World: r3wp

Nice. So you will not postpone it?

Nope, seems that a ELF support large enough to run the demo script 
could be doable tomorrow or in the next days.

Btw, Red/System is now able to do syscalls directly, bypassing the 
libc dependency.

We'll see if this can hold long enough to avoid calling the libc 
at all.

Will that be optional? The libc interface is actually much more portable 
than kernel interfaces

yes

libc support will come as soon as we get support for a loader done

but syscall support is nice (and lean) on it's own :)

It is

So can't Red load dynamic libraries yet?

not on linux/elf, no

Ah

Kaj: I had a quick look at *BSD syscalls, there are pretty much the 
same as Linux (at least for the most used ones). The calling convention 
is different thought, but can be supported by one or two more lines 
of code in Red/System emitter.

When I see how Carl fights with various libc versions...I wonder 
if syscalls are not an easier path even if it can have some minor 
porting costs here and there? Anyway, a libc mapping will be possible 
soon, so all options are open.

I'm not sure the current R3 versions for different GLibC versions 
make any difference. It's R2 that has problems, but usually with 
the higher level X libraries

The syscall calling convention on Syllable is also different. It 
doesn't use SYSENTER/SYSEXIT yet

Red/System uses currently software interrupts for making syscalls, 
I'll add support for SYSENTER/SYSEXIT in the next days.

When I look at R2/Core downloads list, it stills has to ship in two 
flavors for Linux, isn't it because of libc issues?

Yes, but R3 is much better isolated

Anyway, we'll make sure Red runs ok on Syllable, that's the least 
I could do. ;-)

OK :-)

The issue R2 currently has with old Linux systems is with floating 
point. That's in the math library, part of GLibC, not the kernel, 
so there's no way around it unless you use a completely different 
implementation

I'm not denying the libc interface is problematic on GNU systems, 
I'm just saying there's probably no alternative with fewer problems 
:-(

The best strategy is to keep the interface as small as possible, 
like R3

That's more of a Red problem though. Red/System doesn't have floating 
point, just integers.

But it's always good to look to the future :)

Decimal Floating-Point Math Library: http://software.intel.com/en-us/blogs/2008/03/06/intel-decimal-floating-point-math-library/
That's interesting!

If the performances are close to classic floating point math operations, 
that would be a great alternative.

That is not to be expected, because this seems to be a software implementation

Decimal math would still be great to have

Still, the R2 problem with the math library is probably a fluke that 
can be prevented

On the one side you have a chaotic Linux ecosystem where most developers 
refuse to care about binary compatibility, on the other hand you 
have an isolated REBOL cave where Carl refuses to deal with outside 
complexity. This is a recipe for disaster. There must be some middle 
ground - as we have proven in Syllable

I was hoping that such library would use the FPU for low-level maths, 
but can't find any info about that.

I can't even find the download. Is it BSD?

Is it planned for the linker to support static libraries?

Static libs: yes, but not in the short-term (not in the next months, 
unless someone else than me implements it).

Download: that's strange, there was a download button under the title 
of the article, I've got it a few of hours ago, checking the licensing 
terms in the sources archive...

Here's the direct download link: ttp://software.intel.com/file/7242 
(~27MB)

sorry: http://software.intel.com/file/7242

It's 3 clauses BSD.

The library seems to be built upon IEEE 754 standard, so should be 
able to be accelerated by an FPU. Btw, here's the  download page: 
http://software.intel.com/en-us/articles/intel-decimal-floating-point-math-library/

Thanks. Sounds good, except for the size

This IEEE 754R spec is a revision based on IEEE 754, but it seems 
to me that the decimal implementation in this library is in software, 
awaiting new FPUs (IBM already has a few (although already is not 
really a good term for 2011))

Intel means it for compliance ahead of time. I wouldn't expect them 
to compete with the performance of the future FPUs they want to sell 
us

I agree that the library size is too big anyway for a direct inclusion 
in Red's core.

An optimising static linker could make it acceptable :-)

Hi, I added a Red to the Alternatives section in REBOL Wikipedia 
page.

Thanks Ladislav. Btw, could you add links for the alternative projects?

Sure I could, but they were removed by some "purists" telling, that 
the Wikipedia was not a link repository, so I am afraid it might 
be contraproductive.

So, that Intel library is for implementing Red's equivalent of the 
money! type?