Page MenuHomePhabricator

[FPEnv] PowerPC Vector Constrained FP Intrinsics Test
ClosedPublic

Authored by ajwock on May 24 2019, 6:32 AM.

Details

Diff Detail

Repository
rL LLVM

Event Timeline

ajwock created this revision.May 24 2019, 6:32 AM
Herald added a project: Restricted Project. · View Herald TranscriptMay 24 2019, 6:32 AM

Just to clarify, do you currently expect all of these lowerings to be semantically correct, or only to mirror the regular lowering path right now?

kpn added a comment.May 24 2019, 12:14 PM

Drew is working with me here at SAS Institute, Inc.

Just to clarify, do you currently expect all of these lowerings to be semantically correct, or only to mirror the regular lowering path right now?

They should mirror the regular lowering path at least for the most part. Is there a case in this test where that produces semantically incorrect results?

hfinkel accepted this revision.May 24 2019, 12:44 PM
In D62388#1516546, @kpn wrote:

Drew is working with me here at SAS Institute, Inc.

Just to clarify, do you currently expect all of these lowerings to be semantically correct, or only to mirror the regular lowering path right now?

They should mirror the regular lowering path at least for the most part. Is there a case in this test where that produces semantically incorrect results?

No, but I certainly haven't looked through all 10k lines in detail ;) - I worry somewhat about the rounding modes and the lowerings that call libc, that call the runtime library, and that use instruction directly.

In any case, we can move forward with adding this now and then update as need be.

This revision is now accepted and ready to land.May 24 2019, 12:44 PM
This revision was automatically updated to reflect the committed changes.