Re: OpenACC
* Nathan Sidwell: > Targeting PTX, an ISA for use with a single manufacturer's devices, is > not different from targeting the other single-manufacturer ISAs that > GCC already supports. The GCC licensing exception explicitly permits targeting virtual instruction sets, which (to me at least) strongly suggets that even the FSF is not opposed to such a technology.
one idea for next GSoC: libstdc++ locale support based on POSIX 2008 duplocale/uselocale, etc.
Hi. Here is one idea for GSoC: Implement C++ locale support in libstdc++ based on POSIX 2008 uselocale()/duplocale() facilities. This should bring C++ locale support, comparable to that of Glibc platforms, to Darwin and FreeBSD (at least), instead of the current "C" locale only support. -- VZ signature.asc Description: OpenPGP digital signature
Re: one idea for next GSoC: libstdc++ locale support based on POSIX 2008 duplocale/uselocale, etc.
On Sun, Nov 24, 2013 at 11:52:49PM +0100, Václav Zeman wrote: > Here is one idea for GSoC: Implement C++ locale support in libstdc++ > based on POSIX 2008 uselocale()/duplocale() facilities. Doesn't glibc do that since 2002? Jakub
Re: one idea for next GSoC: libstdc++ locale support based on POSIX 2008 duplocale/uselocale, etc.
On 11/25/2013 12:55 AM, Jakub Jelinek wrote: > On Sun, Nov 24, 2013 at 11:52:49PM +0100, Václav Zeman wrote: >> Here is one idea for GSoC: Implement C++ locale support in libstdc++ >> based on POSIX 2008 uselocale()/duplocale() facilities. > > Doesn't glibc do that since 2002? Glibc does seem to have the locale facilities mentioned above. Since when does it have the facilities? I do not know. Are they used in libstdc++'s `std::locale` implementation? Yes and no. The 'gnu' style (`--enable-clocale=gnu` configure option[1]) implementation uses the POSIX 2008 functions but it also uses `typedef __locale_t __c_locale`, i.e., the Glibc internal type instead of `locale_t`. The implementation in `libstdc++-v3/config/locale/gnu` is also riddled with Glibc version checking preprocessor directives. What I am proposing is that somebody takes the pure POSIX 2008 API and implements `std::locale` on top of that. [1]: http://gcc.gnu.org/onlinedocs/gcc-4.8.2/libstdc++/manual/manual/configure.html signature.asc Description: OpenPGP digital signature
Re: OpenACC
On 11/21/13 14:42, Jerome Glisse wrote: What worries me is that no one is thinking about how to bundle the end result ie do you add a new elf section that has ptx code that can then be lower at runtime and also provide fallback CPU code for all those function so that program can start running with the CPU code and start using the GPU code as soon as the runtime is done generating the final GPU code. We have an internal design for this, but I think such review is best left to when patches are available. thank you. nathan -- Nathan Sidwell
