Intel 7.0 Compiler and date_time library
First of all I would like to thank everybody who responded on my previous post. The real showstopper for me at the moment is date_time library. I did a small test -- compiled and run days_till_new_year.cpp against boost 1.29 (patched to not complain about __ICL 700) and current snapshot of boost CVS. Here is what I've got (cpp was renamed to test.cpp for short): ------------------------------------- C:\work\icl>icl Intel(R) C++ Compiler for 32-bit applications, Version 7.0 Build 20021018Z Copyright (C) 1985-2002 Intel Corporation. All rights reserved. 30 DAY EVALUATION LICENSE icl: Command line error: no files specified; for help type "icl -help" C:\work\icl>icl /MD /O2 /DDATE_TIME_INLINE /I..\boost_1_29_0 /nologo test.cpp test.cpp C:\work\icl>test Days till new year: -731982 C:\work\icl>icl /MD /Od /DDATE_TIME_INLINE /I..\boost_1_29_0 /nologo test.cpp test.cpp C:\work\icl>test Days till new year: 329 C:\work\icl>icl /MD /O2 /DDATE_TIME_INLINE /I..\boost /nologo test.cpp test.cpp C:\work\icl>test Days till new year: 3031841 C:\work\icl>icl /MD /Od /DDATE_TIME_INLINE /I..\boost /nologo test.cpp test.cpp C:\work\icl>test Days till new year: 329 ---------------------------------------------- As you can see both versions are broken. What I don't understand is how come that boost regression tests are Ok? Am I using wrong version of Intel Compiler, or am I missing some magic flags to compiler? I will send this report to Intel tech support; however I would appreciate if someone will verify these results, and contact Intel on behalf of Boost community. Regards, Kirill Lapshin.
C:\work\icl>icl /MD /Od /DDATE_TIME_INLINE /I..\boost_1_29_0 /nologo test.cpp
C:\work\icl>test
Days till new year: 329
This one is right. And there is another one that is correct as well. Looks like the /MD /Od settings get the correct result.
...details snipped...
As you can see both versions are broken. What I don't understand is how come that boost regression tests are Ok? Am I using wrong version of Intel Compiler, or am I missing some magic flags to compiler?
Did you run the regression tests (actually the in 1.29 the build runs the tests) or are you counting on the posted regression tests? I'm not sure what settings are used in the online regression... Jeff
--- In Boost-Users@yahoogroups.com, "Jeff Garland"
C:\work\icl>icl /MD /Od /DDATE_TIME_INLINE /I..\boost_1_29_0 /nologo
test.cpp
C:\work\icl>test
Days till new year: 329
This one is right. And there is another one that is correct as well.
Looks like the /MD /Od settings get the correct result.
Yes, basically turning optimization on makes it fail. /Od is fine, /O1 and /O2 fail miserably.
As you can see both versions are broken. What I don't understand is
how come that boost regression tests are Ok? Am I using wrong version
of Intel Compiler, or am I missing some magic flags to compiler?
Did you run the regression tests (actually the in 1.29 the build
runs the tests) or are you counting on the posted regression tests?
I'm not sure what settings are used in the online regression...
bjam build does not work for me, so I can't really run them easily. So yes, I was referring to posted regression tests. I tried to compile and run the tests by hand, and it again worked fine without optimization, and blew up with optimization. So I can see few possibilities, either I am using different version of compiler (mine is Version 7.0 Build 20021018Z), or I am missing certain magic flags, or something is totally wrong with my computer. I am inclined to blaim the compiler. If it is the issue with compiler, then I think we have to bug intel to fix it. I will do it from my end, but more official request from Boost would probably be more effective. - Kirill
If it is the issue with compiler, then I think we have to bug intel to fix it. I will do it from my end, but more official request from Boost would probably be more effective.
Sorry, everyone. Case closed. It turns out that intel shipped me outdated version of the compiler. Upgrdaing it to most recent one solved the problem.
participants (2)
-
Jeff Garland
-
klapshin <klapshin@yahoo.com>