TestNan and TestInf fail on -O3 #3944
Labels
Comments
|
Try -fno-fast-math
fast-math is an optimization where it assumes inf and nan can never occur.
…On Thu, Feb 20, 2020 at 4:22 AM Dee Gelbart ***@***.***> wrote:
To reproduce:
cd src
CXXFLAGS='-g -O3' ./configure
make clean; make depend
cd util
make
make test
Result:
Running text-utils-test .../bin/bash: line 1: 33755 Segmentation fault (core dumped) ./$x > $x.testlog 2>&1
0s... FAIL text-utils-test
Testlog is empty:
ll text-utils-test.testlog
-rw-r--r-- 1 dee dee 0 Feb 19 12:21 text-utils-test.testlog
gdb output:
$ gdb text-utils-test
(gdb) r
Starting program: /mnt/disk1/dee/github/kaldi/src/util/text-utils-test
warning: Error disabling address space randomization: Operation not permitted
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Program received signal SIGSEGV, Segmentation fault.
0x0000562a8de396b5 in std::char_traits<char>::copy (__n=4, __s2=0x7ffca9c92d30 "-nan", __s1=<optimized out>) at /usr/include/c++/7/bits/char_traits.h:350
350 return static_cast<char_type*>(__builtin_memcpy(__s1, __s2, __n));
(gdb) thread apply all bt
Thread 1 (Thread 0x7f3f5d1da740 (LWP 34589)):
#0 0x0000562a8de396b5 in std::char_traits<char>::copy (__n=4, __s2=0x7ffca9c92d30 "-nan", __s1=<optimized out>) at /usr/include/c++/7/bits/char_traits.h:350
#1 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_S_copy (__n=4, __s=0x7ffca9c92d30 "-nan", __d=<optimized out>) at /usr/include/c++/7/bits/basic_string.h:340
#2 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_S_copy_chars (__k2=0x7ffca9c92d34 "", __k1=0x7ffca9c92d30 "-nan", __p=<optimized out>)
at /usr/include/c++/7/bits/basic_string.h:382
#3 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_construct<char*> (__end=0x7ffca9c92d34 "", __beg=0x7ffca9c92d30 "-nan", this=0x7ffca9c92fc0)
at /usr/include/c++/7/bits/basic_string.tcc:225
#4 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_construct_aux<char*> (__end=0x7ffca9c92d34 "", __beg=0x7ffca9c92d30 "-nan", this=0x7ffca9c92fc0)
at /usr/include/c++/7/bits/basic_string.h:236
#5 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_construct<char*> (__end=0x7ffca9c92d34 "", __beg=0x7ffca9c92d30 "-nan", this=0x7ffca9c92fc0)
at /usr/include/c++/7/bits/basic_string.h:255
#6 std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::basic_string<char*, void> (__a=..., __end=0x7ffca9c92d34 "", __beg=0x7ffca9c92d30 "-nan", this=0x7ffca9c92fc0)
at /usr/include/c++/7/bits/basic_string.h:607
#7 __gnu_cxx::__to_xstring<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, char> (__convf=<optimized out>, __n=328, __fmt=0x562a8de54904 "%f", __fmt=0x562a8de54904 "%f",
__n=328, __convf=<optimized out>) at /usr/include/c++/7/ext/string_conversions.h:115
#8 0x0000562a8de4414c in std:: (__val=-nan(0x8000000000000)) at /usr/include/c++/7/bits/basic_string.h:6462
#9 kaldi::TestNan<float> () at text-utils-test.cc:238
#10 0x00007ffca9c92de0 in ?? ()
#11 0x00007ffca9c92e58 in ?? ()
#12 0x00007ffca9c92d00 in ?? ()
#13 0x00007f3f5ca00b50 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#14 0x00007f3f5c73c501 in std::ios_base::~ios_base() () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#15 0x0000562a8de508cf in std::basic_ios<char, std::char_traits<char> >::~basic_ios (this=0x7ffca9c930f8, __in_chrg=<optimized out>) at /usr/include/c++/7/bits/basic_ios.h:282
#16 std::__cxx11::basic_istringstream<char, std::char_traits<char>, std::allocator<char> >::~basic_istringstream (this=0x7ffca9c92de0, __in_chrg=<optimized out>, __vtt_parm=<optimized out>)
at /usr/include/c++/7/sstream:446
#17 kaldi::ConvertStringToReal<double> (str=..., out=0x7ffca9c92fb8) at text-utils.cc:240
#18 0xfff8000000000000 in ?? ()
#19 0x007920302e312078 in ?? ()
#20 0x000031342e362079 in ?? ()
#21 0x00007ffca9c92fd0 in ?? ()
#22 0x0000562a0000000a in ?? ()
#23 0x00007f006e616e2d in ?? ()
#24 0x0000000000000000 in ?? ()
(gdb) quit
valgrind output:
$ valgrind ./text-utils-test
==34652== Memcheck, a memory error detector
==34652== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==34652== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==34652== Command: ./text-utils-test
==34652==
==34652== Invalid write of size 8
==34652== at 0x10E6B5: copy (char_traits.h:350)
==34652== by 0x10E6B5: _S_copy (basic_string.h:340)
==34652== by 0x10E6B5: _S_copy_chars (basic_string.h:382)
==34652== by 0x10E6B5: _M_construct<char*> (basic_string.tcc:225)
==34652== by 0x10E6B5: _M_construct_aux<char*> (basic_string.h:236)
==34652== by 0x10E6B5: _M_construct<char*> (basic_string.h:255)
==34652== by 0x10E6B5: basic_string<char*> (basic_string.h:607)
==34652== by 0x10E6B5: std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > __gnu_cxx::__to_xstring<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, ch
ar>(int (*)(char*, unsigned long, char const*, __va_list_tag*), unsigned long, char const*, ...) [clone .constprop.100] (string_conversions.h:115)
==34652== by 0x11914B: to_string (basic_string.h:6462)
==34652== by 0x11914B: void kaldi::TestNan<float>() (text-utils-test.cc:238)
==34652== Address 0x1fff001000 is not stack'd, malloc'd or (recently) free'd
==34652==
==34652==
==34652== Process terminating with default action of signal 11 (SIGSEGV): dumping core
==34652== Access not within mapped region at address 0x1FFF001000
==34652== at 0x10E6B5: copy (char_traits.h:350)
==34652== by 0x10E6B5: _S_copy (basic_string.h:340)
==34652== by 0x10E6B5: _S_copy_chars (basic_string.h:382)
==34652== by 0x10E6B5: _M_construct<char*> (basic_string.tcc:225)
==34652== by 0x10E6B5: _M_construct_aux<char*> (basic_string.h:236)
==34652== by 0x10E6B5: _M_construct<char*> (basic_string.h:255)
==34652== by 0x10E6B5: basic_string<char*> (basic_string.h:607)
==34652== by 0x10E6B5: std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > __gnu_cxx::__to_xstring<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, ch
ar>(int (*)(char*, unsigned long, char const*, __va_list_tag*), unsigned long, char const*, ...) [clone .constprop.100] (string_conversions.h:115)
==34652== by 0x11914B: to_string (basic_string.h:6462)
==34652== by 0x11914B: void kaldi::TestNan<float>() (text-utils-test.cc:238)
==34652== If you believe this happened as a result of a stack
==34652== overflow in your program's main thread (unlikely but
==34652== possible), you can try to increase the size of the
==34652== main thread stack using the --main-stacksize= flag.
==34652== The main thread stack size used in this run was 8388608.
==34652==
==34652== HEAP SUMMARY:
==34652== in use at exit: 0 bytes in 0 blocks
==34652== total heap usage: 2,092 allocs, 2,092 frees, 203,287 bytes allocated
==34652==
==34652== All heap blocks were freed -- no leaks are possible
==34652==
==34652== For counts of detected and suppressed errors, rerun with: -v
==34652== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Segmentation fault (core dumped)
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#3944?email_source=notifications&email_token=AAZFLO5WQCKIZBC6B4ZI37TRDWIJVA5CNFSM4KYBAXEKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IOYJMUA>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAZFLO6PTBKFA35QDB2AOMTRDWIJVANCNFSM4KYBAXEA>
.
|
|
The error still happens with if I configure with |
|
I assume either you failed to do |
|
I did the clean. Should I take the test failure as a warning not to use kaldi on this optimization setting? Thanks. |
|
Yes, don't use that setting. Maybe 2 will work.
…On Thu, Feb 20, 2020 at 1:56 PM Deanna Gelbart ***@***.***> wrote:
I did the clean. Should I take the test failure as a warning not to use
kaldi on this optimization setting? Thanks.
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub
<#3944?email_source=notifications&email_token=AAZFLO5CYA3KWMRGVCZJMFTRDYLR5A5CNFSM4KYBAXEKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMK27GA#issuecomment-588623768>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAZFLO6PZUU5W4LLN2CPXR3RDYLR5ANCNFSM4KYBAXEA>
.
|
|
Oddly enough, I'm using I think the problematic options, such as |
|
it also depends on version of gcc (they keep optimizing the optimizer) and
the math library. I have had bad experience with OpenBlas several times,
for example.
y.
…On Thu, Feb 20, 2020 at 3:34 AM Ewald Enzinger ***@***.***> wrote:
Oddly enough, I'm using -O3 without problems:
Running const-integer-set-test ... 0s... SUCCESS const-integer-set-test
Running stl-utils-test ... 1s... SUCCESS stl-utils-test
Running text-utils-test ... 0s... SUCCESS text-utils-test
Running edit-distance-test ... 0s... SUCCESS edit-distance-test
Running hash-list-test ... 0s... SUCCESS hash-list-test
Running kaldi-io-test ... 2s... SUCCESS kaldi-io-test
Running parse-options-test ... 0s... SUCCESS parse-options-test
Running kaldi-table-test ... 2s... SUCCESS kaldi-table-test
Running simple-options-test ... 0s... SUCCESS simple-options-test
Running kaldi-thread-test ... 0s... SUCCESS kaldi-thread-test
I think the problematic options, such as -ffast-math get enabled when
using -Ofast (https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html).
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#3944?email_source=notifications&email_token=ACUKYX5WWMM2TORYOMNBMHTRDY6C3A5CNFSM4KYBAXEKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEML5R5Y#issuecomment-588765431>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACUKYX3JN7QRQLWSVBA43NDRDY6C3ANCNFSM4KYBAXEA>
.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
To reproduce:
Result:
Testlog is empty:
gdb output:
valgrind output:
The text was updated successfully, but these errors were encountered: