This simplifies checking of the result (it's just an Error, rather than an
optional<Expected<T>>), and allows T to be deduced rather than requiring that
it be specified.

Reviewed By: dblaikie

Differential Revision: https://reviews.llvm.org/D147583

Some configurations may require `-DCUDAToolkit_ROOT` to find CUDA
properly. This is currently not forwarded to the CMake invocation. This
patch adds a prefix so it will be visible when the runtimes build is
started.

Reviewed By: tra

Differential Revision: https://reviews.llvm.org/D147582

The `clang-offload-packager`. `nvptx-arch`, and `amdgpu-arch` tools are
required for building the GPU target of `libc`. This patch ensures that
we build this tool when directly building `libc` via `ninja libc` or similar.

Reviewed By: tra

Differential Revision: https://reviews.llvm.org/D147581

This patch loads the CUDA driver library directly via its real
`DT_SONAME`. This prevents the filesystem from needing to reload it in
cases when it's already loaded.

Reviewed By: tra

Differential Revision: https://reviews.llvm.org/D147579

This is the first change for FS-AFDO integration with CSSPGO. There are more patches coming.

With pseudo probes, we do not assign FS discriminators to any other instructions since we will be using only probes for profile correlation.

Also call instructions are excluded since their dwarf discriminators are used for other purposes, i.e, storing probe ids. Since they are not getting a FS discriminator, they will also be excluded from MIR profile loading. The corresponding changes will be in the subsequent patches.

Reviewed By: wenlei

Differential Revision: https://reviews.llvm.org/D147286

This makes it possible to use TRAP to make Linux system calls using
inline assembly for instance.

Differential Revision: https://reviews.llvm.org/D147102

Calling symbolization directly from stopTheWorld was causing deadlock.
For libc dep systems, symbolization uses dl_iterate_phdr, which acquire a
dl write lock. It could deadlock if the lock is already acquired by one of
suspended.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D146990

This test doesn't actually depend on being able to launch the process.
This may or may not explain why this test behaves oddly on some of our bots.

This is useful for retriving crash reports of LIT runs when the
temporary folder is not accessible.

Reviewed By: michaelplatings

Differential Revision: https://reviews.llvm.org/D147209

Fixes BPF assembler parsing errors for the following instructions:
- atomic_fetch_add
- atomic_fetch_and
- atomic_fetch_xor
- atomic_fetch_or
- cmpxchg32_32
- cmpxchg_64
- xchg32_32
- xchg_64

Also add a test to verify that all instructions could be assembled and disassembled.

Differential Revision: https://reviews.llvm.org/D147421

IRTranslator lowers switches to [G_SUB] + G_ICMP + G_BRCOND + G_BR
sequences. Since values and destination MBBs are included in the
spv_switch intrinsics, the sequences are not needed for ISel.

Before this commit, the information decoded by these sequences were
added to spv_switch intrinsics in SPIRVPreLegalizer and the sequences
were kept until SPIRVModuleAnalysis where they were marked skipped for
emission.

After this commit, the [G_SUB] + G_ICMP + G_BRCOND + G_BR sequences
and MBBs containing only these MIs are erased in SPIRVPreLegalizer.

Differential Revision: https://reviews.llvm.org/D146923

This code currently assumes that all bulk memory operations occur on
memory 0 which who's type is determined by the wasm32 vs wasm64 target
triple.  Further improvements would be need to support multi-memory.

Differential Revision: https://reviews.llvm.org/D147540

Ironically, MSan copies uninitialized data off the stack into
VAArgTLSCopy in the callee-side handling of va_start. Clamp the copy
size to the actual length of the buffer, and zero-initialize the
remainder.

Differential Revision: https://reviews.llvm.org/D146858

For an expression of the form `&DRE[any]` under an Unspecified
Pointer Context (UPC), we generate a fix-it for it with respect to a
strategy. In case the strategy is `std::span` (it is the only supported
one for now), the fix-it replaces the expression with
`&DRE.data()[any]`.

A UPC includes at least the contexts where
- the expression is being casted to an integer; and
- the expression is an argument of a call to a function that is not marked unsafe.

Reviewed by: NoQ, malavikasamak, t-rasmud, jkorous

Differential revision: https://reviews.llvm.org/D143128

Commit 392d9eb0 added a dependency on
FPE_FLTIDO which was only defined in FreeBSD main on May 19, 2022 and
is not in all supported releases. Just define it if it's missing as we
could use a debugger compiled on an older system to debug a newer one.

Reviewed by: DavidSpickett, emaste, dim

Differential Revision: https://reviews.llvm.org/D147300

This preserves the spirit of the tests even if a follow-up changes only
generates exit values for recurrences if they are actually used.

This is a follow up on D146976.

Reviewed By: stephenpeckham

Differential Revision: https://reviews.llvm.org/D147559

This matches the actual lowering.  The previous costing was "as if" it had been fully scalarized.

Extend the verifier to check if the size of the matrix operands of
matrix.multiply match the sizes specified by the numeric arguments.

Reviewed By: thegameg

Differential Revision: https://reviews.llvm.org/D147466

This issue is caused by incomplete implementation of isKeyOf for DIStringType.

Differential Revision: https://reviews.llvm.org/D147140

This revision implements HLFIR to FIR lowering for complex parts.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D146487

This patch introduces the IRDL dialect, which allow users to represent
dynamic dialect definitions as an MLIR program.

The IRDL dialect defines operations, attributes, and types, using
attribute constraints. For example:

```
module {
  irdl.dialect @cmath {
    irdl.type @complex {
      %0 = irdl.is f32
      %1 = irdl.is f64
      %2 = irdl.any_of(%0, %1)
      irdl.parameters(%2)
    }

    irdl.operation @norm {
      %0 = irdl.any
      %1 = irdl.parametric @complex<%0>
      irdl.operands(%1)
      irdl.results(%0)
    }
}
```

This program will define a new `cmath.complex` type, which expects a single
parameter, which is either an `f32` or an `f64`. It also defines an
`cmath.norm` operation, which expects a single `cmath.complex` type as operand,
and returns a value of the underlying type. Note that like PDL (which IRDL is
heavily inspired from), both uses of `%0` are expected to be of the same attribute.

IRDL handles attributes and types with the same operations, and does this by always
wrapping types in a `TypeAttr`. This is to simplify the language.

Depends on D144690

Reviewed By: rriddle, mehdi_amini

Differential Revision: https://reviews.llvm.org/D144692

This avoids `-Wformat` complains the placeholder type specifier mismatch
on `lld`/`llu`(used for `s64`/`u64`) which have slightly different
interpretation in string_utils.cpp.

Also enable Timer build which was disabled because of the complaining of
`-Wformat`.

Differential Revision: https://reviews.llvm.org/D147496

All the IEEE formats are quite similar, we can merge their code
effectively by writing it parametrically via the fltSemantics object.

We can metaprogram the implementation such that this parametricity is
zero-cost.

The premise here is to allow non-kernel functions to locate external LDS variables without using LDS or extra magic SGPRs to do so.

1/ First it crawls the callgraph to work out which external LDS variables are reachable from a given kernel
2/ Then it creates a new `extern char[0]` variable for each kernel, which will alias all the other extern LDS variables because that's the documented behaviour of these variables
3/ The address of that variable is written to a lookup table. The global variable is tagged with metadata to track what address it was allocated at by codegen
4/ The assembler builds the lookup table using the metadata
5/ Any non-kernel functions use the same magic intrinsic used by table lookups of non-dynamic LDS variables to find the address to use

Heavy overlap with the code paths taken for other lowering, in particular the same intrinsic is used to pass the dynamic scope information through the same sgpr as for table lookups of static LDS.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D144233

Since memory does not have an intrinsic type, we do not need to require value type matching on stores in order to sink them. To facilitate that, this patch finds stores which are sinkable, but have conflicting types, and bitcasts the ValueOperand so they are easily sinkable into a PHINode. Rather than doing fancy analysis to optimally insert the bitcast, we always insert right before the relevant store in the diamond branch. The assumption is that later passes (e.g. GVN, SimplifyCFG) will clean up bitcasts as needed.

Differential Revision: https://reviews.llvm.org/D147348

Also, fix checking of first line in ::parse.  We can't use the
::getLines helper here since that already does comment stripping
internally.

Differential Revision: https://reviews.llvm.org/D147548

This should be implied by --use-reproducer instead as a path is required
for this mode

Differential Revision: https://reviews.llvm.org/D147499

Revert "AMDGPU: Created a subclass for the return address operand in the tail call return instruction"

This reverts commit 461a559b.

This is overly conservative, but at least it's safe.

This is a follow-up to https://reviews.llvm.org/D137707

Currently, the --thinlto-prefix-replace="oldpath;newpath" option is used during
distributed ThinLTO thin links to specify the mapping of the input bitcode object
files' directory tree (oldpath) to the directory tree (newpath) used for both:

1) the output files of the thin link itself (the .thinlto.bc index files and the
optional .imports files)
2) the specified object file paths written to the response file given in the
--thinlto-index-only=${response} option, which is used by the final native
link and must match the paths of the native object files that will be
produced by ThinLTO backend compiles.
This patch expands the --thinlto-prefix-replace option to allow a separate directory
tree mapping to be specified for the object file paths written to the response file
(number 2 above). This is important to support builds and build systems where the
same output directory may not be written by multiple build actions (e.g. the thin link
and the ThinLTO backend compiles).

The new format is: --thinlto-prefix-replace="origpath;outpath[;objpath]"

This replaces the origpath directory tree of the thin link input files with
outpath when writing the thin link index and imports outputs (number 1
above). If objpath is specified it replaces origpath of the input files with
objpath when writing the response file (number 2 above), otherwise it
falls back to the old behavior of using outpath for this as well.

Reviewed By: tejohnson, MaskRay

Differential Revision: https://reviews.llvm.org/D144596

We had this patch downstream for a long time, we need to find the users of the
IR load to guess the bank since with opaque pointers we lost the type information.

The test "big-file-copy.test" introduced in D146189 and constrained to darwin by
D147094, is by this differential further constrained to only those devices with
sufficient resources. Also correct the test to read the environment variable
"result" from the same shell in which it was stored (which may differ on devices).

Reviewed By: thetruestblue

Differential Revision: https://reviews.llvm.org/D147502