From: Hans-Bernhard Broeker Newsgroups: comp.os.msdos.djgpp Subject: Re: DJGPP Compiler with output for x86-64 Date: 20 Feb 2001 14:42:07 GMT Organization: Aachen University of Technology (RWTH) Lines: 36 Message-ID: <96tvnv$plu$1@nets3.rz.RWTH-Aachen.DE> References: <96r96h$q76$1 AT nets3 DOT rz DOT RWTH-Aachen DOT DE> <3A89A375 DOT CC7D883C AT t-online DOT de> NNTP-Posting-Host: acp3bf.physik.rwth-aachen.de X-Trace: nets3.rz.RWTH-Aachen.DE 982680127 26302 137.226.32.75 (20 Feb 2001 14:42:07 GMT) X-Complaints-To: abuse AT rwth-aachen DOT de NNTP-Posting-Date: 20 Feb 2001 14:42:07 GMT Originator: broeker@ To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp Reply-To: djgpp AT delorie DOT com Kai Dietrich wrote: > There is an example how to realize MMX support with compiled > languages: FreePascal. It had to extend the Pascal language, but IMO > these extensions are not that worse. Yes, for a language like Pascal that essentially has no accepted standard in the first place --- and ANSI/ISO Pascal really is seen as a joke, by most Pascal programmers --- you can easily modify and extend the core of the language itself. But C is a language where the ANSI/ISO C standard does play a rather important role in existing compilers, and even more so for a multi-platform compiler suite like GCC. It doesn't even support "quasi-standard"s from the PC industry, like 'near', 'far' and 'huge', and for very good reasons, too. [...] > When MMX support is on, you aren't allowed to > do floating point arithmetic. Which means you're *not* just extending the language, you're also restricting it in very fundamental ways. > IMHO this is a good solution for implementing MMX, SSE, 3DNow!, > AltiVec, ... Support to the C language. Is there a way to use MMX > with DJGPP without using asm? None that I would have heard of. And I'd be quite surprised to hear of any. To outfit C with MMX/3DNow! suppport without tearing the very language to pieces, the only way practicable would be preprocessor macros hiding inline assembly chunks. SSE and AltiVec could be implemented more straightforwardly, as they don't effectively disable the FPU while they're in use, AFAIK. -- Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de) Even if all the snow were burnt, ashes would remain.