This is the mail archive of the
mailing list for the GCC project.
Re: COND_EXPR lowering patch
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- Cc: gcc at gcc dot gnu dot org, law at redhat dot com
- Date: Sat, 23 Aug 2003 20:45:06 -0400
- Subject: Re: COND_EXPR lowering patch
>>>>> Zdenek Dvorak writes:
Zdenek> The main showstopper seems to be the tree-ssa-dce; after this change
Zdenek> it cannot work in the current form (resp. it could, but since most of the
Zdenek> interesting control flow will then be expressed by explicit gotos, it
Zdenek> would be useless).
Zdenek> It was discussed before, but afair there was not any definitive
Zdenek> conclusion. To me the following scheme seems to be the best:
Zdenek> The control structures removal part will only be activated with -O3
Zdenek> and will be done using regular control dependency analysis.
Zdenek> Rationale: this part of the optimization is rarely useful, and even if
Zdenek> it does something, ordinary dce together with jump threading does it
Zdenek> anyway (not completely, but for the most interesting cases it should
Zdenek> work). It is also quite costly, which makes it a clear candidate for
Zdenek> -O3. And this also seems to be the easiest way, since Jeff has the
Zdenek> requiered patch iirc.
Zdenek> Your opinions?
Why are we discussing this in terms of all or nothing control
structure removal instead of progressive lowering? We cannot order the
Tree-SSA passes to handle both requirements?