Migration checklist when moving from the XL-based front end to the Clang-based front end
Aside from invoking the legacy IBM® XL-based compiler front end, you can alternatively invoke the Clang-based front end provided by XL C/C++ by using the xlclang or xlclang++ invocation command. As the Clang-based front end provides enhancements described in Comparison between the XL-based and Clang-based front ends, you might want to move to xlclang/xlclang++. However, you must be aware of these differences during migration.
Binary compatibility of xlc/xlc++ and xlclang/xlclang++
| Category | Case | Required action |
|---|---|---|
| Binary incompatibility case | For C++ source, object files generated by xlc++ and xlclang++ are not directly compatible as the XL-based and Clang-based front ends use different C++ standard libraries, run times, and object models. | You must recompile your legacy programs by using xlclang or xlclang++ to solve such binary incompatibility and link object files successfully. |
| Binary compatibility case | For C source, object files generated by xlc, xlc++, xlclang, and xlclang++ are directly compatible when the symbols names in source files contain only the dollar sign and the characters from the basic character set. | No action is required. You do not need to recompile source files before you link these object files. |
Considerations when merging generated code of C++11 and C++03 or prior
You should treat C++11 as a new language that is not directly interoperable with C++03 or prior language levels.
If you are an expert in the C++ library and runtime state and are willing to manage that
complexity by yourself, provided you take the following considerations, C++03 or prior generated
code can co-exist alongside C++11 generated code:
- There might be two C++ run times active in a single process. Modifying state in the C++ runtime,
such as the
std::localesettings, in both C++03 and C++11 modules results in undefined behavior. - Do not delete a pointer in a C++03 module, which was allocated in a C++11 module, or vice versa.
- Do not use direct C++ language calls between C++11 and C++03 or prior. Use C interfaces instead to call between C++11 and C++03 or prior.
- Do not catch C++ exceptions thrown from C++11 in C++03 or prior, or vice versa.
Discrepancies of option defaults between xlc/xlC and xlclang/xlclang++
| Default on xlc/xlc++ | Default on xlclang/xlclang++ |
|---|---|
| -qnortti | -qrtti |
| -qbitfields=unsigned | Bit fields are treated as signed. |
| -qhalt=s | -Werror (-qhalt=w) is disabled. |
| -qnothreaded | -qthreaded |
| -qtmplinst=auto | -qtmplinst=none |
The following legacy macros are defined: __xlC__,
__xlC_ver__,  __IBMCPP__ ,  __IBMC__, and __xlc__ |
-qnoxlcompatmacros (These legacy macros are not defined.) |
| -qvisibility=unspecified | -qnovisibility (All the visibility pragmas and attributes that are specified in the source are ignored. ) |
Note: Only typical invocations are listed in this table. You can refer
to the full list of compiler
invocations for all basic
invocations and their equivalent special invocations.
Changed predefined macros to identify XL C/C++ when invoked by xlclang/xlclang++
For IBM XL C/C++ for AIX 16.1 that is invoked by xlclang or
xlclang++,
predefined macros to identify the XL C/C++ compiler are changed:
- The formerly predefined legacy macros, namely
__xlC__,__xlC_ver__,__IBMCPP__, __IBMC__, and__xlc__, are no longer predefined by default. - Some new macros are introduced and predefined by default.
To ease your migration efforts, option -qxlcompatmacros is introduced, which defines the five legacy macros. Note that the compiler default for xlclang and xlclang++ invocations is -qnoxlcompatmacros, meaning that the five legacy macros are not predefined. You might need to specify the -qxlcompatmacros option when you migrate programs from IBM XL C/C++ for AIX, V13.1.3 or earlier releases to IBM XL C/C++ for AIX 16.1 that is invoked by xlclang or xlclang++.
Unsupported legacy XL options by xlclang/xlclang++
| Option | xlclang (Compiling C) | xlclang++ (Compiling C++) |
|---|---|---|
| -xc++ | Unsupported | |
| -g3, -g4, -g5, -g6, -g7, -g8, -g9 | Unsupported | Unsupported |
| -ma | Unsupported | Unsupported |
| -qalias=allptrs | noallptrs | global | noglobal | typeptr | notypeptr | Unsupported | Unsupported |
| -qalign | Unsupported | Unsupported |
| -qalignrulefor | Unsupported | Unsupported |
| -qalloca | Unsupported | Unsupported |
| -qasm=stdcpp | Unsupported | Unsupported |
| -qassert | Unsupported | Unsupported |
| -qattr | -qnoattr | Unsupported | Unsupported |
| -qbitfields | Unsupported | Unsupported |
| -qcinc | -qnocinc | Unsupported | Unsupported |
| -qconcurrentupdate | -qnoconcurrentupdate | Unsupported | Unsupported |
| -qcpluscmt | -qnocpluscmt | Unsupported | Unsupported |
| -qc_stdinc | Unsupported | |
| -qcpp_stdinc | Unsupported | |
| -qdbcs | -qnodbcs | Unsupported | Unsupported |
| -qdbxextra | -qnodbxextra | Unsupported | Unsupported |
| -qdfp | -qnodfp | Unsupported | Unsupported |
| -qdigraph | -qnodigraph | Unsupported | Unsupported |
| -qdpcl | -qnodpcl | Unsupported | Unsupported |
| -qfloat=dfpemulate | nodfpemulate | fltint | nofltint | hssngl ｜nohssngl | rndsngl | norndsngl | single | nosingle | Unsupported | Unsupported |
| -qeh | Unsupported | |
| -qenum | Unsupported | Unsupported |
| -qextchk | -qnoextchk | Unsupported | Unsupported |
| -qflag | Unsupported | Unsupported |
| -qformat | -qnoformat | Unsupported | Unsupported |
| -qgenproto | -qnogenproto | Unsupported | Unsupported |
| -qhalt = i | e | s (C only) -qhalt = i | s (C++ only) |
Unsupported | Unsupported |
| -qhaltonmsg | -qnohaltonmsg | Unsupported | Unsupported |
| -qheapdebug | -qnoheapdebug | Unsupported | Unsupported |
| -qignprag | Unsupported | Unsupported |
| -qinfo | -qnoinfo | Unsupported | Unsupported |
| -qinline=autothreshold | Unsupported | Unsupported |
| -qipa=threads | nothreads | Unsupported | Unsupported |
| -qisolated_call | Unsupported | Unsupported |
| -qkeepinlines | Unsupported | |
| -qkeyword | -qnokeyword | Unsupported | Unsupported |
| -qlanglvl=classic | extended | saa | saal2 | feature_suboption (C
only) -qlanglvl=compat366 | strict98 | feature_suboption (C++ only) |
Unsupported | Unsupported |
| -qldbl128 | -qnoldbl128 | Unsupported | Unsupported |
| -qlistfmt | Unsupported | Unsupported |
| -qlistopt | -qnolistopt | Unsupported | Unsupported |
| -qlongdouble | -qnolongdouble | Unsupported | Unsupported |
| -qlonglit | -qnolonglit | Unsupported | Unsupported |
| -qlonglong | -qqnolonglong | Unsupported | Unsupported |
| -qmacpstr | -qnomacpstr | Unsupported | Unsupported |
| -qmakedep | Unsupported | Unsupported |
| -qmaxerr | -qnomaxerr | Unsupported | Unsupported |
| -qmbcs | -qnombcs | Unsupported | Unsupported |
| -qnamemangling | Unsupported | Unsupported |
| -qobjmodel | Unsupported | Unsupported |
| -qoldpassbyvalue | -qnooldpassbyvalue | Unsupported | Unsupported |
| -qoptdebug | -qnooptdebug | Unsupported | Unsupported |
| -qppline | -qnoppline | Unsupported | Unsupported |
| -qprint | -qnoprint | Unsupported | Unsupported |
| -qpriority | Unsupported | |
| -qproto | -qnoproto | Unsupported | Unsupported |
| -qrestrict | -qnorestrict | Unsupported | Unsupported |
| -qroptr | -qnoroptr | Unsupported | Unsupported |
| -qrtti | Unsupported | |
| -qshowinc | -qnoshowinc | Unsupported | Unsupported |
| -qshowmacros = all | pre | nopre | Unsupported | Unsupported |
| -qskipsrc | Unsupported | Unsupported |
| -qsmp | -qnosmp | Unsupported | Unsupported |
| -qsource | -qnosource | Unsupported | Unsupported |
| -qsrcmsg | -qnosrcmsg | Unsupported | Unsupported |
| -qstaticinline | -qnostaticinline | Unsupported | Unsupported |
| -qstatsym | -qnostatsym | Unsupported | Unsupported |
| -qsymtab | Unsupported | Unsupported |
| -qtabsize | Unsupported | Unsupported |
| -qtempinc | -qnotempinc | Unsupported | Unsupported |
| -qtemplatedepth | Unsupported | |
| -qtemplaterecompile | -qnotemplaterecompile | Unsupported | Unsupported |
| -qtemplateregistry | -qnotemplateregistry | Unsupported | Unsupported |
| -qtempmax | Unsupported | Unsupported |
| -qnothreaded | Unsupported | Unsupported |
| -qtmplinst= always | auto | noinline | Unsupported | Unsupported |
| -qtmplparse | Unsupported | Unsupported |
| -qtrigraph | -qnotrigraph | Unsupported | Unsupported |
| -qtwolink | Unsupported | |
| -qupconv | -qnoupconv | Unsupported | Unsupported |
| -qutf | -qnoutf | Unsupported | Unsupported |
| -qwarn0x | -qnowarn0x | Unsupported | Unsupported |
| -qwarn64 | -qnowarn64 | Unsupported | Unsupported |
| -qweaksymbol | -qnoweaksymbol | Unsupported | Unsupported |
| -qxcall | -qnoxcall | Unsupported | Unsupported |
| -qxref | -qnoxref | Unsupported | Unsupported |
For more information, see Supported IBM XL compiler options by different invocations.
Unsupported legacy XL pragmas by xlclang/xlclang++
| IBM pragma | xlclang (Compiling C) | xlclang++ (Compiling C++) |
|---|---|---|
| #pragma alloca (C only) | Unsupported | Unsupported |
| #pragma block_loop | Unsupported | Unsupported |
| #pragma chars | Unsupported | Unsupported |
| #pragma comment | Unsupported | Unsupported |
| #pragma define (C++ only) | Unsupported | Unsupported |
| #pragma instantiate (C++ only) | Unsupported | Unsupported |
| #pragma do_not_instantiate (C++ only) | Unsupported | Unsupported |
| #pragma enum | Unsupported | Unsupported |
| #pragma expected_value | Unsupported | Unsupported |
| #pragma fini (C only) | Unsupported | Unsupported |
| #pragma hashome (C++ only) | Unsupported | Unsupported |
| #pragma ibm iterations | Unsupported | Unsupported |
| #pragma ibm max_iterations | Unsupported | Unsupported |
| #pragma ibm min_iterations | Unsupported | Unsupported |
| #pragma ibm snapshot | Unsupported | Unsupported |
| #pragma implementation (C++ only) | Unsupported | Unsupported |
| #pragma info | Unsupported | Unsupported |
| #pragma init (C only) | Unsupported | Unsupported |
| #pragma ishome (C++ only) | Unsupported | Unsupported |
| #pragma isolated_call | Unsupported | Unsupported |
| #pragma langlvl (C only) | Unsupported | Unsupported |
| #pragma leaves | Unsupported | Unsupported |
| #pragma loopid | Unsupported | Unsupported |
| #pragma map | Unsupported | Unsupported |
| #pragma mc_func | Unsupported | Unsupported |
| #pragma namemangling (C++ only) | Unsupported | Unsupported |
| #pragma namemanglingrule (C++ only) | Unsupported | Unsupported |
| #pragma nofunctrace | Unsupported | Unsupported |
| #pragma nosimd | Unsupported | Unsupported |
| #pragma novector | Unsupported | Unsupported |
| #pragma object_model (C++ only) | Unsupported | Unsupported |
| #pragma operator_new (C++ only) | Unsupported | Unsupported |
| #pragma options | Unsupported | Unsupported |
| #pragma pass_by_value (C++ only) | Unsupported | Unsupported |
| #pragma priority (C++ only) | Unsupported | Unsupported |
| #pragma reg_killed_by | Unsupported | Unsupported |
| #pragma report (C++ only) | Unsupported | Unsupported |
| #pragma simd_level | Unsupported | Unsupported |
| #pragma stream_unroll | Unsupported | Unsupported |
| #pragma strings | Unsupported | Unsupported |
| #pragma unroll | Unsupported | Unsupported |
| #pragma nounroll | Unsupported | Unsupported |
| #pragma unrollandfuse | Unsupported | Unsupported |
| #pragma weak | Unsupported | Unsupported |
| #pragma ibm independent_calls (C only) | Unsupported | Unsupported |
| #pragma ibm permutation (C only) | Unsupported | Unsupported |
| #pragma ibm schedule (C only) | Unsupported | Unsupported |
| #pragma ibm sequential_loop (C only) | Unsupported | Unsupported |
| #pragma omp atomic | Unsupported | Unsupported |
| #pragma omp parallel | Unsupported | Unsupported |
| #pragma omp for | Unsupported | Unsupported |
| #pragma omp ordered | Unsupported | Unsupported |
| #pragma omp parallel for | Unsupported | Unsupported |
| #pragma omp section, #pragma omp sections | Unsupported | Unsupported |
| #pragma omp parallel sections | Unsupported | Unsupported |
| #pragma omp single | Unsupported | Unsupported |
| #pragma omp master | Unsupported | Unsupported |
| #pragma omp critical | Unsupported | Unsupported |
| #pragma omp barrier | Unsupported | Unsupported |
| #pragma omp flush | Unsupported | Unsupported |
| #pragma omp threadprivate | Unsupported | Unsupported |
| #pragma omp task | Unsupported | Unsupported |
| #pragma omp taskyield | Unsupported | Unsupported |
| #pragma omp taskwait | Unsupported | Unsupported |
For more information, see Supported IBM pragmas by different invocations.
Different built-in function names between xlc/xlC and xlclang/xlclang++
The names of the built-in functions supported by IBM XL C/C++ for AIX 16.1 that is
invoked by xlclang/xlclang++
are in the form of __builtin_name while the names supported by
xlc/xlC are in the form of __name. For example, you should use
__builtin_addex when invoking xlclang/xlclang++
while __addex when invoking xlc/xlC. Note that
names of the vector built-in
functions are the same across two front ends.
Unsupported GCC atomic memory access built-in functions by xlclang/xlclang++
IBM XL C/C++ for AIX 16.1 that is invoked by xlclang or xlclang++ does not support the following GCC atomic memory access built-in functions. Use the C++11 atomics features instead.
- Atomic fetch and operation functions
__sync_fetch_and_and__sync_fetch_and_nand__sync_fetch_and_or__sync_fetch_and_xor__sync_fetch_and_add__sync_fetch_and_sub
- Atomic operation and fetch functions
__sync_and_and_fetch__sync_nand_and_fetch__sync_or_and_fetch__sync_xor_and_fetch__sync_add_and_fetch__sync_sub_and_fetch
- Atomic compare and swap functions
__sync_val_compare_and_swap__sync_bool_compare_and_swap
Unsupported decimal floating-point built-in functions by xlclang/xlclang++
IBM XL C/C++ for AIX 16.1 that is invoked by xlclang or xlclang++ does not support decimal floating-point built-in functions.
You can find these built-in functions at Decimal floating-point built-in functions.Changed usage for utility makeC++SharedLib and linkxlC
If you link binaries generated by the Clang-based front end of IBM XL C/C++ for AIX by using makeC++SharedLib or linkxlC, you must also specify -lc++ manually, in addition to the parameters you need, to make the makeC++SharedLib and linkxlC utilities work properly. If you link binaries generated by the XL-based front end, such requirement does not apply.