I am ok with this patch. Thanks! Ruiling
> -----Original Message----- > From: Beignet [mailto:[email protected]] On Behalf Of > Yang Rong > Sent: Thursday, May 26, 2016 1:29 PM > To: [email protected] > Cc: Yang, Rong R <[email protected]> > Subject: [Beignet] [PATCH] GBE: fix a patch JMPI assert. > > If jmpi out of range (+/-32768), should use long jump. > Now decide the long jump in the gbe, use the GEN IR, but > one GEN IR may produce several GEN instructions, especially > long operation pre BDW. So use experiential 8000 as threshold. > But some case still out of range. So decrease the threshold to > 3000. Need move it to context in the futher. > > Fix bug: > https://bugs.freedesktop.org/show_bug.cgi?id=94810 > > Signed-off-by: Yang Rong <[email protected]> > --- > backend/src/backend/gen_insn_selection.cpp | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/backend/src/backend/gen_insn_selection.cpp > b/backend/src/backend/gen_insn_selection.cpp > index fbd9363..a9cd65a 100644 > --- a/backend/src/backend/gen_insn_selection.cpp > +++ b/backend/src/backend/gen_insn_selection.cpp > @@ -1203,7 +1203,7 @@ namespace gbe > } > // FIXME, this longjmp check is too hacky. We need to support instruction > // insertion at code emission stage in the future. > - insn->extra.longjmp = ctx.getFunction().getDistance(start, end) > 8000; > + insn->extra.longjmp = ctx.getFunction().getDistance(start, end) > 3000; > return insn->extra.longjmp ? 2 : 1; > } > > -- > 2.1.4 > > _______________________________________________ > Beignet mailing list > [email protected] > https://lists.freedesktop.org/mailman/listinfo/beignet _______________________________________________ Beignet mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/beignet
