I do. Even then, there are people who use 32bit machine, there are BSDers who use 64bit machine without any 32bit emulation, there are people who use ARM, etc. In most practical tasks you need to use external libraries, but most of those are not written in ASM, and so way to call them in Asm differs with OS calling standard, even on same processor.

These are some practical examples where assembly is inferior to HLLs. If you want your code to be usable for many people out there, you'll have to duplicate A LOT.

I say that should be their problem.

I give off my sources so no one can complain. if they can make it more readable, they're more than welcome. Why is so hard to make it readable? Because I don't know how they think, for me asm and C are easy languages -- others prefer Pascal, etc.. I can't get into their heads so I leave it to them

For me, C has advantages over asm, especially in design phase, or when testing new algorithms (but I still convert every C instruction to asm before, when I think up the algorithm, to know if it's efficient or not prematurely. Of course nothing beats a profiler, but that's just a preview).

That said, they don't exclude each other (C and asm), but complement each other. And assembly is very underrated IMO, especially when it comes to 'difficult to maintain/read/debug' part) It's not hard if you get used to it, but then again, I think everyone who is starting with asm will find it hard, just as they find their first languages very hard and full of bugs

I find it very very very hard to imagine that anyone thinks HLLs in general output good code. (They don't.) Also, not every platform has the latest GCC either. Plus some compilers (Intel) aren't free. And I certainly think a lot of 100% C code could be sped up with little asm bits here and there. Doesn't mean I'm against all HLLs, but for sure they aren't as utopian as some would have you believe. (And is it just me or is GCC kinda slow? Meh.)

rugxulo: HLLs/compilers aren't the end-all-be-all magic silver bullet, the trick is being "good enough" and "comfortable enough to work with" - YMMV might vary wrt. just how good is good enough, but I don't think anybody claims that you can't usually beat a compiler.

And yeah, GCC is terribly slow, at least the versions I've tried. And for MSVC, there's also a noticeable speed hit betwen compiling .c and .cpp files (ie., compiling the same source in C or C++ mode).

Uh ... yeah, sure, this makes perfect sense.

But GCC is already pretty darn good, IMO. I just wish it was even better still (more cpu scheduling, better compile speed, easier to bootstrap) and maybe a bit less weirdly-designed!

As long as this is about porting fasm to different arch, I hope that one day I can see fasm being used in native GPU (ATI/Nvidia) programming.

Am not holding my breath though.

Don't just sit back and wait for it to happen. Sit up and start programming it, that is how things will happen.

penang, before Googling maybe you would like to see this first: http://board.flatassembler.net/topic.php?t=3702

Thanks for the links. I'll soak them up.

In the meantime, the French are doing something about it .....

http://www.google.com/translate?u=http%3A%2F%2Fwww.pcinpact.com%2Factu%2Fnews%2F43165-premier-supercalculateur-GPU-France-Tesla.htm&langpair=fr%7Cen&hl=en&ie=UTF8

For those who understand French ...

http://www.pcinpact.com/actu/news/43165-premier-supercalculateur-GPU-France-Tesla.htm?ca=&vc=1&p=3&#c1596463