Changes

[LoopUnroll] Adjust CostKind query

When TTI was updated to use an explicit cost, TCK_CodeSize was used
although the default implicit cost would have been the hand-wavey
cost of size and latency. So, revert back to this behaviour. This is
not expected to have (much) impact on targets since most (all?) of
them return the same value for SizeAndLatency and CodeSize.

When optimising for size, the logic has been changed to query
CodeSize costs instead of SizeAndLatency.

This patch also adds a testing option in the unroller so that
OptSize thresholds can be specified.

Differential Revision: https://reviews.llvm.org/D85723

[SystemZ/ZOS] Implement computeHostNumPhysicalCores

On z/OS, the information is stored in the Common System Data Area
(CSD). It is the number of CPs allocated to the current LPAR.

Reviewers: aganea, hubert.reinterpertcast, MaskRay

Reviewed By: hubert.reinterpertcast

Differential Revision: https://reviews.llvm.org/D85531

[VectorCombine] add test for Hexagon that would crash; NFC

This test verifies the code change from:
rGb0b95dab1ce2
(although that would not be true if PR47128 is fixed)

[InstCombine] eliminate a pointer cast around insertelement

I'm not sure if this solves PR46839 completely, but reducing the casting should help:
https://bugs.llvm.org/show_bug.cgi?id=46839

Differential Revision: https://reviews.llvm.org/D85647

[ARM] Add additional predicated VFMA tests. NFC

[VectorCombine] add test for x86 target with SSE disabled; NFC

[VectorCombine] early exit if target has no vector registers

Based on post-commit discussion in:
D81766

Other vectorization passes (SLP and Loop) use this TTI API similarly.

Limit Max Vector alignment on COFF targets to 8192.

COFF targets have a max object alignment of 8192, so trying to create
one with a larger size results in an unreachable in WinCOFFObjectWriter.

For the reproducer I have uses thread local storage, however other
alignments are likely affected as well.

This patch sets the MaxVectorAlign for COFF to 8192.  Additionally,
though there is no longer a way to reproduce that I could find, it
correctly sets the MaxTLSAlign for COFF to that value as well, so that
if anyone comes up with a situation where this is true, it will cause an
error.

Differential Revision: https://reviews.llvm.org/D85543

[OPENMP]Fix PR37671: Privatize local(private) variables in untied tasks.

Summary:
In untied tasks, need to allocate the space for local variales, declared
in task region, when the memory for task data is allocated. THe function
can be interrupted and we can exit from the function in untied task
switch. Need to keep the state of the local variables in this case.
Also, the compiler should not call cleanup when exiting in untied task
switch until the real exit out of the declaration scope is met during
 execution.

Reviewers: jdoerfert

Subscribers: yaxunl, guansong, cfe-commits, sstefan1, caomhin

Tags: #clang

Differential Revision: https://reviews.llvm.org/D84457

[MachOYAML] Simplify the section data emitting function. NFC.

This patch helps simplify some codes in writeSectionData() function.

Reviewed By: jhenderson, grimar

Differential Revision: https://reviews.llvm.org/D85821

[DWARFYAML] Make the address size of compilation units optional.

This patch makes the 'AddrSize' field optional. If the address size is
missing, yaml2obj will infer it from the object file.

Reviewed By: jhenderson

Differential Revision: https://reviews.llvm.org/D85805

Revert "[OPENMP]Fix PR37671: Privatize local(private) variables in untied tasks."

This reverts commit ec9563c54ed25e9f9cbe60985399212d50bd801d to
investigate compiler crash revelaed by the buildbots.

AMDGPU: Handle intrinsics in performMemSDNodeCombine

This avoids a possible regression in a future patch

AMDGPU/GlobalISel: Select llvm.amdgcn.global.atomic.fadd

Remove the intermediate transform in the DAG path. I believe this is
the last non-deprecated intrinsic that needs handling.