This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA/RFT] Remove -fcse-skip-blocks


On Wednesday 23 March 2005 14:27, Paolo Bonzini wrote:
> It may be interesting to SPEC-test this (the patch of course is not
> needed for that).

I think it's not so interesting to SPEC test this, I already did
that not too long ago and the results were basically unchanged.
I even tested with path following completely disabled and that
was "SPEC-neutral" too.
But SPEC is not the only thing to look at.  I'd consider at least
the linux kernel and CSiBE as well, and maybe some weird code,
like Brad Lucier's Scheme test cases.

>  common.opt      |    4 -
>  cse.c           |  175
> ++++++--------------------------------------------------
>  doc/invoke.texi |   13 ----
>  opts.c          |    1
>  passes.c        |    6 -
>  5 files changed, 24 insertions(+), 175 deletions(-)
>
> Not a big amount of code, but it is the first step if we want CSE to
> work on EBB's (did I get this right, Steven?).

Obviously, I like the idea ;-)  Without skip-blocks, CSE basically
is already working only on EBBs, just in a not very efficient way.
So from that perspective the patch makes sense.  But we'll have to
build a more convincing case, based on hard numbers, that this is
a good idea.

Gr.
Steven


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]