Update on "Set default qengine to QNNPACK on ARM for quantization tests"


There is no FBGEMM or MKLDNN on ARM so these tests were previously skipped (and QNNPACK is disabled by default for some reason). Not sure if this PR is the right approach, alternatives are probably:
1) keep these tests skipped
2) not disable QNNPACK by default on ARM
3) ?

[ghstack-poisoned]

Update on "Unskip CompositeCompliance tests for ARM"

[ghstack-poisoned]

[NVFuser] Upstream push 0809

Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/

Code changes includes:

- codegen improvements:
  1. removes un-necessary sync from redundant thread compute analysis
  2. symmetric API for BestEffortReplay
  3. support merge on trivial reductions
  4. Ampere async copy improvements
- bug fixes:
  1. vectorization bug fixes
  2. type inference patch : fixes upstream pytorch#81725
  3. segmenter bug fix with deterministic iteration ordering
- parser update
  1. added leaky_relu
- scheduler
  1. normalization scheduler clean up.
  2. simplifies matmul scheduling with new transform propagator
  3. merge all dimensions in PW scheduler
  4. various gemm related improvements
- debuggability
  1. nsight compute support
  2. debug dump for InlinePropagator
  3. Add `UnaryOpType::Print`

Squashed commits to WAR github API
Commits that's actually in this PR from the devel branch:

```
dfe02f3 Merge remote-tracking branch 'csarofeen/devel' into HEAD
1617373 Add `TensorViewBuilder::shape(std::vector<Val*> shape)` (pytorch#1884)
7cfb779 Merge pull request pytorch#1887 from csarofeen/upstream_merge_0803
3399f6d Merge remote-tracking branch 'origin/viable/strict' into HEAD
01208f5 Add `UnaryOpType::Print` which can be helpful for debugging (pytorch#1878)
0646522 Remove redundant TORCH_INTERNAL_ASSERT in lower_magic_zero.cpp (pytorch#1881)
7bc76aa Fix most inlined propagator for mismatched dims (pytorch#1875)
501f4aa Nonaffine swizzle formulation ep.2: Loop swizzle variant. (pytorch#1826)
d863d69 Ampere async copy ep.2: circular buffering extension to support pipelined matmul operand load (pytorch#1827)
e0ae11a Larger sized mma instructions to support full vectorization (pytorch#1824)
9bb4cf7 fragment iteration to support fully unrolled mma ops (pytorch#1823)
a48270a Merge all dims in pointwise scheduler (pytorch#1872)
172fb36 Make MostInlined and BestEffort inline propagation no longer assert replayed (pytorch#1868)
a64462a Allow trivial reduction to be merged (pytorch#1871)
440102b Symmetric API for BestEffortReplay (pytorch#1870)
d1caf33 Some misc cleanups/refactor split out from pytorch#1854 (pytorch#1867)
1013eda Remove some welford specific logic. (pytorch#1864)
51589d3 Some cleanups on tests and heuristics params (pytorch#1866)
a6b3e70 Segmenter bug fix, and deterministic iteration ordering.  (pytorch#1865)
1b665b9 Add nullptr checks to IrBuilder (pytorch#1861)
1cd9451 Simplify matmul scheduling with the new transform propagator.  (pytorch#1817)
bbc1fb9 Add leaky_relu operation (pytorch#1852)
e842a9b Minor cleanup in pointwise scheduler (pytorch#1858)
9ee850c Fix stringstream usage (pytorch#1857)
20a36c1 Improve nsight compute support (pytorch#1855)
4059103 Remove debugging `true ||` from getPointwiseHeuristics (pytorch#1822)
01117bf Misc cleanup (pytorch#1853)
5cc6494 Apply the magic-zero protection to each indexed domain individually for predicate indexing (pytorch#1846)
92e6f02 Cleanup normalization scheduler (pytorch#1845)
db89c65 Type inference patch (pytorch#1848)
102fe93 Add debug dump for InlinePropagator (#1847)
b7a4d93 Redundant thread compute analysis to avoid un-necessary sync insertion (pytorch#1687)
942be5b Upstream ci build fixes (pytorch#1842)
0b83645 Fix vectorization bug introduced in pytorch#1831 (pytorch#1840)
63630f1 Move MaxProducerPosUpdater into InlinePropagator::tearDown (pytorch#1825)
9135a96 Fix transpose benchmark dtype (pytorch#1839)
2c9a6c0 Add extra configurability to `parallelizeAllLike` (pytorch#1831)
```

RUN_TORCHBENCH: nvfuser

[ghstack-poisoned]

Update on "Build MacOS binaries with `-Werror`"

Should prevent proliferating MPS warnings
Fixes pytorch#82966

[ghstack-poisoned]

Update on "[BE] Make MPS compile warning free"


Mostly get rid of unused variables, but also:
- Hide `-[MPSGraphTensorData printNDArray]` behind undefined method
  access
- Add `MPSGraph+CoolMethods.h` that contains definitions of undefined methods used by MPS ops

[ghstack-poisoned]

Update on "Test profiler on ARM"

Just saw that profiler tree tests had to be disabled, and figuring out if it's fundamental.

Differential Revision: [D38487443](https://our.internmc.facebook.com/intern/diff/D38487443/)

[ghstack-poisoned]

Merge remote-tracking branch 'origin/master' into upstream_push_0806

Update on "WIP: Use FindCUDAToolkit to find cuda dependencies"

[ghstack-poisoned]

Disabling the kernel assert UT

Disabling same tests as 11.6