loop optimizations, mark 11

[Richard Henderson]

Things of note in this edition:

  * Incorprorated Bernd Schmidt's suggestions for treatment of benefit
    in and around general_induction_var.

  * Fixed a long-standing bug in express_from that caused the ia32
    code-gen lossage.

  * Fixed a bit in find_mem_givs that was causing -frerun-loop-opt to
    generate worse code on Alpha.  I am now interested to see code
    samples that generate different code when loop is rerun.


r~
-------------- next part --------------
A non-text attachment was scrubbed...
Name: d-loop-11.gz
Type: application/x-gzip
Size: 11170 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/19980602/5f0b629b/attachment.bin>