Simon Pilgrim [Sun, 22 Jan 2023 15:19:17 +0000 (15:19 +0000)]
[X86] avx2-vbroadcast.ll - use X86 check prefix instead of X32
We try to use X32 for tests on gnux32 triples
Markus Böck [Sun, 22 Jan 2023 15:11:27 +0000 (16:11 +0100)]
[mlir][ods] Simplify signature of `custom` printers and parsers of Attributes and Types in presence of default constructible parameters
The vast majority of parameters of C++ types used as parameters for Attributes and Types are likely to be default constructible. Nevertheless, TableGen conservatively generates code for the custom directive, expecting signatures using FailureOr<T> for all parameter types T to accomodate them possibly not being default constructible. This however reduces the ergonomics of the likely case of default constructible parameters.
This patch fixes that issue, while barely changing the generated TableGen code, by using a helper function that is used to pass any parameters into custom parser methods. If the type is default constructible, as deemed by the C++ compiler, a default constructible instance is created and passed into the parser method by reference. In all other cases it is a Noop and a FailureOr is passed as before.
Documentation was also updated to document the new behaviour.
Fixes https://github.com/llvm/llvm-project/issues/60178
Differential Revision: https://reviews.llvm.org/D142301
Simon Pilgrim [Sun, 22 Jan 2023 14:07:20 +0000 (14:07 +0000)]
[X86] commute-3dnow.ll - use X86 check prefix instead of X32
We try to use X32 for tests on gnux32 triples
Simon Pilgrim [Sun, 22 Jan 2023 14:01:04 +0000 (14:01 +0000)]
[X86] avx-vbroadcastf128.ll - use X86 check prefix instead of X32
We try to use X32 for tests on gnux32 triples
Roman Lebedev [Sun, 22 Jan 2023 14:27:17 +0000 (17:27 +0300)]
[NFC][SCEVExpander] `CmpSelCost`: use the cost of the expression, not operand
Currently, for all invocations, it's equivalent, since that is literally
how `SCEVMinMaxExpr::getType()` is defined. But for e.g. `select`,
we'll want to ask about the hand type, and not the type of the operand
that happens to be first.
Roman Lebedev [Sun, 22 Jan 2023 14:35:25 +0000 (17:35 +0300)]
[NFC][SCEV] Reflow `computeSCEVAtScope()` into an exhaustive switch
Roman Lebedev [Sun, 22 Jan 2023 14:15:16 +0000 (17:15 +0300)]
[NFC][SCEV] `getRelevantLoop()`: deduplicate handling
Roman Lebedev [Sun, 22 Jan 2023 14:08:45 +0000 (17:08 +0300)]
[NFC][SCEV] `getBlockDisposition()`: deduplicate handling
Roman Lebedev [Sun, 22 Jan 2023 13:56:28 +0000 (16:56 +0300)]
[NFC][SCEV] `getLoopDisposition()`: deduplicate handling
Roman Lebedev [Sun, 22 Jan 2023 13:32:02 +0000 (16:32 +0300)]
[NFC][SCEV] `computeSCEVAtScope()`: deduplicate handling
Casts and udiv get the exactly the same handling as n-ary,
there is no point in special-handling anything.
Matt Arsenault [Fri, 16 Dec 2022 03:04:36 +0000 (22:04 -0500)]
AMDGPU: Copy a source modifier test for f16/v2f16
This is essentially a modernized copy of
select-fabs-fneg-extract.ll. Stop using kernels with loads and stores,
don't use fsub for fneg, and port the examples to half.
Matt Arsenault [Sun, 18 Dec 2022 12:25:56 +0000 (07:25 -0500)]
AMDGPU: Add modern copy of fneg combines test
Simon Pilgrim [Sun, 22 Jan 2023 13:57:49 +0000 (13:57 +0000)]
[DAG] mergeInsertEltWithShuffle - pull out mergeEltWithShuffle helper. NFCI.
This will allow us to reuse the code to merge an extracted scalar into an updated shuffle in a future patch.
Another step towards fixing some shuffle regressions in D127115.
Roman Lebedev [Sun, 22 Jan 2023 13:14:06 +0000 (16:14 +0300)]
[NFC][X86] Fixup typo in `blend-of-shift.ll`
Roman Lebedev [Sun, 22 Jan 2023 13:13:16 +0000 (16:13 +0300)]
[NFC][X86] Fixup `-mattr=<>` in one runline in `elementwise-store-of-scalar-splat.ll`
Shivam Gupta [Sun, 22 Jan 2023 13:10:58 +0000 (18:40 +0530)]
[NFC] Small indentation fix in lld/ELF/Relocations.cpp
Paul Walker [Fri, 13 Jan 2023 12:00:11 +0000 (12:00 +0000)]
[SVE] Add intrinsics for integer binops that explicitly undefine the result for inactive lanes.
The intent is to lower the clang X form SVE builtins to these
intrinsics. The suffix _x is already in use to signify unpredicated
SVE intrinsics hence my choice to use _u to signify those intrinsics
where the result for inactive lanes is undefined.
Differential Revision: https://reviews.llvm.org/D141937
Jay Foad [Wed, 11 Jan 2023 15:34:37 +0000 (15:34 +0000)]
[Mips] Use MCInstrInfo::get in MipsAsmParser instead of reinventing it. NFC.
Differential Revision: https://reviews.llvm.org/D141503
Owen Pan [Sun, 22 Jan 2023 10:59:23 +0000 (02:59 -0800)]
[clang-format][NFC] Add .clang-format to clang/tools/clang-format/
And reformat ClangFormat.cpp in the directory.
Owen Pan [Sun, 22 Jan 2023 10:37:39 +0000 (02:37 -0800)]
[clang-format][NFC] Set LineEnding to LF in config files
To prevent \r\n line endings from getting into the source files.
Differential Revision: https://reviews.llvm.org/D141098
WANG Xuerui [Sun, 22 Jan 2023 05:24:43 +0000 (13:24 +0800)]
[LoongArch] Allow %pc_lo12 relocs in JIRL's immediate operand position
Currently, gcc-13 will generate such assembly when `-mcmodel=medium`,
which is ostensibly a dirty hack to allow bigger offsets for extern
function calls without having to add more reloc types. This is not the
best way to accomplish the original goal, but such usages will appear
soon and we have to support it anyway.
Example:
```c
extern int foo(int);
int bar(int x) {
return foo(x + 123);
}
```
will produce the following (simplified) assembly when compiled with
`-O2 -mcmodel=medium`:
```
.globl bar
.type bar, @function
bar:
.cfi_startproc
addi.w $r4,$r4,123
pcalau12i $r12,%pc_hi20(foo)
jirl $r0,$r12,%pc_lo12(foo)
.cfi_endproc
```
Reviewed By: SixWeining, wangleiat, MaskRay, xry111
Differential Revision: https://reviews.llvm.org/D142278
Iain Sandoe [Tue, 13 Dec 2022 08:45:08 +0000 (08:45 +0000)]
[C++20][Modules] Fix named module import diagnostics.
We have been incorrectly disallowing imports of named modules in the
global and private module fragments.
This addresses: https://github.com/llvm/llvm-project/issues/59688
Differential Revision: https://reviews.llvm.org/D140927
Benjamin Kramer [Sun, 22 Jan 2023 09:58:47 +0000 (10:58 +0100)]
[bazel] Add missing dependencies for
4f1e244eb5
Johannes Doerfert [Sun, 22 Jan 2023 09:27:41 +0000 (01:27 -0800)]
[OpenMP] Simplify `llvm.assume` operands in device code
Johannes Doerfert [Sun, 22 Jan 2023 09:13:24 +0000 (01:13 -0800)]
[Attributor] Handle constant icmp expressions in AAPotentialValues
A `ConstantExpr` ICmp is pretty much the same thing as an ICmpInst when
we want to simplify it. We just need to be less restrictive wrt. the
type and use the static helper functions directly.
Fixes: https://github.com/llvm/llvm-project/issues/59767
Timm Bäder [Sat, 21 Jan 2023 18:48:37 +0000 (19:48 +0100)]
[clang][Interp][NFCI] Make InitMap::isInitialized() const
Timm Bäder [Sat, 21 Jan 2023 18:32:02 +0000 (19:32 +0100)]
[clang][Interp][NFC] Forward-declare Boolean in PrimTypes.h
We don't need the full header file here.
Timm Bäder [Sat, 21 Jan 2023 16:25:57 +0000 (17:25 +0100)]
[clang][Interp][NFC] Fix header comment file name
Noah Goldstein [Sun, 22 Jan 2023 06:00:14 +0000 (22:00 -0800)]
Transform ctpop(Pow2) -> icmp ne Pow2, 0
This makes folding to 0/1 later on easier and regardless `icmp ne` is
'probably' faster on most targets (especially for vectors).
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D142253
Noah Goldstein [Sun, 22 Jan 2023 05:59:58 +0000 (21:59 -0800)]
Add tests for ctpop(Pow2); NFC
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D142252
Nikolas Klauser [Sat, 21 Jan 2023 07:34:47 +0000 (08:34 +0100)]
[libc++] Rename take_while_view::__sentinel to __take_while_view_sentinel
This makes it easier to specialize traits classes, like __segmented_iterator_traits.
Reviewed By: var-const, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D142276
Yonghong Song [Sun, 22 Jan 2023 06:24:22 +0000 (22:24 -0800)]
[BPF][Clang] Fix func argument pattern in bpf-stack-protector test
Commit
56b038f887f3("[BPF][clang] Ignore stack protector options for BPF
target") added a test for its corresponding functionality.
Douglas Yung found that the test will fail with the release build
buildbot due to different func argument patterns (from %msg
to %0). This patch fixed the issue by using pattern [0-9a-z]+
which allows both %msg and %0.
Johannes Doerfert [Sun, 22 Jan 2023 04:24:43 +0000 (20:24 -0800)]
[OpenMP] Try to fix Flang after new clause was added
Johannes Doerfert [Sun, 22 Jan 2023 03:54:35 +0000 (19:54 -0800)]
[OpenMP][FIX] Split test into amdgpu and nvptx specific ones
This avoids running the test for the host.
Johannes Doerfert [Sun, 22 Jan 2023 03:50:22 +0000 (19:50 -0800)]
[OpenMP][FIX] Add default clause to switch
Johannes Doerfert [Sun, 8 Jan 2023 00:14:48 +0000 (16:14 -0800)]
[OpenMP] Introduce the `ompx_dyn_cgroup_mem(<N>)` clause
Dynamic memory allows users to allocate fast shared memory when a kernel
is launched. We support a single size for all kernels via the
`LIBOMPTARGET_SHARED_MEMORY_SIZE` environment variable but now we can
control it per kernel invocation, hence allow computed values.
Note: Only the nextgen plugins will allocate memory based on the clause,
the old plugins will silently miscompile.
Differential Revision: https://reviews.llvm.org/D141233
Mehdi Amini [Sun, 22 Jan 2023 02:39:49 +0000 (02:39 +0000)]
Add the test dialect as dependent for the "test-legalize-patterns" test pass
Fixes #60183
Mehdi Amini [Sun, 22 Jan 2023 02:26:34 +0000 (02:26 +0000)]
Add missing dependent dialects to "convert-gpu-to-rocdl"
Fixes #60198
Roman Lebedev [Sun, 22 Jan 2023 01:45:38 +0000 (04:45 +0300)]
[NFC][SCEV] `CompareSCEVComplexity`: deduplicate handling
For all but unknown/constant/recurrences, the handling is identical,
there is no point in special-casing anything.
Roman Lebedev [Sun, 22 Jan 2023 01:12:11 +0000 (04:12 +0300)]
[NFC][SCEV] `SCEVTraversal::visitAll()`: deduplicate handling
They don't not do anything different from what we do for n-ary expressions,
there is no point in special-casing them.
Roman Lebedev [Sun, 22 Jan 2023 01:09:39 +0000 (04:09 +0300)]
[NFC][SCEV] `createNodeForSelectOrPHIInstWithICmpInstCond()`: directly take `Type`, not `Instruction`
We don't use the `Instruction` itself, only it's type anyways.
Roman Lebedev [Sun, 22 Jan 2023 01:07:26 +0000 (04:07 +0300)]
[NFC][SCEV] `createNodeForSelectOrPHIInstWithICmpInstCond()`: return optional
We only want about the result if it succeeds, and don't want `SCEVUnknown`.
Noah Goldstein [Sat, 21 Jan 2023 19:33:38 +0000 (11:33 -0800)]
Remove trailing whitespace from comment
Differential Revision: https://reviews.llvm.org/D142289
Kazu Hirata [Sat, 21 Jan 2023 22:48:32 +0000 (14:48 -0800)]
[llvm] Use llvm::bit_width (NFC)
Roman Lebedev [Sat, 21 Jan 2023 22:34:22 +0000 (01:34 +0300)]
[NFC][SCEV] Reflow `getRangeRef()` into an exhaustive switch
Roman Lebedev [Sat, 21 Jan 2023 22:09:13 +0000 (01:09 +0300)]
[NFC][SCEV] Reflow `getRelevantLoop()` into an exhaustive switch
Kazu Hirata [Sat, 21 Jan 2023 21:56:47 +0000 (13:56 -0800)]
[llvm] Use llvm::bit_width (NFC)
Johannes Doerfert [Sat, 21 Jan 2023 21:43:10 +0000 (13:43 -0800)]
[OpenMP][FIX] Runtime args are not kernel args
Clang passes `KernelArgs.NumArgs` to the runtime but not all are kernel
arguments. This ensures we fallback to the old logic. In a follow up we
should introduce a new `KernelArgs.NumKernelArgs` field and set it in
the runtime.
Johannes Doerfert [Sat, 21 Jan 2023 21:23:03 +0000 (13:23 -0800)]
[OpenMP][FIX] Remove version check lines in clang test
We really need a way to make the check line script deal with these
automatically.
Roman Lebedev [Sat, 21 Jan 2023 21:12:27 +0000 (00:12 +0300)]
[X86] `X86TargetLowering`: override `allowsMemoryAccess()`
The baseline `allowsMemoryAccess()` is wrong for X86.
It assumes that aligned memory operations are always allowed,
but that is not true.
For example, We can not perform a 32-byte aligned non-temporal load
of a 32-byte vector, without AVX2 that is, yet `allowsMemoryAccess()`
will say it is allowed, so we may end up merging non-temporal loads,
only to split them up to legalize them, and here we go again.
NOTE: the test changes here are superfluous. The main effect is that without this change,
in D141777, we'd get stuck endlessly merging and splitting non-temporal stores.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D141776
Roman Lebedev [Sat, 21 Jan 2023 20:42:11 +0000 (23:42 +0300)]
[NFC][SCEV] `computeSCEVAtScope()`: reserve vector size upfront
Roman Lebedev [Sat, 21 Jan 2023 20:27:51 +0000 (23:27 +0300)]
[NFC][SCEV] `computeSCEVAtScope()`: `scUnknown`: use early-returns
Roman Lebedev [Sat, 21 Jan 2023 20:23:59 +0000 (23:23 +0300)]
[NFC][SCEV] Reflow `computeSCEVAtScope()` into an exhaustive switch
Otherwise instead of a compile-time error that you forgot to modify it,
you'd get a run-time error, which happened every time i've added new expr.
This is completely NFC, there are no other changes here.
Roman Lebedev [Sat, 21 Jan 2023 20:23:20 +0000 (23:23 +0300)]
[NFC][SCEV] `computeSCEVAtScope()`: clang-format
Argyrios Kyrtzidis [Sat, 21 Jan 2023 19:31:21 +0000 (11:31 -0800)]
[clang/driver] Make sure that `-gno-modules` by itself doesn't enable debug info
Johannes Doerfert [Thu, 19 Jan 2023 21:40:58 +0000 (13:40 -0800)]
[OpenMP] Modernize the kernel launching interface and APIs
We already created a versioned `__tgt_kernel_arguments` struct but it
was only briefly used and its content was passed in isolation anyway.
This makes it hard to add more information in the future. With this
patch we fully embrace the struct as means to pass information from the
compiler to the plugin as part of a kernel launch.
The patch also extends and renames the struct, bumping the version
number to 2. Version 1 entries are auto-upgraded. This is in preparation
for "bare" kernel launches, per kernel dynamic shared memory, CUDA/HIP
lowering, etc.
The `__tgt_target_kernel_nowait` interface was deprecated as it was
unused. Once we actually implement support for something like that, we
can add an appropriate API.
Note: Only plugins with the `launch_kernel` interface are now supported.
That means that a new clang won't be able to use an old runtime.
An old clang can still use the new runtime since the libomptarget
interface did not change.
Differential Revision: https://reviews.llvm.org/D141232
Kazu Hirata [Sat, 21 Jan 2023 18:54:09 +0000 (10:54 -0800)]
[RISCV] Use llvm::bit_width (NFC)
I've verified that the arguments to llvm::bit_width are all of
uint64_t with:
static_assert(std::is_same_v<uint64_t, decltype(Mask)>)
Simon Pilgrim [Sat, 21 Jan 2023 18:23:36 +0000 (18:23 +0000)]
[DAG] Convert static combineABSToABD to DAGCombiner::foldABSToABD. NFCI.
This will make some future legality checks easier.
David Green [Sat, 21 Jan 2023 18:03:24 +0000 (18:03 +0000)]
[ARM] Cortex-M55 Scheduling Model
This adds an Arm Cortex-M55 scheduling model, using the information from
https://developer.arm.com/documentation/102692/latest/
Differential Revision: https://reviews.llvm.org/D141523
Kazu Hirata [Sat, 21 Jan 2023 17:23:39 +0000 (09:23 -0800)]
[AArch64] Simplify isSeveralBitsExtractOpFromShr (NFC)
This patch simplifies isSeveralBitsExtractOpFromShr.
The following statements are equivalent:
unsigned BitWide = 64 - countLeadingOnes(~(AndMask >> SrlImm));
unsigned BitWide = 64 - countLeadingZeros(AndMask >> SrlImm);
Now, consider:
if (BitWide && isMask_64(AndMask >> SrlImm)) {
When isMask_64 returns true, AndMask >> SrlImm and BitWide must be
nonzero. Since BitWide does not contribute to narrowing the
condition, we can simplify the condition as:
if (isMask_64(AndMask >> SrlImm)) {
We can negate the condition for an early exit as recommended by the
LLVM Coding Standards.
Now, all of the following are equivalent if AndMask >> SrlImm is
nonzero:
MSB = BitWide + SrlImm - 1
MSB = (64 - countLeadingZero(AndMask >> SrlImm)) + SrlImm - 1
MSB = (63 - countLeadingZero(AndMask >> SrlImm)) + SrlImm
MSB = 63 - countLeadingZero(AndMask)
MSB = 63 ^ countLeadingZero(AndMask)
MSB = findLastSet(AndMask, ZB_Undefined)
Florian Hahn [Sat, 21 Jan 2023 17:14:13 +0000 (17:14 +0000)]
[VPlan] Consider all recipes in replicate blocks as sink candidates.
Update sinkScalarOperands to consider all operands of recipes in
replicate blocks as sink candidates This enables additional sinking
opportunities and is another step towards retiring LLVM IR-based
sinkScalarOperands.
This enables iterative sinking of operands for successive calls of
sinkScalarOperands.
Depends on D139788.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D139790
Simon Pilgrim [Sat, 21 Jan 2023 15:24:49 +0000 (15:24 +0000)]
[X86] Add DAG tests showing the failure to reassociate IMINMAX nodes to fold constant operands
Test coverage for Issue #58110
Simon Pilgrim [Sat, 21 Jan 2023 14:14:24 +0000 (14:14 +0000)]
[X86] v8i1-masks.ll - add avx512 test coverage and use X86 check prefix instead of X32
We try to use X32 for tests on gnux32 triples
Nikolas Klauser [Thu, 22 Dec 2022 23:23:30 +0000 (00:23 +0100)]
[libc++] Granularize <type_traits> includes in <bit>, <numbers> and <coroutine>
`<coroutine>` seems to be new enough to not be a huge problem.
Reviewed By: Mordante, #libc
Spies: libcxx-commits, ChuanqiXu
Differential Revision: https://reviews.llvm.org/D140600
Nikolas Klauser [Sun, 11 Dec 2022 01:10:31 +0000 (02:10 +0100)]
[libc++] Implement P1413R3 (Deprecate std::aligned_storage and std::aligned_union)
There are no tests for the aliases because clang doesn't diagnose deprecated template aliases currently.
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D127678
ShihPo Hung [Sat, 21 Jan 2023 13:29:05 +0000 (05:29 -0800)]
[Cost] Add CostKind to getVectorInstrCost and its related users
LoopUnroll estimates the loop size via getInstructionCost(),
but getInstructionCost() cannot pass CostKind to getVectorInstrCost().
And so does getShuffleCost() to getBroadcastShuffleOverhead(),
getPermuteShuffleOverhead(), getExtractSubvectorOverhead(),
and getInsertSubvectorOverhead().
To address this, this patch adds an argument CostKind to these
functions.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D142116
Simon Pilgrim [Sat, 21 Jan 2023 13:00:46 +0000 (13:00 +0000)]
[M68k] Fix M68k pipeline order test after
4ece50737d5385fb80cfa23f5297d1111f8eed39
Iain Sandoe [Tue, 17 Jan 2023 10:17:08 +0000 (10:17 +0000)]
[C++20][Modules] Handle defaulted and deleted functions in header units.
Address part of https://github.com/llvm/llvm-project/issues/60079.
Deleted and Defaulted functions are implicitly inline, but that state
is not set at the point that we perform the diagnostic checks for externally-
visible non-inline functions; check the function body type explicitly in the
diagnostic.
Differential Revision: https://reviews.llvm.org/D141908
Jon Chesterfield [Sat, 21 Jan 2023 12:01:13 +0000 (12:01 +0000)]
[openmp] Workaround for HSA in issue 60119
Move plugin initialization to libomptarget initialization.
Removes the call_once control, probably fractionally faster overall.
Fixes issue 60119 because the plugin initialization, which might
try to dlopen unrelated shared libraries, is no longer nested within
a call from application code.
Fixes #60119
Reviewed By: Maetveis, jhuber6
Differential Revision: https://reviews.llvm.org/D142249
Simon Pilgrim [Sat, 21 Jan 2023 11:43:42 +0000 (11:43 +0000)]
[X86] avx2-vperm.ll - use X86 check prefix instead of X32
We try to use X32 for tests on gnux32 triples
Simon Pilgrim [Sat, 21 Jan 2023 11:31:47 +0000 (11:31 +0000)]
[DAG] getNode - add type assertion checks for ISD::ABDS/ABDU
Jay Foad [Sat, 21 Jan 2023 09:26:36 +0000 (09:26 +0000)]
[cmake] Fix some indentation in HandleLLVMOptions.cmake
wanglei [Sat, 21 Jan 2023 05:31:33 +0000 (13:31 +0800)]
[ORC] Add lazy jit support for LoongArch64
This patch adds resolver, indirection and trampoline stubs for
loongarch64, allowing lazy compilation to work.
It assumes hard float feature exists.
Depends on D141036
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D141102
Timm Bäder [Fri, 20 Jan 2023 14:07:25 +0000 (15:07 +0100)]
Re-commit "[clang][Interp] Unify visiting variable declarations"
We often visit the same variable multiple times, e.g. once when checking
its initializer and later when compiling the function. Unify both of
those in visitVarDecl() and do the returning of the value in
visitDecl().
This time, use a VariableScope instead of a DeclScope for local
variables. This way, we don't emit Destroy ops for the local variables
immediately after creating them.
Differential Revision: https://reviews.llvm.org/D136815
Haojian Wu [Sat, 21 Jan 2023 09:09:17 +0000 (10:09 +0100)]
GlobalISel: Fix the broken release build after
c1fc5219cb23d1e14c0115381321f6d8d6a4fc17
Noah Goldstein [Sat, 21 Jan 2023 08:53:14 +0000 (00:53 -0800)]
Make `shouldExpandLogicAtomicRMWInIR` able to match both operands.
Previous logic was buggy and erroneously asserted that I->operand(0) must
be the RMW instruction. This change fixes that and makes it so that the
RMW instruction can be used in operand 0 or 1.
Also update the tests to explicitly test RMW as operand 0/1 (no change
to codegen).
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D142166
Dmitry Vyukov [Fri, 20 Jan 2023 09:26:20 +0000 (10:26 +0100)]
sanmd: refine selection of functions for UAR checking
There are no intrinsic functions that leak arguments.
If the called function does not return, the current function
does not return as well, so no possibility of use-after-return.
Sanitizer function also don't leak or don't return.
It's safe to both pass pointers to local variables to them
and to tail-call them.
Reviewed By: melver
Differential Revision: https://reviews.llvm.org/D142190
hezuoqiang [Sat, 21 Jan 2023 04:28:01 +0000 (23:28 -0500)]
[Clang][test] Avoid FileCheck error when matching `-cc1`
FileCheck patterns consisting of only `-cc1` on a line by itself often
cause mismatches, e.g. with version strings formed from commit hashes
such as "clang-
cc1514432c58". They should be changed to contain more
context and match more precisely.
Differential Revision: https://reviews.llvm.org/D141886
Kazu Hirata [Sat, 21 Jan 2023 04:27:08 +0000 (20:27 -0800)]
[ADT] Add [[nodiscard]] to the functions in bit.h (NFC)
Amir Ayupov [Sat, 21 Jan 2023 03:27:38 +0000 (19:27 -0800)]
[BOLT][NFC] Move getLTOCommonName to Utils
Reuse getLTOCommonName in components other than Profile (to be used in Core)
Reviewed By: #bolt, maksfb
Differential Revision: https://reviews.llvm.org/D142259
wren romano [Thu, 19 Jan 2023 03:11:48 +0000 (19:11 -0800)]
[mlir][sparse] factoring out getRankedTensorType helper function
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D142074
Kazu Hirata [Sat, 21 Jan 2023 03:34:42 +0000 (19:34 -0800)]
[ADT] Add bit_floor, bit_ceil, and bit_width to bit.h
This patch adds C++20-style bit_floor, bit_ceil, and bit_width.
In a subsequent patch, I'm going to define PowerOf2Floor in
MathExtras.h in terms of bit_floor.
Unfortunately, PowerOf2Ceil isn't quite the same as bit_ceil because
PowerOf2Ceil(0) == 0, whereas bit_ceil(0) == 1.
MathExtras.h does not have a function directly corresponding to
bit_width, but Log2_32(X) + 1, which occurs in a few places, can be
replaced with bit_width(X).
Differential Revision: https://reviews.llvm.org/D142179
David Blaikie [Sat, 21 Jan 2023 01:52:20 +0000 (01:52 +0000)]
[bazel][libc] fixes
(Phab review only sent in draft to use bazel presubmit testing -
precommit human review was not required)
Differential Revision: https://reviews.llvm.org/D142267
Amir Ayupov [Fri, 20 Jan 2023 23:50:59 +0000 (15:50 -0800)]
[BOLT] Handle __uniq suffix added by -funique-internal-linkage-names
In profile matching, if `.__uniq` suffix added for internal linkage
symbols with `-funique-internal-linkage-names` prevents BOLT from
matching to a binary function, try to strip the suffix and perform
fuzzy name matching.
Follow-up to D124117.
Reviewed By: #bolt, maksfb
Differential Revision: https://reviews.llvm.org/D142073
Matt Arsenault [Wed, 18 Jan 2023 10:53:18 +0000 (05:53 -0500)]
GlobalISel: Split main function of RegBankSelect up
This will allow for easier overriding of the pass.
Matt Arsenault [Fri, 20 Jan 2023 00:47:18 +0000 (20:47 -0400)]
llvm-reduce: Fix default handling of intermediate format
Bitcode inputs should produce bitcode intermediates by
default.
Matt Arsenault [Tue, 17 Jan 2023 18:35:43 +0000 (13:35 -0500)]
llvm-reduce: Reorganize some function locations
Move things that are naturally methods of ReducerWorkItem to
be methods of ReducerWorkItem in the same source file.
Kazu Hirata [Sat, 21 Jan 2023 02:42:03 +0000 (18:42 -0800)]
Reland: Drop the ZeroBehavior parameter from countLeadingZeros and the like (NFC)
This patch drops the ZeroBehavior parameter from bit counting
functions like countLeadingZeros. ZeroBehavior specifies the behavior
when the input to count{Leading,Trailing}Zeros is zero and when the
input to count{Leading,Trailing}Ones is all ones.
ZeroBehavior was first introduced on May 24, 2013 in commit
eb91eac9fb866ab1243366d2e238b9961895612d. While that patch did not
state the intention, I would guess ZeroBehavior was for performance
reasons. The x86 machines around that time required a conditional
branch to implement countLeadingZero<uint32_t> that returns the 32 on
zero:
test edi, edi
je .LBB0_2
bsr eax, edi
xor eax, 31
.LBB1_2:
mov eax, 32
That is, we can remove the conditional branch if we don't care about
the behavior on zero.
IIUC, Intel's Haswell architecture, launched on June 4, 2013,
introduced several bit manipulation instructions, including lzcnt and
tzcnt, which eliminated the need for the conditional branch.
I think it's time to retire ZeroBehavior as its utility is very
limited. If you care about compilation speed, you should build LLVM
with an appropriate -march= to take advantage of lzcnt and tzcnt.
Even if not, modern host compilers should be able to optimize away
quite a few conditional branches because the input is often known to
be nonzero from dominating conditional branches.
In this iteration, I've moved the forward declarations of
_BitScanForward outside the llvm space to fix builds on Windows.
Differential Revision: https://reviews.llvm.org/D141798
David Blaikie [Sat, 21 Jan 2023 01:26:52 +0000 (01:26 +0000)]
[bazel][libc] fixes
& some aarch64 fixes?
Siva Chandra [Sat, 21 Jan 2023 00:56:07 +0000 (16:56 -0800)]
[libc][Obvious] Restore static keyword of a few static methods.
The static keyword of the concerned methods was accidentally lost in a
previous commit.
Dmitri Gribenko [Sat, 21 Jan 2023 00:22:22 +0000 (01:22 +0100)]
[clang][dataflow] Allow analyzing multiple functions in unit tests
In unit tests for concrete dataflow analyses we typically use the
testonly `checkDataflow()` helper to analyse a free function called
"target". This pattern allows our tests to be uniform and focused on
specific statement- or expression-level C++ features.
As we expand our feature coverage, we want to analyze functions whose
names we don't fully control, like constructors, destructors, operators
etc. In such tests it is often convenient to analyze all functions
defined in the input code, to avoid having to carefully craft an AST
matcher that finds the exact function we're interested in. That can be
easily done by providing `checkDataflow()` with a catch-all matcher like
`functionDecl()`.
It is also often convenient to define multiple special member functions
in a single unit test, for example, multiple constructors, and share the
rest of the class definition code between constructors. As a result, it
makes sense to analyze multiple functions in one unit test.
This change allows `checkDataflow()` to correctly handle AST matchers
that match more than one function. Previously, it would only ever
analyze the first matched function, and silently ignore the rest. Now it
runs dataflow analysis in a loop, and calls `VerifyResults` for each
function that was found in the input and analyzed.
Reviewed By: ymandel, sgatev
Differential Revision: https://reviews.llvm.org/D140859
David Blaikie [Sat, 21 Jan 2023 00:18:19 +0000 (00:18 +0000)]
bazel: libc: Add missing dependency
Mehdi Amini [Sat, 21 Jan 2023 00:16:06 +0000 (01:16 +0100)]
Revert "WIP listeners", totally pushed by mistake!
This reverts commit
2e312a7baec5e8d8f841ee819966ff3f53f274c6.
Should never have been pushed here...
Mehdi Amini [Fri, 20 Jan 2023 23:43:52 +0000 (00:43 +0100)]
WIP listeners
Brad King [Fri, 20 Jan 2023 23:29:11 +0000 (15:29 -0800)]
[clang] Add ElaboratedType sugaring for types on implicit special members
Extend the change from commit
15f3cd6bfc67 ([clang] Implement
ElaboratedType sugaring for types written bare, 2021-10-11, D112374)
to cover types in the signatures of implicit copy-constructor,
copy-assignment, move-constructor, and move-assignment members in
C++ record types.
With this fix, diagnostic messages print types of special members
consistently whether they are explicitly or implicitly defined.
Fixes: https://github.com/llvm/llvm-project/issues/59557
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D141441
Amir Ayupov [Fri, 20 Jan 2023 23:41:37 +0000 (15:41 -0800)]
[BOLT][NFC] Move out ReorderFunctions::printStats
Break out stats-printing code from ReorderFunctions::reorder for brevity.
Reviewed By: #bolt, maksfb
Differential Revision: https://reviews.llvm.org/D142250
Jim Ingham [Fri, 20 Jan 2023 23:40:55 +0000 (15:40 -0800)]
Remove the LINK_COMPONENTS entry from lldb-instr CMakery
The presence of this entry causes the Xcode generator to produce
two references to libLLVMSupport.a, one at the end of the list, and
that causes the build to fail with a link error. The first version
was pulled in when processing the CLANG_LIBS entries in the
lldb/tools/lldb-instr/CMakeLists.txt.
I looked around and the only places where we use LINK_COMPONENTS and
CLANG_LIBS together are various lldb subdirectories that produce .a
files but don't actually link anything.
Anyway, this line doesn't seem to be needed for either the regular
Ninja or the Xcode generator, and removing it makes the xcode build
work again.
Differential Revision: https://reviews.llvm.org/D142241
Joseph Huber [Fri, 20 Jan 2023 23:34:26 +0000 (17:34 -0600)]
[OpenMP] Remove unfinished and unused 'Analyzer' tool
Summary:
This patch removes a tool that was never finished and has no plans of
being picked up again. It does not need to live in LLVM source in an
unusable state.
Joseph Huber [Thu, 19 Jan 2023 17:49:15 +0000 (11:49 -0600)]
[OpenMP] Clean up AMD handling for `-fopenmp-targets=amdgcn` arch inference
Previously we had some special handling here that errored out if
multiple architectures were detected. This isn't a problem anymore as
the runtime can handle multi-archicture binaries automatically. So it's
safe to simply take the first architecture that we know works. If users
use `--offload-arch=native` instead it will build for all the
architectures at the same time rather than just picking one. This patch
makes it consisten with the NVPTX version.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D142138
Eduard Zingerman [Fri, 20 Jan 2023 23:11:00 +0000 (15:11 -0800)]
[BPF][clang] Ignore stack protector options for BPF target
Stack protector builtin functions are not implemented for BPF target,
thus compiling programs with one of the following options would result
in an error:
-fstack-protector
-fstack-protector-all
-fstack-protector-strong
This commit adds logic to ignore these options for BPF target.
Searching through DiagnosticDriverKinds.td shows that all messages for
such kind of behavior are implemented as warnings, this commit follows
the suit.
Here is an example of the diagnostic message:
clang-16: warning: ignoring '-fstack-protector' option as it is not currently supported for target 'bpf' [-Woption-ignored]
Differential Revision: https://reviews.llvm.org/D142046