* [PATCH] return edge in make_eh_edges
@ 2023-10-19 7:58 Alexandre Oliva
2023-10-19 9:09 ` Richard Biener
0 siblings, 1 reply; 8+ messages in thread
From: Alexandre Oliva @ 2023-10-19 7:58 UTC (permalink / raw)
To: gcc-patches; +Cc: hubicka
The need to initialize edge probabilities has made make_eh_edges
undesirably hard to use. I suppose we don't want make_eh_edges to
initialize the probability of the newly-added edge itself, so that the
caller takes care of it, but identifying the added edge in need of
adjustments is inefficient and cumbersome. Change make_eh_edges so
that it returns the added edge.
Regstrapped on x86_64-linux-gnu, and (along with various hardening
patches) on ppc64el-linux-gnu. Also tested on multiple other targets,
on older versions of GCC. The returned value is unused in code already
in the compiler. This is a preparatory patch for uses to be introduced
along with stack scrubbing and control flow redundancy. Ok to install?
for gcc/ChangeLog
* tree-eh.cc (make_eh_edges): Return the new edge.
* tree-eh.h (make_eh_edges): Likewise.
---
gcc/tree-eh.cc | 6 +++---
gcc/tree-eh.h | 2 +-
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/gcc/tree-eh.cc b/gcc/tree-eh.cc
index e8ceff36cc6e7..1cb8e08652909 100644
--- a/gcc/tree-eh.cc
+++ b/gcc/tree-eh.cc
@@ -2274,7 +2274,7 @@ make_eh_dispatch_edges (geh_dispatch *stmt)
/* Create the single EH edge from STMT to its nearest landing pad,
if there is such a landing pad within the current function. */
-void
+edge
make_eh_edges (gimple *stmt)
{
basic_block src, dst;
@@ -2283,14 +2283,14 @@ make_eh_edges (gimple *stmt)
lp_nr = lookup_stmt_eh_lp (stmt);
if (lp_nr <= 0)
- return;
+ return NULL;
lp = get_eh_landing_pad_from_number (lp_nr);
gcc_assert (lp != NULL);
src = gimple_bb (stmt);
dst = label_to_block (cfun, lp->post_landing_pad);
- make_edge (src, dst, EDGE_EH);
+ return make_edge (src, dst, EDGE_EH);
}
/* Do the work in redirecting EDGE_IN to NEW_BB within the EH region tree;
diff --git a/gcc/tree-eh.h b/gcc/tree-eh.h
index 771be50fe9a1d..1382568b7c919 100644
--- a/gcc/tree-eh.h
+++ b/gcc/tree-eh.h
@@ -30,7 +30,7 @@ extern bool remove_stmt_from_eh_lp (gimple *);
extern int lookup_stmt_eh_lp_fn (struct function *, const gimple *);
extern int lookup_stmt_eh_lp (const gimple *);
extern bool make_eh_dispatch_edges (geh_dispatch *);
-extern void make_eh_edges (gimple *);
+extern edge make_eh_edges (gimple *);
extern edge redirect_eh_edge (edge, basic_block);
extern void redirect_eh_dispatch_edge (geh_dispatch *, edge, basic_block);
extern bool operation_could_trap_helper_p (enum tree_code, bool, bool, bool,
--
Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] return edge in make_eh_edges
2023-10-19 7:58 [PATCH] return edge in make_eh_edges Alexandre Oliva
@ 2023-10-19 9:09 ` Richard Biener
2023-10-20 3:34 ` Alexandre Oliva
2023-10-20 6:09 ` [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges) Alexandre Oliva
0 siblings, 2 replies; 8+ messages in thread
From: Richard Biener @ 2023-10-19 9:09 UTC (permalink / raw)
To: Alexandre Oliva; +Cc: gcc-patches, hubicka
On Thu, Oct 19, 2023 at 9:59 AM Alexandre Oliva <oliva@adacore.com> wrote:
>
>
> The need to initialize edge probabilities has made make_eh_edges
> undesirably hard to use. I suppose we don't want make_eh_edges to
> initialize the probability of the newly-added edge itself, so that the
> caller takes care of it, but identifying the added edge in need of
> adjustments is inefficient and cumbersome. Change make_eh_edges so
> that it returns the added edge.
>
> Regstrapped on x86_64-linux-gnu, and (along with various hardening
> patches) on ppc64el-linux-gnu. Also tested on multiple other targets,
> on older versions of GCC. The returned value is unused in code already
> in the compiler. This is a preparatory patch for uses to be introduced
> along with stack scrubbing and control flow redundancy. Ok to install?
OK. Maybe time to do s/make_eh_edges/make_eh_edge/ though.
Richard.
>
> for gcc/ChangeLog
>
> * tree-eh.cc (make_eh_edges): Return the new edge.
> * tree-eh.h (make_eh_edges): Likewise.
> ---
> gcc/tree-eh.cc | 6 +++---
> gcc/tree-eh.h | 2 +-
> 2 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/gcc/tree-eh.cc b/gcc/tree-eh.cc
> index e8ceff36cc6e7..1cb8e08652909 100644
> --- a/gcc/tree-eh.cc
> +++ b/gcc/tree-eh.cc
> @@ -2274,7 +2274,7 @@ make_eh_dispatch_edges (geh_dispatch *stmt)
> /* Create the single EH edge from STMT to its nearest landing pad,
> if there is such a landing pad within the current function. */
>
> -void
> +edge
> make_eh_edges (gimple *stmt)
> {
> basic_block src, dst;
> @@ -2283,14 +2283,14 @@ make_eh_edges (gimple *stmt)
>
> lp_nr = lookup_stmt_eh_lp (stmt);
> if (lp_nr <= 0)
> - return;
> + return NULL;
>
> lp = get_eh_landing_pad_from_number (lp_nr);
> gcc_assert (lp != NULL);
>
> src = gimple_bb (stmt);
> dst = label_to_block (cfun, lp->post_landing_pad);
> - make_edge (src, dst, EDGE_EH);
> + return make_edge (src, dst, EDGE_EH);
> }
>
> /* Do the work in redirecting EDGE_IN to NEW_BB within the EH region tree;
> diff --git a/gcc/tree-eh.h b/gcc/tree-eh.h
> index 771be50fe9a1d..1382568b7c919 100644
> --- a/gcc/tree-eh.h
> +++ b/gcc/tree-eh.h
> @@ -30,7 +30,7 @@ extern bool remove_stmt_from_eh_lp (gimple *);
> extern int lookup_stmt_eh_lp_fn (struct function *, const gimple *);
> extern int lookup_stmt_eh_lp (const gimple *);
> extern bool make_eh_dispatch_edges (geh_dispatch *);
> -extern void make_eh_edges (gimple *);
> +extern edge make_eh_edges (gimple *);
> extern edge redirect_eh_edge (edge, basic_block);
> extern void redirect_eh_dispatch_edge (geh_dispatch *, edge, basic_block);
> extern bool operation_could_trap_helper_p (enum tree_code, bool, bool, bool,
>
>
> --
> Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
> Free Software Activist GNU Toolchain Engineer
> More tolerance and less prejudice are key for inclusion and diversity
> Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] return edge in make_eh_edges
2023-10-19 9:09 ` Richard Biener
@ 2023-10-20 3:34 ` Alexandre Oliva
2023-10-20 6:09 ` [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges) Alexandre Oliva
1 sibling, 0 replies; 8+ messages in thread
From: Alexandre Oliva @ 2023-10-20 3:34 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches, hubicka
On Oct 19, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
> OK. Maybe time to do s/make_eh_edges/make_eh_edge/ though.
Thanks, will do, ideally on top of the already-tested refreshed patches
that I'm going to post shortly.
--
Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges)
2023-10-19 9:09 ` Richard Biener
2023-10-20 3:34 ` Alexandre Oliva
@ 2023-10-20 6:09 ` Alexandre Oliva
2023-10-20 6:34 ` Richard Biener
1 sibling, 1 reply; 8+ messages in thread
From: Alexandre Oliva @ 2023-10-20 6:09 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches, hubicka
On Oct 19, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
> Maybe time to do s/make_eh_edges/make_eh_edge/ though.
How about this? Regstrapped on x86_64-linux-gnu and ppc64le-linux-gnu,
on top of the patches for strub, hardbool, and hardcfr that I've just
posted. Ok to install?
----
Since make_eh_edges creates at most one edge, rename it to
make_eh_edge.
for gcc/ChangeLog
* tree-eh.h (make_eh_edges): Rename to...
(make_eh_edge): ... this.
* tree-eh.cc: Likewise. Adjust all callers.
---
gcc/gimple-harden-conditionals.cc | 2 +-
gcc/gimple-harden-control-flow.cc | 2 +-
gcc/ipa-strub.cc | 2 +-
gcc/tree-cfg.cc | 6 +++---
gcc/tree-eh.cc | 4 ++--
gcc/tree-eh.h | 2 +-
gcc/tree-inline.cc | 2 +-
7 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/gcc/gimple-harden-conditionals.cc b/gcc/gimple-harden-conditionals.cc
index a299b984375cd..1999e827a04ca 100644
--- a/gcc/gimple-harden-conditionals.cc
+++ b/gcc/gimple-harden-conditionals.cc
@@ -580,7 +580,7 @@ pass_harden_compares::execute (function *fun)
if (throwing_compare_p)
{
add_stmt_to_eh_lp (asgnck, lookup_stmt_eh_lp (asgn));
- make_eh_edges (asgnck);
+ make_eh_edge (asgnck);
edge ckeh;
basic_block nbb = split_edge (non_eh_succ_edge
diff --git a/gcc/gimple-harden-control-flow.cc b/gcc/gimple-harden-control-flow.cc
index 5c28fd07f3329..7b69281bffc0b 100644
--- a/gcc/gimple-harden-control-flow.cc
+++ b/gcc/gimple-harden-control-flow.cc
@@ -1285,7 +1285,7 @@ pass_harden_control_flow_redundancy::execute (function *fun)
add_stmt_to_eh_lp (stmt, lp_eh_cleanup);
/* Finally, wire the EH cleanup block into the CFG. */
- edge neeh = make_eh_edges (stmt);
+ edge neeh = make_eh_edge (stmt);
neeh->probability = profile_probability::never ();
gcc_checking_assert (neeh->dest == bb_eh_cleanup);
if (neeh->dest->count.initialized_p ())
diff --git a/gcc/ipa-strub.cc b/gcc/ipa-strub.cc
index 612ece9f18552..e944fe1d1101d 100644
--- a/gcc/ipa-strub.cc
+++ b/gcc/ipa-strub.cc
@@ -2075,7 +2075,7 @@ gsi_insert_finally_seq_after_call (gimple_stmt_iterator gsi, gimple_seq seq)
gsi_insert_before (&gsi, resx, GSI_SAME_STMT);
/* Finally, wire the EH cleanup block into the CFG. */
- edge neeh = make_eh_edges (stmt);
+ edge neeh = make_eh_edge (stmt);
neeh->probability = profile_probability::never ();
gcc_checking_assert (neeh->dest == bb_eh_cleanup);
gcc_checking_assert (!neeh->dest->count.initialized_p ());
diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
index f748ca8508940..d41b80d93b6e7 100644
--- a/gcc/tree-cfg.cc
+++ b/gcc/tree-cfg.cc
@@ -878,7 +878,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
fallthru = false;
break;
case GIMPLE_RESX:
- make_eh_edges (last);
+ make_eh_edge (last);
fallthru = false;
break;
case GIMPLE_EH_DISPATCH:
@@ -894,7 +894,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
/* If this statement has reachable exception handlers, then
create abnormal edges to them. */
- make_eh_edges (last);
+ make_eh_edge (last);
/* BUILTIN_RETURN is really a return statement. */
if (gimple_call_builtin_p (last, BUILT_IN_RETURN))
@@ -911,7 +911,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
/* A GIMPLE_ASSIGN may throw internally and thus be considered
control-altering. */
if (is_ctrl_altering_stmt (last))
- make_eh_edges (last);
+ make_eh_edge (last);
fallthru = true;
break;
diff --git a/gcc/tree-eh.cc b/gcc/tree-eh.cc
index 1cb8e08652909..6a30ed9d2b8a6 100644
--- a/gcc/tree-eh.cc
+++ b/gcc/tree-eh.cc
@@ -2275,7 +2275,7 @@ make_eh_dispatch_edges (geh_dispatch *stmt)
if there is such a landing pad within the current function. */
edge
-make_eh_edges (gimple *stmt)
+make_eh_edge (gimple *stmt)
{
basic_block src, dst;
eh_landing_pad lp;
@@ -4921,7 +4921,7 @@ make_pass_cleanup_eh (gcc::context *ctxt)
#endif
/* Verify that BB containing STMT as the last statement, has precisely the
- edge that make_eh_edges would create. */
+ edge that make_eh_edge would create. */
DEBUG_FUNCTION bool
verify_eh_edges (gimple *stmt)
diff --git a/gcc/tree-eh.h b/gcc/tree-eh.h
index 1382568b7c919..045424fe6ece3 100644
--- a/gcc/tree-eh.h
+++ b/gcc/tree-eh.h
@@ -30,7 +30,7 @@ extern bool remove_stmt_from_eh_lp (gimple *);
extern int lookup_stmt_eh_lp_fn (struct function *, const gimple *);
extern int lookup_stmt_eh_lp (const gimple *);
extern bool make_eh_dispatch_edges (geh_dispatch *);
-extern edge make_eh_edges (gimple *);
+extern edge make_eh_edge (gimple *);
extern edge redirect_eh_edge (edge, basic_block);
extern void redirect_eh_dispatch_edge (geh_dispatch *, edge, basic_block);
extern bool operation_could_trap_helper_p (enum tree_code, bool, bool, bool,
diff --git a/gcc/tree-inline.cc b/gcc/tree-inline.cc
index d63060c9429c4..867f3b22b589a 100644
--- a/gcc/tree-inline.cc
+++ b/gcc/tree-inline.cc
@@ -2615,7 +2615,7 @@ copy_edges_for_bb (basic_block bb, profile_count num, profile_count den,
}
else if (can_throw)
{
- make_eh_edges (copy_stmt);
+ make_eh_edge (copy_stmt);
update_probs = true;
}
--
Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges)
2023-10-20 6:09 ` [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges) Alexandre Oliva
@ 2023-10-20 6:34 ` Richard Biener
2023-10-21 7:17 ` [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge) Alexandre Oliva
2023-10-26 6:23 ` [PATCH] rename make_eh_edges to make_eh_edge Alexandre Oliva
0 siblings, 2 replies; 8+ messages in thread
From: Richard Biener @ 2023-10-20 6:34 UTC (permalink / raw)
To: Alexandre Oliva; +Cc: gcc-patches, hubicka
On Fri, Oct 20, 2023 at 8:09 AM Alexandre Oliva <oliva@adacore.com> wrote:
>
> On Oct 19, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
>
> > Maybe time to do s/make_eh_edges/make_eh_edge/ though.
>
> How about this? Regstrapped on x86_64-linux-gnu and ppc64le-linux-gnu,
> on top of the patches for strub, hardbool, and hardcfr that I've just
> posted. Ok to install?
OK.
Richard.
> ----
>
> Since make_eh_edges creates at most one edge, rename it to
> make_eh_edge.
>
>
> for gcc/ChangeLog
>
> * tree-eh.h (make_eh_edges): Rename to...
> (make_eh_edge): ... this.
> * tree-eh.cc: Likewise. Adjust all callers.
> ---
> gcc/gimple-harden-conditionals.cc | 2 +-
> gcc/gimple-harden-control-flow.cc | 2 +-
> gcc/ipa-strub.cc | 2 +-
> gcc/tree-cfg.cc | 6 +++---
> gcc/tree-eh.cc | 4 ++--
> gcc/tree-eh.h | 2 +-
> gcc/tree-inline.cc | 2 +-
> 7 files changed, 10 insertions(+), 10 deletions(-)
>
> diff --git a/gcc/gimple-harden-conditionals.cc b/gcc/gimple-harden-conditionals.cc
> index a299b984375cd..1999e827a04ca 100644
> --- a/gcc/gimple-harden-conditionals.cc
> +++ b/gcc/gimple-harden-conditionals.cc
> @@ -580,7 +580,7 @@ pass_harden_compares::execute (function *fun)
> if (throwing_compare_p)
> {
> add_stmt_to_eh_lp (asgnck, lookup_stmt_eh_lp (asgn));
> - make_eh_edges (asgnck);
> + make_eh_edge (asgnck);
>
> edge ckeh;
> basic_block nbb = split_edge (non_eh_succ_edge
> diff --git a/gcc/gimple-harden-control-flow.cc b/gcc/gimple-harden-control-flow.cc
> index 5c28fd07f3329..7b69281bffc0b 100644
> --- a/gcc/gimple-harden-control-flow.cc
> +++ b/gcc/gimple-harden-control-flow.cc
> @@ -1285,7 +1285,7 @@ pass_harden_control_flow_redundancy::execute (function *fun)
>
> add_stmt_to_eh_lp (stmt, lp_eh_cleanup);
> /* Finally, wire the EH cleanup block into the CFG. */
> - edge neeh = make_eh_edges (stmt);
> + edge neeh = make_eh_edge (stmt);
> neeh->probability = profile_probability::never ();
> gcc_checking_assert (neeh->dest == bb_eh_cleanup);
> if (neeh->dest->count.initialized_p ())
> diff --git a/gcc/ipa-strub.cc b/gcc/ipa-strub.cc
> index 612ece9f18552..e944fe1d1101d 100644
> --- a/gcc/ipa-strub.cc
> +++ b/gcc/ipa-strub.cc
> @@ -2075,7 +2075,7 @@ gsi_insert_finally_seq_after_call (gimple_stmt_iterator gsi, gimple_seq seq)
> gsi_insert_before (&gsi, resx, GSI_SAME_STMT);
>
> /* Finally, wire the EH cleanup block into the CFG. */
> - edge neeh = make_eh_edges (stmt);
> + edge neeh = make_eh_edge (stmt);
> neeh->probability = profile_probability::never ();
> gcc_checking_assert (neeh->dest == bb_eh_cleanup);
> gcc_checking_assert (!neeh->dest->count.initialized_p ());
> diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
> index f748ca8508940..d41b80d93b6e7 100644
> --- a/gcc/tree-cfg.cc
> +++ b/gcc/tree-cfg.cc
> @@ -878,7 +878,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
> fallthru = false;
> break;
> case GIMPLE_RESX:
> - make_eh_edges (last);
> + make_eh_edge (last);
> fallthru = false;
> break;
> case GIMPLE_EH_DISPATCH:
> @@ -894,7 +894,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
>
> /* If this statement has reachable exception handlers, then
> create abnormal edges to them. */
> - make_eh_edges (last);
> + make_eh_edge (last);
>
> /* BUILTIN_RETURN is really a return statement. */
> if (gimple_call_builtin_p (last, BUILT_IN_RETURN))
> @@ -911,7 +911,7 @@ make_edges_bb (basic_block bb, struct omp_region **pcur_region, int *pomp_index)
> /* A GIMPLE_ASSIGN may throw internally and thus be considered
> control-altering. */
> if (is_ctrl_altering_stmt (last))
> - make_eh_edges (last);
> + make_eh_edge (last);
> fallthru = true;
> break;
>
> diff --git a/gcc/tree-eh.cc b/gcc/tree-eh.cc
> index 1cb8e08652909..6a30ed9d2b8a6 100644
> --- a/gcc/tree-eh.cc
> +++ b/gcc/tree-eh.cc
> @@ -2275,7 +2275,7 @@ make_eh_dispatch_edges (geh_dispatch *stmt)
> if there is such a landing pad within the current function. */
>
> edge
> -make_eh_edges (gimple *stmt)
> +make_eh_edge (gimple *stmt)
> {
> basic_block src, dst;
> eh_landing_pad lp;
> @@ -4921,7 +4921,7 @@ make_pass_cleanup_eh (gcc::context *ctxt)
> #endif
>
> /* Verify that BB containing STMT as the last statement, has precisely the
> - edge that make_eh_edges would create. */
> + edge that make_eh_edge would create. */
>
> DEBUG_FUNCTION bool
> verify_eh_edges (gimple *stmt)
> diff --git a/gcc/tree-eh.h b/gcc/tree-eh.h
> index 1382568b7c919..045424fe6ece3 100644
> --- a/gcc/tree-eh.h
> +++ b/gcc/tree-eh.h
> @@ -30,7 +30,7 @@ extern bool remove_stmt_from_eh_lp (gimple *);
> extern int lookup_stmt_eh_lp_fn (struct function *, const gimple *);
> extern int lookup_stmt_eh_lp (const gimple *);
> extern bool make_eh_dispatch_edges (geh_dispatch *);
> -extern edge make_eh_edges (gimple *);
> +extern edge make_eh_edge (gimple *);
> extern edge redirect_eh_edge (edge, basic_block);
> extern void redirect_eh_dispatch_edge (geh_dispatch *, edge, basic_block);
> extern bool operation_could_trap_helper_p (enum tree_code, bool, bool, bool,
> diff --git a/gcc/tree-inline.cc b/gcc/tree-inline.cc
> index d63060c9429c4..867f3b22b589a 100644
> --- a/gcc/tree-inline.cc
> +++ b/gcc/tree-inline.cc
> @@ -2615,7 +2615,7 @@ copy_edges_for_bb (basic_block bb, profile_count num, profile_count den,
> }
> else if (can_throw)
> {
> - make_eh_edges (copy_stmt);
> + make_eh_edge (copy_stmt);
> update_probs = true;
> }
>
>
>
> --
> Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
> Free Software Activist GNU Toolchain Engineer
> More tolerance and less prejudice are key for inclusion and diversity
> Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge)
2023-10-20 6:34 ` Richard Biener
@ 2023-10-21 7:17 ` Alexandre Oliva
2023-10-23 7:59 ` Richard Biener
2023-10-26 6:23 ` [PATCH] rename make_eh_edges to make_eh_edge Alexandre Oliva
1 sibling, 1 reply; 8+ messages in thread
From: Alexandre Oliva @ 2023-10-21 7:17 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches, hubicka
On Oct 20, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
>> * tree-eh.h (make_eh_edges): Rename to...
>> (make_eh_edge): ... this.
>> * tree-eh.cc: Likewise. Adjust all callers.
Once the above goes in (it depends on the strub monster patch), the
following one should apply as well. Regstrapped on x86_64-linux-gnu.
Ok to install?
Set execution count of EH blocks, and probability of EH edges.
for gcc/ChangeLog
PR tree-optimization/111520
* gimple-harden-conditionals.cc
(pass_harden_compares::execute): Set EH edge probability and
EH block execution count.
for gcc/testsuite/ChangeLog
PR tree-optimization/111520
* g++.dg/torture/harden-comp-pr111520.cc: New.
---
gcc/gimple-harden-conditionals.cc | 12 +++++++++++-
.../g++.dg/torture/harden-comp-pr111520.cc | 17 +++++++++++++++++
2 files changed, 28 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
diff --git a/gcc/gimple-harden-conditionals.cc b/gcc/gimple-harden-conditionals.cc
index 1999e827a04ca..bded288985063 100644
--- a/gcc/gimple-harden-conditionals.cc
+++ b/gcc/gimple-harden-conditionals.cc
@@ -580,11 +580,21 @@ pass_harden_compares::execute (function *fun)
if (throwing_compare_p)
{
add_stmt_to_eh_lp (asgnck, lookup_stmt_eh_lp (asgn));
- make_eh_edge (asgnck);
+ edge eh = make_eh_edge (asgnck);
+ /* This compare looks like it could raise an exception,
+ but it's dominated by the original compare, that
+ would raise an exception first, so the EH edge from
+ this one is never really taken. */
+ eh->probability = profile_probability::never ();
+ if (eh->dest->count.initialized_p ())
+ eh->dest->count += eh->count ();
+ else
+ eh->dest->count = eh->count ();
edge ckeh;
basic_block nbb = split_edge (non_eh_succ_edge
(gimple_bb (asgnck), &ckeh));
+ gcc_checking_assert (eh == ckeh);
gsi_split = gsi_start_bb (nbb);
if (dump_file)
diff --git a/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc b/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
new file mode 100644
index 0000000000000..b4381b4d84ec4
--- /dev/null
+++ b/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-fharden-compares -fsignaling-nans -fnon-call-exceptions" } */
+
+struct S
+{
+ S (bool);
+ ~S ();
+};
+
+float f;
+
+void
+foo ()
+{
+ S a = 0;
+ S b = f;
+}
--
Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge)
2023-10-21 7:17 ` [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge) Alexandre Oliva
@ 2023-10-23 7:59 ` Richard Biener
0 siblings, 0 replies; 8+ messages in thread
From: Richard Biener @ 2023-10-23 7:59 UTC (permalink / raw)
To: Alexandre Oliva; +Cc: gcc-patches, hubicka
On Sat, Oct 21, 2023 at 9:17 AM Alexandre Oliva <oliva@adacore.com> wrote:
>
> On Oct 20, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
>
> >> * tree-eh.h (make_eh_edges): Rename to...
> >> (make_eh_edge): ... this.
> >> * tree-eh.cc: Likewise. Adjust all callers.
>
> Once the above goes in (it depends on the strub monster patch), the
> following one should apply as well. Regstrapped on x86_64-linux-gnu.
> Ok to install?
OK.
> Set execution count of EH blocks, and probability of EH edges.
>
>
> for gcc/ChangeLog
>
> PR tree-optimization/111520
> * gimple-harden-conditionals.cc
> (pass_harden_compares::execute): Set EH edge probability and
> EH block execution count.
>
> for gcc/testsuite/ChangeLog
>
> PR tree-optimization/111520
> * g++.dg/torture/harden-comp-pr111520.cc: New.
> ---
> gcc/gimple-harden-conditionals.cc | 12 +++++++++++-
> .../g++.dg/torture/harden-comp-pr111520.cc | 17 +++++++++++++++++
> 2 files changed, 28 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
>
> diff --git a/gcc/gimple-harden-conditionals.cc b/gcc/gimple-harden-conditionals.cc
> index 1999e827a04ca..bded288985063 100644
> --- a/gcc/gimple-harden-conditionals.cc
> +++ b/gcc/gimple-harden-conditionals.cc
> @@ -580,11 +580,21 @@ pass_harden_compares::execute (function *fun)
> if (throwing_compare_p)
> {
> add_stmt_to_eh_lp (asgnck, lookup_stmt_eh_lp (asgn));
> - make_eh_edge (asgnck);
> + edge eh = make_eh_edge (asgnck);
> + /* This compare looks like it could raise an exception,
> + but it's dominated by the original compare, that
> + would raise an exception first, so the EH edge from
> + this one is never really taken. */
> + eh->probability = profile_probability::never ();
> + if (eh->dest->count.initialized_p ())
> + eh->dest->count += eh->count ();
> + else
> + eh->dest->count = eh->count ();
>
> edge ckeh;
> basic_block nbb = split_edge (non_eh_succ_edge
> (gimple_bb (asgnck), &ckeh));
> + gcc_checking_assert (eh == ckeh);
> gsi_split = gsi_start_bb (nbb);
>
> if (dump_file)
> diff --git a/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc b/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
> new file mode 100644
> index 0000000000000..b4381b4d84ec4
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/torture/harden-comp-pr111520.cc
> @@ -0,0 +1,17 @@
> +/* { dg-do compile } */
> +/* { dg-options "-fharden-compares -fsignaling-nans -fnon-call-exceptions" } */
> +
> +struct S
> +{
> + S (bool);
> + ~S ();
> +};
> +
> +float f;
> +
> +void
> +foo ()
> +{
> + S a = 0;
> + S b = f;
> +}
>
>
> --
> Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
> Free Software Activist GNU Toolchain Engineer
> More tolerance and less prejudice are key for inclusion and diversity
> Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] rename make_eh_edges to make_eh_edge
2023-10-20 6:34 ` Richard Biener
2023-10-21 7:17 ` [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge) Alexandre Oliva
@ 2023-10-26 6:23 ` Alexandre Oliva
1 sibling, 0 replies; 8+ messages in thread
From: Alexandre Oliva @ 2023-10-26 6:23 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches, hubicka
On Oct 20, 2023, Richard Biener <richard.guenther@gmail.com> wrote:
> OK.
Thanks.
>> * tree-eh.h (make_eh_edges): Rename to...
>> (make_eh_edge): ... this.
>> * tree-eh.cc: Likewise. Adjust all callers.
Ugh, "Adjust all callers" is no longer enough, all files need to be
mentioned explicitly. So I'm pushing it with:
* tree-eh.h (make_eh_edges): Rename to...
(make_eh_edge): ... this.
* tree-eh.cc: Likewise. Adjust all callers...
* gimple-harden-conditionals.cc: ... here, ...
* gimple-harden-control-flow.cc: ... here, ...
* tree-cfg.cc: ... here, ...
* tree-inline.cc: ... and here.
The ipa-strub.cc is now part of the strub patch.
--
Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2023-10-26 6:23 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-10-19 7:58 [PATCH] return edge in make_eh_edges Alexandre Oliva
2023-10-19 9:09 ` Richard Biener
2023-10-20 3:34 ` Alexandre Oliva
2023-10-20 6:09 ` [PATCH] rename make_eh_edges to make_eh_edge (was: return edge in make_eh_edges) Alexandre Oliva
2023-10-20 6:34 ` Richard Biener
2023-10-21 7:17 ` [PATCH] [PR111520] set hardcmp eh probs (was: rename make_eh_edges to make_eh_edge) Alexandre Oliva
2023-10-23 7:59 ` Richard Biener
2023-10-26 6:23 ` [PATCH] rename make_eh_edges to make_eh_edge Alexandre Oliva
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).