i see your point. I had forgotten that crossjumping is not really a
pass, it is more of an infectious agent.
On 03/18/2011 01:45 PM, Paolo Bonzini wrote:
On Fri, Mar 18, 2011 at 17:31, Jakub Jelinek<ja...@redhat.com> wrote:
On Fri, Mar 18, 2011 at 12:23:11PM -0400, Kenneth Zadeck wrote:
I believe that this is not the right way to go.
if someone specifies -fcrossjumping, then the pass should turn on
live for the duration of the pass just as ifcvt does. If they ask
for crossjumping you should give them crossjumping and not some
crippled version of it.
Such (untested) patch is in
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48156#c7
But Paolo just added comment there that he prefers this version.
Is the live problem so much more useful for crossjumping than lr problem?
All it wants to prove is if it can safely move a couple of instructions
across some other instructions, and the live_union bitmap that is computed
using the live/lr problem is used to find out if registers set by the
moved instructions are actually live at the end of those instructions
or not. What would be an example where live problem would allow optimizing
more than lr?
LIVE == LR except when you have uninitialized uses.
Unless it is needed for correctness, I see no reason to prefer LIVE to
LR at -O1.
yes, but i think that the reason this is a pr is that it seems to be
needed for correctness.
i certainly am not advocating using live at O1. it is more
expensive. however, we do have the ability to turn on some problem
only in one pass, and that was what i was suggesting. you make
crossjumping require LIVE, and you will get LIVE for that pass, not for
the entire compilation.
Paolo
