https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
Martin Liška changed:
What|Removed |Added
Priority|P3 |P1
Target Milestone|---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #10 from Martin Liška ---
Alternatively, one can see the memory corruption during LTO bootstrap in Go
library build:
$ ../configure --enable-languages=c,c++,lto,go --disable-multilib
--disable-libsanitizer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #9 from Martin Liška ---
... where mod_ref_flags are:
(gdb) p modref_flags
$15 = 247
and
(gdb) p modref_flags
$18 = 22
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #8 from Martin Liška ---
So the only 2 transform happen here:
(gdb) bt
#0 gimple_call_retslot_flags (stmt=0x74543b40) at
/home/marxin/Programming/gcc/gcc/gimple.c:1644
#1 0x00ff6128 in handle_rhs_call
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #7 from Martin Liška ---
$ grep dbgcnt omn*
omnetpp_r.ltrans0.ltrans.140t.pre:***dbgcnt: lower limit 1296 reached for
ipa_mod_ref_pta.***
omnetpp_r.ltrans0.ltrans.140t.pre:***dbgcnt: upper limit 1296 reached for
ipa_mod_ref_pta.***
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #6 from Martin Liška ---
I was able to track that to ltrans0, where one needs 2 optimizations to happen:
-fdbg-cnt=ipa_mod_ref_pta:1296-1296:1482-1482
then optimized dump differs (good vs. bad) in:
void recordScalar (struct
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #5 from Martin Liška ---
This one is real:
Running simulation...
** Event #1 T=0 Elapsed: 0.000s (0m 00s) 0% completed ev/sec=0
** Event #2160226 T=0.003 Elapsed: 0.000s (0m 00s) 100% completed
ev/sec=0
Simulation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #4 from Martin Liška ---
(In reply to Martin Liška from comment #3)
> First valgrind error happens here:
>
> Loading NED files from a SPEC directory 13
> ==22550== Conditional jump or move depends on uninitialised value(s)
>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #3 from Martin Liška ---
First valgrind error happens here:
Loading NED files from a SPEC directory 13
==22550== Conditional jump or move depends on uninitialised value(s)
==22550==at 0x53DF91: Cmdenv::sputn(char const*, int)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
Martin Liška changed:
What|Removed |Added
Last reconfirmed||2021-11-02
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103050
--- Comment #1 from Jan Hubicka ---
Martin,
do we know that the retslot change is the bad commit? I proofread the patch
and did not see anything obviously wrong, however as seen in PR103040 the
previous behaviour that made return slots to
11 matches
Mail list logo