[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #13 from CVS Commits --- The releases/gcc-13 branch has been updated by Richard Biener : https://gcc.gnu.org/g:152400decc8383aeff9a9ad8262b9e7e2fff61e0 commit r13-8096-g152400decc8383aeff9a9ad8262b9e7e2fff61e0 Author: Richard Biener Date: Tue Sep 19 12:36:04 2023 +0200 tree-optimization/111465 - bougs jump threading with no-copy src block The following avoids to forward thread a path with a EDGE_NO_COPY_SRC_BLOCK block that became non-empty due to folding. PR tree-optimization/111465 * tree-ssa-threadupdate.cc (fwd_jt_path_registry::thread_block_1): Cancel the path when a EDGE_NO_COPY_SRC_BLOCK became non-empty. * g++.dg/torture/pr111465.C: New testcase. (cherry picked from commit 564ecb7d5afb0bb4eb39285ce65c631490e37dce)
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 Richard Biener changed: What|Removed |Added Resolution|--- |DUPLICATE Status|ASSIGNED|RESOLVED --- Comment #12 from Richard Biener --- Should be fixed on trunk. Given PR102844 the issue is latent everywhere, marking as duplicate for this. *** This bug has been marked as a duplicate of bug 102844 ***
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #11 from CVS Commits --- The master branch has been updated by Richard Biener : https://gcc.gnu.org/g:564ecb7d5afb0bb4eb39285ce65c631490e37dce commit r14-4128-g564ecb7d5afb0bb4eb39285ce65c631490e37dce Author: Richard Biener Date: Tue Sep 19 12:36:04 2023 +0200 tree-optimization/111465 - bougs jump threading with no-copy src block The following avoids to forward thread a path with a EDGE_NO_COPY_SRC_BLOCK block that became non-empty due to folding. PR tree-optimization/111465 * tree-ssa-threadupdate.cc (fwd_jt_path_registry::thread_block_1): Cancel the path when a EDGE_NO_COPY_SRC_BLOCK became non-empty. * g++.dg/torture/pr111465.C: New testcase.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #10 from Richard Biener --- Testing a patch.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465
Richard Biener changed:
What|Removed |Added
See Also||https://gcc.gnu.org/bugzill
||a/show_bug.cgi?id=102844
CC||law at gcc dot gnu.org
--- Comment #9 from Richard Biener ---
So we simplify
if (iftmp.2_137 <= location_138)
..
else
{
_16 = (signed) prephitmp_15;
if (_16 < 0)
...
to
_192 = (signed) location_138;
if (0)
..
else
{
_16 = (signed) prephitmp_15;
if (_16 < 0)
...
and then thread - actually we have registered the forward thread already
and registered the block with the def of _192 as EDGE_NO_COPY_SRC_BLOCK
(because it was empty before). This is essentially a duplicate of PR102844
which has gone latent.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 Richard Biener changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot gnu.org --- Comment #8 from Richard Biener --- I will have a look.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #7 from Andrew Pinski --- ``` Optimizing block #15 1>>> STMT 1 = prephitmp_58 le_expr location_138 1>>> STMT 1 = prephitmp_58 ge_expr location_138 1>>> STMT 1 = prephitmp_58 eq_expr location_138 1>>> STMT 0 = prephitmp_58 ne_expr location_138 Optimizing statement if (iftmp.2_137 <= location_138) Replaced 'iftmp.2_137' with constant '2147483648' Applying pattern match.pd:6895, gimple-match-8.cc:2512 Applying pattern match.pd:5745, gimple-match-4.cc:2113 gimple_simplified to _192 = (signed int) location_138; if (_192 < 0) Folded to: if (_192 < 0) Registering value_relation (_192 pe32 location_138) (bb15) at _192 = (signed int) location_138; Folding predicate _192 < 0 to 0 Optimizing statement _192 = (signed int) location_138; LKUP STMT _192 = nop_expr location_138 2>>> STMT _192 = nop_expr location_138 Optimizing block #16 Optimizing statement _16 = (signed int) prephitmp_15; Replaced 'prephitmp_15' with variable 'location_138' LKUP STMT _16 = nop_expr location_138 FIND: _192 Replaced redundant expr '(signed int) location_138' with '_192' ASGN _16 = _192 Optimizing statement if (_16 < 0) Replaced '_16' with variable '_192' Folding predicate _192 < 0 to 0 Marking all outgoing edges of unreachable BB 17 as not executable STMT _192 = nop_expr location_138 STMT 0 = prephitmp_58 ne_expr location_138 STMT 1 = prephitmp_58 eq_expr location_138 STMT 1 = prephitmp_58 ge_expr location_138 STMT 1 = prephitmp_58 le_expr location_138 COPY prephitmp_15 = location_138 ``` But _16 is no longer used ...
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #6 from Andrew Pinski --- Created attachment 55928 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55928=edit Reduced even further Still requires `-O2 -fno-exceptions --param=logical-op-non-short-circuit=0` to reproduce.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 Andrew Pinski changed: What|Removed |Added Ever confirmed|0 |1 Last reconfirmed||2023-09-18 Target|powerpc64le-linux-gnu | Build|powerpc64le-linux-gnu | Host|powerpc64le-linux-gnu | Status|UNCONFIRMED |NEW --- Comment #5 from Andrew Pinski --- (In reply to Andrew Pinski from comment #4) > Created attachment 55927 [details] > reduced somewhat > > Compile with `-O2 -fno-exceptions`. This reduced testcase fails on x86_64-linux-gnu with `-O2 -fno-exceptions --param=logical-op-non-short-circuit=0` Confirmed.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap from r14-4089-gd45ddc2c04e471d0dcee01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #4 from Andrew Pinski --- Created attachment 55927 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55927=edit reduced somewhat Compile with `-O2 -fno-exceptions`.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #3 from Andrew Pinski --- Reproduced, reducing ...
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 --- Comment #2 from seurer at gcc dot gnu.org --- Created attachment 55926 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55926=edit preprocessed .ii file Yup, that looks like the culprit. Preprocessed file attached.
[Bug tree-optimization/111465] [14 regression] stage 3 ICE kills bootstrap
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111465 Andrew Pinski changed: What|Removed |Added Component|bootstrap |tree-optimization Target Milestone|--- |14.0 Keywords||build, ice-checking, ||ice-on-valid-code --- Comment #1 from Andrew Pinski --- I suspect this was caused by/exposed by r14-4089-gd45ddc2c04e471d0dcee01 . Do you have the preprocessed source which you could attach?
