ed/raddebugger
1
0
Fork 0
mirror of https://github.com/Ed94/raddebugger.git synced 2026-08-14 17:28:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

tweak thread broadcasting and balance symbol inputs

Browse Source
This commit is contained in:
Nikita Smith
2026-04-06 12:04:33 -07:00
parent 77ece3193e
commit f69532e11d
3 changed files with 98 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/linker/lnk_debug_info.c
+80 -58
View File
@@ -512,7 +512,24 @@ lnk_make_code_view_input(TP_Context *tp, TP_Arena *tp_arena, LNK_IO_Flags io_fla
}
}
input.symbol_input_ranges = tp_divide_work(tp_arena->v[0], input.symbol_input_count, tp->worker_count);
ProfBegin("Make Ranges");
U64 total_input_size = 0;
for EachIndex(i, input.symbol_input_count) { total_input_size += input.symbol_inputs[i].raw_symbols.size; }
U64 max_weight = CeilIntegerDiv(total_input_size, tp->worker_count);
U64 cursor = 0;
input.symbol_input_ranges = push_array(tp_arena->v[0], Rng1U64, tp->worker_count);
for EachIndex(i, tp->worker_count) {
if (cursor >= input.symbol_input_count) { break; }
U64 begin = cursor;
U64 weight = 0;
for (; cursor < input.symbol_input_count; cursor += 1) {
if (weight >= max_weight) { break; }
weight += input.symbol_inputs[cursor].raw_symbols.size;
}
input.symbol_input_ranges[i] = r1u64(begin, cursor);
}
ProfEnd();
}
ProfEnd();
@@ -1174,9 +1191,9 @@ THREAD_POOL_TASK_FUNC(lnk_cv_patcher_symbols_task)
{
ProfBeginFunction();
LNK_MergeTypes *task = raw_task;
for EachInRange(i, task->input->symbol_input_ranges[task_id]) {
Rng1U64 range = task->input->symbol_input_ranges[task_id];
for EachInRange(i, range) {
LNK_SymbolInput symbols = task->input->symbol_inputs[i];
for (U64 cursor = 0; cursor + sizeof(CV_SymbolHeader) <= symbols.raw_symbols.size; ) {
Temp temp = temp_begin(task->fixed_arenas[task_id]);
@@ -1199,8 +1216,9 @@ THREAD_POOL_TASK_FUNC(lnk_cv_patcher_inlines_task)
LNK_MergeTypes *task = raw_task;
U64 obj_idx = task_id;
String8List inlinee_lines = cv_sub_section_from_debug_s(task->input->debug_s_arr[obj_idx], CV_C13SubSectionKind_InlineeLines);
Arena *fixed_arena = task->fixed_arenas[worker_id];
for EachNode(inline_data_n, String8Node, inlinee_lines.first) {
Temp temp = temp_begin(task->fixed_arenas[worker_id]);
Temp temp = temp_begin(fixed_arena);
CV_TypeIndexInfoList ti_info_list = cv_get_inlinee_type_index_offsets(temp.arena, inline_data_n->string);
lnk_fixup_cv_type_indices(task, obj_idx, inline_data_n->string, ti_info_list);
temp_end(temp);
@@ -1213,8 +1231,10 @@ THREAD_POOL_TASK_FUNC(lnk_cv_patcher_leaves_task)
{
ProfBeginFunction();
LNK_MergeTypes *task = raw_task;
for EachInRange(leaf_ref_idx, task->ranges[task_id]) {
Temp temp = temp_begin(task->fixed_arenas[task_id]);
Rng1U64 range = task->ranges[task_id];
Arena *fixed_arena = task->fixed_arenas[task_id];
for EachInRange(leaf_ref_idx, range) {
Temp temp = temp_begin(fixed_arena);
LNK_LeafRef *patch = task->unique_leaf_refs_arr[task->ti_source].v[leaf_ref_idx];
CV_DebugT *debug_t = &task->input->debug_t_arr[patch->obj_idx];
CV_Leaf leaf = cv_debug_t_get_leaf(debug_t, patch->leaf_idx);
@@ -1229,7 +1249,8 @@ internal
THREAD_POOL_TASK_FUNC(lnk_unbucket_raw_leaves_task)
{
LNK_MergeTypes *task = raw_task;
for EachInRange(i, task->ranges[task_id]) {
Rng1U64 range = task->ranges[task_id];
for EachInRange(i, range) {
LNK_LeafRef leaf_ref = *task->unique_leaf_refs_arr[task->ti_source].v[i];
CV_DebugT *debug_t = &task->input->debug_t_arr[leaf_ref.obj_idx];
String8 raw_leaf = cv_debug_t_get_raw_leaf(debug_t, leaf_ref.leaf_idx);
@@ -1751,19 +1772,11 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
PDB_PsiContext *psi = task->pdb->psi;
U32Array obj_indices = task->obj_indices[task_id];
//
// global symbols
//
ProfBegin("Global Symbols");
{
// collect global data and global typedefs
U64 *global_symbol_count = 0;
if (task_id == 0) {
global_symbol_count = push_array(scratch.arena, U64, 1);
}
tp_broadcast(&global_symbol_count);
VoidList global_symbols = {0};
for EachInRange(i, task->cv->symbol_input_ranges[task_id]) {
Rng1U64 symbol_input_range = task->cv->symbol_input_ranges[task_id];
for EachInRange(i, symbol_input_range) {
LNK_SymbolInput symbols = task->cv->symbol_inputs[i];
for (U64 cursor = 0, depth = 0; cursor + sizeof(CV_SymbolHeader) <= symbols.raw_symbols.size; ) {
CV_Symbol symbol = {0};
@@ -1787,14 +1800,15 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
}
}
}
ins_atomic_u64_add_eval(global_symbol_count, global_symbols.count);
barrier_wait(tp->barrier);
U64 *bucket_cap;
// collect global data and global typedefs
U64 global_symbol_count = tp_sum_u64(tp, task_id, global_symbols.count);
U64 bucket_cap;
void **buckets;
if (task_id == 0) {
bucket_cap = push_u64(scratch.arena, *global_symbol_count * 13 / 10);
buckets = push_array(scratch.arena, void *, *bucket_cap);
bucket_cap = global_symbol_count * 13 / 10;
buckets = push_array(scratch.arena, void *, bucket_cap);
}
tp_broadcast(&bucket_cap);
tp_broadcast(&buckets);
@@ -1803,31 +1817,30 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
for EachNode(n, VoidNode, global_symbols.first) {
String8 raw = cv_raw_from_symbol(n->v);
U64 hash = u64_hash_from_str8(raw);
cv_symbol_deduper_insert_or_update(buckets, *bucket_cap, hash, n->v);
cv_symbol_deduper_insert_or_update(buckets, bucket_cap, hash, n->v);
}
barrier_wait(tp->barrier);
U64 *symbol_count;
void **symbol_arr;
Rng1U64 *symbol_ranges; // [worker_count]
U32 *symbol_hashes; // [symbol_count]
U64 symbol_count = 0;
void **symbol_arr = 0; // [symbol_count]
Rng1U64 *symbol_ranges = 0; // [worker_count]
U32 *symbol_hashes = 0; // [symbol_count]
if (task_id == 0) {
// compact buckets
ProfBeginV("Compact Buckets [bucket_cap %llu]", bucket_cap);
{
U64 k = 0;
for (U64 i = 1; i < *bucket_cap; i += 1) {
if (buckets[i] != 0 && buckets[i-1] == 0) {
while (buckets[k] != 0) { k += 1; }
buckets[k] = buckets[i];
buckets[i] = 0;
U64 dst = 0;
for EachIndex(src, bucket_cap) {
if (buckets[src] != 0) {
buckets[dst++] = buckets[src];
}
}
symbol_count = push_u64(scratch.arena, k);
symbol_count = dst;
}
ProfEnd();
symbol_arr = buckets;
symbol_ranges = tp_divide_work(scratch.arena, *symbol_count, tp->worker_count);
symbol_hashes = push_array_no_zero(scratch.arena, U32, *symbol_count);
symbol_ranges = tp_divide_work(scratch.arena, symbol_count, tp->worker_count);
symbol_hashes = push_array_no_zero(scratch.arena, U32, symbol_count);
}
tp_broadcast(&symbol_count);
tp_broadcast(&symbol_arr);
@@ -1835,7 +1848,8 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
tp_broadcast(&symbol_hashes);
// hash symbols
for EachInRange(i, symbol_ranges[task_id]) {
Rng1U64 symbol_range = symbol_ranges[task_id];
for EachInRange(i, symbol_range) {
CV_Symbol symbol = cv_symbol_from_ptr(symbol_arr[i]);
String8 name = cv_name_from_symbol(symbol.kind, symbol.data);
symbol_hashes[i] = gsi_hash(gsi, name);
@@ -1844,8 +1858,8 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
// push global symbols
if (task_id == 0) {
CV_SymbolNode *nodes = push_array_no_zero(gsi->arena, CV_SymbolNode, *symbol_count);
for EachIndex(i, *symbol_count) {
CV_SymbolNode *nodes = push_array_no_zero(gsi->arena, CV_SymbolNode, symbol_count);
for EachIndex(i, symbol_count) {
CV_SymbolNode *n = &nodes[i];
n->prev = n->next = 0;
n->data = cv_symbol_from_ptr(symbol_arr[i]);
@@ -1853,10 +1867,9 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
}
}
}
ProfEnd();
//
// proc refs
//
ProfBegin("Proc Refs");
{
U64 *proc_ref_sizes = 0;
U64 *proc_ref_counts = 0;
@@ -1908,6 +1921,7 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
tp_broadcast(&proc_ref_arenas);
tp_broadcast(&proc_ref_nodes);
Arena *proc_ref_arena = proc_ref_arenas[task_id];
U64 proc_ref_idx = proc_ref_indices[task_id];
for EachIndex(i, obj_indices.count) {
U64 obj_idx = obj_indices.v[i];
@@ -1923,7 +1937,7 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
if (symbol.kind == CV_SymKind_GPROC32 || symbol.kind == CV_SymKind_LPROC32) {
String8 name = cv_name_from_symbol(symbol.kind, symbol.data);
proc_ref_nodes[proc_ref_idx].data = cv_make_proc_ref(proc_ref_arenas[task_id], imod, symbol_cursor, name, cv_is_lproc(symbol));
proc_ref_nodes[proc_ref_idx].data = cv_make_proc_ref(proc_ref_arena, imod, symbol_cursor, name, cv_is_lproc(symbol));
proc_ref_hashes[proc_ref_idx] = hash_from_cv_symbol(&proc_ref_nodes[proc_ref_idx].data);
proc_ref_idx += 1;
}
@@ -1941,10 +1955,9 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
}
barrier_wait(tp->barrier);
}
ProfEnd();
//
// public symbols
//
ProfBegin("Public Symbols");
{
U64 *public_symbol_sizes = 0; // [worker_count]
U64 *public_symbol_node_counts = 0; // [worker_count]
@@ -1956,7 +1969,10 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
tp_broadcast(&public_symbol_node_counts);
// compute buffer size for CV public symbols
for EachNode(chunk, LNK_SymbolHashTrieChunk, task->symtab->chunks[task_id].first) {
LNK_SymbolHashTrieChunkList symbol_chunks = task->symtab->chunks[task_id];
U64 public_symbol_size = 0;
U64 public_symbol_count = 0;
for EachNode(chunk, LNK_SymbolHashTrieChunk, symbol_chunks.first) {
for EachIndex(i, chunk->count) {
LNK_Symbol *symbol = chunk->v[i].symbol;
LNK_ObjSymbolRef symbol_ref = lnk_ref_from_symbol(symbol);
@@ -1966,12 +1982,13 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
COFF_SymbolValueInterpType symbol_interp = coff_interp_from_parsed_symbol(symbol_parsed);
if (symbol_interp != COFF_SymbolValueInterp_Regular) { continue; }
public_symbol_sizes[task_id] += sizeof(CV_SymPub32);
public_symbol_sizes[task_id] += symbol->name.size + 1;
public_symbol_sizes[task_id] = AlignPow2(public_symbol_sizes[task_id], sizeof(void *));
public_symbol_size += AlignPow2(sizeof(CV_SymPub32) + symbol->name.size + 1, sizeof(void *));
public_symbol_count += 1;
public_symbol_node_counts[task_id] += 1;
}
}
public_symbol_sizes [task_id] += public_symbol_size;
public_symbol_node_counts[task_id] += public_symbol_count;
barrier_wait(tp->barrier);
Arena **public_symbol_arenas = 0;
@@ -1991,7 +2008,10 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
tp_broadcast(&public_symbol_hashes);
// make CV public symbols
for EachNode(chunk, LNK_SymbolHashTrieChunk, task->symtab->chunks[task_id].first) {
Arena *public_symbol_arena = public_symbol_arenas [task_id];
Arena *public_symbol_node_arena = public_symbol_node_arenas[task_id];
CV_SymbolList *public_symbol_list = &public_symbols [task_id];
for EachNode(chunk, LNK_SymbolHashTrieChunk, symbol_chunks.first) {
for EachIndex(i, chunk->count) {
LNK_Symbol *symbol = chunk->v[i].symbol;
LNK_ObjSymbolRef symbol_ref = lnk_ref_from_symbol(symbol);
@@ -2004,20 +2024,21 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
CV_Pub32Flags flags = COFF_SymbolType_IsFunc(symbol_parsed.type) ? CV_Pub32Flag_Function : 0;
ISectOff sc = lnk_sc_from_symbol(symbol);
CV_Symbol pub_symbol = cv_make_pub32(public_symbol_arenas[task_id], flags, safe_cast_u32(sc.off), safe_cast_u16(sc.isect), symbol->name);
cv_symbol_list_push(public_symbol_node_arenas[task_id], &public_symbols[task_id], pub_symbol);
CV_Symbol pub_symbol = cv_make_pub32(public_symbol_arena, flags, safe_cast_u32(sc.off), safe_cast_u16(sc.isect), symbol->name);
cv_symbol_list_push(public_symbol_node_arena, public_symbol_list, pub_symbol);
}
}
barrier_wait(tp->barrier);
// hash public symbols
{
U64 n_idx = 0;
public_symbol_hashes[task_id] = push_array(scratch.arena, U32, public_symbols[task_id].count);
U64 hash_idx = 0;
U32 *hashes = push_array(scratch.arena, U32, public_symbols[task_id].count);
for EachNode(n, CV_SymbolNode, public_symbols[task_id].first) {
String8 name = cv_name_from_symbol(n->data.kind, n->data.data);
public_symbol_hashes[task_id][n_idx++] = gsi_hash(gsi, name);
hashes[hash_idx++] = gsi_hash(gsi, name);
}
public_symbol_hashes[task_id] = hashes;
}
barrier_wait(tp->barrier);
@@ -2034,6 +2055,7 @@ THREAD_POOL_TASK_FUNC(lnk_move_global_symbols_to_gsi)
}
barrier_wait(tp->barrier);
}
ProfEnd();
scratch_end(scratch);
}
src/linker/thread_pool/thread_pool.c
+6 -5
View File
@@ -246,18 +246,19 @@ tp_divide_work(Arena *arena, U64 item_count, U32 worker_count)
return range_arr;
}
internal void *
tp_broadcast_(TP_Context *tp, U64 task_id, void *ptr)
internal void
tp_broadcast_(TP_Context *tp, U64 task_id, void *ptr, U64 ptr_size)
{
if (task_id == 0) {
tp->broadcast = ptr;
tp->broadcast_size = ptr_size;
}
barrier_wait(tp->barrier);
void *result = tp->broadcast;
if (task_id != 0) {
MemoryCopy(ptr, tp->broadcast, tp->broadcast_size);
}
barrier_wait(tp->barrier);
return result;
}
internal U64
src/linker/thread_pool/thread_pool.h
+3 -2
View File
@@ -4,7 +4,7 @@
#pragma once
struct TP_Context;
#define THREAD_POOL_TASK_FUNC(name) void name(Arena *arena, volatile U64 worker_id, volatile U64 task_id, void *raw_task, struct TP_Context *tp)
#define THREAD_POOL_TASK_FUNC(name) void name(Arena *arena, U64 worker_id, U64 task_id, void *raw_task, struct TP_Context *tp)
typedef THREAD_POOL_TASK_FUNC(TP_TaskFunc);
typedef struct TP_Arena
@@ -34,6 +34,7 @@ typedef struct TP_Context
Semaphore main_semaphore;
Barrier barrier;
void *broadcast;
U64 broadcast_size;
U64 sum;
U32 worker_count;
@@ -56,5 +57,5 @@ internal void tp_temp_end(TP_Temp temp);
#define tp_for_parallel_prof(pool, arena, task_count, task_func, task_data, zone_name) ProfBegin(zone_name); tp_for_parallel(pool, arena, task_count, task_func, task_data); ProfEnd();
internal void tp_for_parallel(TP_Context *pool, TP_Arena *arena, U64 task_count, TP_TaskFunc *task_func, void *task_data);
internal Rng1U64 * tp_divide_work(Arena *arena, U64 item_count, U32 worker_count);
#define tp_broadcast(p) *(p) = tp_broadcast_(tp, task_id, *(p))
#define tp_broadcast(p) tp_broadcast_(tp, task_id, p, sizeof(*p))