Address comments.

Simplify conditional and add comment. Add test with KILL.

Use isMetaInstruction; move func entry waitcnt change to separate patch.

It seems like there is an issue with defining exactly which bundles should be allowed to be modified and what passes should be allowed to modify them. There is already a pass that will do unbundling and that seems like the best place to do that if we enabled it.

In D96336#2553008, @critson wrote:

If every SGPR spill that goes to scratch has to do an extra store+load (or multiple) then is that not potentially worse than the performance hit of reserving an entire VGPR for spilling in the case that we know we are going to have to use one? (I guess perhaps we have no way of knowing we need one?)

Rebase.

Rebase. Address comments.

Rebase after 560d7e04113bf.

Don't use Optional type for TargetID.

Rebase.

I'm working on eliminating some AGPR->VGPR->AGPR copies which commonly cause problems in loops where the VGPR is considered live across iterations by the register allocator even if the VGPR copy should be sunk. Why do we need this temp VGPR?

Add M0 to defs.

In D91048#2438757, @rampitec wrote:

JBTW, having a need to def M0 probably defeats the idea of potential rescheduling which justifies the separation of the indirect access methods.

ping

Fix test.

Rebase. Move some utility functions to header.

EPSDB is passing.

Address comments.

In D91048#2383872, @rampitec wrote:

In D91048#2383678, @kerbowa wrote:

In D91048#2383420, @rampitec wrote:

Is there a practical reason to have different pseudos for set_idx and movrel methods and not just expand a common pseudo accordingly post RA?

It's a good question but in the end, I chose not to since m0 initializations can be rescheduled, and combing the pseudo's would disallow this. The other issue is that the movrel pseudo includes the scalar version which is used on targets with VGPR indexing as well. I'm open to trying it if you think it seems more practical.

Both methods use M0, aren't they?

In D91048#2383420, @rampitec wrote:

Is there a practical reason to have different pseudos for set_idx and movrel methods and not just expand a common pseudo accordingly post RA?

Remove dead code.

Update, rebase, and merge with change of default patch.

In D89753#2342419, @rampitec wrote:

Do we properly count number of pre-existing wait states if we have s_nop > 0?

Address comments.

Add test.

In D89077#2320507, @t-tye wrote:

Is this now running after the waitcnt insertion pass? That would avoid the NOPs currently being inserted to split memory clauses that are not necessary as the waitcnt instructions will split the clauses.

In D89077#2320507, @t-tye wrote:

Is this now running after the waitcnt insertion pass? That would avoid the NOPs currently being inserted to split memory clauses that are not necessary as the waitcnt instructions will split the clauses.

Wt -> Wr

Rename "Default" to "Any". Address comments.

Re-add DoesNotSupport features. Use enum class. No fatal error on setting mismatch.

Fix typo.

As discussed offline with Mahesha, this and https://reviews.llvm.org/D85767 are fixing the same issue, we decided I would take the fix forward.