Difference between revisions of "Porting/Macros"

From RCS Wiki
Jump to navigation Jump to search
m
(C/C++ preprocessor macros for PPC64 porting)
 
(7 intermediate revisions by the same user not shown)
Line 12: Line 12:
 
#endif
 
#endif
 
</pre>
 
</pre>
 +
 +
Details discussed on #talos-workstation:
 +
 +
Apple's gcc (4.0.1) in MacOS 10.5, ppc32 headers define:
 +
 +
<pre>__POWERPC__ and __ppc__</pre>
 +
 +
with -m64, the following are defined:
 +
 +
<pre>__POWERPC__, __ppc__, and __ppc64__</pre>
 +
 +
GCC compiling for the Linux kernel defines, for ppc32:
 +
 +
<pre>__PPC__ and __powerpc__</pre> [https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config/rs6000/linux.h;h=8dc2a89684a29ff05f6f5a17a280eb29129db229;hb=HEAD#l57 gcc/config/rs6000/linux.h]
 +
 +
and for ppc64:
 +
 +
<pre>__PPC__, __powerpc__, __PPC64__ and __powerpc64__</pre> [https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config/rs6000/linux64.h;h=98b7255c95f1b55ec5c97726cfcb9dc6fb9ebf56;hb=HEAD#l293 gcc/config/rs6000/linux64.h]
 +
 +
''Example from Debian ppc64le package: libtbb-dev, version 2020.3-1, /usr/include/tbb/machine/gcc_arm.h:''
 +
 +
<pre>
 +
#if __powerpc64__ || __ppc64__
 +
    // IBM XL documents __powerpc64__ (and __PPC64__).
 +
    // Apple documents __ppc64__ (with __ppc__ only on 32-bit).
 +
    #define __TBB_WORDSIZE 8
 +
#else
 +
    #define __TBB_WORDSIZE 4
 +
#endif
 +
</pre>
 +
 +
''Example from GCC/libstdc++-v3 source code: libstdc++-v3/src/c++17/fast_float/fast_float.h''
 +
 +
<pre>
 +
[...] || (defined(__ppc64__) || defined(__PPC64__) || defined(__ppc64le__) || defined(__PPC64LE__)) )
 +
</pre>
 +
 +
[https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libstdc%2B%2B-v3/src/c%2B%2B17/fast_float/fast_float.h;h=7551c4f89ef7b89c638995852e086109c1f43b4f;hb=HEAD#l101 libstdc++-v3 fast_float.h]
 +
 +
That refers to "ppc64le" variants, but those are not canonical.  It's better to check endianess separately.
 +
 +
[[Category:Ports]]

Latest revision as of 09:54, 14 September 2023

Some projects have been ported to Apple ppc64. In that case, they may test macros like this:

#if defined(__ppc64__)
#endif

GCC does not define __ppc64__ (lowercase). It defines __PPC64__ (uppercase). To make the above code also build on Debian ppc64le or ppc64 (big endian) change the above code to:

#if defined(__ppc64__) || defined(__PPC64__)
#endif

Details discussed on #talos-workstation:

Apple's gcc (4.0.1) in MacOS 10.5, ppc32 headers define:

__POWERPC__ and __ppc__

with -m64, the following are defined:

__POWERPC__, __ppc__, and __ppc64__

GCC compiling for the Linux kernel defines, for ppc32:

__PPC__ and __powerpc__

gcc/config/rs6000/linux.h

and for ppc64:

__PPC__, __powerpc__, __PPC64__ and __powerpc64__

gcc/config/rs6000/linux64.h

Example from Debian ppc64le package: libtbb-dev, version 2020.3-1, /usr/include/tbb/machine/gcc_arm.h:

#if __powerpc64__ || __ppc64__
    // IBM XL documents __powerpc64__ (and __PPC64__).
    // Apple documents __ppc64__ (with __ppc__ only on 32-bit).
    #define __TBB_WORDSIZE 8
#else
    #define __TBB_WORDSIZE 4
#endif

Example from GCC/libstdc++-v3 source code: libstdc++-v3/src/c++17/fast_float/fast_float.h

[...] || (defined(__ppc64__) || defined(__PPC64__) || defined(__ppc64le__) || defined(__PPC64LE__)) )

libstdc++-v3 fast_float.h

That refers to "ppc64le" variants, but those are not canonical. It's better to check endianess separately.