Hello Miller and everyone on the list!
i have finally got around testing pd compiled to use double precision floats..
what i did was this:
pulled the sources from miller's git repo, in fact i didn't realise there was a master tree like this! (if i understand it correct it's the current version repository? the version that i can see in Help->About is reported to be 0.42.5 or may be it's .42.5+ ? please correct me on that ..)
so i just used env CFLAGS="-pipe -O2 -march=athlon64 -DPD_FLOATTYPE=double' ./configure
CFLAGS="-pipe -O2 -march=athlon64" is what i always use ..tought all versions of pd i have compiled before were compiled with this CFLAGS.
then make all whent fine, but failed to compile sigmund~, that's ok
as haven't used sigmund~ yet ..))
then i ran pd and tryied the testtone.pd ..and that's what was weired - it output noise+sine ..hm maybe to do with conversion ..i use RME Hammerfall card .. so there is conversion heppening somewhere ..
ok, made another patch, basica [osc~ 200]--[dac~ 1] and that gave DC into the soundcard. using number-[osc~] WORKED and number-[vline~]-[osc~] DID NOT..((
cannot tell if that was a pure sine for 100%, but somewhere most likely pure :)
so trying to use a phasor~ seems to be failing too ..
has anyone else tested this propertly?
as Miller says in CHANGES.txt: " 0.42.0
changed definition of t_float, t_sample, t_floatarg so that they can be set via #defines (PD_FLOATTYPE, etc). " ..hm what does "etc" mean?? maybe i need to define something extra?
or [osc~] is all we got working with doubles so far??
well, i checked again ..and it's actually not that near to a sinewave ..it's got some harmonics ..hm
i can test anything else ..but there is a general issue to be resolved. can anyone help? it would be very nice to get this working ..hopefully 64bit DAC's are not to far from the general public market :)
oh, i forgot to give some details: gcc (Gentoo 4.3.4 p1.1, pie-10.1.5) 4.3.4 so basically i am on gentoo, it's a current stable branch ..sort of
but i think it doesn't matter really..it's to do with general porting i reckon!
On 2010-03-17 16:33, errordeveloper@gmail.com wrote:
or [osc~] is all we got working with doubles so far??
i guess most of the double-precision code is still the one i wrote. which means, that the lookup-table generators (phasor~, cos~, osc~) are not implemented at all. soundfiles should work though.
masdrt IOhannes
ok ..so there should be no output issues, just the ugen's need to be implemented, rigth?
IOhannes, could you give a brief idea on what porting to double float is generally about? i suppose there should be some generic patern, right?
i would consider trying to crack on with this ..if it's just a boring job with a lot of testing etc.. as my experience is quite little so far.
also from what i can figure there is a little dilema - weather we should make all object to use any float size OR just work around what to do if the size is double ..hm, what if in a short time we would need bigger floats for some crazy reason..
On Wed, Mar 17, 2010 at 05:13:55PM +0100, IOhannes m zmoelnig wrote:
On 2010-03-17 16:33, errordeveloper@gmail.com wrote:
or [osc~] is all we got working with doubles so far??
i guess most of the double-precision code is still the one i wrote. which means, that the lookup-table generators (phasor~, cos~, osc~) are not implemented at all. soundfiles should work though.
masdrt IOhannes
Did you have any luck with this?
.hc
On Mar 19, 2010, at 6:49 PM, errordeveloper@gmail.com wrote:
ok ..so there should be no output issues, just the ugen's need to be implemented, rigth?
IOhannes, could you give a brief idea on what porting to double
float is generally about? i suppose there should be some generic patern, right?i would consider trying to crack on with this ..if it's just a boring job with a lot of testing etc.. as my experience is quite little so
far.also from what i can figure there is a little dilema - weather we
should make all object to use any float size OR just work around what to do
if the size is double ..hm, what if in a short time we would need bigger floats for some crazy reason..On Wed, Mar 17, 2010 at 05:13:55PM +0100, IOhannes m zmoelnig wrote:
On 2010-03-17 16:33, errordeveloper@gmail.com wrote:
or [osc~] is all we got working with doubles so far??
i guess most of the double-precision code is still the one i wrote. which means, that the lookup-table generators (phasor~, cos~, osc~)
are not implemented at all. soundfiles should work though.masdrt IOhannes
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
"Making boring techno music is really easy with modern tools, but with
live coding, boring techno is much harder." - Chris McCormick
-
errordeveloper＠gmail.com -
Hans-Christoph Steiner -
IOhannes m zmoelnig