[LLVMdev] Additional Optimization I'm Missing?

[me22]

On Mon, Mar 31, 2008 at 7:46 PM, Bobby Powers <bobbypowers at gmail.com> wrote:
>
> The thing is, in forbody it is still doing:
> subtmp = population + (population * .3) - (population * .1)
>
>
> ideally I would love to see it reduced to:
> subtmp = 1.2 * population
>

I expect that such a transformation is not safe in general thanks to
rounding errors, and as such isn't applied.  AFAIK, compilers usually
don't reorganize floating point operations.

Have you tried it on other compilers?  I expect it'll never happen
without some imprecise math option (like gcc's -ffast-math).