[PATCH] More rewrites of re_node_set_*

[Paolo Bonzini]

This patch achieves an improvement of 3-5%.  It is useless to do binary searches and
other fancy things if the function does a possibly quadratic memmove.  memmove itself
is a bottleneck because we know we are transferring 32-bit quantities, because we
know in which direction they should be transferred, and on x86 because "rep movsb" is
bad on many processors.

Paolo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: regex-rewrite-node-set-funcs.patch
Type: application/octet-stream
Size: 8340 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20040103/9735b301/attachment.obj>