Re: [PATCH] Follow-up to PR middle-end/13392

[Roger Sayle <roger at eyesopen dot com>]

On 12 Jan 2004, Gabriel Dos Reis wrote:
> | The patch below is a follow-up to Jakub Jelinek's patch to solve
> | PR middle-end/13392.  I was investigating an alternate version of
> | Jakub's original patch, when his latest revision was posted and
> | approved by Richard Henderson.  Our final solutions ended up being
> | almost identical, with the exception of the logic clean-up and
> | enhancement below.  Although this patch no longer fixes any bugs,
> | I thought it worthwhile to post this patch so it doesn't get lost,
> | and while the issues are still fresh in everyone's memories.
> |
> | The following patch has been tested on i686-pc-linux-gnu with a full
> | bootstrap, all languages except Ada and treelang, and regression
> | tested with a top-level "make -k check" with no new failures.
> |
> | Ok for mainline?
>
> And your assessment for gcc-3_3-branch?

I don't think this change qualifies for the gcc-3_3-branch.  Its
mainly a code clean-up (getting rid of a unneccessary goto) and an
additional optimization that I suspect is rare in practice.  Jakub's
patch, as already committed to gcc-3_3-branch, is sufficient to resolve
the PR and already improves the quality of the code we generate for
__builtin_expect over previous releases.

This patch was already borderline for mainline.

Roger
--