Hal, thanks -- I've committed this as r210384. The patch essentially looks like the whole test has changed, but very little of note actually has. Having two different CHECK sets meant that the existing formatting wasn't usable. For powerpc-unknown-unknown, nothing has changed. For powerpc64-unknown-unknown, the tests work identically to powerpc-unknown-unknown. For powerpc64le-unknown-unknown, almost all the tests are the same, but there are changes for all of the vec_pack and vec_mul[eo] variants.
Thanks, Bill On Fri, 2014-06-06 at 11:33 -0500, Hal Finkel wrote: > ----- Original Message ----- > > From: "Bill Schmidt" <[email protected]> > > To: "Hal Finkel" <[email protected]> > > Cc: [email protected] > > Sent: Friday, June 6, 2014 11:32:29 AM > > Subject: Re: r210340 - [PPC64LE] Implement little-endian semantics for > > vec_pack family > > > > Hm, that probably isn't true. I suppose the test will "pass" by > > producing the wrong (big-endian) code generation, so this is a > > reasonable approach. > > > > I'll look into it. > > Okay, thanks. Having the tests with the commits is almost always better, so > if it is feasible, then I'd prefer it. > > Thanks again, > Hal > > > _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
