This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Some middle-end improvements for bitfield handling
On Jun 30, 2004, at 06:43, Jakub Jelinek wrote:
Not sure if it wouldn't be better to expose bitfields operations
alreary during gimplification or somewhere in the middle of tree-ssa
passes though, i.e. rtl expanders would only see shifting/masking etc.
and tree-ssa passes could already optimize.
I don't think so, at least it shouldn't be done before optimizations
such as SRA, which are extremely useful for records with bitfields.
You'll lose all per-field information as soon as everything
gets smashed down to shift/mask so that it will become extremely
hard to do trivial things such as constant propagation.
We may want to lower bitfields in some cases, for example if
the vectorizer determines that accesses in a loop can be vectorized.
However, I think for any such choices we'll need to use target
information for choosing the right expansion and otherwise we'll
be better off leaving this to RTL for now.
-Geert