Csound › Csound - Dev

i(k1 OP k2)

Classic

List

Threaded

10 messages Options

Tito Latini

i(k1 OP k2)

I have found a solution for the i() problem in csound list.
This is worth for the general case

ivar = FN(k1 OP k2)

This is what currently happens:

i(k1 `op` k2)

-> k0 op.kk k1 k2
i(k0)

but in entry.c, all OP.kk don't have a init function. For example:

{ "add.kk", S(AOP), 2, "k", "kk", NULL, addkk }

therefore k0 is 0 during the first init pass and fn(k0) fails.
The simplest solution is

{ "add.kk", S(AOP), 3, "k", "kk", addkk, addkk }

Idem for FN.k

In this way we can write

i(k1 + k3 * abs(k4 / k6) OP FN(k7 ...) ...)

This solution costs an operation during the init pass with the normal
k-rate operations.

Besides, after the changes, it'is simple to add this syntax

ivar = k1 OP k2 OP .. FN(...) ...

The alternative is to change `create_expression` in
csound_orc_expression.c, OP.ii and FN.i if the function is i(), or if
the left side of `=' is an i-variable, but it requires more changes.

tito

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Remove Ads

Steven Yi

Re: i(k1 OP k2)

This is interesting. I'm trying to imagine if there are any backwards
incompatibilities. A side effect of this is that it would seem then
that one wouldn't need to to use 'init' for k- and a-vars and could
use 'kval = 0' and have it be valid in the initial i-pass.

I'm wondering though, the init pass is a curious thing, especially
with if-then's. I'm wondering about something like:

kstate init 0

if (kstate == 0) then
...do something...
kstate = 1

else
kstate = 2

endif

if there is a change to do i-time assignment and expression changes,
wouldn't kstate = 2 after the init-pass? I'm thinking that k-rate
if-then's would have their opcodes still initialized in init pass.

On Sat, Jan 28, 2012 at 9:26 AM, Tito Latini <[hidden email]> wrote:

> I have found a solution for the i() problem in csound list.
> This is worth for the general case
>
> ivar = FN(k1 OP k2)
>
> This is what currently happens:
>
> i(k1 `op` k2)
>
> -> k0 op.kk k1 k2
> i(k0)
>
> but in entry.c, all OP.kk don't have a init function. For example:
>
> { "add.kk", S(AOP), 2, "k", "kk", NULL, addkk }
>
> therefore k0 is 0 during the first init pass and fn(k0) fails.
> The simplest solution is
>
> { "add.kk", S(AOP), 3, "k", "kk", addkk, addkk }
>
> Idem for FN.k
>
> In this way we can write
>
> i(k1 + k3 * abs(k4 / k6) OP FN(k7 ...) ...)
>
> This solution costs an operation during the init pass with the normal
> k-rate operations.
>
> Besides, after the changes, it'is simple to add this syntax
>
> ivar = k1 OP k2 OP .. FN(...) ...
>
> The alternative is to change `create_expression` in
> csound_orc_expression.c, OP.ii and FN.i if the function is i(), or if
> the left side of `=' is an i-variable, but it requires more changes.
>
> tito
>
> ------------------------------------------------------------------------------
> Try before you buy = See our experts in action!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-dev2
> _______________________________________________
> Csound-devel mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/csound-devel

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Tito Latini

Re: i(k1 OP k2)

After the `add.kk' change, with this example

<CsoundSynthesizer>
<CsInstruments>

instr 1
kstate init 1
if (kstate == 0) then
istate1 = i(kstate + kstate)
kstate = 10
else
kstate = 100
endif

istate2 = i(kstate + kstate)

print istate1
print istate2

endin
</CsInstruments>
<CsScore>
i1 0 1
</CsScore>
</CsoundSynthesizer>

csound test.csd
...
instr 1: istate1 = 2.000
instr 1: istate2 = 2.000
...

Wrong; istate1 initialized even if "kstate == 0".
No problems for k-vars. So, the initial idea doesn't work.

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Tito Latini

Re: i(k1 OP k2)

In reply to this post by Steven Yi

Ah no, it seems correct because

<CsoundSynthesizer>
<CsInstruments>

instr 1
kstate init 1
if (kstate == 0) then
istate1 = 1
kstate = 10
else
kstate = 100
endif

print istate1

endin
</CsInstruments>
<CsScore>
i1 0 1
</CsScore>
</CsoundSynthesizer>

csound test.csd
...
instr 1: istate1 = 1.000
...

In general, we have to also modify "=.k" (minit for init and krate)

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Steven Yi

Re: i(k1 OP k2)

I think this is still not correct. The tricky thing is that it seems
correct, but the problem is being masked. All of the opcodes in the
chain get init methods called if there are krate if-then's. The if
doesn't get performed until k-time (I believe). You're printing
istate1 but that's not the variable that's really important here, but
the kstate. If after i-pass, with your changes, I imagine kstate
would be 100, due to it having an init method now.

On Sat, Jan 28, 2012 at 10:47 AM, Tito Latini <[hidden email]> wrote:

> Ah no, it seems correct because
>
> <CsoundSynthesizer>
> <CsInstruments>
>
> instr 1
> kstate init 1
> if (kstate == 0) then
> istate1 = 1
> kstate = 10
> else
> kstate = 100
> endif
>
> print istate1
>
> endin
> </CsInstruments>
> <CsScore>
> i1 0 1
> </CsScore>
> </CsoundSynthesizer>
>
> csound test.csd
> ...
> instr 1: istate1 = 1.000
> ...
>
> In general, we have to also modify "=.k" (minit for init and krate)
>
> ------------------------------------------------------------------------------
> Try before you buy = See our experts in action!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-dev2
> _______________________________________________
> Csound-devel mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/csound-devel

jpff

Re: i(k1 OP k2)

In reply to this post by Tito Latini

I am very unhappy about this. it is a general rule that k-rate values are
not valued in the i-pass. It is true that gk variables MAY have a value
but that is not guaranteed. This is a semantic change that seems wrong
top me
==John

Tito Latini

Re: i(k1 OP k2)

Ok, I now have a doubt; is this behavior normal?
(code without changes)

kval init 1

if (kval == 0) then
ivar = 123
endif

print ivar

result:
...
instr 1: ivar = 123.000
...

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Steven Yi

Re: i(k1 OP k2)

Hi Tito,

Yes, it is normal, but certainly awkward. The general advice I give
is never to mix i-time code with k-time code when using if-then. In
practice, I think most people do this, so this scenario doesn't come
up often, but it's incredibly awkward when it does come up.

On the other hand, I need to review about when if-then's are evaluated
at i-time and when they are not. The manual (from
http://csounds.com/manual/html/if.html
) says:

Note
Note that if the condition uses a k-rate variable (for instance, “if
kval > 0”), the if...goto or if...then statement will be ignored
during the i-time pass. This allows for opcode initialization, even if
the k-rate variable has already been assigned an appropriate value by
an earlier init statement.

I had to do a strange the thing with a UDO I was working on a while
back using an i-rate if-then together with a k-rate if-then just to
initialize a bank of recursive UDO's and optionally have adjustable
recursion at performance time. It's not very transparent stuff
unfortunately.

On Sat, Jan 28, 2012 at 11:12 AM, Tito Latini <[hidden email]> wrote:

> Ok, I now have a doubt; is this behavior normal?
> (code without changes)
>
> kval init 1
>
> if (kval == 0) then
> ivar = 123
> endif
>
> print ivar
>
>
> result:
> ...
> instr 1: ivar = 123.000
> ...
>
> ------------------------------------------------------------------------------
> Try before you buy = See our experts in action!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-dev2
> _______________________________________________
> Csound-devel mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/csound-devel

Tito Latini

Re: i(k1 OP k2)

If-then uses ckgoto with k-condition and

kval init 1
ckgoto kval == 1, cont
ivar = 123
cont:
print ivar

result: ivar = 123.000

`cggoto' also fails

`goto' works:

kval init 1
goto cont
ivar = 123
cont:
print ivar

result: ivar = 0.000

`goto' works because it's without condition,
therefore I think that the problem is here (OOps/goto_ops.c)

int icgoto(CSOUND *csound, CGOTO *p)
{
if (*p->cond)
csound->ids = p->lblblk->prvi;
return OK;
}

int kcgoto(CSOUND *csound, CGOTO *p)
{
if (*p->cond)
csound->pds = p->lblblk->prvp;
return OK;
}

I now understand the difficulties of relinking with if-then.
When `create_goto_token' (csound_orc_expression.c) uses ckgoto,
if-then fails with the irate variables. It seems a bug to fix.

Returning to i() topic. Redefining "=.k" is wrong, I know it, it was
a first (wrong) reaction to the if-then problem. I have wanted only to
change the operations, like "add", "sub" etc. like the `goto' opcode

{ "goto", S(GOTO), 3, "", "l", igoto, kgoto }

To add two kvars

add k1 k2

doesn't change the state of `k1' or `k2'. The only sacrifice is
a calculation during the init pass. I have tried add.kk in the
modified version

{ "add.kk", S(AOP), 2, "k", "kk", addkk, addkk }

and I don't see problems yet; in

kfoo = k1 + k2

the value of kfoo is the value of k1 + the value of k2 to the same k-cycle.
The init pass is gone (=.k is krate). It works fine with i(). Probably
the idea is wrong, but I want to see my errors in the practice, so
I'll make other tests with the hope to find a more convincing solution.
Sorry for all this noise.

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel

Andres Cabrera

Re: i(k1 OP k2)

In reply to this post by jpff

Hi John,

I think this is good, and especially it should be consistent, otherwise it can be very confusing as it is now.

But I'm wondering, this would mean that many k-rate only opcodes (like invalue) would need to have a version added that works at init time.

In any case I think this should definitely be defined and fixed. It really makes doing complex work in Csound much harder and error prone than it could be. I don't mind if it is chosen to have all k-rate opcodes update values at initialization, or none of them.

Another option might be having some sort of syntax to run the initialization and a single control pass for an opcode and store the values in i-type variables. But this might be overkill.

Cheers,
Andrés

On Sat, Jan 28, 2012 at 4:03 PM, <[hidden email]> wrote:

I am very unhappy about this. it is a general rule that k-rate values are
not valued in the i-pass. It is true that gk variables MAY have a value
but that is not guaranteed. This is a semantic change that seems wrong
top me
==John

> Ah no, it seems correct because
>
> <CsoundSynthesizer>
> <CsInstruments>
>
> instr 1
> kstate init 1
> if (kstate == 0) then
> istate1 = 1
> kstate = 10
> else
> kstate = 100
> endif
>
> print istate1
>
> endin
> </CsInstruments>
> <CsScore>
> i1 0 1
> </CsScore>
> </CsoundSynthesizer>
>
> csound test.csd
> ...
> instr 1: istate1 = 1.000
> ...
>
> In general, we have to also modify "=.k" (minit for init and krate)
>
> ------------------------------------------------------------------------------
> Try before you buy = See our experts in action!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-dev2
> _______________________________________________
> Csound-devel mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>
>
>

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Csound-devel mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/csound-devel