This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
[patch] Improve handling of file names with .gdb_index support
- From: dje at google dot com (Doug Evans)
- To: gdb-patches at sourceware dot org
- Date: Fri, 10 Dec 2010 14:25:04 -0800 (PST)
- Subject: [patch] Improve handling of file names with .gdb_index support
Hi.
This patch is a follow-on to
http://sourceware.org/ml/gdb-patches/2010-12/msg00090.html
and
http://sourceware.org/ml/gdb-patches/2010-12/msg00091.html
It creates a hash table of per-CU/TU file names.
TUs can share the same line_header data as CUs,
and in a program with 5000 CUs and 95000 TUs,
that's a lot of sharing!
Plus the "quick" functions only need the file names, they no longer need
line_header once the file names are obtained, so this patch
frees line_header once done.
Regression tested on amd64-linux with
bash$ runtest \
CC_FOR_TARGET="/bin/sh $srcdir/cc-with-index.sh gcc" \
CXX_FOR_TARGET="/bin/sh $srcdir/cc-with-index.sh g++"
I plan to check this in tomorrow.
2010-12-10 Doug Evans <dje@google.com>
* dwarf2read.c (dwarf2_per_objfile): New member quick_file_names_table.
(dwarf2_per_cu_quick_data): Move definition down to "quick functions"
section. Delete members lines full_names. Rename read_lines to
no_file_data.
(quick_file_names): New struct.
(hash_file_name_entry, eq_file_name_entry): New functions.
(delete_file_name_entry, create_quick_file_names_table): New functions.
(dwarf2_read_index): Initialize quick_file_names_table.
(dw2_get_file_names): Renamed from dw2_require_line_header.
Return pointer to quick_file_names entry. All callers updated.
(dw2_get_real_path): Renamed from dw2_require_full_path.
Replace per_cu arg with pointer to quick_file_names entry.
All callers updated.
(dw2_free_cached_file_names): New function.
(dw2_forget_cached_source_info): Call it (via htab_traverse_noresize).
(dwarf2_initialize_objfile): Initialize quick_file_names_table in
the OBJF_READNOW case.
(dwarf2_free_objfile): Rewrite freeing of quick file names tables.
Index: dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.488
diff -u -p -r1.488 dwarf2read.c
--- dwarf2read.c 10 Dec 2010 22:00:00 -0000 1.488
+++ dwarf2read.c 10 Dec 2010 22:03:36 -0000
@@ -211,6 +211,12 @@ struct dwarf2_per_objfile
/* The mapped index, or NULL if .gdb_index is missing or not being used. */
struct mapped_index *index_table;
+ /* When using index_table, this keeps track of all quick_file_names entries.
+ TUs can share line table entries with CUs or other TUs, and there can be
+ a lot more TUs than unique line tables, so we maintain a separate table
+ of all line table entries to support the sharing. */
+ htab_t quick_file_names_table;
+
/* Set during partial symbol reading, to prevent queueing of full
symbols. */
int reading_partial_symbols;
@@ -391,32 +397,6 @@ struct dwarf2_cu
unsigned int has_namespace_info : 1;
};
-/* When using the index (and thus not using psymtabs), each CU has an
- object of this type. This is used to hold information needed by
- the various "quick" methods. */
-struct dwarf2_per_cu_quick_data
-{
- /* The line table. This can be NULL if there was no line table. */
- struct line_header *lines;
-
- /* The file names from the line table. */
- const char **file_names;
- /* The file names from the line table after being run through
- gdb_realpath. */
- const char **full_names;
-
- /* The corresponding symbol table. This is NULL if symbols for this
- CU have not yet been read. */
- struct symtab *symtab;
-
- /* A temporary mark bit used when iterating over all CUs in
- expand_symtabs_matching. */
- unsigned int mark : 1;
-
- /* True if we've tried to read the line table. */
- unsigned int read_lines : 1;
-};
-
/* Persistent data held for a compilation unit, even when not
processing it. We put a pointer to this structure in the
read_symtab_private field of the psymtab. If we encounter
@@ -1644,6 +1624,102 @@ dwarf2_get_section_info (struct objfile
}
+/* DWARF quick_symbols_functions support. */
+
+/* TUs can share .debug_line entries, and there can be a lot more TUs than
+ unique line tables, so we maintain a separate table of all .debug_line
+ derived entries to support the sharing.
+ All the quick functions need is the list of file names. We discard the
+ line_header when we're done and don't need to record it here. */
+struct quick_file_names
+{
+ /* The offset in .debug_line of the line table. We hash on this. */
+ unsigned int offset;
+
+ /* The number of entries in file_names, real_names. */
+ unsigned int num_file_names;
+
+ /* The file names from the line table, after being run through
+ file_full_name. */
+ const char **file_names;
+
+ /* The file names from the line table after being run through
+ gdb_realpath. These are computed lazily. */
+ const char **real_names;
+};
+
+/* When using the index (and thus not using psymtabs), each CU has an
+ object of this type. This is used to hold information needed by
+ the various "quick" methods. */
+struct dwarf2_per_cu_quick_data
+{
+ /* The file table. This can be NULL if there was no file table
+ or it's currently not read in.
+ NOTE: This points into dwarf2_per_objfile->quick_file_names_table. */
+ struct quick_file_names *file_names;
+
+ /* The corresponding symbol table. This is NULL if symbols for this
+ CU have not yet been read. */
+ struct symtab *symtab;
+
+ /* A temporary mark bit used when iterating over all CUs in
+ expand_symtabs_matching. */
+ unsigned int mark : 1;
+
+ /* True if we've tried to read the file table and found there isn't one.
+ There will be no point in trying to read it again next time. */
+ unsigned int no_file_data : 1;
+};
+
+/* Hash function for a quick_file_names. */
+
+static hashval_t
+hash_file_name_entry (const void *e)
+{
+ const struct quick_file_names *file_data = e;
+
+ return file_data->offset;
+}
+
+/* Equality function for a quick_file_names. */
+
+static int
+eq_file_name_entry (const void *a, const void *b)
+{
+ const struct quick_file_names *ea = a;
+ const struct quick_file_names *eb = b;
+
+ return ea->offset == eb->offset;
+}
+
+/* Delete function for a quick_file_names. */
+
+static void
+delete_file_name_entry (void *e)
+{
+ struct quick_file_names *file_data = e;
+ int i;
+
+ for (i = 0; i < file_data->num_file_names; ++i)
+ {
+ xfree ((void*) file_data->file_names[i]);
+ if (file_data->real_names)
+ xfree ((void*) file_data->real_names[i]);
+ }
+
+ /* The space for the struct itself lives on objfile_obstack,
+ so we don't free it here. */
+}
+
+/* Create a quick_file_names hash table. */
+
+static htab_t
+create_quick_file_names_table (unsigned int nr_initial_entries)
+{
+ return htab_create_alloc (nr_initial_entries,
+ hash_file_name_entry, eq_file_name_entry,
+ delete_file_name_entry, xcalloc, xfree);
+}
/* Read in the symbols for PER_CU. OBJFILE is the objfile from which
this CU came. */
@@ -1993,6 +2069,8 @@ dwarf2_read_index (struct objfile *objfi
dwarf2_per_objfile->index_table = map;
dwarf2_per_objfile->using_index = 1;
+ dwarf2_per_objfile->quick_file_names_table =
+ create_quick_file_names_table (dwarf2_per_objfile->n_comp_units);
return 1;
}
@@ -2010,12 +2088,12 @@ dw2_setup (struct objfile *objfile)
/* A helper for the "quick" functions which attempts to read the line
table for THIS_CU. */
-static void
-dw2_require_line_header (struct objfile *objfile,
- struct dwarf2_per_cu_data *this_cu)
+static struct quick_file_names *
+dw2_get_file_names (struct objfile *objfile,
+ struct dwarf2_per_cu_data *this_cu)
{
bfd *abfd = objfile->obfd;
- struct line_header *lh = NULL;
+ struct line_header *lh;
struct attribute *attr;
struct cleanup *cleanups;
struct die_info *comp_unit_die;
@@ -2026,10 +2104,15 @@ dw2_require_line_header (struct objfile
unsigned int bytes_read, buffer_size;
struct die_reader_specs reader_specs;
char *name, *comp_dir;
+ void **slot;
+ struct quick_file_names *qfn;
+ unsigned int line_offset;
- if (this_cu->v.quick->read_lines)
- return;
- this_cu->v.quick->read_lines = 1;
+ if (this_cu->v.quick->file_names != NULL)
+ return this_cu->v.quick->file_names;
+ /* If we know there is no line data, no point in looking again. */
+ if (this_cu->v.quick->no_file_data)
+ return NULL;
init_one_comp_unit (&cu, objfile);
cleanups = make_cleanup (free_stack_comp_unit, &cu);
@@ -2064,52 +2147,73 @@ dw2_require_line_header (struct objfile
info_ptr = read_full_die (&reader_specs, &comp_unit_die, info_ptr,
&has_children);
+ lh = NULL;
+ slot = NULL;
+ line_offset = 0;
attr = dwarf2_attr (comp_unit_die, DW_AT_stmt_list, &cu);
if (attr)
{
- unsigned int line_offset = DW_UNSND (attr);
+ struct quick_file_names find_entry;
+
+ line_offset = DW_UNSND (attr);
+
+ /* We may have already read in this line header (TU line header sharing).
+ If we have we're done. */
+ find_entry.offset = line_offset;
+ slot = htab_find_slot (dwarf2_per_objfile->quick_file_names_table,
+ &find_entry, INSERT);
+ if (*slot != NULL)
+ {
+ do_cleanups (cleanups);
+ this_cu->v.quick->file_names = *slot;
+ return *slot;
+ }
+
lh = dwarf_decode_line_header (line_offset, abfd, &cu);
}
if (lh == NULL)
{
do_cleanups (cleanups);
- return;
+ this_cu->v.quick->no_file_data = 1;
+ return NULL;
}
- find_file_and_directory (comp_unit_die, &cu, &name, &comp_dir);
+ qfn = obstack_alloc (&objfile->objfile_obstack, sizeof (*qfn));
+ qfn->offset = line_offset;
+ gdb_assert (slot != NULL);
+ *slot = qfn;
- this_cu->v.quick->lines = lh;
+ find_file_and_directory (comp_unit_die, &cu, &name, &comp_dir);
- this_cu->v.quick->file_names
- = obstack_alloc (&objfile->objfile_obstack,
- lh->num_file_names * sizeof (char *));
+ qfn->num_file_names = lh->num_file_names;
+ qfn->file_names = obstack_alloc (&objfile->objfile_obstack,
+ lh->num_file_names * sizeof (char *));
for (i = 0; i < lh->num_file_names; ++i)
- this_cu->v.quick->file_names[i] = file_full_name (i + 1, lh, comp_dir);
+ qfn->file_names[i] = file_full_name (i + 1, lh, comp_dir);
+ qfn->real_names = NULL;
+ free_line_header (lh);
do_cleanups (cleanups);
+
+ this_cu->v.quick->file_names = qfn;
+ return qfn;
}
/* A helper for the "quick" functions which computes and caches the
- real path for a given file name from the line table.
- dw2_require_line_header must have been called before this is
- invoked. */
+ real path for a given file name from the line table. */
static const char *
-dw2_require_full_path (struct objfile *objfile,
- struct dwarf2_per_cu_data *per_cu,
- int index)
-{
- if (!per_cu->v.quick->full_names)
- per_cu->v.quick->full_names
- = OBSTACK_CALLOC (&objfile->objfile_obstack,
- per_cu->v.quick->lines->num_file_names,
- sizeof (char *));
-
- if (!per_cu->v.quick->full_names[index])
- per_cu->v.quick->full_names[index]
- = gdb_realpath (per_cu->v.quick->file_names[index]);
+dw2_get_real_path (struct objfile *objfile,
+ struct quick_file_names *qfn, int index)
+{
+ if (qfn->real_names == NULL)
+ qfn->real_names = OBSTACK_CALLOC (&objfile->objfile_obstack,
+ qfn->num_file_names, sizeof (char *));
+
+ if (qfn->real_names[index] == NULL)
+ qfn->real_names[index] = gdb_realpath (qfn->file_names[index]);
- return per_cu->v.quick->full_names[index];
+ return qfn->real_names[index];
}
static struct symtab *
@@ -2122,28 +2226,34 @@ dw2_find_last_source_symtab (struct objf
return dw2_instantiate_symtab (objfile, dw2_get_cu (index));
}
-static void
-dw2_forget_cached_source_info (struct objfile *objfile)
+/* Traversal function for dw2_forget_cached_source_info. */
+
+static int
+dw2_free_cached_file_names (void **slot, void *info)
{
- int i;
+ struct quick_file_names *file_data = (struct quick_file_names *) *slot;
- dw2_setup (objfile);
- for (i = 0; i < (dwarf2_per_objfile->n_comp_units
- + dwarf2_per_objfile->n_type_comp_units); ++i)
+ if (file_data->real_names)
{
- struct dwarf2_per_cu_data *per_cu = dw2_get_cu (i);
+ int i;
- if (per_cu->v.quick->full_names)
+ for (i = 0; i < file_data->num_file_names; ++i)
{
- int j;
-
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
- {
- xfree ((void *) per_cu->v.quick->full_names[j]);
- per_cu->v.quick->full_names[j] = NULL;
- }
+ xfree ((void*) file_data->real_names[i]);
+ file_data->real_names[i] = NULL;
}
}
+
+ return 1;
+}
+
+static void
+dw2_forget_cached_source_info (struct objfile *objfile)
+{
+ dw2_setup (objfile);
+
+ htab_traverse_noresize (dwarf2_per_objfile->quick_file_names_table,
+ dw2_free_cached_file_names, NULL);
}
static int
@@ -2162,17 +2272,18 @@ dw2_lookup_symtab (struct objfile *objfi
{
int j;
struct dwarf2_per_cu_data *per_cu = dw2_get_cu (i);
+ struct quick_file_names *file_data;
if (per_cu->v.quick->symtab)
continue;
- dw2_require_line_header (objfile, per_cu);
- if (!per_cu->v.quick->lines)
+ file_data = dw2_get_file_names (objfile, per_cu);
+ if (file_data == NULL)
continue;
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
+ for (j = 0; j < file_data->num_file_names; ++j)
{
- const char *this_name = per_cu->v.quick->file_names[j];
+ const char *this_name = file_data->file_names[j];
if (FILENAME_CMP (name, this_name) == 0)
{
@@ -2186,11 +2297,11 @@ dw2_lookup_symtab (struct objfile *objfi
if (full_path != NULL)
{
- const char *this_full_name = dw2_require_full_path (objfile,
- per_cu, j);
+ const char *this_real_name = dw2_get_real_path (objfile,
+ file_data, j);
- if (this_full_name
- && FILENAME_CMP (full_path, this_full_name) == 0)
+ if (this_real_name != NULL
+ && FILENAME_CMP (full_path, this_real_name) == 0)
{
*result = dw2_instantiate_symtab (objfile, per_cu);
return 1;
@@ -2199,11 +2310,11 @@ dw2_lookup_symtab (struct objfile *objfi
if (real_path != NULL)
{
- const char *this_full_name = dw2_require_full_path (objfile,
- per_cu, j);
+ const char *this_real_name = dw2_get_real_path (objfile,
+ file_data, j);
- if (this_full_name != NULL
- && FILENAME_CMP (real_path, this_full_name) == 0)
+ if (this_real_name != NULL
+ && FILENAME_CMP (real_path, this_real_name) == 0)
{
*result = dw2_instantiate_symtab (objfile, per_cu);
return 1;
@@ -2337,17 +2448,18 @@ dw2_expand_symtabs_with_filename (struct
{
int j;
struct dwarf2_per_cu_data *per_cu = dw2_get_cu (i);
+ struct quick_file_names *file_data;
if (per_cu->v.quick->symtab)
continue;
- dw2_require_line_header (objfile, per_cu);
- if (!per_cu->v.quick->lines)
+ file_data = dw2_get_file_names (objfile, per_cu);
+ if (file_data == NULL)
continue;
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
+ for (j = 0; j < file_data->num_file_names; ++j)
{
- const char *this_name = per_cu->v.quick->file_names[j];
+ const char *this_name = file_data->file_names[j];
if (FILENAME_CMP (this_name, filename) == 0)
{
dw2_instantiate_symtab (objfile, per_cu);
@@ -2362,6 +2474,7 @@ dw2_find_symbol_file (struct objfile *ob
{
struct dwarf2_per_cu_data *per_cu;
offset_type *vec;
+ struct quick_file_names *file_data;
dw2_setup (objfile);
@@ -2380,11 +2493,11 @@ dw2_find_symbol_file (struct objfile *ob
/* vec[0] is the length, which must always be >0. */
per_cu = dw2_get_cu (MAYBE_SWAP (vec[1]));
- dw2_require_line_header (objfile, per_cu);
- if (!per_cu->v.quick->lines)
+ file_data = dw2_get_file_names (objfile, per_cu);
+ if (file_data == NULL)
return NULL;
- return per_cu->v.quick->file_names[per_cu->v.quick->lines->num_file_names - 1];
+ return file_data->file_names[file_data->num_file_names - 1];
}
static void
@@ -2423,18 +2536,19 @@ dw2_expand_symtabs_matching (struct objf
{
int j;
struct dwarf2_per_cu_data *per_cu = dw2_get_cu (i);
+ struct quick_file_names *file_data;
per_cu->v.quick->mark = 0;
if (per_cu->v.quick->symtab)
continue;
- dw2_require_line_header (objfile, per_cu);
- if (!per_cu->v.quick->lines)
+ file_data = dw2_get_file_names (objfile, per_cu);
+ if (file_data == NULL)
continue;
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
+ for (j = 0; j < file_data->num_file_names; ++j)
{
- if (file_matcher (per_cu->v.quick->file_names[j], data))
+ if (file_matcher (file_data->file_names[j], data))
{
per_cu->v.quick->mark = 1;
break;
@@ -2541,19 +2655,20 @@ dw2_map_symbol_filenames (struct objfile
{
int j;
struct dwarf2_per_cu_data *per_cu = dw2_get_cu (i);
+ struct quick_file_names *file_data;
if (per_cu->v.quick->symtab)
continue;
- dw2_require_line_header (objfile, per_cu);
- if (!per_cu->v.quick->lines)
+ file_data = dw2_get_file_names (objfile, per_cu);
+ if (file_data == NULL)
continue;
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
+ for (j = 0; j < file_data->num_file_names; ++j)
{
- const char *this_full_name = dw2_require_full_path (objfile, per_cu,
- j);
- (*fun) (per_cu->v.quick->file_names[j], this_full_name, data);
+ const char *this_real_name = dw2_get_real_path (objfile, file_data,
+ j);
+ (*fun) (file_data->file_names[j], this_real_name, data);
}
}
}
@@ -2603,6 +2718,8 @@ dwarf2_initialize_objfile (struct objfil
dwarf2_per_objfile->using_index = 1;
create_all_comp_units (objfile);
create_debug_types_hash_table (objfile);
+ dwarf2_per_objfile->quick_file_names_table =
+ create_quick_file_names_table (dwarf2_per_objfile->n_comp_units);
for (i = 0; i < (dwarf2_per_objfile->n_comp_units
+ dwarf2_per_objfile->n_type_comp_units); ++i)
@@ -14536,30 +14653,8 @@ dwarf2_free_objfile (struct objfile *obj
/* Cached DIE trees use xmalloc and the comp_unit_obstack. */
free_cached_comp_units (NULL);
- if (dwarf2_per_objfile->using_index)
- {
- int i;
-
- for (i = 0; i < dwarf2_per_objfile->n_comp_units; ++i)
- {
- int j;
- struct dwarf2_per_cu_data *per_cu =
- dwarf2_per_objfile->all_comp_units[i];
-
- if (!per_cu->v.quick->lines)
- continue;
-
- for (j = 0; j < per_cu->v.quick->lines->num_file_names; ++j)
- {
- if (per_cu->v.quick->file_names)
- xfree ((void *) per_cu->v.quick->file_names[j]);
- if (per_cu->v.quick->full_names)
- xfree ((void *) per_cu->v.quick->full_names[j]);
- }
-
- free_line_header (per_cu->v.quick->lines);
- }
- }
+ if (dwarf2_per_objfile->quick_file_names_table)
+ htab_delete (dwarf2_per_objfile->quick_file_names_table);
/* Everything else should be on the objfile obstack. */
}