Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On 01/27/2016 05:14 PM, David Edelsohn wrote: On Wed, Jan 27, 2016 at 6:36 PM, Jeff Law wrote: On 01/27/2016 12:39 PM, David Edelsohn wrote: The new sra-17.c and sra-18.c tests fail on AIX because the regex is too restrictive -- AIX labels don't have exactly the same format. On AIX, the labels in the dumps look like "LC..0" instead of ".LC0". This patch adds "*" and ".*" so that the "." prepended to LC is optional and to allow characters between the "LC" and the "0". I needed extra escapes for the sra-17.c line that matches multiple times - for no apparent reason. The joys of expect/tcl. I just keep escaping until the regex that I developed outside the suite works. I have been trying to get away from using .* though. The longest match nature sometimes gives surprising results. In theory .*? ought to work better, but I haven't tried using it much. Anyway, the change looks fine to me. Segher pointed out to me that my revised regex was matching multiple lines, so it was not triggering multiple times without the restriction on the pattern. A revised, tighter patch uses "?" Yup, that's precisely why I've moved away from .* :-) jeff
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On Wed, Jan 27, 2016 at 6:36 PM, Jeff Law wrote:
> On 01/27/2016 12:39 PM, David Edelsohn wrote:
>>
>> The new sra-17.c and sra-18.c tests fail on AIX because the regex is
>> too restrictive -- AIX labels don't have exactly the same format. On
>> AIX, the labels in the dumps look like "LC..0" instead of ".LC0".
>>
>> This patch adds "*" and ".*" so that the "." prepended to LC is
>> optional and to allow characters between the "LC" and the "0".
>>
>> I needed extra escapes for the sra-17.c line that matches multiple
>> times - for no apparent reason.
>
> The joys of expect/tcl. I just keep escaping until the regex that I
> developed outside the suite works. I have been trying to get away from
> using .* though. The longest match nature sometimes gives surprising
> results. In theory .*? ought to work better, but I haven't tried using it
> much.
>
> Anyway, the change looks fine to me.
Segher pointed out to me that my revised regex was matching multiple
lines, so it was not triggering multiple times without the restriction
on the pattern.
A revised, tighter patch uses "?"
Index: sra-17.c
===
--- sra-17.c(revision 232904)
+++ sra-17.c(working copy)
@@ -15,5 +15,5 @@
abort ();
}
-/* { dg-final { scan-tree-dump-times "Removing load: a = \\\*.LC0;" 1
"esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ = \\\*.LC0\\\[" 4
"esra" } } */
+/* { dg-final { scan-tree-dump-times "Removing load: a =
\\\*\\.?LC\\.?\\.?0;" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR\\.\[0-9_\]+ =
\\\*\\.?LC\\.?\\.?0\\\[" 4 "esra" } } */
Index: sra-18.c
===
--- sra-18.c(revision 232904)
+++ sra-18.c(working copy)
@@ -21,8 +21,8 @@
abort ();
}
-/* { dg-final { scan-tree-dump-times "Removing load: a = \\\*.LC0;" 1
"esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[0\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[0\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[1\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[1\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "Removing load: a =
\\\*\\.?LC\\.?\\.?0;" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR\\.\[0-9_\]+ =
\\\*\\.?LC\\.?\\.?0\\.b\\\[0\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR\\.\[0-9_\]+ =
\\\*\\.?LC\\.?\\.?0\\.b\\\[0\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR\\.\[0-9_\]+ =
\\\*\\.?LC\\.?\\.?0\\.b\\\[1\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR\\.\[0-9_\]+ =
\\\*\\.?LC\\.?\\.?0\\.b\\\[1\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On 01/27/2016 12:39 PM, David Edelsohn wrote: The new sra-17.c and sra-18.c tests fail on AIX because the regex is too restrictive -- AIX labels don't have exactly the same format. On AIX, the labels in the dumps look like "LC..0" instead of ".LC0". This patch adds "*" and ".*" so that the "." prepended to LC is optional and to allow characters between the "LC" and the "0". I needed extra escapes for the sra-17.c line that matches multiple times - for no apparent reason. The joys of expect/tcl. I just keep escaping until the regex that I developed outside the suite works. I have been trying to get away from using .* though. The longest match nature sometimes gives surprising results. In theory .*? ought to work better, but I haven't tried using it much. Anyway, the change looks fine to me. jeff
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
The new sra-17.c and sra-18.c tests fail on AIX because the regex is
too restrictive -- AIX labels don't have exactly the same format. On
AIX, the labels in the dumps look like "LC..0" instead of ".LC0".
This patch adds "*" and ".*" so that the "." prepended to LC is
optional and to allow characters between the "LC" and the "0".
I needed extra escapes for the sra-17.c line that matches multiple
times - for no apparent reason.
Okay?
Thanks, David
Index: sra-17.c
===
--- sra-17.c(revision 232872)
+++ sra-17.c(working copy)
@@ -15,5 +15,5 @@
abort ();
}
-/* { dg-final { scan-tree-dump-times "Removing load: a = \\\*.LC0;" 1
"esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ = \\\*.LC0\\\[" 4
"esra" } } */
+/* { dg-final { scan-tree-dump-times "Removing load: a =
\\\*.*LC.*0;" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*\\.*LC\\.*0\\\[" 4 "esra" } } */
Index: sra-18.c
===
--- sra-18.c(revision 232872)
+++ sra-18.c(working copy)
@@ -21,8 +21,8 @@
abort ();
}
-/* { dg-final { scan-tree-dump-times "Removing load: a = \\\*.LC0;" 1
"esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[0\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[0\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[1\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
-/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.LC0\\.b\\\[1\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "Removing load: a =
\\\*.*LC.*0;" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.*LC.*0\\.b\\\[0\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.*LC.*0\\.b\\\[0\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.*LC.*0\\.b\\\[1\\\]\\.f\\\[0\\\]\\.x" 1 "esra" } } */
+/* { dg-final { scan-tree-dump-times "SR.\[0-9_\]+ =
\\\*.*LC.*0\\.b\\\[1\\\]\\.f\\\[1\\\]\\.x" 1 "esra" } } */
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On Fri, Jan 15, 2016 at 11:27 AM, Alan Lawrence
wrote:
> On 24/12/15 11:53, Alan Lawrence wrote:
>>
>> Here's a new version that fixes the gcc.dg/guality/pr54970.c failures seen
>> on
>> aarch64 and powerpc64. Prior to SRA handling constant pool decls,
>> -fdump-tree-esra-details (at -O1 -g) had shown:
>>:
>>a = *.LC0;
>># DEBUG a$0 => MEM[(int[3] *)&*.LC0]
>>a$4_3 = MEM[(int[3] *)&*.LC0 + 4B];
>># DEBUG a$4 => a$4_3
>>a$8_4 = MEM[(int[3] *)&*.LC0 + 8B];
>>
>> The previous patch changed this to:
>>:
>>SR.5_3 = *.LC0[0];
>>SR.6_4 = *.LC0[1];
>>SR.7_19 = *.LC0[2];
>>SR.8_21 = *.LC1[0];
>>SR.9_22 = *.LC1[1];
>>SR.10_23 = *.LC1[2];
>># DEBUG a$0 => NULL // Note here
>>a$4_24 = SR.6_4;
>># DEBUG a$4 => a$4_24
>>a$8_25 = SR.7_19;
>>
>> Turns out the DEBUG a$0 => NULL was produced in
>> load_assign_lhs_subreplacements:
>>
>> if (lacc && lacc->grp_to_be_debug_replaced)
>> {
>> gdebug *ds;
>> tree drhs;
>> struct access *racc = find_access_in_subtree (sad->top_racc,
>> offset,
>> lacc->size);
>>
>> if (racc && racc->grp_to_be_replaced)
>> {
>> if (racc->grp_write)
>> drhs = get_access_replacement (racc);
>> else
>> drhs = NULL; // <=== HERE
>> }
>> ...
>> ds = gimple_build_debug_bind (get_access_replacement (lacc),
>> drhs, gsi_stmt
>> (sad->old_gsi));
>>
>> Prior to the patch, we'd skipped around load_assign_lhs_subreplacements,
>> because
>> access_has_children_p(racc) (for racc = *.LC0) didn't hold in
>> sra_modify_assign.
>>
>> I also added a constant_decl_p function, combining the two checks, plus
>> some
>> testcase fixes.
>>
>> Bootstrapped + check-gcc,g++ on x86_64, ARM, AArch64,
>>also on powerpc64{,le}-none-linux-gnu *in combination with the other
>> patches
>>in the series* (I haven't tested the individual patches on PPC),
>>plus Ada on ARM and x86_64.
>>
>> gcc/ChangeLog:
>>
>> PR target/63679
>> * tree-sra.c (disqualified_constants, constant_decl_p): New.
>> (sra_initialize): Allocate disqualified_constants.
>> (sra_deinitialize): Free disqualified_constants.
>> (disqualify_candidate): Update disqualified_constants when
>> appropriate.
>> (create_access): Scan for constant-pool entries as we go along.
>> (scalarizable_type_p): Add check against
>> type_contains_placeholder_p.
>> (maybe_add_sra_candidate): Allow constant-pool entries.
>> (load_assign_lhs_subreplacements): Bind debug for constant pool
>> vars.
>> (initialize_constant_pool_replacements): New.
>> (sra_modify_assign): Avoid mangling assignments created by
>> previous,
>> and don't generate writes into constant pool.
>> (sra_modify_function_body): Call
>> initialize_constant_pool_replacements.
>>
>> gcc/testsuite/ChangeLog:
>>
>> * gcc.dg/tree-ssa/sra-17.c: New.
>> * gcc.dg/tree-ssa/sra-18.c: New.
>
>
> Ping.
Ok.
Thanks,
Richard.
> (The next bit is false, unless you force SRA to happen more widely, but all
> the above stands)
>
>> This also fixes a bunch of other guality tests on AArch64 that were
>> failing
>> prior to the patch series, and another bunch on PowerPC64 (bigendian
>> -m32), listed below.
>
>>
>>
>> Tests fixed on aarch64-none-linux-gnu:
>>
>> gcc.dg/guality/pr54970.c -O1 line 15 *p == 3
>> gcc.dg/guality/pr54970.c -O1 line 15 *q == 2
>> gcc.dg/guality/pr54970.c -O1 line 20 *p == 13
>> gcc.dg/guality/pr54970.c -O1 line 20 *q == 2
>> gcc.dg/guality/pr54970.c -O1 line 25 *p == 13
>> gcc.dg/guality/pr54970.c -O1 line 25 *q == 12
>> gcc.dg/guality/pr54970.c -O1 line 31 *p == 6
>> gcc.dg/guality/pr54970.c -O1 line 31 *q == 5
>> gcc.dg/guality/pr54970.c -O1 line 36 *p == 26
>> gcc.dg/guality/pr54970.c -O1 line 36 *q == 5
>> gcc.dg/guality/pr54970.c -O1 line 45 *p == 26
>> gcc.dg/guality/pr54970.c -O1 line 45 *q == 25
>> gcc.dg/guality/pr54970.c -O1 line 45 p[-1] == 25
>> gcc.dg/guality/pr54970.c -O1 line 45 p[-2] == 4
>> gcc.dg/guality/pr54970.c -O1 line 45 q[-1] == 4
>> gcc.dg/guality/pr54970.c -O1 line 45 q[1] == 26
>> gcc.dg/guality/pr56154-1.c -O1 line pr56154-1.c:20 x.a == 6
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.f == 5.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.g == 6.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.f == 0.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.g == 6.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.f == 5.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.g == 6.0
>> gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s2.f == 5
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On 24/12/15 11:53, Alan Lawrence wrote:
Here's a new version that fixes the gcc.dg/guality/pr54970.c failures seen on
aarch64 and powerpc64. Prior to SRA handling constant pool decls,
-fdump-tree-esra-details (at -O1 -g) had shown:
:
a = *.LC0;
# DEBUG a$0 => MEM[(int[3] *)&*.LC0]
a$4_3 = MEM[(int[3] *)&*.LC0 + 4B];
# DEBUG a$4 => a$4_3
a$8_4 = MEM[(int[3] *)&*.LC0 + 8B];
The previous patch changed this to:
:
SR.5_3 = *.LC0[0];
SR.6_4 = *.LC0[1];
SR.7_19 = *.LC0[2];
SR.8_21 = *.LC1[0];
SR.9_22 = *.LC1[1];
SR.10_23 = *.LC1[2];
# DEBUG a$0 => NULL // Note here
a$4_24 = SR.6_4;
# DEBUG a$4 => a$4_24
a$8_25 = SR.7_19;
Turns out the DEBUG a$0 => NULL was produced in load_assign_lhs_subreplacements:
if (lacc && lacc->grp_to_be_debug_replaced)
{
gdebug *ds;
tree drhs;
struct access *racc = find_access_in_subtree (sad->top_racc,
offset,
lacc->size);
if (racc && racc->grp_to_be_replaced)
{
if (racc->grp_write)
drhs = get_access_replacement (racc);
else
drhs = NULL; // <=== HERE
}
...
ds = gimple_build_debug_bind (get_access_replacement (lacc),
drhs, gsi_stmt (sad->old_gsi));
Prior to the patch, we'd skipped around load_assign_lhs_subreplacements, because
access_has_children_p(racc) (for racc = *.LC0) didn't hold in sra_modify_assign.
I also added a constant_decl_p function, combining the two checks, plus some
testcase fixes.
Bootstrapped + check-gcc,g++ on x86_64, ARM, AArch64,
also on powerpc64{,le}-none-linux-gnu *in combination with the other patches
in the series* (I haven't tested the individual patches on PPC),
plus Ada on ARM and x86_64.
gcc/ChangeLog:
PR target/63679
* tree-sra.c (disqualified_constants, constant_decl_p): New.
(sra_initialize): Allocate disqualified_constants.
(sra_deinitialize): Free disqualified_constants.
(disqualify_candidate): Update disqualified_constants when appropriate.
(create_access): Scan for constant-pool entries as we go along.
(scalarizable_type_p): Add check against type_contains_placeholder_p.
(maybe_add_sra_candidate): Allow constant-pool entries.
(load_assign_lhs_subreplacements): Bind debug for constant pool vars.
(initialize_constant_pool_replacements): New.
(sra_modify_assign): Avoid mangling assignments created by previous,
and don't generate writes into constant pool.
(sra_modify_function_body): Call initialize_constant_pool_replacements.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/sra-17.c: New.
* gcc.dg/tree-ssa/sra-18.c: New.
Ping.
(The next bit is false, unless you force SRA to happen more widely, but all the
above stands)
This also fixes a bunch of other guality tests on AArch64 that were failing
prior to the patch series, and another bunch on PowerPC64 (bigendian -m32),
listed below.
>
Tests fixed on aarch64-none-linux-gnu:
gcc.dg/guality/pr54970.c -O1 line 15 *p == 3
gcc.dg/guality/pr54970.c -O1 line 15 *q == 2
gcc.dg/guality/pr54970.c -O1 line 20 *p == 13
gcc.dg/guality/pr54970.c -O1 line 20 *q == 2
gcc.dg/guality/pr54970.c -O1 line 25 *p == 13
gcc.dg/guality/pr54970.c -O1 line 25 *q == 12
gcc.dg/guality/pr54970.c -O1 line 31 *p == 6
gcc.dg/guality/pr54970.c -O1 line 31 *q == 5
gcc.dg/guality/pr54970.c -O1 line 36 *p == 26
gcc.dg/guality/pr54970.c -O1 line 36 *q == 5
gcc.dg/guality/pr54970.c -O1 line 45 *p == 26
gcc.dg/guality/pr54970.c -O1 line 45 *q == 25
gcc.dg/guality/pr54970.c -O1 line 45 p[-1] == 25
gcc.dg/guality/pr54970.c -O1 line 45 p[-2] == 4
gcc.dg/guality/pr54970.c -O1 line 45 q[-1] == 4
gcc.dg/guality/pr54970.c -O1 line 45 q[1] == 26
gcc.dg/guality/pr56154-1.c -O1 line pr56154-1.c:20 x.a == 6
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.f == 0.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s2.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s2.g == 6.0
gcc.dg/guality/sra-1.c -O1 line 21 a.j == 14
gcc.dg/guality/sra-1.c -O1 line 32 a[1] == 14
gcc.dg/guality/sra-1.c -O1 line 43 a.i == 4
gcc.dg/guality/sra-1.c -O1 line 43 a.j == 14
(and other optimization levels)
On ppc64 bigendian, with the rest of the series (but I expect it's this patch):
unix/-m32: gcc.dg/guality/pr54970.c -O1 line 15 a[0] == 1
u
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
On 24/12/15 11:53, Alan Lawrence wrote:
Here's a new version that fixes the gcc.dg/guality/pr54970.c failures seen on
aarch64 and powerpc64.
[snip]
This also fixes a bunch of other guality tests on AArch64 that were failing
prior to the patch series, and another bunch on PowerPC64 (bigendian -m32),
listed below.
Ach, sorry, not quite. That version avoids any regressions (e.g. in pr54970.c),
but does not fix all those other tests, unless you also have this hunk
(https://gcc.gnu.org/ml/gcc-patches/2015-08/msg01483.html):
diff --git a/gcc/tree-sra.c b/gcc/tree-sra.c
index a3ff2df..2a741b8 100644
--- a/gcc/tree-sra.c
+++ b/gcc/tree-sra.c
@@ -2651,7 +2651,8 @@ analyze_all_variable_accesses (void)
&& scalarizable_type_p (TREE_TYPE (var)))
{
if (tree_to_uhwi (TYPE_SIZE (TREE_TYPE (var)))
- <= max_scalarization_size)
+ <= max_scalarization_size
+ || DECL_IN_CONSTANT_POOL (var))
{
create_total_scalarization_access (var);
completely_scalarize (var, TREE_TYPE (var), 0, var);
...which I was using to increase test coverage of the SRA changes.
(Alternatively, you can "fix" the tests by running the testsuite with a forced
--param sra-max-scalarization-size. But this is only saying that the dwarf info
now generated by scalarizing constant-pools, is better than whatever dwarf was
being generated by whatever other part of the compiler before.)
--Alan
Re: [PATCH 1/4] Make SRA scalarize constant-pool loads
Here's a new version that fixes the gcc.dg/guality/pr54970.c failures seen on
aarch64 and powerpc64. Prior to SRA handling constant pool decls,
-fdump-tree-esra-details (at -O1 -g) had shown:
:
a = *.LC0;
# DEBUG a$0 => MEM[(int[3] *)&*.LC0]
a$4_3 = MEM[(int[3] *)&*.LC0 + 4B];
# DEBUG a$4 => a$4_3
a$8_4 = MEM[(int[3] *)&*.LC0 + 8B];
The previous patch changed this to:
:
SR.5_3 = *.LC0[0];
SR.6_4 = *.LC0[1];
SR.7_19 = *.LC0[2];
SR.8_21 = *.LC1[0];
SR.9_22 = *.LC1[1];
SR.10_23 = *.LC1[2];
# DEBUG a$0 => NULL // Note here
a$4_24 = SR.6_4;
# DEBUG a$4 => a$4_24
a$8_25 = SR.7_19;
Turns out the DEBUG a$0 => NULL was produced in load_assign_lhs_subreplacements:
if (lacc && lacc->grp_to_be_debug_replaced)
{
gdebug *ds;
tree drhs;
struct access *racc = find_access_in_subtree (sad->top_racc,
offset,
lacc->size);
if (racc && racc->grp_to_be_replaced)
{
if (racc->grp_write)
drhs = get_access_replacement (racc);
else
drhs = NULL; // <=== HERE
}
...
ds = gimple_build_debug_bind (get_access_replacement (lacc),
drhs, gsi_stmt (sad->old_gsi));
Prior to the patch, we'd skipped around load_assign_lhs_subreplacements, because
access_has_children_p(racc) (for racc = *.LC0) didn't hold in sra_modify_assign.
I also added a constant_decl_p function, combining the two checks, plus some
testcase fixes.
This also fixes a bunch of other guality tests on AArch64 that were failing
prior to the patch series, and another bunch on PowerPC64 (bigendian -m32),
listed below.
Bootstrapped + check-gcc,g++ on x86_64, ARM, AArch64,
also on powerpc64{,le}-none-linux-gnu *in combination with the other patches
in the series* (I haven't tested the individual patches on PPC),
plus Ada on ARM and x86_64.
gcc/ChangeLog:
PR target/63679
* tree-sra.c (disqualified_constants, constant_decl_p): New.
(sra_initialize): Allocate disqualified_constants.
(sra_deinitialize): Free disqualified_constants.
(disqualify_candidate): Update disqualified_constants when appropriate.
(create_access): Scan for constant-pool entries as we go along.
(scalarizable_type_p): Add check against type_contains_placeholder_p.
(maybe_add_sra_candidate): Allow constant-pool entries.
(load_assign_lhs_subreplacements): Bind debug for constant pool vars.
(initialize_constant_pool_replacements): New.
(sra_modify_assign): Avoid mangling assignments created by previous,
and don't generate writes into constant pool.
(sra_modify_function_body): Call initialize_constant_pool_replacements.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/sra-17.c: New.
* gcc.dg/tree-ssa/sra-18.c: New.
Tests fixed on aarch64-none-linux-gnu:
gcc.dg/guality/pr54970.c -O1 line 15 *p == 3
gcc.dg/guality/pr54970.c -O1 line 15 *q == 2
gcc.dg/guality/pr54970.c -O1 line 20 *p == 13
gcc.dg/guality/pr54970.c -O1 line 20 *q == 2
gcc.dg/guality/pr54970.c -O1 line 25 *p == 13
gcc.dg/guality/pr54970.c -O1 line 25 *q == 12
gcc.dg/guality/pr54970.c -O1 line 31 *p == 6
gcc.dg/guality/pr54970.c -O1 line 31 *q == 5
gcc.dg/guality/pr54970.c -O1 line 36 *p == 26
gcc.dg/guality/pr54970.c -O1 line 36 *q == 5
gcc.dg/guality/pr54970.c -O1 line 45 *p == 26
gcc.dg/guality/pr54970.c -O1 line 45 *q == 25
gcc.dg/guality/pr54970.c -O1 line 45 p[-1] == 25
gcc.dg/guality/pr54970.c -O1 line 45 p[-2] == 4
gcc.dg/guality/pr54970.c -O1 line 45 q[-1] == 4
gcc.dg/guality/pr54970.c -O1 line 45 q[1] == 26
gcc.dg/guality/pr56154-1.c -O1 line pr56154-1.c:20 x.a == 6
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s1.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.f == 0.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:17 s2.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s1.g == 6.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s2.f == 5.0
gcc.dg/guality/pr59776.c -O1 line pr59776.c:20 s2.g == 6.0
gcc.dg/guality/sra-1.c -O1 line 21 a.j == 14
gcc.dg/guality/sra-1.c -O1 line 32 a[1] == 14
gcc.dg/guality/sra-1.c -O1 line 43 a.i == 4
gcc.dg/guality/sra-1.c -O1 line 43 a.j == 14
(and other optimization levels)
On ppc64 bigendian, with the rest of the series (but I expect it's this patch):
unix/-m32: gcc.dg/guality/pr54970.c -O1 line 15 a[0] == 1
unix/-m32: gcc.dg/guality/pr54970.c -O1 line 20 a[0] == 1
unix/-m32: gcc.dg/guality/pr54970.c -O1 line 25 a[0] == 1
unix/-m32: gcc.dg/guality/pr54970.c -O1 line 31
