LGTM

LGTM.

Adding David Stuttard as I think he worked on dmask originally.

Added David Stuttard as reviewer since this was originally his.

https://reviews.llvm.org/D69287 covers this.

Ping.

Upstreamed.

Added condition in foldOperands to allow tied subreg folding
for V_FMAC and V_MAC operations

In D58918#1443108, @hakzsam wrote:

Hi there,
Do you have the cmpswap fix somewhere?
Thanks

Ping.

Forgot to capitalize var name.

Set bool in condition.

Made some changes to code.

Updating.

Added test case

This has been copied over and extended on in https://reviews.llvm.org/D62802 so I'm abandoning this revision.

Hmm... looks like AMDGPUISelLowering.cpp has changed quite a bit in regards to the calling convention.

ping

Remove underline to ref hyperlink from invariant.group

Removed preceding underline to ref hyperlink.

Added hyperlinks to LangRef.rst

Changed LangRef.rst: Moved !dereferenceable and !dereferenceable_or_null to their
own section in Metadata and pointed inttoptr and load to that section.

Added IntToPtrInst to case statement in Value::getPointerDereferenceableBytes

Changed LangRef.rst to reflect !dereferenceable is applicable to pointer types not
just load types.

Updated error comment and case error test

Changed error message in AnalyzeFormalArguments to be more generic.
Added comments.

Fixed test to use llvm-as and to just check for passing.

Added !dereferenceable_or_null to syntax of inttoptr in LangRef

Added tests and updated LangRef

Added test case. Changed assert to report_fatal_error when checking
arguments and returns.

Fixed suggestions

Fixed idents.

Suggested changes: Added InstrMapping. Three test files. Lower-cased suffixes for MI defs. Changed multiclass suffix.

Changes per suggestions

In D63494#1548986, @rampitec wrote:

In D63494#1548974, @rtaylor wrote:

In D63494#1548719, @rampitec wrote:

Can we have a disasm test? I want to see that we do not mess with decoding.

Something other than the -disassemble checks?

Ah, missed it. Thanks, this is sufficient.

In D63494#1548719, @rampitec wrote:

Can we have a disasm test? I want to see that we do not mess with decoding.

Added test for SGPR tuple

More detailed explanation of why LCSSA is needed

Added StackProtectorID and changed LCSSA to use it instead of StackProtector directly.

Fixed 105 SGPR to 106 SGPR, increased limit count for allocateSGPRTuple

Preserving StackProtector in LCSSA to avoid pass scheduling conflict
Removed explicit call of LCSSA in AMDGPUTargetMachine

Scratch that last commit, this is still broken.

I was misunderstanding the dependency issues, this should fix it.

[AMDGPU] Increased SGPR Count for Calling Convention

In D60834#1475522, @arsenm wrote:

Also needs a comment explaining why LCSSA is needed

Moved LCSSA call to after the sinking pass.

In D60834#1470764, @arsenm wrote:

In D60834#1470763, @arsenm wrote:

It sort of intuitively makes sense to me that the control flow lowering would like LCSSA. However, this should not be handled by adding it directly to the pass pipeline. You can add this as a dependency, e.g. AU.addRequiredID(LCSSAID);

I would also like to see the an IR->IR testcase showing LCSSA was implicitly run

Actually, what really requires LCSSA? Is it DivergenceAnalysis or StructurizeCFG directly?

In D60834#1470734, @arsenm wrote:

This is a workaround. The structurizer / annotator must be correct without relying on another pass to hide situations they don't handle correctly

We have a test case such that a value that is uniform in the loop is used outside the loop where threads might have diverged.