Re: OptimizeExpression
- To: mathgroup at smc.vnet.net
- Subject: [mg86193] Re: OptimizeExpression
- From: dh <dh at metrohm.ch>
- Date: Wed, 5 Mar 2008 03:41:14 -0500 (EST)
- References: <fqisvi$mct$1@smc.vnet.net>
Hi Evan,
the problem your are facing is to prevent evaluation of a semicolon
separated list.
I see two approaches to this, either you replace the semicolons by
commas or you wrap the expression in Hold[..]. As C also uses
semicolons, we will try the second approach. By ReplaceRepeated you can
extract the lines of code and wrap them by Hold. Assume t1 is your
expression Experimental`OptimizeExpression[...], then:
t2=t1/.{HoldPattern[_[Block[{__}, x___]]]->Hold[x]}
this gives you the lines of code wrapped in Hold. To get C code:
t2 // CForm
now you only need to remove Hold.
hope this helps, Daniel
Evan Drumwright wrote:
> Hi everyone,
>
> I'm trying to send a very large (200+MB) expression to a C file using CForm, which, as you might imagine, crashes my compiler. From other discussions on this list, I have seen that you can use the OptimizeExpression[] function to eliminate common subexpressions, and I think that this will work. However, I don't understand how to process the resulting compound expression. For instance:
>
>
> Experimental`OptimizeExpression[(3+3*a^2 + Sqrt[5 + 6*a + 5*a^2] + a*(4 + Sqrt[5 + 6*a + 5*a^2]))/6]
>
> generates
>
> Out[1]= Experimental`OptimizedExpression[Block[{$$11, $$13, $$14, $$15,
>
> 2
>> $$16}, $$11 = a ; $$13 = 6 a; $$14 = 5 $$11; $$15 = 5 + $$13 + $$14;
>
> 3 + 3 $$11 + $$16 + a (4 + $$16)
>> $$16 = Sqrt[$$15]; --------------------------------]]
> 6
>
> Now, I'd like to get out the individual set of statements:
>
> $$11 = a;
> $$13 = 6a;
> $$14 = 5 $$11;
> $$15 = 5 + $$13 + $$14;
> $$16 = Sqrt[$$15];
> (3 + 3$$11 + $$16 + a(4+$$16)/6
>
>
> But, if I were to look at the function in the optimized expression, I get the original function:
>
> In[2]:= %[[1]]
>
> 2 2 2
> 3 + 3 a + Sqrt[5 + 6 a + 5 a ] + a (4 + Sqrt[5 + 6 a + 5 a ])
> Out[2]= --------------------------------------------------------------
> 6
>
>
> I suppose part of the difficulty is that I don't see a way to treat a CompoundExpression as a list of statements to be CForm'd. Can anyone tell me how I'd do this?
>
>
> Thanks,
> Evan
>
>