* [PATCH] ipa: Avoid IPA confusing scalar values and single-field aggregates (PR 108679)
@ 2023-02-14 14:50 Martin Jambor
2023-02-14 15:08 ` Richard Biener
0 siblings, 1 reply; 2+ messages in thread
From: Martin Jambor @ 2023-02-14 14:50 UTC (permalink / raw)
To: GCC Patches; +Cc: Jan Hubicka, Jan Hubicka
Hi,
PR 108679 testcase shows a situation when IPA-CP is able to track a
scalar constant in a single-field structure that is part of a bigger
structure. This smaller struture is however also passed in a few calls
to other functions, but the two same-but-different entities, originally
places at the same offset and with the same size, and this confuses the
mechanism that takes care of handling call statements after IPA-SRA.
I think that in stage 4 it is best to revert to GCC 12 behavior in this
particular case (when IPA-CP detects a constant in a single-field
structure or a single element array that is part of a bigger aggregate)
and the patch below does that. If accepted, I plan to file a
missed-optimization bug to track that we could use the IPA-CP propagated
value to re-construct the small aggregate arguments.
Bootstrapped and tested and LTO bootstrapped on x86_64-linux. OK for
master?
Thanks,
Martin
gcc/ChangeLog:
2023-02-13 Martin Jambor <mjambor@suse.cz>
PR ipa/108679
* ipa-sra.cc (push_param_adjustments_for_index): Do not omit
creation of non-scalar replacements even if IPA-CP knows their
contents.
gcc/testsuite/ChangeLog:
2023-02-13 Martin Jambor <mjambor@suse.cz>
PR ipa/108679
* gcc.dg/ipa/pr108679.c: New test.
---
gcc/ipa-sra.cc | 2 +-
gcc/testsuite/gcc.dg/ipa/pr108679.c | 25 +++++++++++++++++++++++++
2 files changed, 26 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/gcc.dg/ipa/pr108679.c
diff --git a/gcc/ipa-sra.cc b/gcc/ipa-sra.cc
index 0495f446bf4..3de7d426b7e 100644
--- a/gcc/ipa-sra.cc
+++ b/gcc/ipa-sra.cc
@@ -3989,7 +3989,7 @@ push_param_adjustments_for_index (isra_func_summary *ifs, unsigned base_index,
{
ipa_argagg_value_list avl (ipcp_ts);
tree value = avl.get_value (base_index, pa->unit_offset);
- if (value)
+ if (value && !AGGREGATE_TYPE_P (pa->type))
{
if (dump_file)
fprintf (dump_file, " - omitting component at byte "
diff --git a/gcc/testsuite/gcc.dg/ipa/pr108679.c b/gcc/testsuite/gcc.dg/ipa/pr108679.c
new file mode 100644
index 00000000000..b1ed50bb831
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/ipa/pr108679.c
@@ -0,0 +1,25 @@
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+
+struct S1 {
+ signed f0;
+};
+struct S2 {
+ struct S1 f2;
+ short f8;
+} g_18;
+void safe_lshift_func_int16_t_s_u();
+void safe_unary_minus_func_uint64_t_u();
+int safe_mul_func_uint8_t_u_u(int, struct S1 p_14);
+int g_732, func_6_l_17;
+static int *func_12();
+static int func_6(struct S2 p_7) { func_12(func_6_l_17, p_7.f2, g_18, 0); }
+static int *func_12(int, struct S1 p_14) {
+ safe_lshift_func_int16_t_s_u();
+ safe_unary_minus_func_uint64_t_u();
+ g_732 = safe_mul_func_uint8_t_u_u(0, p_14);
+}
+int main() {
+ struct S2 l_10 = {3};
+ func_6(l_10);
+}
--
2.39.1
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] ipa: Avoid IPA confusing scalar values and single-field aggregates (PR 108679)
2023-02-14 14:50 [PATCH] ipa: Avoid IPA confusing scalar values and single-field aggregates (PR 108679) Martin Jambor
@ 2023-02-14 15:08 ` Richard Biener
0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2023-02-14 15:08 UTC (permalink / raw)
To: Martin Jambor; +Cc: GCC Patches, Jan Hubicka, Jan Hubicka
On Tue, Feb 14, 2023 at 3:50 PM Martin Jambor <mjambor@suse.cz> wrote:
>
> Hi,
>
> PR 108679 testcase shows a situation when IPA-CP is able to track a
> scalar constant in a single-field structure that is part of a bigger
> structure. This smaller struture is however also passed in a few calls
> to other functions, but the two same-but-different entities, originally
> places at the same offset and with the same size, and this confuses the
> mechanism that takes care of handling call statements after IPA-SRA.
>
> I think that in stage 4 it is best to revert to GCC 12 behavior in this
> particular case (when IPA-CP detects a constant in a single-field
> structure or a single element array that is part of a bigger aggregate)
> and the patch below does that. If accepted, I plan to file a
> missed-optimization bug to track that we could use the IPA-CP propagated
> value to re-construct the small aggregate arguments.
>
> Bootstrapped and tested and LTO bootstrapped on x86_64-linux. OK for
> master?
OK.
Richard.
> Thanks,
>
> Martin
>
>
>
>
> gcc/ChangeLog:
>
> 2023-02-13 Martin Jambor <mjambor@suse.cz>
>
> PR ipa/108679
> * ipa-sra.cc (push_param_adjustments_for_index): Do not omit
> creation of non-scalar replacements even if IPA-CP knows their
> contents.
>
> gcc/testsuite/ChangeLog:
>
> 2023-02-13 Martin Jambor <mjambor@suse.cz>
>
> PR ipa/108679
> * gcc.dg/ipa/pr108679.c: New test.
> ---
> gcc/ipa-sra.cc | 2 +-
> gcc/testsuite/gcc.dg/ipa/pr108679.c | 25 +++++++++++++++++++++++++
> 2 files changed, 26 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/gcc.dg/ipa/pr108679.c
>
> diff --git a/gcc/ipa-sra.cc b/gcc/ipa-sra.cc
> index 0495f446bf4..3de7d426b7e 100644
> --- a/gcc/ipa-sra.cc
> +++ b/gcc/ipa-sra.cc
> @@ -3989,7 +3989,7 @@ push_param_adjustments_for_index (isra_func_summary *ifs, unsigned base_index,
> {
> ipa_argagg_value_list avl (ipcp_ts);
> tree value = avl.get_value (base_index, pa->unit_offset);
> - if (value)
> + if (value && !AGGREGATE_TYPE_P (pa->type))
> {
> if (dump_file)
> fprintf (dump_file, " - omitting component at byte "
> diff --git a/gcc/testsuite/gcc.dg/ipa/pr108679.c b/gcc/testsuite/gcc.dg/ipa/pr108679.c
> new file mode 100644
> index 00000000000..b1ed50bb831
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/ipa/pr108679.c
> @@ -0,0 +1,25 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2" } */
> +
> +struct S1 {
> + signed f0;
> +};
> +struct S2 {
> + struct S1 f2;
> + short f8;
> +} g_18;
> +void safe_lshift_func_int16_t_s_u();
> +void safe_unary_minus_func_uint64_t_u();
> +int safe_mul_func_uint8_t_u_u(int, struct S1 p_14);
> +int g_732, func_6_l_17;
> +static int *func_12();
> +static int func_6(struct S2 p_7) { func_12(func_6_l_17, p_7.f2, g_18, 0); }
> +static int *func_12(int, struct S1 p_14) {
> + safe_lshift_func_int16_t_s_u();
> + safe_unary_minus_func_uint64_t_u();
> + g_732 = safe_mul_func_uint8_t_u_u(0, p_14);
> +}
> +int main() {
> + struct S2 l_10 = {3};
> + func_6(l_10);
> +}
> --
> 2.39.1
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-02-14 15:08 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-14 14:50 [PATCH] ipa: Avoid IPA confusing scalar values and single-field aggregates (PR 108679) Martin Jambor
2023-02-14 15:08 ` Richard Biener
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).