Just a minor nit I noticed.

Really, really nice find here. A bunch of comments below, but they're all pretty minor on the whole.

Thanks for the comment, I've updated the diff with the change from Mb to Mib.

In D35621#845117, @avt77 wrote:

Coby, Reid, is eveything OK? Could you get LGTM?

• Addressed review comment.

After the FIXME about record layout is addressed, i guess the next action item is to make sure the trivial constructor for the empty base is not evaluated conservatively (doh!). Because whatever bindings we make, for now they'd be blown away immediately during "construction".

I've run this across SPEC CPU 2006 and the LLVM test suite on my Haswell system.

• Addressed review comment.

LGTM. Maybe elaborate a bit more on the summary part

Coby, Reid, is eveything OK? Could you get LGTM?

Skip the default binding for empty bases.

Ping

In D36795#844373, @uweigand wrote:

Oh, and just comment on this:

BTW, I don't quite understand the purpose of updateRegAllocHint(). It seems to be run during coalescing, but that's before getRegAllocationHints() has ever been called. So what is being updated?

I understand this is about something else: maintaining the MachineRegisterInfo set/getRegAllocationHint data. It seems the ARM target sets up the MRI hints in a special pass ahead of time, and then *uses* the MRI hints in its implementation of getRegAllocationHints. However, for this to be useful the MRI hints need to be updated for the effects of register coalescing; the updateRegAllocHint allows the target to do just that.

Since your implementation of getRegAllocationHints doesn't actually make any use of the MRI hints, you don't need to implement updateRegAllocHint either.

Rebase

Rebase

LGTM

In D36717#844930, @Meinersbur wrote:

One more remark: The 'ref' dataset of 638.imagick_s on all the computers I tried on took between 2 and 3 hours. Submitted results to SPEC are in a similar range. Unfortunately timeit.py has a 7200 seconds (2 hours) limit hardcoded (for --limit-cpu and --timeout). I had to remove that limit to be able to have a successful run.

Can we make the limit configurable or set higher?

This is the same as what i said on D36682 :)

TL;DR This statement is slightyl different, but also has no effect.
It is also a statement that is more or less required by statute/regulation.

Thanks, @pcc -- PTAL?

The statements are, AFAIK, best viewed as statements of provenance, and are fairly normal for government released works (some are even required by statute)
They do not change or affect the license in any way.

Address comments.