gcc on Intel x86 and failure of -ffloat-store option

Mark Mitchell mark@codesourcery.com
Fri Dec 31 20:54:00 GMT 1999


>>>>> "Stephen" == Stephen L Moshier <moshier@mediaone.net> writes:

    Stephen> On Fri, 10 Dec 1999, Mark Mitchell wrote:
    >> Shouldn't you do this only for mems that are written to?

    Stephen> That is an easy speed question but perhaps a hard IEEE
    Stephen> question.  At first I thought maybe the change should
    Stephen> affect only variables involved in a loop exit test.
    Stephen> Anyway here is a version for the ones that have been
    Stephen> written.

    Stephen>         * loop.c (load_mems): Don't hoist written
    Stephen> floating point mem if -ffloat-store.

Thanks; I installed this change.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com



More information about the Gcc-bugs mailing list