Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
29253 Discussions

internal compiler error, possibly a known bug

nooj
Beginner
1,028 Views
Hi, I didn't see any reports of this problem. It's an internal abort during ipo attempts. I think the problem is me having changed the Makefile between compilation and linking; perhaps some object files were not compiled compatibly: For instance, I recently removed the -parallel flags, but note that it is attempting parallelization.

I recompiled from scratch and the problem went away.


parallel loop: line 347
shared : { "var$2_dv_template.addr_a0_V$378" "var$2_dv_template.dim_info.lower_bound_V$381" "var$2_dv_template.addr_a0_V$38e" "var$2_dv_template.dim_info.lower_bound_V$397" "isuppq" "var$532_V$774.91.73" }
private : { "p" "t264" "var$540_V$778.91.73" "var$536_V$776.91.73" "var$534_V$775.91.73" }
first priv.: { "delt" "integral_of_q" }
reductions : { }
backend signals

fortcom: Severe: **Internal compiler error: internal abort** Please report this error along with the circumstances in which it occurred in a Software Problem Report. Note: File and line given may not be explicit cause of this error.

ifort: error #10014: problem during multi-file optimization compilation (code 3)
make: *** [obj/mac_laptop2/ipo/AneurysmElast_speedup.ipo.mac_laptop2] Error 3


ifort (IFORT) 11.1 20100401
OSX 10.6.4
Xcode 3.2.1

ifort options:
-B/opt/intel/Compiler/11.1/088/Frameworks/mkl
-I/opt/intel/Compiler/11.1/088/Frameworks/mkl/include
-Iobj/mac_laptop2/ipo
-L/opt/intel/Compiler/11.1/088/Frameworks/mkl/lib/em64t
-Lobj/mac_laptop2/ipo
-L/lib
-lguide
-lpthread
-lmkl_solver_lp64
-lmkl_intel_lp64
-lmkl_intel_thread
-lmkl_sequential
-lmkl_core obj/mac_laptop2/ipo/LIB_VTK_IO.a
-ipo
-ipo-jobs4
-module obj/mac_laptop2/ipo
-o obj/mac_laptop2/ipo/AneurysmElast_speedup.ipo.mac_laptop2

the offending code
(yes, that is a continuation star instead of an ampersand.
no, it's not the problem):
do p=0,isuppq
bigQ(p) =
& ( .5d0*q(p) + sum(q(p+1:isuppq-1)) + .5d0*q(isuppq) )
* *Delt
& /
& integral_of_q
enddo !p <---------LINE 347



0 Kudos
1 Solution
mecej4
Honored Contributor III
1,028 Views
Since it appears doubtful that even you can reproduce the ICE (unless you made a detailed record of the state of the project before you recompiled), there is little that anyone can do other than to acknowledge that an ICE is always a sign of a bug.

View solution in original post

0 Kudos
5 Replies
nooj
Beginner
1,028 Views
PS: Happy Halloween!
0 Kudos
mecej4
Honored Contributor III
1,029 Views
Since it appears doubtful that even you can reproduce the ICE (unless you made a detailed record of the state of the project before you recompiled), there is little that anyone can do other than to acknowledge that an ICE is always a sign of a bug.
0 Kudos
nooj
Beginner
1,028 Views
Yeah. I only reported it because it specifically asked me to, and I didn't realize the cause/solution until after I wrote the post. I went ahead and submitted the post because this kind of ICE (the kind caused by changing my compile/link flags and only re-making half of the project) has happened before. Maybe you have a sheet of paper somewhere that reads along the top, "Reports of ICEs caused by linking incompatible .os", which needs another tickmark.
0 Kudos
Xiaoping_D_Intel
Employee
1,028 Views
"internal compiler error" shows there is a bug in the compiler. I would like to submit a bug report for it. Is it possible for you to provide a test case which can be used to reproduce the error? If the code can't be public you can set the thread to be private. Another way is to submit the error with test case to Intel Premier Support if you have an account.

I saw the compiler you are using is not the latest version 11.1 update 7 (11.1.089 build 20100806). Maybe you can try this new version to see if it has been fixed. It can be downloaded from Intel Registration Center.



Thanks,
Xiaoping Duan
Intel Customer Support
0 Kudos
nooj
Beginner
1,028 Views
I have not forgotten about your request; I simply haven't run across the bug again (though I haven't been trying to find it, either).
0 Kudos
Reply