If there are others experiencing the same error then I can push my fix: https://github.com/llvm/llvm-project/pull/67696 if there are no other suggestions.

In D157624#4583164, @RKSimon wrote:

@mbrkusanin This is causing failures on EXPENSIVE_CHECKS buildbots: https://lab.llvm.org/buildbot/#/builders/16/builds/52934 - please can you take a look?

• add test case which needs kimm16 and kimm32 to be not inlinable
• remove some now redundant checks

• Update tablegen comments

• Set FixedSize for FMAMK and FMAAK instructions

LGTM

• Rebase + Ping
• Do we still want a solution that works as a new repair kind in regbankselect? Alternatively we could make a regbank combiner.

• Rebase + Ping

• gfx11 has no x2/f64 insts so global_atomic_*_x2 patterns are no longer in isGFX10Plus group (new test for gfx10 only: fp-min-max-global-atomics-f64.ll)
• updated tests so they would better match with common check lines

• new test: llvm/test/CodeGen/AMDGPU/verify-image-partial-nsa.mir

• Removed unnecessary if checks
• Updated SIShrinkInstructions to handle partial NSA

• Rebase

In D144034#4138543, @arsenm wrote:

Do we need verifier checks for the NSA restrictions on gfx11?

Since PartialNSA is now a SubtargetFeature it can be turned off with -mattr=+partial-nsa-encoding to get old behavior.

• new feature FeaturePartialNSAEncoding
• NSAMaxSize is no longer a tablegen feature

Same could be done for GFX10, however differences between GFX10.1 and GFX10.3 cause inconveniences.
MaxNSA size for 10.1 is 5 and for 10.3 is 13 so _V6, _V7,... opcodes for GFX10 already exist where every vaddr is a VGPR_32.
We would need new versions exclusive for non-10.3.

rG43924cbd29a32dc391f36850405d6c26384acaa1

• reverted name change
• rebase

Nothing that I can see.

In D133012#3933895, @Joe_Nash wrote:

In D133012#3933756, @mbrkusanin wrote:

• By overriding opName (or PseudoInstr of tablegen record) for _strict pseudo we can have both pseudos select into same real instruction, so now we don't need two.

This is a creative use of the opName, but I think it probably hinders readability. No other instructions are doing this, so when I look at definition of the Real instruction it seems to refer to the non-strict pseudo.

Well, it refers to both in a way. I changed the one it actually uses for initialization to be _strict pseudo which in turns uses same name as non strict one.

In D116441#3932143, @arsenm wrote:

Oops, this was at least mostly re-implemented. Is there anything here main doesn't have already?

In D133012#3932090, @arsenm wrote:

Is this still relevant? I thought there was a discussion that this wasn't actually broken?

• By overriding opName (or PseudoInstr of tablegen record) for _strict pseudo we can have both pseudos select into same real instruction, so now we don't need two.
• Updated use of features/predicate so now it works properly with -mattr=-mad-intra-fwd-bug which was the point of this change.