On Mon, 2014-05-05 at 13:33 -0500, Peter Bergner wrote: > Currently, the IBM long double routines in libgcc use a union to construct > a long double from two double values. This causes horrific code generation > that copies the two double from the FP registers over to GPRs and back > again, giving us two loads and two stores, which leads to two load-hit-store > hazzards. The following patch makes use of the new __builtin_pack_longdouble > builtin to construct the long double giving us at worse, one or two fmrs. > > Is this ok for mainline once my bootstrap and regtesting are complete?
Ok, bootstrapping and regtesting have completed with no regressions. Peter
