcpy(rdim_shared->baked_strings.string_data + dst_off, src->string.str, src->string.size); } } } } } lane_sync(); RDIM_StringBakeResult baked_strings = rdim_shared->baked_strings; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake idx runs // if(need_index_runs) ProfScope("bake idx runs") { // rjf: set up if(lane_idx() == 0) { rdim_shared->baked_idx_runs.idx_count = bake_idx_runs->slots_base_idxs[bake_idx_runs->slots_count]; rdim_shared->baked_idx_runs.idx_runs = push_array(arena, RDI_U32, rdim_shared->baked_idx_runs.idx_count); } lane_sync(); // rjf: wide fill { Rng1U64 range = lane_range(bake_idx_runs->slots_count); for EachInRange(slot_idx, range) { RDI_U64 off = bake_idx_runs->slots_base_idxs[slot_idx]; for EachNode(n, RDIM_BakeIdxRunChunkNode, bake_idx_runs->slots[slot_idx].first) { StaticAssert(sizeof(rdim_shared->baked_idx_runs.idx_runs[0]) == sizeof(n->v[0].idxes[0]), idx_run_size_check); for EachIndex(n_idx, n->count) { rdim_memcpy(rdim_shared->baked_idx_runs.idx_runs + off, n->v[n_idx].idxes, sizeof(n->v[n_idx].idxes[0]) * n->v[n_idx].count); off += n->v[n_idx].count; } } } } } lane_sync(); RDIM_IndexRunBakeResult baked_idx_runs = rdim_shared->baked_idx_runs; ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake name maps // ProfScope("bake name maps") { // rjf: count unique names in all name maps; lay out baked nodes ProfScope("count unique names in all name maps; lay out baked nodes") { if(lane_idx() == 0) { for EachNonZeroEnumVal(RDI_NameMapKind, k) { rdim_shared->lane_name_map_node_counts[k] = push_array(arena, U64, lane_count()); rdim_shared->lane_name_map_node_offs[k] = push_array(arena, U64, lane_count()); } } lane_sync(); for EachNonZeroEnumVal(RDI_NameMapKind, k) { if(!name_maps_need_build[k]) { continue; } RDIM_BakeNameMapTopology *top = &bake_name_maps_tops[k]; RDIM_BakeNameMap *map = bake_name_maps[k]; Rng1U64 range = lane_range(top->slots_count); for EachInRange(idx, range) { if(map->slots[idx] != 0) { U64 total_unique_name_count = 0; U64 last_hash = 0; for EachNode(n, RDIM_BakeNameChunkNode, map->slots[idx]->first) { for EachIndex(n_idx, n->count) { if(n->v[n_idx].hash != last_hash) { total_unique_name_count += 1; last_hash = n->v[n_idx].hash; } } } rdim_shared->lane_name_map_node_counts[k][lane_idx()] += total_unique_name_count; } } } lane_sync(); if(lane_idx() == 0) { for EachNonZeroEnumVal(RDI_NameMapKind, k) { RDI_U64 node_off = 0; for EachIndex(l_idx, lane_count()) { rdim_shared->name_map_node_counts[k] += rdim_shared->lane_name_map_node_counts[k][l_idx]; rdim_shared->lane_name_map_node_offs[k][l_idx] = node_off; node_off += rdim_shared->lane_name_map_node_counts[k][l_idx]; } rdim_shared->total_name_map_node_count += rdim_shared->name_map_node_counts[k]; } } } lane_sync(); // rjf: setup ProfScope("setup") { if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->baked_top_level_name_maps.name_maps_count = RDI_NameMapKind_COUNT; rdim_shared->baked_top_level_name_maps.name_maps = push_array(arena, RDI_NameMap, rdim_shared->baked_top_level_name_maps.name_maps_count); RDI_U32 bucket_off = 0; RDI_U32 node_off = 0; for EachNonZeroEnumVal(RDI_NameMapKind, k) { rdim_shared->baked_top_level_name_maps.name_maps[k].bucket_base_idx = bucket_off; rdim_shared->baked_top_level_name_maps.name_maps[k].node_base_idx = node_off; rdim_shared->baked_top_level_name_maps.name_maps[k].bucket_count = (RDI_U32)bake_name_maps_tops[k].slots_count; // TODO(rjf): @u64_to_u32 rdim_shared->baked_top_level_name_maps.name_maps[k].node_count = (RDI_U32)rdim_shared->name_map_node_counts[k]; // TODO(rjf): @u64_to_u32 bucket_off += rdim_shared->baked_top_level_name_maps.name_maps[k].bucket_count; node_off += rdim_shared->baked_top_level_name_maps.name_maps[k].node_count; } rdim_shared->baked_name_maps.buckets_count = bucket_off; rdim_shared->baked_name_maps.buckets = push_array(arena, RDI_NameMapBucket, rdim_shared->baked_name_maps.buckets_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->baked_name_maps.nodes_count = rdim_shared->total_name_map_node_count; rdim_shared->baked_name_maps.nodes = push_array(arena, RDI_NameMapNode, rdim_shared->baked_name_maps.nodes_count); } } lane_sync(); // rjf: wide fill baked name maps ProfScope("wide fill baked name maps") { for EachNonZeroEnumVal(RDI_NameMapKind, k) ProfScope("wide fill (%.*s)", str8_varg(rdi_string_from_name_map_kind(k))) { if(!name_maps_need_build[k]) { continue; } RDI_U64 write_node_off = rdim_shared->lane_name_map_node_offs[k][lane_idx()]; RDIM_BakeNameMapTopology *top = &bake_name_maps_tops[k]; U64 slots_count = top->slots_count; RDIM_BakeNameMap *src_map = bake_name_maps[k]; RDI_NameMap *dst_map = &rdim_shared->baked_top_level_name_maps.name_maps[k]; RDI_NameMapBucket *dst_buckets = rdim_shared->baked_name_maps.buckets + dst_map->bucket_base_idx; RDI_NameMapNode *dst_nodes = rdim_shared->baked_name_maps.nodes + dst_map->node_base_idx; Rng1U64 slot_range = lane_range(slots_count); for EachInRange(slot_idx, slot_range) { RDIM_BakeNameChunkList *src_slot = src_map->slots[slot_idx]; if(src_slot == 0) { continue; } RDI_NameMapBucket *dst_bucket = &dst_buckets[slot_idx]; dst_bucket->first_node = write_node_off; { Temp scratch = scratch_begin(&arena, 1); typedef struct IdxRunNode IdxRunNode; struct IdxRunNode { IdxRunNode *next; RDI_U64 idx; }; IdxRunNode *first_idx_run_node = 0; IdxRunNode *last_idx_run_node = 0; U64 active_idx_count = 0; String8 active_string = {0}; RDIM_BakeNameChunkNode *n = src_slot->first; U64 n_idx = 0; for(;;) { // rjf: advance chunk if(n != 0 && n_idx >= n->count) { n = n->next; n_idx = 0; } // rjf: grab next element U64 idx = 0; String8 string = {0}; if(n != 0) { idx = n->v[n_idx].idx; string = n->v[n_idx].string; } // rjf: next element doesn't match the active list? -> push index run, clear active list, start new list if(!str8_match(active_string, string, 0)) { // rjf: has active run -> flatten & serialize if(active_string.size != 0) { // rjf: flatten idxes RDI_U64 idxs_count = active_idx_count; RDI_U32 *idxs = rdim_push_array(scratch.arena, RDI_U32, idxs_count); { U64 write_idx = 0; for EachNode(idx_run_n, IdxRunNode, first_idx_run_node) { idxs[write_idx] = (RDI_U32)idx_run_n->idx; // TODO(rjf): @u64_to_u32 write_idx += 1; } } // rjf: serialize node RDI_NameMapNode *dst_node = &dst_nodes[write_node_off]; dst_node->string_idx = rdim_bake_idx_from_string(bake_strings, active_string); dst_node->match_count = idxs_count; if(dst_node->match_count == 1) { dst_node->match_idx_or_idx_run_first = idxs[0]; } else if(dst_node->match_count > 1) { dst_node->match_idx_or_idx_run_first = rdim_bake_idx_from_idx_run(bake_idx_runs, idxs, idxs_count); } dst_bucket->node_count += 1; write_node_off += 1; } // rjf: start new list active_string = string; first_idx_run_node = 0; last_idx_run_node = 0; active_idx_count = 0; temp_end(scratch); } // rjf: hash matches the active list -> push if(active_string.size != 0 && str8_match(active_string, string, 0)) { IdxRunNode *idx_run_n = push_array(scratch.arena, IdxRunNode, 1); idx_run_n->idx = idx; SLLQueuePush(first_idx_run_node, last_idx_run_node, idx_run_n); active_idx_count += 1; } // rjf: advance index n_idx += 1; // rjf: end on zero node if(n == 0) { break; } } scratch_end(scratch); } } } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage gather line-bucketed src line map data // ProfScope("gather line-bucketed src line map data") { if(lane_idx() == 0) { rdim_shared->bake_src_line_maps = push_array(arena, RDIM_BakeSrcLineMap, params->src_files.total_count); } lane_sync(); { for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { U64 file_idx = n->base_idx + n_idx; RDIM_BakeSrcLineMap *map = &rdim_shared->bake_src_line_maps[file_idx]; // rjf: set up map map->slots_count = n->v[n_idx].total_line_count; map->slots = push_array(arena, RDIM_BakeSrcLineMapSlot, map->slots_count); // rjf: gather line-bucketed info for EachNode(frag, RDIM_SrcFileLineMapFragment, n->v[n_idx].first_line_map_fragment) { RDIM_LineSequence *seq = frag->seq; for EachIndex(idx, seq->line_count) { RDI_U32 line_num = seq->line_nums[idx]; RDI_U64 voff_first = seq->voffs[idx]; RDI_U64 voff_opl = seq->voffs[idx+1]; RDI_U64 slot_idx = line_num%map->slots_count; // rjf: find existing line node RDIM_BakeSrcLineMapNode *line_node = 0; { for EachNode(line_n, RDIM_BakeSrcLineMapNode, map->slots[slot_idx].first) { if(line_n->line_num == line_num) { line_node = line_n; break; } } } // rjf: construct new node if unseen if(line_node == 0) { line_node = push_array(arena, RDIM_BakeSrcLineMapNode, 1); SLLQueuePush(map->slots[slot_idx].first, map->slots[slot_idx].last, line_node); line_node->line_num = line_num; map->line_count += 1; } // rjf: push this voff range RDIM_Rng1U64 voff_range = {voff_first, voff_opl}; rdim_rng1u64_list_push(arena, &line_node->voff_ranges, voff_range); map->voff_range_count += 1; } } } } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage sort line-bucketed src line map data // ProfScope("sort line-bucketed src line map data") { U64 map_count = params->src_files.total_count; if(lane_idx() == 0) { rdim_shared->bake_src_line_map_keys = push_array(arena, RDIM_SortKey *, map_count); } lane_sync(); for(;;) { U64 map_num = ins_atomic_u64_inc_eval(&rdim_shared->bake_src_line_map_take_counter); if(map_num < 1 || map_count < map_num) { break; } U64 map_idx = map_num-1; RDIM_BakeSrcLineMap *map = &rdim_shared->bake_src_line_maps[map_idx]; // rjf: gather keys rdim_shared->bake_src_line_map_keys[map_idx] = push_array_no_zero(arena, RDIM_SortKey, map->line_count); RDIM_SortKey *keys = rdim_shared->bake_src_line_map_keys[map_idx]; { U64 key_idx = 0; for EachIndex(slot_idx, map->slots_count) { for EachNode(n, RDIM_BakeSrcLineMapNode, map->slots[slot_idx].first) { keys[key_idx].key = n->line_num; keys[key_idx].val = n; key_idx += 1; } } } // rjf: sort keys { radsort(keys, map->line_count, rdim_sort_key_is_before); } } } ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage compute src file / src file line map layout // ProfScope("compute src file / src file line map layout") { // rjf: set up if(lane_idx() == 0) { rdim_shared->lane_chunk_src_file_num_counts = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_voff_counts = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_map_counts = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_num_offs = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_voff_offs = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_map_offs = push_array(arena, U64, lane_count()*params->src_files.chunk_count); for EachEnumVal(RDI_ChecksumKind, k) { rdim_shared->lane_chunk_src_file_checksum_counts[k] = push_array(arena, U64, lane_count()*params->src_files.chunk_count); rdim_shared->lane_chunk_src_file_checksum_offs[k] = push_array(arena, U64, lane_count()*params->src_files.chunk_count); } } lane_sync(); // rjf: wide count { U64 chunk_idx = 0; for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { Rng1U64 range = lane_range(n->count); U64 slot_idx = lane_idx()*params->src_files.chunk_count + chunk_idx; for EachInRange(idx, range) { RDIM_BakeSrcLineMap *map = &rdim_shared->bake_src_line_maps[n->base_idx + idx]; rdim_shared->lane_chunk_src_file_num_counts[slot_idx] += map->line_count; rdim_shared->lane_chunk_src_file_voff_counts[slot_idx] += map->voff_range_count; rdim_shared->lane_chunk_src_file_map_counts[slot_idx] += !!map->line_count; RDI_ChecksumKind k = n->v[idx].checksum_kind; String8 val = n->v[idx].checksum; if(RDI_ChecksumKind_NULL < k && k < RDI_ChecksumKind_COUNT && val.size != 0) { rdim_shared->lane_chunk_src_file_checksum_counts[k][slot_idx] += 1; } } chunk_idx += 1; } } lane_sync(); // rjf: layout if(lane_idx() == 0) { U64 chunk_idx = 0; U64 num_layout_off = 0; U64 voff_layout_off = 0; U64 map_layout_off = 1; U64 checksum_layout_offs[RDI_ChecksumKind_COUNT] = {0}; for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->src_files.chunk_count + chunk_idx; rdim_shared->lane_chunk_src_file_num_offs[slot_idx] = num_layout_off; rdim_shared->lane_chunk_src_file_voff_offs[slot_idx] = voff_layout_off; rdim_shared->lane_chunk_src_file_map_offs[slot_idx] = map_layout_off; num_layout_off += rdim_shared->lane_chunk_src_file_num_counts[slot_idx]; voff_layout_off += rdim_shared->lane_chunk_src_file_voff_counts[slot_idx]; map_layout_off += rdim_shared->lane_chunk_src_file_map_counts[slot_idx]; for EachEnumVal(RDI_ChecksumKind, k) { rdim_shared->lane_chunk_src_file_checksum_offs[k][slot_idx] = checksum_layout_offs[k]; checksum_layout_offs[k] += rdim_shared->lane_chunk_src_file_checksum_counts[k][slot_idx]; } } chunk_idx += 1; } rdim_shared->total_src_map_line_count = num_layout_off; rdim_shared->total_src_map_voff_count = voff_layout_off; for EachEnumVal(RDI_ChecksumKind, k) { rdim_shared->total_checksum_counts[k] = checksum_layout_offs[k]; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake src files // ProfScope("bake src files") { //- rjf: set up if(lane_idx() == 0) { rdim_shared->baked_src_files.source_files_count = params->src_files.total_count+1; rdim_shared->baked_src_files.source_files = push_array(arena, RDI_SourceFile, rdim_shared->baked_src_files.source_files_count); rdim_shared->baked_src_files.source_line_maps_count = params->src_files.source_line_map_count+1; rdim_shared->baked_src_files.source_line_maps = push_array(arena, RDI_SourceLineMap, rdim_shared->baked_src_files.source_line_maps_count); rdim_shared->baked_src_files.source_line_map_nums_count = rdim_shared->total_src_map_line_count; rdim_shared->baked_src_files.source_line_map_nums = push_array(arena, RDI_U32, rdim_shared->baked_src_files.source_line_map_nums_count); rdim_shared->baked_src_files.source_line_map_rngs_count = rdim_shared->total_src_map_line_count + rdim_shared->baked_src_files.source_line_maps_count; rdim_shared->baked_src_files.source_line_map_rngs = push_array(arena, RDI_U32, rdim_shared->baked_src_files.source_line_map_rngs_count); rdim_shared->baked_src_files.source_line_map_voffs_count = rdim_shared->total_src_map_voff_count; rdim_shared->baked_src_files.source_line_map_voffs = push_array(arena, RDI_U64, rdim_shared->baked_src_files.source_line_map_voffs_count); } lane_sync(); //- rjf: bake U64 chunk_idx = 0; for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { Rng1U64 range = lane_range(n->count); U64 slot_idx = lane_idx()*params->src_files.chunk_count + chunk_idx; U64 dst_num_off = rdim_shared->lane_chunk_src_file_num_offs[slot_idx]; U64 dst_map_off = rdim_shared->lane_chunk_src_file_map_offs[slot_idx]; U64 dst_voff_off = rdim_shared->lane_chunk_src_file_voff_offs[slot_idx]; U64 dst_rng_off = dst_num_off + dst_map_off; U64 dst_checksums_off[RDI_ChecksumKind_COUNT] = {0}; for EachEnumVal(RDI_ChecksumKind, k) { dst_checksums_off[k] = 1 + rdim_shared->lane_chunk_src_file_checksum_offs[k][slot_idx]; } for EachInRange(idx, range) { RDIM_BakeSrcLineMap *map = &rdim_shared->bake_src_line_maps[n->base_idx + idx]; RDIM_SortKey *sorted_map_keys = rdim_shared->bake_src_line_map_keys[n->base_idx + idx]; RDIM_SrcFile *src = &n->v[idx]; RDI_SourceFile *dst = &rdim_shared->baked_src_files.source_files[n->base_idx + idx + 1]; RDI_SourceLineMap *dst_map = &rdim_shared->baked_src_files.source_line_maps[dst_map_off]; RDI_U32 *dst_nums = &rdim_shared->baked_src_files.source_line_map_nums[dst_num_off]; RDI_U32 *dst_rngs = &rdim_shared->baked_src_files.source_line_map_rngs[dst_rng_off]; RDI_U64 *dst_voffs = &rdim_shared->baked_src_files.source_line_map_voffs[dst_voff_off]; //- rjf: fill file info Temp scratch = scratch_begin(&arena, 1); String8 normalized_path = rdim_lower_from_str8(scratch.arena, src->path); B32 has_checksum = (RDI_ChecksumKind_NULL < src->checksum_kind && src->checksum_kind < RDI_ChecksumKind_COUNT && src->checksum.size != 0); dst->file_path_node_idx = rdim_bake_path_node_idx_from_string(path_tree, src->path); dst->normal_full_path_string_idx = rdim_bake_idx_from_string(bake_strings, normalized_path); dst->source_line_map_idx = src->total_line_count ? dst_map_off : 0; dst->checksum_kind = src->checksum_kind; dst->checksum_idx = has_checksum ? dst_checksums_off[dst->checksum_kind] : 0; scratch_end(scratch); //- rjf: advance checksum offset for this kind if(has_checksum) { dst_checksums_off[dst->checksum_kind] += 1; } //- rjf: fill map info if(src->total_line_count != 0) { dst_map->line_count = (RDI_U32)map->line_count; // TODO(rjf): @u64_to_u32 dst_map->voff_count = (RDI_U32)map->voff_range_count; // TODO(rjf): @u64_to_u32 dst_map->line_map_nums_base_idx = (RDI_U32)dst_num_off; // TODO(rjf): @u64_to_u32 dst_map->line_map_range_base_idx = (RDI_U32)dst_rng_off; // TODO(rjf): @u64_to_u32 dst_map->line_map_voff_base_idx = (RDI_U32)dst_voff_off; // TODO(rjf): @u64_to_u32 dst_map_off += 1; } //- rjf: fill nums/ranges/voffs info if(src->total_line_count != 0) { U64 *dst_voff_ptr = dst_voffs; for EachIndex(line_num_idx, map->line_count) { dst_nums[line_num_idx] = (RDI_U32)sorted_map_keys[line_num_idx].key; // TODO(rjf): @u64_to_u32 dst_rngs[line_num_idx] = (RDI_U32)(dst_voff_ptr - dst_voffs); // TODO(rjf): @u64_to_u32 RDIM_BakeSrcLineMapNode *node = (RDIM_BakeSrcLineMapNode *)sorted_map_keys[line_num_idx].val; for EachNode(rng_n, RDIM_Rng1U64Node, node->voff_ranges.first) { dst_voff_ptr[0] = rng_n->v.min; dst_voff_ptr += 1; } } dst_rngs[map->line_count] = (RDI_U32)map->voff_range_count; // TODO(rjf): @u64_to_u32 dst_num_off += map->line_count; dst_rng_off += map->line_count+1; dst_voff_off += map->voff_range_count; } } chunk_idx += 1; } } ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake checksums // ProfScope("bake checksums") { // rjf: allocate if(lane_idx() == 0) { rdim_shared->baked_checksums.md5s_count = rdim_shared->total_checksum_counts[RDI_ChecksumKind_MD5] + 1; rdim_shared->baked_checksums.sha1s_count = rdim_shared->total_checksum_counts[RDI_ChecksumKind_SHA1] + 1; rdim_shared->baked_checksums.sha256s_count = rdim_shared->total_checksum_counts[RDI_ChecksumKind_SHA256] + 1; rdim_shared->baked_checksums.timestamps_count = rdim_shared->total_checksum_counts[RDI_ChecksumKind_Timestamp] + 1; rdim_shared->baked_checksums.md5s = push_array(arena, RDI_MD5, rdim_shared->baked_checksums.md5s_count); rdim_shared->baked_checksums.sha1s = push_array(arena, RDI_SHA1, rdim_shared->baked_checksums.sha1s_count); rdim_shared->baked_checksums.sha256s = push_array(arena, RDI_SHA256, rdim_shared->baked_checksums.sha256s_count); rdim_shared->baked_checksums.timestamps = push_array(arena, RDI_U64, rdim_shared->baked_checksums.timestamps_count); } lane_sync(); // rjf: fill { U64 chunk_idx = 0; for EachNode(n, RDIM_SrcFileChunkNode, params->src_files.first) { Rng1U64 range = lane_range(n->count); U64 slot_idx = lane_idx()*params->src_files.chunk_count + chunk_idx; U64 dst_offs[RDI_ChecksumKind_COUNT] = {0}; for EachEnumVal(RDI_ChecksumKind, k) { dst_offs[k] = 1 + rdim_shared->lane_chunk_src_file_checksum_offs[k][slot_idx]; } for EachInRange(n_idx, range) { RDI_ChecksumKind k = n->v[n_idx].checksum_kind; String8 val = n->v[n_idx].checksum; if(RDI_ChecksumKind_NULL < k && k < RDI_ChecksumKind_COUNT && val.size != 0) { switch((RDI_ChecksumKindEnum)k) { case RDI_ChecksumKind_NULL: case RDI_ChecksumKind_COUNT: {}break; #define Case(name, table_name) case RDI_ChecksumKind_##name:{MemoryCopy(&rdim_shared->baked_checksums.table_name[dst_offs[k]], val.str, Min(val.size, sizeof(rdim_shared->baked_checksums.table_name[0])));}break Case(MD5, md5s); Case(SHA1, sha1s); Case(SHA256, sha256s); Case(Timestamp, timestamps); #undef Case } dst_offs[k] += 1; } } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage compute lane UDT member/enum-val layouts // ProfScope("compute lane UDT member/enum-val layouts") { // rjf: allocate if(lane_idx() == 0) { rdim_shared->member_chunk_lane_counts = push_array(arena, U64, lane_count() * params->udts.chunk_count); rdim_shared->member_chunk_lane_offs = push_array(arena, U64, lane_count() * params->udts.chunk_count); rdim_shared->enum_val_chunk_lane_counts = push_array(arena, U64, lane_count() * params->udts.chunk_count); rdim_shared->enum_val_chunk_lane_offs = push_array(arena, U64, lane_count() * params->udts.chunk_count); } lane_sync(); // rjf: count { U64 chunk_idx = 0; for EachNode(n, RDIM_UDTChunkNode, params->udts.first) { Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { U64 slot_idx = lane_idx()*params->udts.chunk_count + chunk_idx; rdim_shared->member_chunk_lane_counts[slot_idx] += n->v[idx].member_count; rdim_shared->enum_val_chunk_lane_counts[slot_idx] += n->v[idx].enum_val_count; } chunk_idx += 1; } } lane_sync(); // rjf: layout if(lane_idx() == 0) { U64 member_layout_off = 1; U64 enum_val_layout_off = 1; U64 chunk_idx = 0; for EachNode(n, RDIM_UDTChunkNode, params->udts.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->udts.chunk_count + chunk_idx; rdim_shared->member_chunk_lane_offs[slot_idx] = member_layout_off; rdim_shared->enum_val_chunk_lane_offs[slot_idx] = enum_val_layout_off; member_layout_off += rdim_shared->member_chunk_lane_counts[slot_idx]; enum_val_layout_off += rdim_shared->enum_val_chunk_lane_counts[slot_idx]; } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake UDTs // ProfScope("bake UDTs") { //- rjf: set up ProfScope("set up") { if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->baked_udts.udts_count = params->udts.total_count+1; rdim_shared->baked_udts.udts = push_array(arena, RDI_UDT, rdim_shared->baked_udts.udts_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->baked_udts.members_count = params->udts.total_member_count+1; rdim_shared->baked_udts.members = push_array(arena, RDI_Member, rdim_shared->baked_udts.members_count); } if(lane_idx() == lane_from_task_idx(2)) { rdim_shared->baked_udts.enum_members_count = params->udts.total_enum_val_count+1; rdim_shared->baked_udts.enum_members = push_array(arena, RDI_EnumMember, rdim_shared->baked_udts.enum_members_count); } } lane_sync(); //- rjf: bake UDTs ProfScope("bake UDTs") { U64 chunk_idx = 0; for EachNode(n, RDIM_UDTChunkNode, params->udts.first) { Rng1U64 range = lane_range(n->count); U64 layout_slot_idx = lane_idx()*params->udts.chunk_count + chunk_idx; U64 member_layout_off = rdim_shared->member_chunk_lane_offs[layout_slot_idx]; U64 enum_val_layout_off = rdim_shared->enum_val_chunk_lane_offs[layout_slot_idx]; for EachInRange(n_idx, range) { RDIM_UDT *src_udt = &n->v[n_idx]; RDI_UDT *dst_udt = &rdim_shared->baked_udts.udts[n->base_idx + n_idx + 1]; //- rjf: fill basics dst_udt->self_type_idx = (RDI_U32)rdim_idx_from_type(src_udt->self_type); // TODO(rjf): @u64_to_u32 dst_udt->file_idx = (RDI_U32)rdim_idx_from_src_file(src_udt->src_file); // TODO(rjf): @u64_to_u32 dst_udt->line = src_udt->line; dst_udt->col = src_udt->col; //- rjf: fill member info if(src_udt->first_member != 0) { U64 member_off_first = member_layout_off; for EachNode(src_member, RDIM_UDTMember, src_udt->first_member) { RDI_Member *dst_member = &rdim_shared->baked_udts.members[member_layout_off]; dst_member->kind = src_member->kind; dst_member->name_string_idx = rdim_bake_idx_from_string(bake_strings, src_member->name); dst_member->type_idx = (RDI_U32)rdim_idx_from_type(src_member->type); // TODO(rjf): @u64_to_u32 dst_member->off = src_member->off; member_layout_off += 1; } U64 member_off_opl = member_layout_off; dst_udt->member_first = (RDI_U32)member_off_first; // TODO(rjf): @u64_to_u32 dst_udt->member_count = (RDI_U32)(member_off_opl - member_off_first); // TODO(rjf): @u64_to_u32 } //- rjf: fill enum val info else if(src_udt->first_enum_val != 0) { U64 enum_val_off_first = enum_val_layout_off; for EachNode(src_enum_val, RDIM_UDTEnumVal, src_udt->first_enum_val) { RDI_EnumMember *dst_member = &rdim_shared->baked_udts.enum_members[enum_val_layout_off]; dst_member->name_string_idx = rdim_bake_idx_from_string(bake_strings, src_enum_val->name); dst_member->val = src_enum_val->val; enum_val_layout_off += 1; } U64 enum_val_off_opl = enum_val_layout_off; dst_udt->flags |= RDI_UDTFlag_EnumMembers; dst_udt->member_first = (RDI_U32)enum_val_off_first; // TODO(rjf): @u64_to_u32 dst_udt->member_count = (RDI_U32)(enum_val_off_opl - enum_val_off_first); // TODO(rjf): @u64_to_u32 } } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage compute lane location block layout // U64 total_location_case_chunk_count = (params->scopes.chunk_count + params->procedures.chunk_count); ProfScope("compute lane location block layout") { // rjf: set up if(lane_idx() == 0) { rdim_shared->location_case_chunk_lane_counts = push_array(arena, RDI_U64, lane_count() * total_location_case_chunk_count); rdim_shared->location_case_chunk_lane_offs = push_array(arena, RDI_U64, lane_count() * total_location_case_chunk_count); } lane_sync(); // rjf: per-chunk-lane count of location cases { // rjf: count location cases in scopes U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { U64 slot_idx = lane_idx() * total_location_case_chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { for EachNode(local, RDIM_Local, n->v[idx].first_local) { rdim_shared->location_case_chunk_lane_counts[slot_idx] += local->location_cases.count; } } chunk_idx += 1; } // rjf: count location cases in procedures for EachNode(n, RDIM_SymbolChunkNode, params->procedures.first) { U64 slot_idx = lane_idx() * total_location_case_chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { rdim_shared->location_case_chunk_lane_counts[slot_idx] += n->v[idx].location_cases.count; } chunk_idx += 1; } } lane_sync(); // rjf: lay out location case offsets if(lane_idx() == 0) { U64 chunk_idx = 0; U64 location_case_layout_off = 1; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx * total_location_case_chunk_count + chunk_idx; rdim_shared->location_case_chunk_lane_offs[slot_idx] = location_case_layout_off; location_case_layout_off += rdim_shared->location_case_chunk_lane_counts[slot_idx]; } chunk_idx += 1; } for EachNode(n, RDIM_SymbolChunkNode, params->procedures.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx * total_location_case_chunk_count + chunk_idx; rdim_shared->location_case_chunk_lane_offs[slot_idx] = location_case_layout_off; location_case_layout_off += rdim_shared->location_case_chunk_lane_counts[slot_idx]; } chunk_idx += 1; } rdim_shared->total_location_case_count = location_case_layout_off; } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake location blocks // ProfScope("bake location blocks") { // rjf: set up if(lane_idx() == 0) { rdim_shared->baked_location_blocks.location_blocks_count = rdim_shared->total_location_case_count; rdim_shared->baked_location_blocks.location_blocks = push_array(arena, RDI_LocationBlock, rdim_shared->baked_location_blocks.location_blocks_count); } lane_sync(); // rjf: wide fill from scopes U64 chunk_idx = 0; ProfScope("wide fill from scopes") { for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { U64 layout_slot_idx = lane_idx() * total_location_case_chunk_count + chunk_idx; U64 layout_off = rdim_shared->location_case_chunk_lane_offs[layout_slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { for EachNode(local, RDIM_Local, n->v[idx].first_local) { for EachNode(src, RDIM_LocationCase, local->location_cases.first) { RDI_LocationBlock *dst = &rdim_shared->baked_location_blocks.location_blocks[layout_off]; dst->scope_off_first = (RDI_U32)src->voff_range.min; // TODO(rjf): @u64_to_u32 dst->scope_off_opl = (RDI_U32)src->voff_range.max; // TODO(rjf): @u64_to_u32 dst->location_data_off = (RDI_U32)rdim_off_from_location(src->location); // TODO(rjf): @u64_to_u32 layout_off += 1; } } } chunk_idx += 1; } } // rjf: wide fill from procedures ProfScope("wide fill from procedures") { for EachNode(n, RDIM_SymbolChunkNode, params->procedures.first) { U64 layout_slot_idx = lane_idx() * total_location_case_chunk_count + chunk_idx; U64 layout_off = rdim_shared->location_case_chunk_lane_offs[layout_slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { for EachNode(src, RDIM_LocationCase, n->v[idx].location_cases.first) { RDI_LocationBlock *dst = &rdim_shared->baked_location_blocks.location_blocks[layout_off]; dst->scope_off_first = (RDI_U32)src->voff_range.min; // TODO(rjf): @u64_to_u32 dst->scope_off_opl = (RDI_U32)src->voff_range.max; // TODO(rjf): @u64_to_u32 dst->location_data_off = (RDI_U32)rdim_off_from_location(src->location); // TODO(rjf): @u64_to_u32 layout_off += 1; } } chunk_idx += 1; } } } ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake locations // ProfScope("bake locations") { if(lane_idx() == 0) { rdim_shared->baked_locations.location_data_size = params->locations.total_encoded_size+1; rdim_shared->baked_locations.location_data = push_array(arena, RDI_U8, rdim_shared->baked_locations.location_data_size); } lane_sync(); for EachNode(n, RDIM_LocationChunkNode, params->locations.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Location *loc = &n->v[n_idx]; RDI_U8 *dst = &rdim_shared->baked_locations.location_data[n->base_encoding_off + loc->relative_encoding_off + 1]; switch((RDI_LocationKindEnum)loc->info.kind) { case RDI_LocationKind_NULL:{}break; case RDI_LocationKind_AddrBytecodeStream: case RDI_LocationKind_ValBytecodeStream: { MemoryCopy(dst+0, &loc->info.kind, sizeof(loc->info.kind)); RDI_U64 write_off = sizeof(loc->info.kind); for EachNode(op_node, RDIM_EvalBytecodeOp, loc->info.bytecode.first_op) { MemoryCopy(dst + write_off, &op_node->op, 1); write_off += 1; MemoryCopy(dst + write_off, &op_node->p, op_node->p_size); write_off += op_node->p_size; } dst[write_off] = 0; }break; case RDI_LocationKind_AddrRegPlusU16: case RDI_LocationKind_AddrAddrRegPlusU16: { RDI_LocationRegPlusU16 baked = {loc->info.kind, loc->info.reg_code, loc->info.offset}; MemoryCopy(dst, &baked, sizeof(baked)); }break; case RDI_LocationKind_ValReg: { RDI_LocationReg baked = {loc->info.kind, loc->info.reg_code}; MemoryCopy(dst, &baked, sizeof(baked)); }break; } } } } ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage compute layout for scope sub-lists (locals / voffs) // ProfScope("compute layout for scope sub-lists (locals / voffs)") { // rjf: set up if(lane_idx() == 0) { rdim_shared->scope_local_chunk_lane_counts = push_array(arena, RDI_U64, lane_count() * params->scopes.chunk_count); rdim_shared->scope_local_chunk_lane_offs = push_array(arena, RDI_U64, lane_count() * params->scopes.chunk_count); rdim_shared->scope_voff_chunk_lane_counts = push_array(arena, RDI_U64, lane_count() * params->scopes.chunk_count); rdim_shared->scope_voff_chunk_lane_offs = push_array(arena, RDI_U64, lane_count() * params->scopes.chunk_count); } lane_sync(); // rjf: count per-lane-chunk { U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { U64 num_locals_in_this_lane_and_node = 0; U64 num_voffs_in_this_lane_and_node = 0; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { num_locals_in_this_lane_and_node += n->v[n_idx].local_count; num_voffs_in_this_lane_and_node += n->v[n_idx].voff_ranges.count*2; } rdim_shared->scope_local_chunk_lane_counts[lane_idx()*params->scopes.chunk_count + chunk_idx] = num_locals_in_this_lane_and_node; rdim_shared->scope_voff_chunk_lane_counts[lane_idx()*params->scopes.chunk_count + chunk_idx] = num_voffs_in_this_lane_and_node; chunk_idx += 1; } } lane_sync(); // rjf: lay out each lane's range if(lane_idx() == 0) { U64 local_layout_off = 1; U64 voff_layout_off = 1; U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->scopes.chunk_count + chunk_idx; rdim_shared->scope_local_chunk_lane_offs[slot_idx] = local_layout_off; rdim_shared->scope_voff_chunk_lane_offs[slot_idx] = voff_layout_off; local_layout_off += rdim_shared->scope_local_chunk_lane_counts[slot_idx]; voff_layout_off += rdim_shared->scope_voff_chunk_lane_counts[slot_idx]; } chunk_idx += 1; } } lane_sync(); } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake scopes // ProfScope("bake scopes") { //- rjf: setup outputs if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->baked_scopes.scopes_count = params->scopes.total_count+1; rdim_shared->baked_scopes.scopes = push_array(arena, RDI_Scope, rdim_shared->baked_scopes.scopes_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->baked_scopes.scope_voffs_count = params->scopes.scope_voff_count+1; rdim_shared->baked_scopes.scope_voffs = push_array(arena, RDI_U64, rdim_shared->baked_scopes.scope_voffs_count); } if(lane_idx() == lane_from_task_idx(2)) { rdim_shared->baked_scopes.locals_count = params->scopes.local_count+1; rdim_shared->baked_scopes.locals = push_array(arena, RDI_Local, rdim_shared->baked_scopes.locals_count); } lane_sync(); //- rjf: wide fill { U64 chunk_idx = 0; for EachNode(n, RDIM_ScopeChunkNode, params->scopes.first) { Rng1U64 range = lane_range(n->count); U64 scope_chunk_lane_slot_idx = lane_idx()*params->scopes.chunk_count + chunk_idx; U64 chunk_local_off = rdim_shared->scope_local_chunk_lane_offs[scope_chunk_lane_slot_idx]; U64 chunk_voff_off = rdim_shared->scope_voff_chunk_lane_offs[scope_chunk_lane_slot_idx]; U64 location_block_chunk_lane_slot_idx = lane_idx() * total_location_case_chunk_count + chunk_idx; U64 chunk_location_block_off = rdim_shared->location_case_chunk_lane_offs[location_block_chunk_lane_slot_idx]; for EachInRange(n_idx, range) { U64 dst_idx = 1 + n->base_idx + n_idx; RDIM_Scope *src_scope = &n->v[n_idx]; RDI_Scope *dst_scope = &rdim_shared->baked_scopes.scopes[dst_idx]; //- rjf: fill voff ranges U64 voff_idx_first = chunk_voff_off; for EachNode(rng_n, RDIM_Rng1U64Node, src_scope->voff_ranges.first) { rdim_shared->baked_scopes.scope_voffs[chunk_voff_off+0] = rng_n->v.min; rdim_shared->baked_scopes.scope_voffs[chunk_voff_off+1] = rng_n->v.max; chunk_voff_off += 2; } U64 voff_idx_opl = chunk_voff_off; //- rjf: fill locals U64 local_idx_first = chunk_local_off; for EachNode(src_local, RDIM_Local, src_scope->first_local) { RDI_Local *dst_local = &rdim_shared->baked_scopes.locals[chunk_local_off]; dst_local->kind = src_local->kind; dst_local->name_string_idx = rdim_bake_idx_from_string(bake_strings, src_local->name); dst_local->type_idx = (RDI_U32)rdim_idx_from_type(src_local->type); // TODO(rjf): @u64_to_u32 if(src_local->location_cases.count != 0) { dst_local->location_first = chunk_location_block_off; dst_local->location_opl = chunk_location_block_off + src_local->location_cases.count; chunk_location_block_off += src_local->location_cases.count; } chunk_local_off += 1; } U64 local_idx_opl = chunk_local_off; //- rjf: fill scope dst_scope->proc_idx = (RDI_U32)rdim_idx_from_symbol(src_scope->symbol); // TODO(rjf): @u64_to_u32 dst_scope->parent_scope_idx = (RDI_U32)rdim_idx_from_scope(src_scope->parent_scope); // TODO(rjf): @u64_to_u32 dst_scope->first_child_scope_idx = (RDI_U32)rdim_idx_from_scope(src_scope->first_child); // TODO(rjf): @u64_to_u32 dst_scope->next_sibling_scope_idx = (RDI_U32)rdim_idx_from_scope(src_scope->next_sibling); // TODO(rjf): @u64_to_u32 dst_scope->voff_range_first = (RDI_U32)voff_idx_first; // TODO(rjf): @u64_to_u32 dst_scope->voff_range_opl = (RDI_U32)voff_idx_opl; // TODO(rjf): @u64_to_u32 dst_scope->local_first = (RDI_U32)local_idx_first; // TODO(rjf): @u64_to_u32 dst_scope->local_count = (RDI_U32)(local_idx_opl - local_idx_first); // TODO(rjf): @u64_to_u32 dst_scope->inline_site_idx = (RDI_U32)rdim_idx_from_inline_site(src_scope->inline_site); // TODO(rjf): @u64_to_u32 } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake procedures // ProfScope("bake procedures") { if(lane_idx() == 0) { rdim_shared->baked_procedures.procedures_count = params->procedures.total_count+1; rdim_shared->baked_procedures.procedures = push_array(arena, RDI_Procedure, rdim_shared->baked_procedures.procedures_count); } lane_sync(); { U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->procedures.first) { U64 location_block_layout_slot_idx = lane_idx()*total_location_case_chunk_count + params->scopes.chunk_count + chunk_idx; U64 location_block_off = rdim_shared->location_case_chunk_lane_offs[location_block_layout_slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Symbol *src = &n->v[n_idx]; RDI_Procedure *dst = &rdim_shared->baked_procedures.procedures[n->base_idx + n_idx + 1]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->link_name_string_idx = rdim_bake_idx_from_string(bake_strings, src->link_name); if(src->is_extern) { dst->link_flags |= RDI_LinkFlag_External; } if(src->container_type != 0) { dst->link_flags |= RDI_LinkFlag_TypeScoped; dst->container_idx = src->container_type ? (RDI_U32)rdim_idx_from_udt(src->container_type->udt) : 0; // TODO(rjf): @u64_to_u32 } else if(src->container_symbol != 0) { dst->link_flags |= RDI_LinkFlag_ProcScoped; dst->container_idx = (RDI_U32)rdim_idx_from_symbol(src->container_symbol); // TODO(rjf): @u64_to_u32 } dst->type_idx = (RDI_U32)rdim_idx_from_type(src->type); // TODO(rjf): @u64_to_u32 dst->root_scope_idx = (RDI_U32)rdim_idx_from_scope(src->root_scope); // TODO(rjf): @u64_to_u32 if(src->location_cases.count != 0) { dst->frame_base_location_first = location_block_off; dst->frame_base_location_opl = location_block_off + src->location_cases.count; location_block_off += src->location_cases.count; } } chunk_idx += 1; } } } ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage compute layout for constant data // ProfScope("compute layout for constant data") { // rjf: set up if(lane_idx() == 0) { rdim_shared->constant_data_chunk_lane_counts = push_array(arena, U64, lane_count() * params->constants.chunk_count); rdim_shared->constant_data_chunk_lane_offs = push_array(arena, U64, lane_count() * params->constants.chunk_count); } lane_sync(); // rjf: count { U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->constants.first) { U64 slot_idx = lane_idx()*params->constants.chunk_count + chunk_idx; Rng1U64 range = lane_range(n->count); for EachInRange(idx, range) { rdim_shared->constant_data_chunk_lane_counts[slot_idx] += n->v[idx].value_data.size; } chunk_idx += 1; } } lane_sync(); // rjf: layout if(lane_idx() == 0) { U64 chunk_idx = 0; U64 layout_off = 0; for EachNode(n, RDIM_SymbolChunkNode, params->constants.first) { for EachIndex(l_idx, lane_count()) { U64 slot_idx = l_idx*params->constants.chunk_count + chunk_idx; rdim_shared->constant_data_chunk_lane_offs[slot_idx] = layout_off; layout_off += rdim_shared->constant_data_chunk_lane_counts[slot_idx]; } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake constants // ProfScope("bake constants") { // rjf: set up if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->baked_constants.constant_values_count = params->constants.total_count+1; rdim_shared->baked_constants.constant_values = push_array(arena, RDI_U32, rdim_shared->baked_constants.constant_values_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->baked_constants.constant_value_data_size = params->constants.total_value_data_size; rdim_shared->baked_constants.constant_value_data = push_array(arena, RDI_U8, rdim_shared->baked_constants.constant_value_data_size); } if(lane_idx() == lane_from_task_idx(2)) { rdim_shared->baked_constants.constants_count = params->constants.total_count+1; rdim_shared->baked_constants.constants = push_array(arena, RDI_Constant, rdim_shared->baked_constants.constants_count); } lane_sync(); // rjf: wide bake { U64 chunk_idx = 0; for EachNode(n, RDIM_SymbolChunkNode, params->constants.first) { U64 slot_idx = lane_idx()*params->constants.chunk_count + chunk_idx; U64 value_data_off = rdim_shared->constant_data_chunk_lane_offs[slot_idx]; Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Symbol *src = &n->v[n_idx]; RDI_Constant *dst = &rdim_shared->baked_constants.constants[1 + n->base_idx + n_idx]; RDI_U32 *dst_value_off = &rdim_shared->baked_constants.constant_values[1 + n->base_idx + n_idx]; RDI_U8 *dst_value_data = rdim_shared->baked_constants.constant_value_data + value_data_off; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->type_idx = (RDI_U32)rdim_idx_from_type(src->type); // TODO(rjf): @u64_to_u32 dst->constant_value_idx = 1 + n->base_idx + n_idx; dst_value_off[0] = (RDI_U32)value_data_off; // TODO(rjf): @u64_to_u32 rdim_memcpy(dst_value_data, src->value_data.str, src->value_data.size); value_data_off += src->value_data.size; } chunk_idx += 1; } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake units, symbols, types, UDTs // { //- rjf: setup outputs if(lane_idx() == lane_from_task_idx(0)) { rdim_shared->baked_units.units_count = params->units.total_count+1; rdim_shared->baked_units.units = push_array(arena, RDI_Unit, rdim_shared->baked_units.units_count); } if(lane_idx() == lane_from_task_idx(1)) { rdim_shared->baked_type_nodes.type_nodes_count = params->types.total_count+1; rdim_shared->baked_type_nodes.type_nodes = push_array(arena, RDI_TypeNode, rdim_shared->baked_type_nodes.type_nodes_count); } if(lane_idx() == lane_from_task_idx(2)) { rdim_shared->baked_global_variables.global_variables_count = params->global_variables.total_count+1; rdim_shared->baked_global_variables.global_variables = push_array(arena, RDI_GlobalVariable, rdim_shared->baked_global_variables.global_variables_count); } if(lane_idx() == lane_from_task_idx(3)) { rdim_shared->baked_thread_variables.thread_variables_count = params->thread_variables.total_count+1; rdim_shared->baked_thread_variables.thread_variables = push_array(arena, RDI_ThreadVariable, rdim_shared->baked_thread_variables.thread_variables_count); } if(lane_idx() == lane_from_task_idx(4)) { rdim_shared->baked_inline_sites.inline_sites_count = params->inline_sites.total_count+1; rdim_shared->baked_inline_sites.inline_sites = push_array(arena, RDI_InlineSite, rdim_shared->baked_inline_sites.inline_sites_count); } lane_sync(); //- rjf: bake units ProfScope("bake units") { for EachNode(n, RDIM_UnitChunkNode, params->units.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Unit *src = &n->v[n_idx]; RDI_Unit *dst = &rdim_shared->baked_units.units[n->base_idx + n_idx + 1]; dst->unit_name_string_idx = rdim_bake_idx_from_string(bake_strings, src->unit_name); dst->compiler_name_string_idx = rdim_bake_idx_from_string(bake_strings, src->compiler_name); dst->source_file_path_node = rdim_bake_path_node_idx_from_string(path_tree, src->source_file); dst->object_file_path_node = rdim_bake_path_node_idx_from_string(path_tree, src->object_file); dst->archive_file_path_node = rdim_bake_path_node_idx_from_string(path_tree, src->archive_file); dst->build_path_node = rdim_bake_path_node_idx_from_string(path_tree, src->build_path); dst->language = src->language; dst->line_table_idx = (RDI_U32)rdim_idx_from_line_table(src->line_table); // TODO(rjf): @u64_to_u32 } } } //- rjf: bake type nodes ProfScope("bake type nodes") { for EachNode(n, RDIM_TypeChunkNode, params->types.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Type *src = &n->v[n_idx]; RDI_TypeNode *dst = &rdim_shared->baked_type_nodes.type_nodes[n->base_idx + n_idx + 1]; //- rjf: fill shared type node info dst->kind = src->kind; dst->flags = (RDI_U16)src->flags; // TODO(rjf): @u32_to_u16 dst->byte_size = src->byte_size; //- rjf: fill built-in-only type node info if(RDI_TypeKind_FirstBuiltIn <= dst->kind && dst->kind <= RDI_TypeKind_LastBuiltIn) { dst->built_in.name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); } //- rjf: fill array sizes else if(dst->kind == RDI_TypeKind_Array) { U64 direct_byte_size = 1; if(src->direct_type && src->direct_type->byte_size > 0) { direct_byte_size = src->direct_type->byte_size; } dst->constructed.direct_type_idx = (RDI_U32)rdim_idx_from_type(src->direct_type); dst->constructed.count = src->byte_size / direct_byte_size; } //- rjf: fill constructed type node info else if(RDI_TypeKind_FirstConstructed <= dst->kind && dst->kind <= RDI_TypeKind_LastConstructed) { dst->constructed.direct_type_idx = (RDI_U32)rdim_idx_from_type(src->direct_type); // TODO(rjf): @u64_to_u32 dst->constructed.count = src->count; if(dst->kind == RDI_TypeKind_Function || dst->kind == RDI_TypeKind_Method) { RDI_U32 param_idx_run_count = src->count; RDI_U32 *param_idx_run = rdim_push_array_no_zero(arena, RDI_U32, param_idx_run_count); for(RDI_U32 idx = 0; idx < param_idx_run_count; idx += 1) { param_idx_run[idx] = (RDI_U32)rdim_idx_from_type(src->param_types[idx]); // TODO(rjf): @u64_to_u32 } dst->constructed.param_idx_run_first = rdim_bake_idx_from_idx_run(bake_idx_runs, param_idx_run, param_idx_run_count); } else if(dst->kind == RDI_TypeKind_MemberPtr) { // TODO(rjf): member pointers not currently supported. } } //- rjf: fill user-defined-type info else if(RDI_TypeKind_FirstUserDefined <= dst->kind && dst->kind <= RDI_TypeKind_LastUserDefined) { dst->user_defined.name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->user_defined.udt_idx = (RDI_U32)rdim_idx_from_udt(src->udt); // TODO(rjf): @u64_to_u32 dst->user_defined.direct_type_idx = (RDI_U32)rdim_idx_from_type(src->direct_type); // TODO(rjf): @u64_to_u32 } //- rjf: fill bitfield info else if(dst->kind == RDI_TypeKind_Bitfield) { dst->bitfield.direct_type_idx = (RDI_U32)rdim_idx_from_type(src->direct_type); // TODO(rjf): @u64_to_u32 dst->bitfield.off = src->off; dst->bitfield.size = src->count; } } } } //- rjf: bake global variables ProfScope("bake global variables") { for EachNode(n, RDIM_SymbolChunkNode, params->global_variables.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Symbol *src = &n->v[n_idx]; RDI_GlobalVariable *dst = &rdim_shared->baked_global_variables.global_variables[n->base_idx + n_idx + 1]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->voff = src->offset; dst->type_idx = (RDI_U32)rdim_idx_from_type(src->type); // TODO(rjf): @u64_to_u32 if(src->is_extern) { dst->link_flags |= RDI_LinkFlag_External; } if(src->container_type != 0) { dst->link_flags |= RDI_LinkFlag_TypeScoped; dst->container_idx = src->container_type ? (RDI_U32)rdim_idx_from_udt(src->container_type->udt) : 0; // TODO(rjf): @u64_to_u32 } else if(src->container_symbol != 0) { dst->link_flags |= RDI_LinkFlag_ProcScoped; dst->container_idx = (RDI_U32)rdim_idx_from_symbol(src->container_symbol); // TODO(rjf): @u64_to_u32 } } } } //- rjf: bake thread variables ProfScope("bake thread variables") { for EachNode(n, RDIM_SymbolChunkNode, params->thread_variables.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDIM_Symbol *src = &n->v[n_idx]; RDI_ThreadVariable *dst = &rdim_shared->baked_thread_variables.thread_variables[n->base_idx + n_idx + 1]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->tls_off = (RDI_U32)src->offset; // TODO(rjf): @u64_to_u32 dst->type_idx = (RDI_U32)rdim_idx_from_type(src->type); if(src->is_extern) { dst->link_flags |= RDI_LinkFlag_External; } if(src->container_type != 0) { dst->link_flags |= RDI_LinkFlag_TypeScoped; dst->container_idx = src->container_type ? (RDI_U32)rdim_idx_from_udt(src->container_type->udt) : 0; // TODO(rjf): @u64_to_u32 } else if(src->container_symbol != 0) { dst->link_flags |= RDI_LinkFlag_ProcScoped; dst->container_idx = (RDI_U32)rdim_idx_from_symbol(src->container_symbol); // TODO(rjf): @u64_to_u32 } } } } //- rjf: bake inline sites ProfScope("bake inline sites") { for EachNode(n, RDIM_InlineSiteChunkNode, params->inline_sites.first) { Rng1U64 range = lane_range(n->count); for EachInRange(n_idx, range) { RDI_InlineSite *dst = &rdim_shared->baked_inline_sites.inline_sites[n->base_idx + n_idx + 1]; RDIM_InlineSite *src = &n->v[n_idx]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->type_idx = (RDI_U32)rdim_idx_from_type(src->type); // TODO(rjf): @u64_to_u32 dst->owner_type_idx = (RDI_U32)rdim_idx_from_type(src->owner); // TODO(rjf): @u64_to_u32 dst->line_table_idx = (RDI_U32)rdim_idx_from_line_table(src->line_table); // TODO(rjf): @u64_to_u32 } } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage bake file paths // ProfScope("bake file paths") { // rjf: set up if(lane_idx() == 0) { rdim_shared->baked_file_paths.nodes_count = path_tree->count; rdim_shared->baked_file_paths.nodes = push_array(arena, RDI_FilePathNode, rdim_shared->baked_file_paths.nodes_count); rdim_shared->baked_file_path_src_nodes = push_array(arena, RDIM_BakePathNode *, rdim_shared->baked_file_paths.nodes_count); { U64 idx = 0; for(RDIM_BakePathNode *n = path_tree->first; n != 0; n = n->next_order) { rdim_shared->baked_file_path_src_nodes[idx] = n; idx += 1; } } } lane_sync(); // rjf: fill { Rng1U64 range = lane_range(rdim_shared->baked_file_paths.nodes_count); for EachInRange(idx, range) { RDIM_BakePathNode *src = rdim_shared->baked_file_path_src_nodes[idx]; RDI_FilePathNode *dst = &rdim_shared->baked_file_paths.nodes[idx]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->source_file_idx = rdim_idx_from_src_file(src->src_file); if(src->parent != 0) { dst->parent_path_node = src->parent->idx; } if(src->first_child != 0) { dst->first_child = src->first_child->idx; } if(src->next_sibling != 0) { dst->next_sibling = src->next_sibling->idx; } } } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage do small final baking tasks // ProfScope("do small final baking tasks") { if(lane_idx() == lane_from_task_idx(0)) ProfScope("bake top level info") { rdim_shared->baked_top_level_info.top_level_info.arch = params->top_level_info.arch; rdim_shared->baked_top_level_info.top_level_info.exe_name_string_idx = rdim_bake_idx_from_string(bake_strings, params->top_level_info.exe_name); rdim_shared->baked_top_level_info.top_level_info.exe_hash = params->top_level_info.exe_hash; rdim_shared->baked_top_level_info.top_level_info.voff_max = params->top_level_info.voff_max; rdim_shared->baked_top_level_info.top_level_info.guid = params->top_level_info.guid; rdim_shared->baked_top_level_info.top_level_info.producer_name_string_idx = rdim_bake_idx_from_string(bake_strings, params->top_level_info.producer_name); } if(lane_idx() == lane_from_task_idx(1)) ProfScope("bake binary sections") { RDIM_BinarySectionList *src = ¶ms->binary_sections; RDI_BinarySection *dst_base = rdim_push_array(arena, RDI_BinarySection, src->count+1); U64 dst_idx = 1; for(RDIM_BinarySectionNode *src_n = src->first; src_n != 0; src_n = src_n->next, dst_idx += 1) { RDIM_BinarySection *src = &src_n->v; RDI_BinarySection *dst = &dst_base[dst_idx]; dst->name_string_idx = rdim_bake_idx_from_string(bake_strings, src->name); dst->flags = src->flags; dst->voff_first = src->voff_first; dst->voff_opl = src->voff_opl; dst->foff_first = src->foff_first; dst->foff_opl = src->foff_opl; } rdim_shared->baked_binary_sections.binary_sections = dst_base; rdim_shared->baked_binary_sections.binary_sections_count = dst_idx; } } lane_sync(); ////////////////////////////////////////////////////////////// //- rjf: @rdim_bake_stage package results // RDIM_BakeResults result = {0}; { result.top_level_info = rdim_shared->baked_top_level_info; result.binary_sections = rdim_shared->baked_binary_sections; result.units = rdim_shared->baked_units; result.unit_vmap = rdim_shared->baked_unit_vmap; result.src_files = rdim_shared->baked_src_files; result.checksums = rdim_shared->baked_checksums; result.line_tables = rdim_shared->baked_line_tables; result.type_nodes = rdim_shared->baked_type_nodes; result.udts = rdim_shared->baked_udts; result.global_variables = rdim_shared->baked_global_variables; result.global_vmap = rdim_shared->baked_global_vmap; result.thread_variables = rdim_shared->baked_thread_variables; result.constants = rdim_shared->baked_constants; result.procedures = rdim_shared->baked_procedures; result.scopes = rdim_shared->baked_scopes; result.inline_sites = rdim_shared->baked_inline_sites; result.scope_vmap = rdim_shared->baked_scope_vmap; result.top_level_name_maps = rdim_shared->baked_top_level_name_maps; result.name_maps = rdim_shared->baked_name_maps; result.file_paths = rdim_shared->baked_file_paths; result.strings = rdim_shared->baked_strings; result.idx_runs = rdim_shared->baked_idx_runs; result.locations = rdim_shared->baked_locations; result.location_blocks = rdim_shared->baked_location_blocks; } return result; } internal RDIM_SerializedSectionBundle rdim_compress(Arena *arena, RDIM_SerializedSectionBundle *in) { Temp scratch = scratch_begin(&arena, 1); RDIM_SerializedSectionBundle out_ = {0}; RDIM_SerializedSectionBundle *out = &out_; lane_sync_u64(&out, 0); //- rjf: set up compression context rr_lzb_simple_context ctx = {0}; ctx.m_tableSizeBits = 14; ctx.m_hashTable = push_array(scratch.arena, U16, 1<sections[k]; RDIM_SerializedSection *dst = &out->sections[k]; MemoryCopyStruct(dst, src); if(src->encoded_size != 0) { MemoryZero(ctx.m_hashTable, sizeof(U16)*(1<data = push_array_no_zero(arena, U8, src->encoded_size); dst->encoded_size = rr_lzb_simple_encode_veryfast(&ctx, src->data, src->encoded_size, dst->data); dst->unpacked_size = src->encoded_size; dst->encoding = RDI_SectionEncoding_LZB; } } lane_sync(); scratch_end(scratch); return *out; }