[RISCV][NFC] Broke ReadVRGatherVV into ReadVRGatherVV_data and ReadVRGatherVV_index to separate the reads for VRGatherVV

Differential Revision: https://reviews.llvm.org/D145406

Differential Revision: https://reviews.llvm.org/D145402

The original -show-graph option dumped the LinkGraph for all graphs loaded into
the session, but can make it difficult to see small graphs (e.g. reduced test
cases) among the surrounding larger files (especially the ORC runtime).

The new -show-graphs option takes a regex and dumps only those graphs matching
the regex. This allows testcases to specify exactly which graphs to dump.

This flag causes the toolchain distribution to be built using LLVM
CMake's multiple distribution feature. The distribution* family of CMake
targets would be replaced with the toolchain-distribution* family.

This shouldn't otherwise affect the semantics of the build, but it sets
up the ability to split out the LLDB build from the main distribution
used by Fuchsia.

Reviewed By: phosek

Differential Revision: https://reviews.llvm.org/D146665

Depends On D146674

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146676

There were added when I removed the swig interface files in
662548c8. However, they mostly meant for
me to better track the differences between the existing API headers and
bindings interfaces. There's nothing actionable about these so I remove
them.

When the input types don't match the accumulator type in named
convolution ops there is supposed to be a conversion to the accumulator
type before the multiply and accumulate.

Differential Revision: https://reviews.llvm.org/D146824

For singleton transfer reads, we allow hoisting them out
of the enclosing loop if its users are either transfer reads
or memory effect free.

Differential Revision: https://reviews.llvm.org/D146828

In the next few commits I will be converting the various Merger identifier typedefs into newtypes; and once that's done, the `kInvalidId` constant will only be used internally and therefore does not need to be part of the public `mlir::sparse_tensor` namespace.

Depends On D146673

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146674

Depends On D146562

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146673

* Moving the `Children` class to be nested under `TensorExp`.
* Marking `TensorExp`, `TensorExp::Children`, and `LatPoint` as final.

Depends On D146083

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146562

This helps the `Merger` maintain invariants, as well as clarifying the immutability of the underlying objects (with the one exception of `TensorExp::val`).

Depends On: D146559

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D146083

When printing a value, allow the root value's name to be elided, without omiting the
names of child values.

At the API level, this adds `SetHideRootName()`, which joins the existing
`SetHideName()` function.

This functionality is used by `dwim-print` and `expression`.

Fixes an issue identified by @jgorbe in https://reviews.llvm.org/D145609.

Differential Revision: https://reviews.llvm.org/D146783

The option -fno-omit-frame-pointer was accidentally added to the x86_64
longjmp target. This change not only removes it, but makes it
-fomit-frame-pointer.

X. Sun et al. (https://dl.acm.org/doi/10.5555/3454287.3454728) published
a paper showing that an FP format with 4 bits of exponent, 3 bits of
significand and an exponent bias of 11 would work quite well for ML
applications.

Google hardware supports a variant of this format where 0x80 is used to
represent NaN, as in the Float8E4M3FNUZ format. Just like the
Float8E4M3FNUZ format, this format does not support -0 and values which
would map to it will become +0.

This format is proposed for inclusion in OpenXLA's StableHLO dialect: https://github.com/openxla/stablehlo/pull/1308

As part of inclusion in that dialect, APFloat needs to know how to
handle this format.

Differential Revision: https://reviews.llvm.org/D146441

This patch fixes:

  llvm/unittests/Support/ScopedPrinterTest.cpp:519:20: error: unused
  variable 'InfDouble' [-Werror,-Wunused-variable]

  llvm/unittests/Support/ScopedPrinterTest.cpp:520:16: error: unused
  variable 'NaNDouble' [-Werror,-Wunused-variable]

  llvm/unittests/Support/ScopedPrinterTest.cpp:516:15: error: unused
  variable 'NaNFloat' [-Werror,-Wunused-variable]

  llvm/unittests/Support/ScopedPrinterTest.cpp:515:19: error: unused
  variable 'InfFloat' [-Werror,-Wunused-variable]

Since commit fa56e362 has temporarily
disabled tests involving these constants, this patch simply comments
them out instead of removing them.

The issue was reported with compiler output here: https://reviews.llvm.org/D144083#4219383

Folding of the tosa.transpose operation is both time and memory
intensive as the underlying ElementsAttr is processed as a sequence of
Attributes. This change attempts operate on the underlying raw data of
the ElementsAttr.

In an example resnet50 network, this change reduces the time spent in
folding transpose ops from 35s to 1.5s.

Reviewed By: GeorgeARM, rsuderman, stellaraccident

Differential Revision: https://reviews.llvm.org/D146526

Summary:
A recent patch allowed us to emit a callable kernel from freestanding
NVPTX code. This allows us to move away from using the CUDA language.
This has several advantages in that it works around an entire assortment
of errors I was seeing while implementing RPC for Nvidia.

Added a test to https://reviews.llvm.org/D146657, make sure the guid and name are computed using the debug info name.

Reviewed By: hoy, wenlei

Differential Revision: https://reviews.llvm.org/D146826

As of c5bfa3da, REPL.h no longer has a
private implementation header in it. This TODO and the thing it marks
cdan be removed.

We may want to be able to mark certain regions as kernels even without
being in an accepted CUDA or OpenCL language mode. This patch introduces
a new attribute limited to `nvptx` targets called `nvptx_kernel` which
will perform the same metadata action as the existing CUDA ones. This
closely mimics the behaviour of the `amdgpu_kernel` attribute. This
allows for making executable NVPTX device images without using an
existing offloading language model.

I was unsure how to do this, I could potentially re-use all the CUDA
attributes and just replace the `CUDA` language requirement with an
`NVPTX` architecture requirement. Also I don't know if I should add more
than just this attribute.

Reviewed By: tra

Differential Revision: https://reviews.llvm.org/D140226

Missing sign extension.

Reviewed By: rsuderman

Differential Revision: https://reviews.llvm.org/D145744

Summary:
These messages have been wrong for quite some time. Update them to be
more descriptive of why the tests weren't built.

Fix linking ClangdTests to specify the dependency on the private
clangTesting library via target_link_libraries() rather than
clang_target_link_libraries().  The latter uses libclang-cpp when
CLANG_LINK_CLANG_DYLIB is used, and clangTesting is not included
in this library.

This fixes d60d3455.

Differential Revision: https://reviews.llvm.org/D146427

This patch implements setjmp and longjmp in riscv using inline asm. The
following changes were required:

* Omit frame pointer: otherwise gcc won't allow us to use s0
* Use __attribute__((naked)): otherwise both gcc and clang will generate
function prologue and epilogue in both functions. This doesn't happen
in x86_64, so we guard it to only riscv

Furthermore, using __attribute__((naked)) causes two problems: we
can't use `return 0` (both gcc and clang) and the function arguments in
the function body (clang only), so we had to use a0 and a1 directly.

Reviewed By: sivachandra

Differential Revision: https://reviews.llvm.org/D145584

The logic in ConstraintElimination should trivially apply to GEP
constant expressions as well, so update code to deal with GEPOperator
instead.

When computing symbol hashes in BinarySection::hash, we try to find relocations
in the section which reference the passed BinaryData. We do so by doing
lower_bound on data begin offset and upper_bound on data end offset. Since
offsets are relative to the current section, if it is a data from the previous
section, we get underflow when computing offset and lower_bound returns
Relocations.end(). If this data also ends where current section begins,
upper_bound on zero offset will return some valid iterator if we have any
relocations after the first byte. Then we'll try to iterate from lower_bound to
upper_bound, since they're not equal, which in that case means we'll dereference
Relocations.end(), increment it, and try to do so until we reach the second
valid iterator. Of course we reach segfault earlier. In this patch we stop BOLT
from searching relocations for symbols outside of the current section.

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D146620

Support bitcasting between int/fp/vector values and 'r'/'f'/'v' inline
assembly operands. This is intended to match GCCs beahvior.

Reviewed By: Ulrich Weigand

Differential Revision: https://reviews.llvm.org/D146059

This reverts commit 72073fc9.

This is still breaking on some platforms. The underlying implementation
doesn't seem to be the cause, rather the test is not robust across
platforms. So, we'll just disable this for the time being, to unblock
builds until we have a proper fix.

Reviewed By: abhina.sreeskantharajan

Differential Revision: https://reviews.llvm.org/D146834

Review: https://reviews.llvm.org/D145831

* Avoid unnecessary frame & tag push/pops if memory access is ignored
* Rename function and add comment to make it clearer what the code does
* Make helper functions static and move inside `#if !SANITIZER_GO`

Differential Revision: https://reviews.llvm.org/D146670

Without this the function will be use an Arm subtarget, meaning the
instructions in it will be invalid for the current subtarget.

Differential Revision: https://reviews.llvm.org/D144733

Put the value into A0 instead of data registers. And remove the
redundant `RetCC_M68kCommon` as there aren't many rules shared between
existing CCs other than the pointer one.
This change is tested by existing tests.

I accidentally broke the FreeBSD lldb-server build in 0c5cee77 because it
now depends on PlatformFreeBSD. PlatformFreeBSD depends on PlatformPOSIX
but this dependency was not explicitly tracked in CMake. As a result,
the FreeBSD lldb-server build broke.

Credit to John F. Carr <jfc@mit.edu> for pointing out the issue and
providing a fix.