Re: [Valgrind-developers] JCC counting

[Julian S. <js...@ac...>]

> (2) Correct it and show how to do it the right way
>
> Solution (2) complicates lackey (a little?), but if it is also
> a tutorial about branch counting, it should be correct.

I agree.  Sounds like a good solution to me.

> So perhaps a third solution would be:
> (3) Not invert the meaning of conditions in the "pre-instr IR optimisation"
>     phase

That is hard to avoid given that IR optimisation does some simple
loop unrolling for single-bb loops.  eg, it will unroll your

   top:
       add    %eax,%edx
       add    $0x1,%eax
       cmp    $0xf4240,%eax
       jl     top

producing IR as if you had written

   top:
       add    %eax,%edx
       add    $0x1,%eax
       cmp    $0xf4240,%eax
       jnl    after
       add    %eax,%edx
       add    $0x1,%eax
       cmp    $0xf4240,%eax
       jl     top
    after:

See, the middle exit is now inverted.

Also, if we ever do more aggressive block formation which speculatively
chases across conditional branches ("trace formation"), then some branch
conditions will have to be inverted - it is unavoidable.

J