On 8/20/2012 11:29, jhu wrote:
> In the ARM world, everything but the main integer execution core is optional.

In part, this is because the ARM is a core, while the x86 machines are chips.

A "core" is a hardware design that you integrate onto a die along with the 
other stuff you want. So a core vs chip is like a chip vs a board. As far as 
I know, ARM does not actually sell chips themselves, only VHDL 
specifications for chips.

I'm wondering why nobody is building Raspberry.pi like devices from these 
http://opencores.org/

-- 
Darren New, San Diego CA, USA (PST)
   "They're the 1-800-#-GORILA of the telecom business."

Post a reply to this message

Am 18.08.2012 00:45, schrieb clipka:

> That's all nice to read, but /something/ must be going wrong, and I
> still suspect the floating point arithmetics (if only because of the
> difference that openEXR makes).
>
> I've just ordered one of those tiny little brats myself to further
> investigate the issue. Don't hold your breath though - delivery is
> estimated to take 15 weeks.

Here's what I get /with/ OpenEXR enabled. Note that it's exactly what 
Dave reported /without/ OpenEXR.

I swear I'm gonna find out what's going on here...

Post a reply to this message

Attachments:
Download 'biscuit.png' (595 KB)

Preview of image 'biscuit.png'

Am 12.12.2012 20:30, schrieb clipka:

> I swear I'm gonna find out what's going on here...

Apparently there's something fishy happening with the phong highlighting.

Post a reply to this message

Am 12.12.2012 20:56, schrieb clipka:
> Am 12.12.2012 20:30, schrieb clipka:
>
>> I swear I'm gonna find out what's going on here...
>
> Apparently there's something fishy happening with the phong highlighting.

Further investigations suggest that the Pi isn't happy with denormalized 
values, at least not in single precision.

I'll try to throw together some simple C test program, and if that 
confirms my observations I'll file it as a bug to the Pi people.

Post a reply to this message

Am 12.12.2012 22:03, schrieb clipka:
> Am 12.12.2012 20:56, schrieb clipka:
>> Am 12.12.2012 20:30, schrieb clipka:
>>
>>> I swear I'm gonna find out what's going on here...
>>
>> Apparently there's something fishy happening with the phong highlighting.
>
> Further investigations suggest that the Pi isn't happy with denormalized
> values, at least not in single precision.
>
> I'll try to throw together some simple C test program, and if that
> confirms my observations I'll file it as a bug to the Pi people.

Apparently the problems are related to the -ffast-math compiler option; 
I'll try whether turning that off will help.

Post a reply to this message

Am 13.12.2012 00:24, schrieb clipka:

> Apparently the problems are related to the -ffast-math compiler option;
> I'll try whether turning that off will help.

Well, no, not really. No success, except that now the color of the 
artifacts is inverted.

Post a reply to this message

Attachments:
Download 'biscuit.png' (164 KB)

Preview of image 'biscuit.png'

Am 13.12.2012 08:52, schrieb clipka:
> Am 13.12.2012 00:24, schrieb clipka:
>
>> Apparently the problems are related to the -ffast-math compiler option;
>> I'll try whether turning that off will help.
>
> Well, no, not really. No success, except that now the color of the
> artifacts is inverted.

Turning off /all/ optimizations does give the expected results. I don't 
consider this an option though, so I guess we'll have to wait for 
feedback from the raspbian team.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 13.12.2012 08:52, schrieb clipka:
> > Am 13.12.2012 00:24, schrieb clipka:
> >
> >> Apparently the problems are related to the -ffast-math compiler option;
> >> I'll try whether turning that off will help.
> >
> > Well, no, not really. No success, except that now the color of the
> > artifacts is inverted.
>
> Turning off /all/ optimizations does give the expected results. I don't
> consider this an option though, so I guess we'll have to wait for
> feedback from the raspbian team.

I just compiled 3.6 on the Raspberry Pi and it has the same issues.

Post a reply to this message

Am 01.02.2013 17:14, schrieb calhjh:
> clipka <ano### [at] anonymousorg> wrote:
>> Am 13.12.2012 08:52, schrieb clipka:
>>> Am 13.12.2012 00:24, schrieb clipka:
>>>
>>>> Apparently the problems are related to the -ffast-math compiler option;
>>>> I'll try whether turning that off will help.
>>>
>>> Well, no, not really. No success, except that now the color of the
>>> artifacts is inverted.
>>
>> Turning off /all/ optimizations does give the expected results. I don't
>> consider this an option though, so I guess we'll have to wait for
>> feedback from the raspbian team.
>
> I just compiled 3.6 on the Raspberry Pi and it has the same issues.

I would have expected that. The Pi seems to have a general problem with 
denormalized (i.e. very small) single-precision floating point values.

Post a reply to this message

Am 13.12.2012 14:48, schrieb clipka:

> Turning off /all/ optimizations does give the expected results. I don't
> consider this an option though, so I guess we'll have to wait for
> feedback from the raspbian team.

I've received feedback at last: The problem /may/ be solved with 
gcc-4.7; I'm currently recompiling POV-Ray on my Pi.

Raspbian Wheezy's standard gcc package is still at 4.6, so I had to 
manually install the gcc-4.7 and g++-4.7 packages, and force the 
./configure script to actually use gcc-4.7 via the "CXX=gcc-4.7" option.

We shall see in a couple of hours whether that actually solves the issue.

Post a reply to this message