| Age | Commit message (Collapse) | Author |
|---|
|
function elimination is off
|
|
class of {Function, GlobalVariable, IFunc} (#125757)
This is a split of https://github.com/llvm/llvm-project/pull/125756
|
|
When a test depends on a new debugserver feature/fix, the API test must
be marked @skipIfOutOfTreeDebugserver because the macOS CI bots test
using the latest Xcode release debugserver. But over time all of these
fixes & new features are picked up in the Xcode debugserver and these
skips can be removed.
We may see unexpected test failures from removing all of these 1+ year
old skips, but that's likely a separate reason the test is failing that
is being papered over by this skip.
|
|
For 0d vector type the rewrite crashes.
|
|
Implement HLSLElementwiseCast excluding support for splat cases
Do not support casting types that contain bitfields.
Partly closes #100609 and partly closes #100619
|
|
Summary:
Currently the `-Xarch` argument needs to re-parse the option, which goes
through every single registered argument. This causes errors when trying
to pass `-O1` through it because it thinks it's a DXC option. This patch
changes the behavior to only allow `clang` options. Concievably we could
detect the driver mode to make this more robust, but I don't know if
there are other users for this.
Fixes: https://github.com/llvm/llvm-project/issues/110325
|
|
Summary:
This patch cleans up how we query the offloading toolchain. We create a
single that is more similar to the existing `getToolChain` driver
function and make all the offloading handlers use it.
|
|
da053415d214d6a66ff5f8c69eb35b2c9ada9caf
|
|
The DAG will now receive a callback whenever a new instruction is
created and will update itself accordingly.
|
|
(#125962)
|
|
Export all symbols from both EC and native symbol tables. If an explicit
export is present in either symbol table, auto-export is disabled for
both.
|
|
This patch is a followup of the previous one: #115922, It adds an option
to turn on emitting non-atomic rmw code sequence instead of atomic rmw.
|
|
I missed a few places to tidy up from before using the tablengen files
directly for the builtins. I didn't remove all of the modulemap entries
and there were two small `.def` files left lingering. This should clean
all of that up. I went through to cross check the list of files and it
looks correct now.
|
|
We were previously checking this after recursing on all callers, but if
we already have a single allocation type there is no need to even look
at any callers. Didn't show a significant improvement overall, but it
does reduce the count of times we enter the identifyClones and do other
checks.
|
|
Fixes: #125102
|
|
Add interface for `sinpi`, `cospi` and `sincospi` and also expose
`sincosf`
|
|
This patch implements generic associative container benchmarks for
containers with unique keys. In doing so, it replaces the existing
std::map benchmarks which were based on the cartesian product
infrastructure and were too slow to execute.
These new benchmarks aim to strike a balance between exhaustive coverage
of all operations in the most interesting case, while executing fairly
rapidly (~40s on my machine).
This bumps the requirement for the map benchmarks from C++17 to C++20
because the common header that provides associative container benchmarks
requires support for C++20 concepts.
|
|
Turns out there are users who use gcc to compile compiler-rt. Using the
clang-specific builtin function `__builtin_readcyclecounter()` does not
work in this case.
Solution is to use inline assembly using the stckf instruction in case
the compiler is not clang.
|
|
Add support for expanding `%b` in `LLVM_PROFILE_FILE` to the binary ID
(build ID). It can be used with `%m` to avoid its signature collisions.
This is supported on all platforms where writing binary IDs into
profiles is implemented, as the `__llvm_write_binary_ids` function is
used.
Fixes #51560.
|
|
Remove some indirection when matching recipe and matcher operands by
directly using fold over parameter pack.
|
|
The file descriptor of the first opened file is not necessarily 3, so we
change the assertion so that it's >= 0 (i.e. not an error.)
Fixes #126106
|
|
The non-GTest library will be shared by unittests of Flang and Flang-RT.
Promote it as a regular library for use by both projects.
In the long term, we may want to convert these to regular GTest checks
to avoid having multiple testing frameworks.
|
|
These tests are for frame handling code with push/pop. To increase
coverage of CFI/Unwind info, this removes the `nounwind` annotations and
regenerates the checks for this test.
|
|
(#125933)
Add initial parsing/sema support for new assumption clause so clause can
be specified. For now, it's ignored, just like the others.
Added support for 'no_openmp_construct' to release notes.
Testing
- Updated appropriate LIT tests.
- Testing: check-all
|
|
73478708839fad8b02b3cfc84959d64a15ba93ca introduced a textual header but
did not update clang's module map. This PR adds the header to the module
map.
|
|
This commit upgrades our npm dependencies to the latest available
version.
I was prompted to this change because `npm run package` failed for me
with an error. The error disappeared after upgrading `@vscode/vsce`. I
also upgraded the other dependencies because I think it's generally
preferable to stay up-to-date.
I did not bump the `@types/vscode` and `@types/node` versions, since
this would effectively make older VS-Code versions unsupported. I also
changed `@types/vscode` to be a precise version match, since we are
claiming compatibility with that version via the `enginges.vscode`
property.
|
|
This patch fixes a bug in the dependency node iterators that would
incorrectly not skip nodes that are not in the current DAG. This
resulted in iterators returning nullptr when dereferenced.
The fix is to update the existing "skip" function to not only skip
non-instruction values but also to skip instructions not in the DAG.
|
|
utility. (#125925)
1. Our static functions are a bit spread out in this file. I am
gathering them in an anonymous namespace
2. Moving the code to get the `target` attribute on a `fir.global` into
its own utility.
|
|
Summary:
This is supposed to be `__llvm_rpc_client` but I screwed it up and
didn't notice at the time. Will need to be backported.
|
|
Suboptimally handled by visitInstruction: llvm.aarch64.neon.
- fcvtas, fcvtau
- fcvtms, fcvtmu
- fcvtns, fcvtnu
- fcvtps, fcvtpu
- fcvtzs, fcvtzu
- fcvtxn
- vcvtfp2fxs, vcvtfp2fxu
- vcvtfxs2fp, vcvtfxu2fp
Forked from llvm/test/CodeGen/AArch64/arm64-{cvt,vcvt}.ll
|
|
Tested call-graph matching on some of Meta's large services, it works to
reuse some renamed function profiles, no negative perf or significant
build speed regression observed. Turned it on by default for CSSPGO
mode.
|
|
This follows suite with disabling float printing.
|
|
The current check in writeFileDefinition() is incorrect, and prevents us
from ever emitting the URL from the clang-doc tool. The unit tests do
test this, but call the API directly circumventing the check.
This is the first step towards addressing #59814.
|
|
|
|
Fix clang-with-thin-lto-ubuntu - failed build
|
|
|
|
|
|
Goals:
1. To add syntax and semantic to 'batch_matmul' without changing any of
the existing syntax expectations for current usage. batch_matmul is
still just batch_matmul.
2. Move the definition of batch_matmul from linalg OpDsl to tablegen ODS
infra.
Scope of this patch:
To expose broadcast and transpose semantics on the 'batch_matmul'.
The broadcast and transpose semantic are as follows:
By default, 'linalg.batch_matmul' behavior will remain as is. Broadcast
and Transpose semantics can be applied by specifying the explicit
attribute 'indexing_maps' as shown below. This is a list attribute, so
the list must include all the maps if specified.
Example Transpose:
```
linalg.batch_matmul indexing_maps = [
affine_map< (d0, d1, d2, d3) -> (d0, d3, d1)>, //transpose
affine_map< (d0, d1, d2, d3) -> (d0, d3, d2)>,
affine_map< (d0, d1, d2, d3) -> (d0, d1, d2)>
]
ins (%arg0, %arg1: memref<2x5x3xf32>,memref<2x5x7xf32>)
outs (%arg2: memref<2x3x7xf32>)
```
Example Broadcast:
```
linalg.batch_matmul indexing_maps = [
affine_map< (d0, d1, d2, d3) -> (d3)>, //broadcast
affine_map< (d0, d1, d2, d3) -> (d0, d3, d2)>,
affine_map< (d0, d1, d2, d3) -> (d0, d1, d2)>
]
ins (%arg0, %arg1: memref<5xf32>,memref<2x5x7xf32>)
outs (%arg2: memref<2x3x7xf32>)
```
Example Broadcast and transpose:
```
linalg.batch_matmul indexing_maps = [
affine_map< (d0, d1, d2, d3) -> (d1, d3)>, //broadcast
affine_map< (d0, d1, d2, d3) -> (d0, d2, d3)>, //transpose
affine_map< (d0, d1, d2, d3) -> (d0, d1, d2)>
]
ins (%arg0, %arg1: memref<3x5xf32>, memref<2x7x5xf32>)
outs (%arg2: memref<2x3x7xf32>)
```
RFCs and related PR:
https://discourse.llvm.org/t/rfc-linalg-opdsl-constant-list-attribute-definition/80149
https://discourse.llvm.org/t/rfc-op-explosion-in-linalg/82863
https://discourse.llvm.org/t/rfc-mlir-linalg-operation-tree/83586
https://github.com/llvm/llvm-project/pull/115319
|
|
We test the `--repository` output in the unit tests, but that option fails to
change the HTML output in the end-to-end tests. Upcoming patches will address
the incorrect behavior.
|
|
The fix requires more investigation, and it's a test issue so reverting
the product changes should not be necessary.
Signed-off-by: Sarnie, Nick <nick.sarnie@intel.com>
|
|
ShouldSkipZero is not set (#126044)
I ran into this while working on a different patch where I'm emitting a
zero-valued DWARF enum field which shouldn't be skipped.
This patch checks the (currently unused) `ShouldSkipZero` before
deciding to skip printing this field. Based on git history this seems
like an oversight from the initial refactor that introduced this. We
have a similar check in `printInt`.
Wasn't sure how to best test this, but tests in an upcoming patch rely
on this functionality (see
https://github.com/llvm/llvm-project/pull/126045).
Currently the only place `ShouldSkipZero` is set to `false` is when
emitting the `DW_LANG_` enum. But the language codes start at `0x1`. So
it never exercised this codepath (and we should probably just make it
not pass this parameter).
|
|
LLVM itself is generally moving away from using `undef` and towards
using `poison`, to the point of having a lint that caches new uses of
`undef` in tests.
In order to not trip the lint on new patterns and to conform to the
evolution of LLVM
- Rename valious ::undef() methods on StructBuilder subclasses to
::poison()
- Audit the uses of UndefOp in the MLIR libraries and replace almost all
of them with PoisonOp
The remaining uses of `undef` are initializing `uninitialized` memrefs,
explicit conversions to undef from SPIR-V, and a few cases in
AMDGPUToROCDL where usage like
%v = insertelement <M x iN> undef, iN %v, i32 0
%arg = bitcast <M x iN> %v to i(M * N)
is used to handle "i32" arguments that are are really packed vectors of
smaller types that won't always be fully initialized.
|
|
NaryReassociate would crash on expressions like the one in the added
test that involved pointers where the size of the type was greater than
the index width of the pointer, causing calls to SCEV's zext expression
on types that didn't need to be zero-extended.
This commit fixes the issue.
|
|
Since line zero means "no line information", when symbolizing a location
(an address or an inline frame associated with the address) that has a
line zero location, we shouldn't include other irrelevant data (like
filename) in the result.
|
|
- Use `Emitter::OptClass` to invoke `InstrInfoEmitter::run` and
eliminate the `EmitInstrInfo` function.
|
|
|
|
|
|
|
|
|
|
z) + z even with Zicond. (#125772)"
With the test changes.
Original message:
The Zicond version of this requires an li instruction and an
additional register.
Without Zicond we match this in a DAGCombine on RISCVISD::SELECT_CC.
This PR has 2 commits. I'll pre-commit the test change if this looks
good.
|
