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: [tree-ssa] PATCH: Feedback-based prefetching


As I understand it, the tree-ssa branch is the current branch for "new development". I
can't add new optimizations to the 3.4 branch any more, and I have been told that
tree-ssa will become the 3.5 compiler.


I will take a look at the branch you suggest. Are there any plans to merge the hammer-3_3
branch into the "main" FSF compiler any time soon?



On Dec 11, 2003, at 4:13 PM, Zdenek Dvorak wrote:


Hello,

This patch adds a new optimization to the tree-ssa branch:

why tree-ssa branch? I don't see how this is relevant.


You might also want to check hammer-3_3-branch where I have something
very similar (see vpt.c:speculative_prefetching*).

feedback-based prefetching.  The
optimization works on the assumption there will be two compilation
passes, with a training
run in between.  During the first pass, we make good guesses as to
places that are likely to
benefit from the insertion of prefetches, and we insert prefetches in
all those places (mostly
based on induction variables, and the current gcc induction-variable
based prefetching
mechanism).  In addition to inserting lots of prefetch instructions,
the first pass also inserts
profile data collection code, to collect information during the
training run about the
effectiveness of the inserted prefetches.  During the training run,
this data is collected, and
written to a file, "TRACE_FILE", which goes into the current working
directory.

You should use infrastructure provided by value-prof.c for this.


Zdenek



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