Intel® C++ Compiler

error: pointer to incomplete class type is not allowed

Dear Sirs,

I am compiling part of a program (PAVE, from ). I have some problems during the compilation, and I don't know how to solve them.

To help you to find the problem I attach the files implied in the compilation. In concret, there is the makefile which I compile, the called Makeinclude, the output in a .log file and the file with the function with errors.

Also I write below fuctions belonging to diferent includes, to help to understand or solve the error:

/** netdb.h **/

Problems compiling with icc9.0


I'm currently redeveloping an existing application that was developed on MS Visual Studio .NET 2003. When it is built with the Microsoft compiler the app executes as expected.

Due to the fact that I'm adding speech processing to the app, I'm using IPPv5.0 and icc9.0. When recompiling, it compiles fine, and executes with unexpected results.

I changed the command line optionsin attempt to makeicc compile the same as ms compiler with no luck. I also tried disabling optimization of the icc compiler but stillno luck.

Any help will be gratefully appreciated.

icpc complains over missing gcc __builtin_xxxxx functions


I had this problem with Mandriva Linux 2006, which ships with a gcc 4.0.1 version:

icpc -static-libcxa -c -g -DLINUX -DDEBUG -DWHOAMI="burger_savitri" -DHOST_TYPE="2.6.12-12mdksmp" -DUSE_LITTLE_ENDIAN -I. -I/mnt/heimann/vsheet2/hitrax/include/ -O0 -ip -DUSE_LIBXML2 -I/usr/include/libxml2/ -I../other -fPIC eigsrt.cpp -o
/usr/include/c++/4.0.1/cmath(361): error: identifier "__builtin_powil" is undefined
{ return __builtin_powil(__x, __n); }

Questions on Intel's STL implementation


Does anyone know if there is available documentation on Intel's implementation of STL?

Some of my questions are:

Are the non-standard hash containers available in the implementation?

How is the STL allocator implemented? Is there more than one I can choose? (as in other implementations - pooled memory, standard malloc/free, etc).

Is the map a red-black tree (as Microsoft's STL is, log(N) searches and inserts) or a hash?

Options for Core Duo

Which compiler flags should I use for Core Duo under Linux?
In particular, shall I use -xB, because Core Duo
is similar to the Pentium M family?

Alternatively, will a future compiler have explicit
optimizations for Core Duo?

Message Edited by hnaparst on 01-30-2006 08:03 AM

Vector conditional


How would I go about efficiently changing each value of a vector based on a conditional? For example, I have a vector filled with doubles that should either have a value added to and multiplied by it, or should get set to 0.

If it were a scalar, I would do the following (using "lt." because actual character breaks html):

foo = (foo lt. bar) ? (foo+1.1)*2.2 : 0;

Basically I'm trying to avoid:

for (int i=0; i lt. size; ++i)


if (*(foo + i) lt. bar)

foo = (foo+1.1)*2.2;


foo = 0;

ICL9 bad performance?

I'm testing some algorithms which deal with sorting range, and I constantly find that at full optimizations, ICL executable is inferior to Visual C++.

I'm running these tests on a Pentium III, not on a P4.

I run tests on a randomly shuffled vector of length 10000000, working both with raw pointers (i.e. using begin=&V[0], end=&V[0]+V.size()) and on vector iterators.

Execution times are measured via QueryPerformanceCounter.

Here are some stats:

I list in the table average time and (max-min) time:

icl9(ptr) 0,4460 0,0367

Intel processors not optimal for executing java programs?


I know this is a bit off-topic but it also has to do with software development.
Recently there was a discussion about executing java programs on different platforms and which platform performance-wise is best for executing such highly optimized hotspot-generated code.

Subscribe to Intel® C++ Compiler