Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
Thanks Ivan! That does work. Pixel perfect.
Ivan Busquets wrote:
I'm not sure that Posterize would give the desired result in this
case, since it will still throw values that are not possible to
represent as a half precision float.
You'd probably want to re-create a 16 bit half-float value by breaking
the 32 bit float into its exponent and mantissa(significant) values,
and then truncating those.
The following expression might work (not thoroughly tested, though)
set cut_paste_input [stack 0]
version 7.0 v4
push $cut_paste_input
add_layer {alpha alpha.red alpha.beta}
Expression {
expr0 "exponent(r) > 16 ? r*inf: ldexp(rint(mantissa(r)*
(2**11))/(2**11), exponent(r))"
expr1 "exponent(g) > 16 ? g*inf: ldexp(rint(mantissa(g)*
(2**11))/(2**11), exponent(g))"
expr2 "exponent(b) > 16 ? b*inf: ldexp(rint(mantissa(b)*
(2**11))/(2**11), exponent(b))"
channel3 alpha
name Float_To_Half
selected true
xpos 292
ypos 226
}
Hope that helps.
PS. Mark, hope you're doing well! :)
On Sun, Feb 17, 2013 at 5:35 AM, Shailendra Pandey > wrote:
well actually 281474976710656 -1
281474976710655
Hope that helps
On Sun, Feb 17, 2013 at 9:25 PM, Shailendra Pandey
mailto:shail...@gmail.com>> wrote:
Hi Mark
You can use a posterize node
with a value of 281474976710656
which is 2 to the power(16*3)
Cheers
Shail
On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton
mailto:mnettle...@ilm.com>> wrote:
I'm generating an ST map within Nuke, that needs to line
up with 16 bit half float ST map images on disk.
Is there a way I can generate 16bit half float values
within nuke? Or convert 32 bit values to 16 bit half?
(without writing to disk and reading back)
Thanks
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk
,
http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk
,
http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
Good Stuff!
thanks for that Ivan :D
On Mon, Feb 18, 2013 at 1:04 PM, Michael Garrett wrote:
> add_layer {alpha alpha.red alpha.beta}
>
> nice
>
>
> On 17 February 2013 13:55, Ivan Busquets wrote:
>
>> I'm not sure that Posterize would give the desired result in this case,
>> since it will still throw values that are not possible to represent as a
>> half precision float.
>>
>> You'd probably want to re-create a 16 bit half-float value by breaking
>> the 32 bit float into its exponent and mantissa(significant) values, and
>> then truncating those.
>>
>> The following expression might work (not thoroughly tested, though)
>>
>> set cut_paste_input [stack 0]
>> version 7.0 v4
>> push $cut_paste_input
>> add_layer {alpha alpha.red alpha.beta}
>> Expression {
>> expr0 "exponent(r) > 16 ? r*inf: ldexp(rint(mantissa(r)*
>> (2**11))/(2**11), exponent(r))"
>> expr1 "exponent(g) > 16 ? g*inf: ldexp(rint(mantissa(g)*
>> (2**11))/(2**11), exponent(g))"
>> expr2 "exponent(b) > 16 ? b*inf: ldexp(rint(mantissa(b)*
>> (2**11))/(2**11), exponent(b))"
>> channel3 alpha
>> name Float_To_Half
>> selected true
>> xpos 292
>> ypos 226
>> }
>>
>> Hope that helps.
>>
>> PS. Mark, hope you're doing well! :)
>>
>>
>>
>>
>> On Sun, Feb 17, 2013 at 5:35 AM, Shailendra Pandey wrote:
>>
>>> well actually 281474976710656 -1
>>> 281474976710655
>>> Hope that helps
>>>
>>>
>>>
>>> On Sun, Feb 17, 2013 at 9:25 PM, Shailendra Pandey
>>> wrote:
>>>
Hi Mark
You can use a posterize node
with a value of 281474976710656
which is 2 to the power(16*3)
Cheers
Shail
On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton wrote:
> **
> I'm generating an ST map within Nuke, that needs to line up with 16
> bit half float ST map images on disk.
>
> Is there a way I can generate 16bit half float values within nuke? Or
> convert 32 bit values to 16 bit half? (without writing to disk and reading
> back)
>
> Thanks
>
> ___
> Nuke-users mailing list
> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
>>>
>>> ___
>>> Nuke-users mailing list
>>> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
>>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>>
>>
>>
>> ___
>> Nuke-users mailing list
>> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>
>
> ___
> Nuke-users mailing list
> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
add_layer {alpha alpha.red alpha.beta}
nice
On 17 February 2013 13:55, Ivan Busquets wrote:
> I'm not sure that Posterize would give the desired result in this case,
> since it will still throw values that are not possible to represent as a
> half precision float.
>
> You'd probably want to re-create a 16 bit half-float value by breaking the
> 32 bit float into its exponent and mantissa(significant) values, and then
> truncating those.
>
> The following expression might work (not thoroughly tested, though)
>
> set cut_paste_input [stack 0]
> version 7.0 v4
> push $cut_paste_input
> add_layer {alpha alpha.red alpha.beta}
> Expression {
> expr0 "exponent(r) > 16 ? r*inf: ldexp(rint(mantissa(r)*
> (2**11))/(2**11), exponent(r))"
> expr1 "exponent(g) > 16 ? g*inf: ldexp(rint(mantissa(g)*
> (2**11))/(2**11), exponent(g))"
> expr2 "exponent(b) > 16 ? b*inf: ldexp(rint(mantissa(b)*
> (2**11))/(2**11), exponent(b))"
> channel3 alpha
> name Float_To_Half
> selected true
> xpos 292
> ypos 226
> }
>
> Hope that helps.
>
> PS. Mark, hope you're doing well! :)
>
>
>
>
> On Sun, Feb 17, 2013 at 5:35 AM, Shailendra Pandey wrote:
>
>> well actually 281474976710656 -1
>> 281474976710655
>> Hope that helps
>>
>>
>>
>> On Sun, Feb 17, 2013 at 9:25 PM, Shailendra Pandey wrote:
>>
>>> Hi Mark
>>>
>>> You can use a posterize node
>>> with a value of 281474976710656
>>> which is 2 to the power(16*3)
>>>
>>>
>>>
>>> Cheers
>>> Shail
>>>
>>> On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton wrote:
>>>
**
I'm generating an ST map within Nuke, that needs to line up with 16 bit
half float ST map images on disk.
Is there a way I can generate 16bit half float values within nuke? Or
convert 32 bit values to 16 bit half? (without writing to disk and reading
back)
Thanks
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>>
>>>
>>
>> ___
>> Nuke-users mailing list
>> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>
>
> ___
> Nuke-users mailing list
> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
I'm not sure that Posterize would give the desired result in this case,
since it will still throw values that are not possible to represent as a
half precision float.
You'd probably want to re-create a 16 bit half-float value by breaking the
32 bit float into its exponent and mantissa(significant) values, and then
truncating those.
The following expression might work (not thoroughly tested, though)
set cut_paste_input [stack 0]
version 7.0 v4
push $cut_paste_input
add_layer {alpha alpha.red alpha.beta}
Expression {
expr0 "exponent(r) > 16 ? r*inf: ldexp(rint(mantissa(r)* (2**11))/(2**11),
exponent(r))"
expr1 "exponent(g) > 16 ? g*inf: ldexp(rint(mantissa(g)* (2**11))/(2**11),
exponent(g))"
expr2 "exponent(b) > 16 ? b*inf: ldexp(rint(mantissa(b)* (2**11))/(2**11),
exponent(b))"
channel3 alpha
name Float_To_Half
selected true
xpos 292
ypos 226
}
Hope that helps.
PS. Mark, hope you're doing well! :)
On Sun, Feb 17, 2013 at 5:35 AM, Shailendra Pandey wrote:
> well actually 281474976710656 -1
> 281474976710655
> Hope that helps
>
>
>
> On Sun, Feb 17, 2013 at 9:25 PM, Shailendra Pandey wrote:
>
>> Hi Mark
>>
>> You can use a posterize node
>> with a value of 281474976710656
>> which is 2 to the power(16*3)
>>
>>
>>
>> Cheers
>> Shail
>>
>> On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton wrote:
>>
>>> **
>>> I'm generating an ST map within Nuke, that needs to line up with 16 bit
>>> half float ST map images on disk.
>>>
>>> Is there a way I can generate 16bit half float values within nuke? Or
>>> convert 32 bit values to 16 bit half? (without writing to disk and reading
>>> back)
>>>
>>> Thanks
>>>
>>> ___
>>> Nuke-users mailing list
>>> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
>>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>>
>>
>>
>
> ___
> Nuke-users mailing list
> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
___
Nuke-users mailing list
Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
well actually 281474976710656 -1 281474976710655 Hope that helps On Sun, Feb 17, 2013 at 9:25 PM, Shailendra Pandey wrote: > Hi Mark > > You can use a posterize node > with a value of 281474976710656 > which is 2 to the power(16*3) > > > > Cheers > Shail > > On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton wrote: > >> ** >> I'm generating an ST map within Nuke, that needs to line up with 16 bit >> half float ST map images on disk. >> >> Is there a way I can generate 16bit half float values within nuke? Or >> convert 32 bit values to 16 bit half? (without writing to disk and reading >> back) >> >> Thanks >> >> ___ >> Nuke-users mailing list >> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> > > ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] 32 bit float -> 16 bit half float expression?
Hi Mark You can use a posterize node with a value of 281474976710656 which is 2 to the power(16*3) Cheers Shail On Sun, Feb 17, 2013 at 6:25 AM, Mark Nettleton wrote: > ** > I'm generating an ST map within Nuke, that needs to line up with 16 bit > half float ST map images on disk. > > Is there a way I can generate 16bit half float values within nuke? Or > convert 32 bit values to 16 bit half? (without writing to disk and reading > back) > > Thanks > > ___ > Nuke-users mailing list > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
[Nuke-users] 32 bit float -> 16 bit half float expression?
I'm generating an ST map within Nuke, that needs to line up with 16 bit half float ST map images on disk. Is there a way I can generate 16bit half float values within nuke? Or convert 32 bit values to 16 bit half? (without writing to disk and reading back) Thanks ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
