https://gcc.gnu.org/g:91c35e3c317686b9c45cdfc72f0c4e7103c1fccd
commit 91c35e3c317686b9c45cdfc72f0c4e7103c1fccd
Author: Tobias Burnus <tbur...@baylibre.com>
Date: Tue Sep 3 12:02:23 2024 +0200
LTO/WPA: Ensure that output_offload_tables only writes table once [PR116535]
When ltrans was written concurrently, e.g. via -flto=N (N > 1, assuming
sufficient partiations, e.g., via -flto-partition=max),
output_offload_tables
wrote the output tables once per fork.
PR lto/116535
gcc/ChangeLog:
* lto-cgraph.cc (output_offload_tables): Remove offload_ frees.
* lto-streamer-out.cc (lto_output): Make call to it depend on
lto_get_out_decl_state ()->output_offload_tables_p.
* lto-streamer.h (struct lto_out_decl_state): Add
output_offload_tables_p field.
* tree-pass.h (ipa_write_optimization_summaries): Add bool argument.
* passes.cc (ipa_write_summaries_1): Add bool
output_offload_tables_p arg.
(ipa_write_summaries): Update call.
(ipa_write_optimization_summaries): Accept output_offload_tables_p.
gcc/lto/ChangeLog:
* lto.cc (stream_out): Update call to
ipa_write_optimization_summaries to pass true for first partition.
(cherry picked from commit 2fcccf21a34f92ea060b492c9b2aecb56cd5d167)
Diff:
---
gcc/ChangeLog.omp | 17 +++++++++++++++++
gcc/lto-cgraph.cc | 10 ----------
gcc/lto-streamer-out.cc | 3 ++-
gcc/lto-streamer.h | 3 +++
gcc/lto/ChangeLog.omp | 8 ++++++++
gcc/lto/lto.cc | 2 +-
gcc/passes.cc | 11 ++++++++---
gcc/tree-pass.h | 3 ++-
8 files changed, 41 insertions(+), 16 deletions(-)
diff --git a/gcc/ChangeLog.omp b/gcc/ChangeLog.omp
index 481899b4d73e..938926d8f85d 100644
--- a/gcc/ChangeLog.omp
+++ b/gcc/ChangeLog.omp
@@ -1,3 +1,20 @@
+2024-12-18 Thomas Schwinge <tschwi...@baylibre.com>
+
+ Backported from trunk:
+ 2024-09-03 Tobias Burnus <tbur...@baylibre.com>
+
+ PR lto/116535
+ * lto-cgraph.cc (output_offload_tables): Remove offload_ frees.
+ * lto-streamer-out.cc (lto_output): Make call to it depend on
+ lto_get_out_decl_state ()->output_offload_tables_p.
+ * lto-streamer.h (struct lto_out_decl_state): Add
+ output_offload_tables_p field.
+ * tree-pass.h (ipa_write_optimization_summaries): Add bool argument.
+ * passes.cc (ipa_write_summaries_1): Add bool
+ output_offload_tables_p arg.
+ (ipa_write_summaries): Update call.
+ (ipa_write_optimization_summaries): Accept output_offload_tables_p.
+
2024-08-01 Tobias Burnus <tbur...@baylibre.com>
Backported from master:
diff --git a/gcc/lto-cgraph.cc b/gcc/lto-cgraph.cc
index a3dc76f219f8..768e009c100a 100644
--- a/gcc/lto-cgraph.cc
+++ b/gcc/lto-cgraph.cc
@@ -1135,16 +1135,6 @@ output_offload_tables (void)
streamer_write_uhwi_stream (ob->main_stream, 0);
lto_destroy_simple_output_block (ob);
-
- /* In WHOPR mode during the WPA stage the joint offload tables need to be
- streamed to one partition only. That's why we free offload_funcs and
- offload_vars after the first call of output_offload_tables. */
- if (flag_wpa)
- {
- vec_free (offload_funcs);
- vec_free (offload_vars);
- vec_free (offload_ind_funcs);
- }
}
/* Verify the partitioning of NODE. */
diff --git a/gcc/lto-streamer-out.cc b/gcc/lto-streamer-out.cc
index 42456d348236..897f528bac74 100644
--- a/gcc/lto-streamer-out.cc
+++ b/gcc/lto-streamer-out.cc
@@ -2835,7 +2835,8 @@ lto_output (void)
statements using the statement UIDs. */
output_symtab ();
- output_offload_tables ();
+ if (lto_get_out_decl_state ()->output_offload_tables_p)
+ output_offload_tables ();
if (flag_checking)
{
diff --git a/gcc/lto-streamer.h b/gcc/lto-streamer.h
index 895aa27692b7..3962900f0472 100644
--- a/gcc/lto-streamer.h
+++ b/gcc/lto-streamer.h
@@ -535,6 +535,9 @@ struct lto_out_decl_state
/* True if decl state is compressed. */
bool compressed;
+
+ /* True if offload tables should be output. */
+ bool output_offload_tables_p;
};
typedef struct lto_out_decl_state *lto_out_decl_state_ptr;
diff --git a/gcc/lto/ChangeLog.omp b/gcc/lto/ChangeLog.omp
new file mode 100644
index 000000000000..c43b9fc540b1
--- /dev/null
+++ b/gcc/lto/ChangeLog.omp
@@ -0,0 +1,8 @@
+2024-12-18 Thomas Schwinge <tschwi...@baylibre.com>
+
+ Backported from trunk:
+ 2024-09-03 Tobias Burnus <tbur...@baylibre.com>
+
+ PR lto/116535
+ * lto.cc (stream_out): Update call to
+ ipa_write_optimization_summaries to pass true for first partition.
diff --git a/gcc/lto/lto.cc b/gcc/lto/lto.cc
index 91aa2fbddb4e..ce704189d095 100644
--- a/gcc/lto/lto.cc
+++ b/gcc/lto/lto.cc
@@ -176,7 +176,7 @@ stream_out (char *temp_filename, lto_symtab_encoder_t
encoder, int part)
gcc_assert (!dump_file);
streamer_dump_file = dump_begin (TDI_lto_stream_out, NULL, part);
- ipa_write_optimization_summaries (encoder);
+ ipa_write_optimization_summaries (encoder, part == 0);
free (CONST_CAST (char *, file->filename));
diff --git a/gcc/passes.cc b/gcc/passes.cc
index e9d18b9d22e1..74a34dd19851 100644
--- a/gcc/passes.cc
+++ b/gcc/passes.cc
@@ -2829,11 +2829,13 @@ ipa_write_summaries_2 (opt_pass *pass, struct
lto_out_decl_state *state)
summaries. SET is the set of nodes to be written. */
static void
-ipa_write_summaries_1 (lto_symtab_encoder_t encoder)
+ipa_write_summaries_1 (lto_symtab_encoder_t encoder,
+ bool output_offload_tables_p)
{
pass_manager *passes = g->get_passes ();
struct lto_out_decl_state *state = lto_new_out_decl_state ();
state->symtab_node_encoder = encoder;
+ state->output_offload_tables_p = output_offload_tables_p;
lto_output_init_mode_table ();
lto_push_out_decl_state (state);
@@ -2896,7 +2898,8 @@ ipa_write_summaries (void)
if (vnode->need_lto_streaming)
lto_set_symtab_encoder_in_partition (encoder, vnode);
- ipa_write_summaries_1 (compute_ltrans_boundary (encoder));
+ ipa_write_summaries_1 (compute_ltrans_boundary (encoder),
+ flag_generate_offload);
free (order);
if (streamer_dump_file)
@@ -2951,10 +2954,12 @@ ipa_write_optimization_summaries_1 (opt_pass *pass,
NULL, write out all summaries of all nodes. */
void
-ipa_write_optimization_summaries (lto_symtab_encoder_t encoder)
+ipa_write_optimization_summaries (lto_symtab_encoder_t encoder,
+ bool output_offload_tables_p)
{
struct lto_out_decl_state *state = lto_new_out_decl_state ();
state->symtab_node_encoder = encoder;
+ state->output_offload_tables_p = output_offload_tables_p;
lto_output_init_mode_table ();
lto_push_out_decl_state (state);
diff --git a/gcc/tree-pass.h b/gcc/tree-pass.h
index 29267589eeb3..a2c8f685bf3c 100644
--- a/gcc/tree-pass.h
+++ b/gcc/tree-pass.h
@@ -680,7 +680,8 @@ extern void emergency_dump_function (void);
extern void print_current_pass (FILE *);
extern void debug_pass (void);
extern void ipa_write_summaries (void);
-extern void ipa_write_optimization_summaries (struct lto_symtab_encoder_d *);
+extern void ipa_write_optimization_summaries (struct lto_symtab_encoder_d *,
+ bool);
extern void ipa_read_summaries (void);
extern void ipa_read_optimization_summaries (void);
extern void register_one_dump_file (opt_pass *);
