plug in codeview-conversion-backed pdb converison

This commit is contained in:
Ryan Fleury
2026-06-18 08:54:41 -07:00
parent d56cd83cac
commit a441a5eef3
9 changed files with 187 additions and 27 deletions
+1 -2
View File
@@ -324,7 +324,7 @@ pdb_tpi_from_data(Arena *arena, String8 data)
} }
internal PDB_TpiHashParsed* internal PDB_TpiHashParsed*
pdb_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, String8 data, String8 aux_data) pdb_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, String8 data)
{ {
ProfBeginFunction(); ProfBeginFunction();
PDB_TpiHashParsed *result = push_array(arena, PDB_TpiHashParsed, 1); PDB_TpiHashParsed *result = push_array(arena, PDB_TpiHashParsed, 1);
@@ -442,7 +442,6 @@ pdb_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, Str
// fill result // fill result
result->data = data; result->data = data;
result->aux_data = aux_data;
result->buckets = buckets; result->buckets = buckets;
result->bucket_count = bucket_count; result->bucket_count = bucket_count;
result->bucket_mask = bucket_mask; result->bucket_mask = bucket_mask;
+1 -7
View File
@@ -163,8 +163,6 @@ typedef struct PDB_TpiHashBlock
typedef struct PDB_TpiHashParsed typedef struct PDB_TpiHashParsed
{ {
String8 data; String8 data;
String8 aux_data;
PDB_TpiHashBlock **buckets; PDB_TpiHashBlock **buckets;
U32 bucket_count; U32 bucket_count;
U32 bucket_mask; U32 bucket_mask;
@@ -198,11 +196,7 @@ internal PDB_Strtbl* pdb_strtbl_from_data(Arena *arena, String8 strtbl_
internal PDB_DbiParsed* pdb_dbi_from_data(Arena *arena, String8 dbi_data); internal PDB_DbiParsed* pdb_dbi_from_data(Arena *arena, String8 dbi_data);
internal PDB_TpiParsed* pdb_tpi_from_data(Arena *arena, String8 tpi_data); internal PDB_TpiParsed* pdb_tpi_from_data(Arena *arena, String8 tpi_data);
internal PDB_TpiHashParsed* pdb_tpi_hash_from_data(Arena *arena, internal PDB_TpiHashParsed* pdb_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, String8 tpi_hash_data);
PDB_Strtbl *strtbl,
PDB_TpiParsed *tpi,
String8 tpi_hash_data,
String8 tpi_hash_aux_data);
internal PDB_GsiParsed* pdb_gsi_from_data(Arena *arena, String8 gsi_data); internal PDB_GsiParsed* pdb_gsi_from_data(Arena *arena, String8 gsi_data);
internal U64 pdb_gsi_symbol_from_string(PDB_GsiParsed *gsi, String8 symbol_data, String8 string); internal U64 pdb_gsi_symbol_from_string(PDB_GsiParsed *gsi, String8 symbol_data, String8 string);
+7
View File
@@ -874,8 +874,15 @@ rb_thread_entry_point(void *p)
convert_params.subset_flags = subset_flags; convert_params.subset_flags = subset_flags;
convert_params.deterministic = cmd_line_has_flag(cmdline, str8_lit("deterministic")); convert_params.deterministic = cmd_line_has_flag(cmdline, str8_lit("deterministic"));
} }
if(cmd_line_has_flag(cmdline, s("a")))
{
ProfScope("convert") pdb_bake_params = p2r_convert(arena, &convert_params); ProfScope("convert") pdb_bake_params = p2r_convert(arena, &convert_params);
} }
else
{
ProfScope("convert") pdb_bake_params = p2r_convert2(arena, &convert_params);
}
}
//- rjf: RDI inputs => RDI joining //- rjf: RDI inputs => RDI joining
if(input_files_from_format_table[RB_FileFormat_RDI].count > 1) if(input_files_from_format_table[RB_FileFormat_RDI].count > 1)
+2
View File
@@ -406,6 +406,7 @@
#include "unwind/unwind_inc.h" #include "unwind/unwind_inc.h"
#include "rdi_from_coff/rdi_from_coff.h" #include "rdi_from_coff/rdi_from_coff.h"
#include "rdi_from_elf/rdi_from_elf.h" #include "rdi_from_elf/rdi_from_elf.h"
#include "rdi_from_codeview/rdi_from_codeview.h"
#include "rdi_from_pdb/rdi_from_pdb.h" #include "rdi_from_pdb/rdi_from_pdb.h"
#include "rdi_from_dwarf/rdi_from_dwarf.h" #include "rdi_from_dwarf/rdi_from_dwarf.h"
#include "radbin/radbin.h" #include "radbin/radbin.h"
@@ -466,6 +467,7 @@
#include "unwind/unwind_inc.c" #include "unwind/unwind_inc.c"
#include "rdi_from_coff/rdi_from_coff.c" #include "rdi_from_coff/rdi_from_coff.c"
#include "rdi_from_elf/rdi_from_elf.c" #include "rdi_from_elf/rdi_from_elf.c"
#include "rdi_from_codeview/rdi_from_codeview.c"
#include "rdi_from_pdb/rdi_from_pdb.c" #include "rdi_from_pdb/rdi_from_pdb.c"
#include "rdi_from_dwarf/rdi_from_dwarf.c" #include "rdi_from_dwarf/rdi_from_dwarf.c"
#include "radbin/radbin.c" #include "radbin/radbin.c"
+9 -1
View File
@@ -331,7 +331,15 @@ cv2r_itypes_from_name(Arena *arena, CV2R_TPIHash *tpi_hash, CV_LeafParsed *leaf,
internal CV_TypeId internal CV_TypeId
cv2r_first_itype_from_name(CV2R_TPIHash *tpi_hash, CV_LeafParsed *tpi_leaf, String8 name, B32 compare_unique_name) cv2r_first_itype_from_name(CV2R_TPIHash *tpi_hash, CV_LeafParsed *tpi_leaf, String8 name, B32 compare_unique_name)
{ {
Temp scratch = scratch_begin(0, 0);
CV_TypeIdArray array = cv2r_itypes_from_name(scratch.arena, tpi_hash, tpi_leaf, name, compare_unique_name, 1);
CV_TypeId result = 0;
if(array.count > 0)
{
result = array.itypes[0];
}
scratch_end(scratch);
return result;
} }
//////////////////////////////// ////////////////////////////////
@@ -49,7 +49,6 @@ typedef struct CV2R_TPIHash CV2R_TPIHash;
struct CV2R_TPIHash struct CV2R_TPIHash
{ {
String8 data; String8 data;
String8 aux_data;
CV2R_TPIHashBlock **buckets; CV2R_TPIHashBlock **buckets;
U32 bucket_count; U32 bucket_count;
U32 bucket_mask; U32 bucket_mask;
+159 -15
View File
@@ -550,12 +550,11 @@ p2r_convert(Arena *arena, P2R_ConvertParams *params)
if(lane_idx() == lane_from_task_idx(1)) ProfScope("parse TPI hash") if(lane_idx() == lane_from_task_idx(1)) ProfScope("parse TPI hash")
{ {
String8 hash_data = msf_data_from_stream(msf, tpi->hash_sn); String8 hash_data = msf_data_from_stream(msf, tpi->hash_sn);
String8 aux_data = msf_data_from_stream(msf, tpi->hash_sn_aux);
if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs))) if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs)))
{ {
hash_data = aux_data = str8_zero(); hash_data = str8_zero();
} }
tpi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, tpi, hash_data, aux_data); tpi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, tpi, hash_data);
} }
if(lane_idx() == lane_from_task_idx(2)) ProfScope("parse TPI leaf") if(lane_idx() == lane_from_task_idx(2)) ProfScope("parse TPI leaf")
{ {
@@ -565,12 +564,11 @@ p2r_convert(Arena *arena, P2R_ConvertParams *params)
if(lane_idx() == lane_from_task_idx(3)) ProfScope("parse IPI hash") if(lane_idx() == lane_from_task_idx(3)) ProfScope("parse IPI hash")
{ {
String8 hash_data = msf_data_from_stream(msf, ipi->hash_sn); String8 hash_data = msf_data_from_stream(msf, ipi->hash_sn);
String8 aux_data = msf_data_from_stream(msf, ipi->hash_sn_aux);
if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs))) if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs)))
{ {
hash_data = aux_data = str8_zero(); hash_data = str8_zero();
} }
ipi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, ipi, hash_data, aux_data); ipi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, ipi, hash_data);
} }
if(lane_idx() == lane_from_task_idx(4)) ProfScope("parse IPI leaf") if(lane_idx() == lane_from_task_idx(4)) ProfScope("parse IPI leaf")
{ {
@@ -5137,6 +5135,136 @@ p2r_convert(Arena *arena, P2R_ConvertParams *params)
return result; return result;
} }
////////////////////////////////
//~ rjf: PDB -> (Codeview -> RDI) TPI Hash Table Building
internal CV2R_TPIHash *
p2r_cv2r_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, String8 data)
{
CV2R_TPIHash *result = push_array(arena, CV2R_TPIHash, 1);
U32 stride = tpi->hash_key_size;
U32 bucket_count = tpi->hash_bucket_count;
if(1 <= stride && stride <= 8 && bucket_count > 0 && data.size != 0)
{
// allocate buckets
CV2R_TPIHashBlock **buckets = push_array(arena, CV2R_TPIHashBlock*, bucket_count);
// extract "hash" array
U8 *hashes = data.str + tpi->hash_vals_off;
U8 *hash_opl = hashes + tpi->hash_vals_size;
// for each index in the array...
CV_TypeId itype = tpi->itype_first;
U8 *hash_cursor = hashes;
for(;hash_cursor + stride <= hash_opl;)
{
// read index
U64 bucket_idx = 0;
MemoryCopy(&bucket_idx, hash_cursor, stride);
// save to map
if(bucket_idx < bucket_count)
{
CV2R_TPIHashBlock *block = buckets[bucket_idx];
if(block == 0 || block->local_count == ArrayCount(block->itypes))
{
block = push_array(arena, CV2R_TPIHashBlock, 1);
SLLStackPush(buckets[bucket_idx], block);
}
if(block->local_count != 0)
{
MemoryCopy(block->itypes+1, block->itypes, sizeof(CV_TypeId)*block->local_count);
}
block->itypes[0] = itype;
block->local_count += 1;
}
// advance cursor
hash_cursor += stride;
itype += 1;
}
//- rjf: compute bucket mask
U32 bucket_mask = 0;
if(IsPow2OrZero(bucket_count))
{
bucket_mask = bucket_count-1;
}
//- rjf: apply hash adjustments, to pull correct type IDs to the front of
// the chains
if(tpi->hash_adj_size != 0)
{
// NOTE(rjf): this table is laid out in the following format:
//
// pair_count: U32 -> # of name_index/type_index pairs
// slot_count: U32 -> # of slots in this hash table
// present_bit_array_count: U32 -> count for next array
// present_bit_array: U32[present_bit_array_count] -> 1 bit per slot, "is present"
// deleted_bit_array_count: U32 -> count for next array
// deleted_bit_array: U32[deleted_bit_array_count] -> 1 bit per slot, "is deleted"
// (U32, U32)[pair_count] -> array of name_index/type_index pairs
//
U8 *adjs = data.str + tpi->hash_adj_off;
U8 *adjs_opl = adjs + tpi->hash_adj_size;
U8 *adjs_cursor = adjs;
U32 pair_count = *(U32 *)adjs_cursor;
adjs_cursor += sizeof(U32);
U32 slot_count = *(U32 *)adjs_cursor;
adjs_cursor += sizeof(U32);
U32 present_bit_array_count = *(U32 *)adjs_cursor; // skip present_bit_array
adjs_cursor += sizeof(U32);
adjs_cursor += present_bit_array_count*sizeof(U32);
U32 deleted_bit_array_count = *(U32 *)adjs_cursor; // skip deleted_bit_array
adjs_cursor += sizeof(U32);
adjs_cursor += deleted_bit_array_count*sizeof(U32);
U32 adjs_stride = sizeof(U32)*2;
U32 pair_idx = 0;
for(;adjs_cursor < adjs_opl && pair_idx < pair_count;
adjs_cursor += adjs_stride, pair_idx += 1)
{
U32 name_off = ((U32 *)adjs_cursor)[0];
CV_TypeId type_id = ((CV_TypeId *)adjs_cursor)[1];
String8 string = pdb_strtbl_string_from_off(strtbl, name_off);
U32 hash = pdb_hash_v1(string);
U32 bucket_idx = ((bucket_mask != 0) ? hash&bucket_mask : hash%bucket_count);
CV2R_TPIHashBlock *prev_block = 0;
for(CV2R_TPIHashBlock *block = buckets[bucket_idx];
block != 0;
prev_block = block, block = block->next)
{
for(U32 local_idx = 0;
local_idx < block->local_count && local_idx < ArrayCount(block->itypes);
local_idx += 1)
{
if(block->itypes[local_idx] == type_id)
{
if(prev_block != 0)
{
prev_block->next = block->next;
block->next = buckets[bucket_idx];
buckets[bucket_idx] = block;
}
if(local_idx != 0)
{
Swap(CV_TypeId, block->itypes[0], block->itypes[local_idx]);
}
break;
}
}
}
}
}
// fill result
result->data = data;
result->buckets = buckets;
result->bucket_count = bucket_count;
result->bucket_mask = bucket_mask;
}
return result;
}
//////////////////////////////// ////////////////////////////////
//~ rjf: Top-Level Conversion Entry Point (Revised) //~ rjf: Top-Level Conversion Entry Point (Revised)
@@ -5282,9 +5410,9 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////
//- rjf: hash EXE, parse TPI/IPI hash/leaf & global symbol stream & comp units //- rjf: hash EXE, parse TPI/IPI hash/leaf & global symbol stream & comp units
// //
PDB_TpiHashParsed *tpi_hash = 0; CV2R_TPIHash *tpi_hash = 0;
CV_LeafParsed *tpi_leaf = 0; CV_LeafParsed *tpi_leaf = 0;
PDB_TpiHashParsed *ipi_hash = 0; CV2R_TPIHash *ipi_hash = 0;
CV_LeafParsed *ipi_leaf = 0; CV_LeafParsed *ipi_leaf = 0;
PDB_CompUnitArray *comp_units = 0; PDB_CompUnitArray *comp_units = 0;
PDB_CompUnitContributionArray *comp_unit_contributions = 0; PDB_CompUnitContributionArray *comp_unit_contributions = 0;
@@ -5293,12 +5421,11 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
if(lane_idx() == lane_from_task_idx(0)) ProfScope("parse TPI hash") if(lane_idx() == lane_from_task_idx(0)) ProfScope("parse TPI hash")
{ {
String8 hash_data = msf_data_from_stream(msf, tpi->hash_sn); String8 hash_data = msf_data_from_stream(msf, tpi->hash_sn);
String8 aux_data = msf_data_from_stream(msf, tpi->hash_sn_aux);
if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs))) if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs)))
{ {
hash_data = aux_data = str8_zero(); hash_data = str8_zero();
} }
tpi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, tpi, hash_data, aux_data); tpi_hash = p2r_cv2r_tpi_hash_from_data(scratch.arena, strtbl, tpi, hash_data);
} }
if(lane_idx() == lane_from_task_idx(1)) ProfScope("parse TPI leaf") if(lane_idx() == lane_from_task_idx(1)) ProfScope("parse TPI leaf")
{ {
@@ -5308,12 +5435,11 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
if(lane_idx() == lane_from_task_idx(2)) ProfScope("parse IPI hash") if(lane_idx() == lane_from_task_idx(2)) ProfScope("parse IPI hash")
{ {
String8 hash_data = msf_data_from_stream(msf, ipi->hash_sn); String8 hash_data = msf_data_from_stream(msf, ipi->hash_sn);
String8 aux_data = msf_data_from_stream(msf, ipi->hash_sn_aux);
if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs))) if(!(params->subset_flags & (RDIM_SubsetFlag_Types|RDIM_SubsetFlag_UDTs)))
{ {
hash_data = aux_data = str8_zero(); hash_data = str8_zero();
} }
ipi_hash = pdb_tpi_hash_from_data(scratch.arena, strtbl, ipi, hash_data, aux_data); ipi_hash = p2r_cv2r_tpi_hash_from_data(scratch.arena, strtbl, ipi, hash_data);
} }
if(lane_idx() == lane_from_task_idx(3)) ProfScope("parse IPI leaf") if(lane_idx() == lane_from_task_idx(3)) ProfScope("parse IPI leaf")
{ {
@@ -5424,7 +5550,7 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
{ {
cv2r_comp_units[idx].obj_name = comp_units->units[idx]->obj_name; cv2r_comp_units[idx].obj_name = comp_units->units[idx]->obj_name;
cv2r_comp_units[idx].group_name = comp_units->units[idx]->group_name; cv2r_comp_units[idx].group_name = comp_units->units[idx]->group_name;
cv2r_comp_units[idx].ranges = unit_ranges[idx-1]; cv2r_comp_units[idx].ranges = unit_ranges[idx+1];
} }
lane_sync(); lane_sync();
} }
@@ -5478,6 +5604,19 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
lane_sync(); lane_sync();
} }
//////////////////////////////////////////////////////////////
//- rjf: produce cv2r string table
//
CV2R_StringTable cv2r_strtbl = {0};
{
cv2r_strtbl.data = strtbl->data;
cv2r_strtbl.bucket_count = strtbl->bucket_count;
cv2r_strtbl.strblock_min = strtbl->strblock_min;
cv2r_strtbl.strblock_max = strtbl->strblock_max;
cv2r_strtbl.buckets_min = strtbl->buckets_min;
cv2r_strtbl.buckets_max = strtbl->buckets_max;
}
////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////
//- rjf: bundle codeview conversion parameters //- rjf: bundle codeview conversion parameters
// //
@@ -5495,8 +5634,13 @@ p2r_convert2(Arena *arena, P2R_ConvertParams *params)
cv2r_params.comp_unit_contributions = cv2r_comp_unit_contributions; cv2r_params.comp_unit_contributions = cv2r_comp_unit_contributions;
cv2r_params.sections_count = cv2r_sections_count; cv2r_params.sections_count = cv2r_sections_count;
cv2r_params.sections = cv2r_sections; cv2r_params.sections = cv2r_sections;
cv2r_params.strtbl = &cv2r_strtbl;
cv2r_params.tpi_leaf = tpi_leaf; cv2r_params.tpi_leaf = tpi_leaf;
cv2r_params.ipi_leaf = ipi_leaf; cv2r_params.ipi_leaf = ipi_leaf;
cv2r_params.tpi_hash = tpi_hash;
cv2r_params.ipi_hash = ipi_hash;
cv2r_params.subset_flags = params->subset_flags;
cv2r_params.deterministic = params->deterministic;
} }
////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////
+5
View File
@@ -134,6 +134,11 @@ internal void p2r_location_case_list_push_over_lvar_addr_range(Arena *arena, RDI
internal RDIM_BakeParams p2r_convert(Arena *arena, P2R_ConvertParams *params); internal RDIM_BakeParams p2r_convert(Arena *arena, P2R_ConvertParams *params);
////////////////////////////////
//~ rjf: PDB -> (Codeview -> RDI) TPI Hash Table Building
internal CV2R_TPIHash *p2r_cv2r_tpi_hash_from_data(Arena *arena, PDB_Strtbl *strtbl, PDB_TpiParsed *tpi, String8 data);
//////////////////////////////// ////////////////////////////////
//~ rjf: Top-Level Conversion Entry Point (Revised) //~ rjf: Top-Level Conversion Entry Point (Revised)
+2
View File
@@ -66,6 +66,7 @@
#include "llvm/llvm.h" #include "llvm/llvm.h"
#include "rdi_from_coff/rdi_from_coff.h" #include "rdi_from_coff/rdi_from_coff.h"
#include "rdi_from_elf/rdi_from_elf.h" #include "rdi_from_elf/rdi_from_elf.h"
#include "rdi_from_codeview/rdi_from_codeview.h"
#include "rdi_from_pdb/rdi_from_pdb.h" #include "rdi_from_pdb/rdi_from_pdb.h"
#include "rdi_from_dwarf/rdi_from_dwarf.h" #include "rdi_from_dwarf/rdi_from_dwarf.h"
#include "obj/obj.h" #include "obj/obj.h"
@@ -145,6 +146,7 @@
#include "llvm/llvm.c" #include "llvm/llvm.c"
#include "rdi_from_coff/rdi_from_coff.c" #include "rdi_from_coff/rdi_from_coff.c"
#include "rdi_from_elf/rdi_from_elf.c" #include "rdi_from_elf/rdi_from_elf.c"
#include "rdi_from_codeview/rdi_from_codeview.c"
#include "rdi_from_pdb/rdi_from_pdb.c" #include "rdi_from_pdb/rdi_from_pdb.c"
#include "rdi_from_dwarf/rdi_from_dwarf.c" #include "rdi_from_dwarf/rdi_from_dwarf.c"
#include "obj/obj.c" #include "obj/obj.c"