Weird common subexpression simplification

[Dimas Abreu Dutra]

Hello all,

I'm using sympy for automatic code generation and differentiation. Before code generation I perform common subexpression simplification. I'm getting weird results in one of my expressions, though. I get a substitution for (_cse1, x0**3) and then in my expression I get _cse1**(2/3) instead of x0**2. Although both expressions are mathematically equivalent, numerically x0**2 is much more efficient and robust. Bellow is a code snippet of this particular expression.

>>> import sympy

>>> variables = sympy.var('x:10 hyperpar:10 u:10 u_dot:10 dt theta:10')

>>> e = 0.5*dt**2*hyperpar9**2*x1 + 0.5*dt**2*hyperpar9*theta0*x0**3 + 0.5*dt**2*hyperpar9*theta1*x0 - 0.5*dt**2*hyperpar9*u0 - 1.5*dt**2*theta0*x0**2*x1 - 0.5*dt**2*theta1*x1 + 0.5*dt**2*u_dot0 - dt*hyperpar9*x1 - dt*theta0*x0**3 - dt*theta1*x0 + dt*u0 + x1

>>> sympy.cse(e, sympy.numbered_symbols('_cse'))

([(_cse0, dt**2), (_cse1, x0**3), (_cse2, theta1*x0), (_cse3, 0.5*_cse0), (_cse4, _cse1*theta0)], [-1.5*_cse0*_cse1**(2/3)*theta0*x1 + _cse2*_cse3*hyperpar9 - _cse2*dt + _cse3*_cse4*hyperpar9 + _cse3*hyperpar9**2*x1 - _cse3*hyperpar9*u0 - _cse3*theta1*x1 + _cse3*u_dot0 - _cse4*dt - dt*hyperpar9*x1 + dt*u0 + x1])

Does anyone have any idea how to prevent such substitutions in cse?

Thanks in advance,

~Dimas

[Aaron Meurer]

Maybe it would be possible to do this using either the optimizations
or postprocess options to cse().

Aaron Meurer

> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To view this discussion on the web visit
> https://groups.google.com/d/msg/sympy/-/JfSiGL3LTMUJ.
> To post to this group, send email to sy...@googlegroups.com.
> To unsubscribe from this group, send email to
> sympy+un...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/sympy?hl=en.