4.1 ten times slower when using utf8 charset

[David Shanahan]

My application needs to search through a lot of (Japanese) text records.

Ideally, I would like to use fulltext indexes but that won't work with Japanese text.
since Japanese doesn't have any whitespace between words.

Anyway with MySQL 4.0.22 I was getting quite fast query times, around 700 milliseconds,
which was fast enough for the application.

Experimenting with 4.1, the same queries on the same hardware with the same buffer
and log settings takes about 7 seconds, a 10 times slowdown.

If I use the latin1 charset on 4.1 the queries run fast again but the Japanese text gets
destroyed. Under 4.0, I could get away with using latin1 charset with utf8 data.

Is there any compile or runtime options to make utf8 charsets work as fast as latin1 ?
Would it help to compile utf8 in as the only charset ?