I would like advice on what to do.

The current state is that in standard build moogladder is unchanged,
calls tan many times per sample and is slow.  If built with my
experimental macro it uses an approximation to tanh and is
significantly faster, with no sonic difference in the manual example.

There were objections to making the fast version standard; so options are

a) Continue with existing code
b) Replace by less accurate but faster version
c) Have a new opcode (moogladder2) for the fast approx
d) add an optional argument to select fast version (but actually a pita
   to code and loses some speed)
e) Use a lookup table and interpolation for tanh (I tried and it was
   not easy)
f) Something else.

Any views?

The moogladder2 is the best solution.

Victor Lazzarini
Dean of Arts, Celtic Studies, and Philosophy
Maynooth University
Ireland

> On 19 Apr 2016, at 20:23, jpff  wrote:
> 
> I would like advice on what to do.
> 
> The current state is that in standard build moogladder is unchanged,
> calls tan many times per sample and is slow.  If built with my
> experimental macro it uses an approximation to tanh and is
> significantly faster, with no sonic difference in the manual example.
> 
> There were objections to making the fast version standard; so options are
> 
> a) Continue with existing code
> b) Replace by less accurate but faster version
> c) Have a new opcode (moogladder2) for the fast approx
> d) add an optional argument to select fast version (but actually a pita
>   to code and loses some speed)
> e) Use a lookup table and interpolation for tanh (I tried and it was
>   not easy)
> f) Something else.
> 
> Any views?

+1 to moogladder2

On Tue, Apr 19, 2016 at 3:38 PM, Victor Lazzarini
 wrote:
> The moogladder2 is the best solution.
>
> Victor Lazzarini
> Dean of Arts, Celtic Studies, and Philosophy
> Maynooth University
> Ireland
>
>> On 19 Apr 2016, at 20:23, jpff  wrote:
>>
>> I would like advice on what to do.
>>
>> The current state is that in standard build moogladder is unchanged,
>> calls tan many times per sample and is slow.  If built with my
>> experimental macro it uses an approximation to tanh and is
>> significantly faster, with no sonic difference in the manual example.
>>
>> There were objections to making the fast version standard; so options are
>>
>> a) Continue with existing code
>> b) Replace by less accurate but faster version
>> c) Have a new opcode (moogladder2) for the fast approx
>> d) add an optional argument to select fast version (but actually a pita
>>   to code and loses some speed)
>> e) Use a lookup table and interpolation for tanh (I tried and it was
>>   not easy)
>> f) Something else.
>>
>> Any views?

Done, commited, documented

On Tue, 19 Apr 2016, Steven Yi wrote:

> +1 to moogladder2
>
> On Tue, Apr 19, 2016 at 3:38 PM, Victor Lazzarini
>  wrote:
>> The moogladder2 is the best solution.
>>
>> Victor Lazzarini
>> Dean of Arts, Celtic Studies, and Philosophy
>> Maynooth University
>> Ireland
>>
>>> On 19 Apr 2016, at 20:23, jpff  wrote:
>>>
>>> I would like advice on what to do.
>>>
>>> The current state is that in standard build moogladder is unchanged,
>>> calls tan many times per sample and is slow.  If built with my
>>> experimental macro it uses an approximation to tanh and is
>>> significantly faster, with no sonic difference in the manual example.
>>>
>>> There were objections to making the fast version standard; so options are
>>>
>>> a) Continue with existing code
>>> b) Replace by less accurate but faster version
>>> c) Have a new opcode (moogladder2) for the fast approx
>>> d) add an optional argument to select fast version (but actually a pita
>>>   to code and loses some speed)
>>> e) Use a lookup table and interpolation for tanh (I tried and it was
>>>   not easy)
>>> f) Something else.
>>>
>>> Any views?
>>> ==John ffitch