Jonas Devlieghere [Thu, 13 Apr 2023 21:19:39 +0000 (14:19 -0700)]
[lldb] Use the host's target triple in TestObjectFileJSON
Use the target's triple when adding JSON modules in TestObjectFileJSON.
This should fix the test failure on non-Darwin bots.
Jonas Devlieghere [Thu, 13 Apr 2023 20:02:45 +0000 (13:02 -0700)]
[lldb] Make ObjectFileJSON loadable as a module
This patch adds support for creating modules from JSON object files.
This is necessary for the crashlog use case where we don't have either a
module or a symbol file. In that case the ObjectFileJSON serves as both.
The patch adds support for an object file type (i.e. executable, shared
library, etc). It also adds the ability to specify sections, which is
necessary in order specify symbols by address. Finally, this patch
improves error handling and fixes a bug where we wouldn't read more than
the initial 512 bytes in GetModuleSpecifications.
Differential revision: https://reviews.llvm.org/D148062
Florian Hahn [Thu, 13 Apr 2023 21:00:51 +0000 (22:00 +0100)]
[VPlan] Switch to checking sinking legality for recurrences in VPlan.
Building on D142885 and D142589, retire the SinkAfter map from the
recurrence handling code. It is replaced by checking whether it is
possible to sink all users of a recurrence directly in VPlan. This
results in simpler code overall and allows to handle additional cases
(see the improvements in @test_crash).
Depends on D142885.
Depends on D142589.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D142886
Leandro Lupori [Fri, 27 Jan 2023 18:34:00 +0000 (18:34 +0000)]
Reland "[compiler-rt] Fix scudo build on ARM"
This reverts commit
5abef0bdbe9237b5728215bb7cd915ffb960e5c3 and
fixes Fuchsia AArch64 cross-compile.
Nicolas Vasilache [Thu, 13 Apr 2023 15:27:40 +0000 (08:27 -0700)]
[mlir][Vector] Add a masked vectorization of tensor.pad
This revision takes advantage of masking support to introduce a vectorized
version of pad that does not require lowering to lower-level form.
Lowering to lower-level form (if/else + generate + fill + copy + insert_slice)
creates unnecessary complexity that can be completely sidestepped by using
masked vectorization properly.
Differential Revision: https://reviews.llvm.org/D148261
Fangrui Song [Thu, 13 Apr 2023 19:40:55 +0000 (12:40 -0700)]
[AArch64] Fix -Wunused-variable in -DLLVM_ENABLE_ASSERTIONS=off builds
Florian Hahn [Thu, 13 Apr 2023 19:23:56 +0000 (20:23 +0100)]
[LV] Add exit users for recurrences in test, fix names.
Add users of exit values for recurrences to make sure exit value
generation will be checked in a follow-up change.
Also adjusts/fixes naming in the test.
Nicolas Vasilache [Thu, 13 Apr 2023 19:19:22 +0000 (12:19 -0700)]
[mlir][Vector] Add a vector.materialize_masks transform operation
Craig Topper [Thu, 13 Apr 2023 18:53:43 +0000 (11:53 -0700)]
[LoopIdiomRecognize] Remove NUW flag from SCEV in getTripCount.
Based on the conversation in D147355.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D148170
Matt Arsenault [Tue, 11 Apr 2023 18:20:05 +0000 (14:20 -0400)]
Attributor: Add baseline tests for nofpclass changes
Matt Arsenault [Thu, 13 Apr 2023 15:42:35 +0000 (11:42 -0400)]
ValueTracking: Add some more uitofp/sitofp tests
Before computeKnownFPClass had a use, these were
only covered by unit tests and didn't cover vectors.
Ivan Trofimov [Fri, 3 Feb 2023 08:29:00 +0000 (00:29 -0800)]
[runtimes][asan] Fix swapcontext interception
Resetting oucp's stack to zero in swapcontext interception is incorrect,
since it breaks ucp cleanup after swapcontext returns in some cases:
Say we have two contexts, A and B, and we swapcontext from A to B, do
some work on Bs stack and then swapcontext back from B to A. At this
point shadow memory of Bs stack is in arbitrary state, but since we
can't know whether B will ever swapcontext-ed to again we clean up it's
shadow memory, because otherwise it remains poisoned and blows in
completely unrelated places when heap-allocated memory of Bs context
gets reused later (see https://github.com/llvm/llvm-project/issues/58633
for example). swapcontext prototype is swapcontext(ucontext* oucp,
ucontext* ucp), so in this example A is oucp and B is ucp, and i refer
to the process of cleaning up Bs shadow memory as ucp cleanup.
About how it breaks:
Take the same example with A and B: when we swapcontext back from B to A
the oucp parameter of swapcontext is actually B, and current trunk
resets its stack in a way that it becomes "uncleanupable" later. It
works fine if we do A->B->A, but if we do A->B->A->B->A no cleanup is
performed for Bs stack after B "returns" to A second time. That's
exactly what happens in the test i provided, and it's actually a pretty
common real world scenario.
Instead of resetting oucp's we make use of uc_stack.ss_flags to mark
context as "cleanup-able" by storing stack specific hash. It should be
safe since this field is not used in [get|make|swap]context functions
and is hopefully never meaningfully used in real-world scenarios (and i
haven't seen any).
Fixes https://github.com/llvm/llvm-project/issues/58633
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D137654
Matt Arsenault [Sat, 8 Apr 2023 23:13:56 +0000 (19:13 -0400)]
ValueTracking: Add baseline test for computeKnownFPClass for sin/cos
Matt Arsenault [Wed, 12 Apr 2023 14:35:13 +0000 (10:35 -0400)]
ValueTracking: Add cannotBeOrderedLessThanZero to KnownFPClass
Eventually we should be able to replace the existing
CannotBeOrderedLessThanZero.
Matt Arsenault [Sat, 8 Apr 2023 19:24:33 +0000 (15:24 -0400)]
ValueTracking: Address todo for nan fmul handling in computeKnownFPClass
If both operands can't be zero or nan, the result can't be nan.
Matt Arsenault [Sat, 8 Apr 2023 18:35:15 +0000 (14:35 -0400)]
ValueTracking: Handle no-nan check for computeKnownFPClass for fmul
Copy the logic from isKnownNeverNaN for fadd/fsub. Leave the
extension to handle the zero case for a future change.
Matt Arsenault [Sat, 8 Apr 2023 22:37:16 +0000 (18:37 -0400)]
ValueTracking: Add baseline test for fmul computeKnownFPClass handling
Simon Pilgrim [Thu, 13 Apr 2023 18:39:47 +0000 (19:39 +0100)]
[X86] Only fold PTEST->TESTP on AVX targets
While PTEST is a SSE41 instruction, TESTPS/TESTPD was only added for AVX
Simon Pilgrim [Thu, 13 Apr 2023 18:31:17 +0000 (19:31 +0100)]
[X86] Split 128 and 256-bit PTEST combines
We are missing SSE41 test coverage (and in fact have a bug with a PTEST->TESTP fold that is only valid on AVX).
Alexey Bataev [Thu, 13 Apr 2023 18:30:25 +0000 (11:30 -0700)]
[SLP][NFC]Remove extra semicolons after function definitions, NFC
Valentin Clement [Thu, 13 Apr 2023 18:32:35 +0000 (11:32 -0700)]
[flang][openacc] Add proper TODO for the block construct
Trigger the not yet implemented message for
block constructs that are not lowered.
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D148253
Ziqing Luo [Thu, 13 Apr 2023 18:22:45 +0000 (11:22 -0700)]
[-Wunsafe-buffer-usage] A follow-up fix to
762af11d4c5d0bd1e76f23a07087773db09ef17d
Add -triple=arm-apple to the test `SemaCXX/warn-unsafe-buffer-usage-fixits-pre-increment.cpp`
Fangrui Song [Thu, 13 Apr 2023 18:17:35 +0000 (11:17 -0700)]
Revert "[Modules] Remove unnecessary check when generating name lookup table in ASTWriter"
This reverts commit
bc95f27337c7ed77c28e713c855272848f01802a, originally
db987b9589be1eb604fcb74c85b410469e31485f.
clang/test/Modules/pr61065.cppm is retained to make relands show less diff.
There are other module-related issues that were not caught, related to
false positive errors like
"error: no matching constructor for initialization of 'union (anonymous union at ..."
Reopen #61065
Ellis Hoag [Thu, 13 Apr 2023 17:58:29 +0000 (10:58 -0700)]
Fix warnings in InstrProfTest.cpp
The warnings were introduced in https://reviews.llvm.org/D148150
Reviewed By: snehasish
Differential Revision: https://reviews.llvm.org/D148259
Blue Gaston [Thu, 13 Apr 2023 18:10:21 +0000 (11:10 -0700)]
[Test][Sanitizer][atos] Disable atos-symbolized-recovery test
This test tests uses undefined behavior and is proving to be
very flakey. I am disabling for now.
Radar to add a new test: rdar://
108003900
rdar://
107846128
Balaji V. Iyer [Thu, 13 Apr 2023 17:58:14 +0000 (17:58 +0000)]
[mlir][math] Expand math.round to truncate, compare and increment.
Round functions are pushed directly to libm. This is problematic for
situations where libm is not available. This patch will decompose the
roundf function by adding 0.5 to positive number to input
(subtracting for negative) following by a truncate.
Reviewed By: rsuderman
Differential Revision: https://reviews.llvm.org/D148026
Simon Pilgrim [Thu, 13 Apr 2023 18:00:35 +0000 (19:00 +0100)]
[CostModel][X86] Add latency/code-size/size-latency test coverage for integer add/sub saturation intrinsics
Florian Hahn [Thu, 13 Apr 2023 18:00:30 +0000 (19:00 +0100)]
[Matrix] Fix IsSupported check in lowerDotProduct.
The check incorrectly checks the RHS while LHS is transformed later.
Update to check LHS, which fixes a crash in the newly added test cases.
Vitaly Buka [Thu, 13 Apr 2023 17:58:33 +0000 (10:58 -0700)]
[nfc][asan] Reformat the file
Fangrui Song [Thu, 13 Apr 2023 17:55:15 +0000 (10:55 -0700)]
[lldb] Fix library layering after D145574
Craig Topper [Thu, 13 Apr 2023 17:39:47 +0000 (10:39 -0700)]
[lld][RISCV] Implement GP relaxation for R_RISCV_HI20/R_RISCV_LO12_I/R_RISCV_LO12_S.
This implements support for relaxing these relocations to use the GP
register to compute addresses of globals in the .sdata and .sbss
sections.
This feature is off by default and must be enabled by passing
--relax-gp to the linker.
The GP register might not always be the "global pointer". It can
be used for other purposes. See discussion here
https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/371
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D143673
Simon Pilgrim [Thu, 13 Apr 2023 17:42:06 +0000 (18:42 +0100)]
[CostModel][X86] Add SSE2 test coverage to integer add/sub saturation tests
Ellis Hoag [Wed, 12 Apr 2023 17:21:38 +0000 (10:21 -0700)]
[InstrProf][Temporal] Add weight field to traces
As discussed in [0], add a `weight` field to temporal profiling traces found in profiles. This allows users to use the `--weighted-input=` flag in the `llvm-profdata merge` command to weight traces from different scenarios differently.
Note that this is a breaking change, but since [1] landed very recently and there is no way to "use" this trace data, there should be no users of this feature. We believe it is acceptable to land this change without bumping the profile format version.
[0] https://reviews.llvm.org/D147812#4259507
[1] https://reviews.llvm.org/D147287
Reviewed By: snehasish
Differential Revision: https://reviews.llvm.org/D148150
Sam Clegg [Thu, 13 Apr 2023 02:33:12 +0000 (19:33 -0700)]
[lld][WebAssembly] Trace export of symbols when specified with --trace-symbol. NFC
Differential Revision: https://reviews.llvm.org/D148190
Valentin Clement [Thu, 13 Apr 2023 17:35:37 +0000 (10:35 -0700)]
[mlir][openacc] Add acc.serial operation
The acc.serial operation models the OpenACC serial construct.
The serial construct defines a region of a program that is to be
executed sequentially on the current device.
The operation is modelled on the acc.parallel operation and will
receive similar updates when the data operands operations will
be implemented.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D148250
Fangrui Song [Thu, 13 Apr 2023 17:28:58 +0000 (10:28 -0700)]
Internalize AllocationBegin functions after D147005
Reviewed By: thurston
Differential Revision: https://reviews.llvm.org/D148195
Alexey Bataev [Wed, 12 Apr 2023 17:36:57 +0000 (10:36 -0700)]
[SLP]Introduce gather cost estimation function.
Introduced BoUpSLP::ShuffleCostEstimator::gather function as an initial
implementation of the gather/buildvector cost estimation for buildvector
nodes. It will allow to use general codegen infrastructure for better
cost estimation + it improves the cost estimation for the
gathers/buildvectors.
Improved part of D110978.
Differential Revision: https://reviews.llvm.org/D148174
Valentin Clement [Thu, 13 Apr 2023 17:14:37 +0000 (10:14 -0700)]
[mlir][openacc][NFC] Use assembly format for acc.parallel
Remove the custoom parser and printer for the acc.parallel
operation and use the assembly format directly.
Reviewed By: PeteSteinfeld, razvanlupusoru
Differential Revision: https://reviews.llvm.org/D148183
Simon Pilgrim [Thu, 13 Apr 2023 17:07:03 +0000 (18:07 +0100)]
[CostModel][X86] Add latency/code-size/size-latency target costs for minnum/maxnum intrinsics
Using the latest version of the script from D103695 to compare costmodel vs llvm-mca statistics.
Avoids using the default costs, which was assuming libm calls.
ruturaj4 [Thu, 13 Apr 2023 12:48:30 +0000 (13:48 +0100)]
[clang][ExtractAPI] Complete declaration fragments for TagDecl types defined in a typedef
enums and structs declared inside typedefs have incorrect declaration fragments, where the typedef keyword and other syntax is missing.
For the following struct:
typedef struct Test {
int hello;
} Test;
The produced declaration is:
"declarationFragments": [
{
"kind": "keyword",
"spelling": "struct"
},
{
"kind": "text",
"spelling": " "
},
{
"kind": "identifier",
"spelling": "Test"
}
],
instead the declaration fragments should represent the following
typedef struct Test {
…
} Test;
This patch removes the condition in SymbolGraphSerializer.cpp file and completes declaration fragments
Reviewed By: dang
Differential Revision: https://reviews.llvm.org/D146385
Ivan Butygin [Wed, 12 Apr 2023 15:35:50 +0000 (17:35 +0200)]
[mlir][scf] Make whileOp builder funcs optional
Create empty block without the terminator in this case.
Differential Revision: https://reviews.llvm.org/D148137
Peter Klausler [Wed, 12 Apr 2023 19:32:37 +0000 (12:32 -0700)]
[flang][runtime] Reset the left tab limit when flushing output
When flushing output to a non-positionable tty or socket file, reset the
left tab limit. Otherwise, non-advancing output to that file will contain
an increasing amount of leading spaces in each flush. Also, detect
newline characters in stream output, and treat them as record
advancement.
Differential Revision: https://reviews.llvm.org/D148157
Nick Desaulniers [Thu, 13 Apr 2023 16:30:19 +0000 (09:30 -0700)]
[StackProtector] don't check stack protector before calling nounwind functions
https://reviews.llvm.org/rGd656ae28095726830f9beb8dbd4d69f5144ef821
introduced a additional checks before calling noreturn functions in
response to this security paper related to Catch Handler Oriented
Programming (CHOP):
https://download.vusec.net/papers/chop_ndss23.pdf
See also:
https://bugs.chromium.org/p/llvm/issues/detail?id=30
This causes stack canaries to be inserted in C code which was
unexpected; we noticed certain Linux kernel trees stopped booting after
this (in functions trying to initialize the stack canary itself).
https://github.com/ClangBuiltLinux/linux/issues/1815
There is no point checking the stack canary like this when exceptions
are disabled (-fno-exceptions or function is marked noexcept) or for C
code. The GCC patch for this issue does something similar:
https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=
a25982ada523689c8745d7fb4b1b93c8f5dab2e7
Android measured a 2% regression in RSS as a result of
d656ae280957 and
undid it globally:
https://android-review.googlesource.com/c/platform/build/soong/+/2524336
Reviewed By: xiangzhangllvm
Differential Revision: https://reviews.llvm.org/D147975
Simon Pilgrim [Thu, 13 Apr 2023 16:21:15 +0000 (17:21 +0100)]
[CostModel][X86] Add latency/code-size test coverage for minnum/maxnum intrinsics
And improve test coverage for the existing size-latency tests to match
Antoine Moynault [Thu, 13 Apr 2023 16:27:13 +0000 (16:27 +0000)]
[llvm-readobj] fix unit test failure on 32bit machines
Several bots are failing on 32-bit since https://reviews.llvm.org/D145761 was merged
https://lab.llvm.org/buildbot/#/builders/178/builds/4384
It seems due to the use of uintptr_t (32bit here) for storing 64 bit values.
Issue is fixed by replacing to uint64_t (as suggested by DavidSpickett).
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D148208
Arthur Eubanks [Thu, 13 Apr 2023 16:09:18 +0000 (09:09 -0700)]
Revert "[COFF] Add MC support for emitting IMAGE_WEAK_EXTERN_ANTI_DEPENDENCY symbols"
This reverts commit
fffdb7eac58b4efde5e23c1281e7a7f93a42d280.
Causes crashes, see https://reviews.llvm.org/D145208
Balaji V. Iyer [Thu, 13 Apr 2023 15:54:21 +0000 (15:54 +0000)]
[mlir][math] Expand math.exp2 to use math.exp.
Exp2 functions are pushed directly to libm. This is problematic for
situations where libm is not available. This patch will expand the exp2
function to use exp2 with the input multiplied by ln2 (natural log).
Reviewed By: rsuderman
Differential Revision: https://reviews.llvm.org/D148064
Matt Arsenault [Thu, 13 Apr 2023 16:05:05 +0000 (12:05 -0400)]
InstSimplify: Regenerate checks in a test
For some reason this was missing the "Assertions have been autogenerated"
comment so were getting spurious diffs in a future change.
Simon Pilgrim [Thu, 13 Apr 2023 15:59:35 +0000 (16:59 +0100)]
[SLP] Compute min/max scalar reduction costs using min/max intrinsics instead of expanded cmp+sel
By default these will expand back to cmp/sel, but some targets (X86) has optimized costs for scalar integer min/max patterns which are lower than the default expansion (pre-SSE41 is particularly weak for vector min/max support).
Differential Revision: [SLP] Compute min/max scalar reduction costs using min/max intrinsics instead of expanded cmp+sel
Jay Foad [Thu, 13 Apr 2023 14:53:09 +0000 (15:53 +0100)]
[AMDGPU] Allow use of TTMP registers in AMDGPUResourceUsageAnalysis
With architected SGPRs, workgroup IDs are passed into a compute shader
in TTMP registers. Allow for this in AMDGPUResourceUsageAnalysis instead
of failing an assertion.
Differential Revision: https://reviews.llvm.org/D148239
Jay Foad [Thu, 13 Apr 2023 15:52:03 +0000 (16:52 +0100)]
[AMDGPU] Avoid else-after-return in isLegalAddressingMode. NFC.
Lawrence Benson [Thu, 13 Apr 2023 15:44:09 +0000 (16:44 +0100)]
[AArch64] Add more efficient vector bitcast for AArch64
Adds a DAG combine checks for vector comparisons followed by a bitcast to a
scalar value. Previously, this resulted in an expand. Now, this is done with a
constant number of instructions that take one bit per vector value (via an AND
mask) and perfom a horizontal add to get a single value. This is especially
useful for Clang's __builtin_convertvector() to a bool vector.
Issue: https://github.com/llvm/llvm-project/issues/59829
Differential Revision: https://reviews.llvm.org/D145301
Simon Pilgrim [Thu, 13 Apr 2023 15:40:37 +0000 (16:40 +0100)]
[IR] llvm::createMinMaxOp - create integer min/max intrinsics instead of icmp/sel
Based off D148215, when expanding a min/max reduction we should be creating min/max intrinsics directly instead of relying on instcombine to fold them back together.
This patch handles integer min/max cases. Hopefully we can add floating point support soon (at least for fastmath/nnan cases) - but we're missing some of the plumbing to pass the correct FMF to the intrinsic at the moment.
Differential Revision: https://reviews.llvm.org/D148221
Caroline Tice [Wed, 12 Apr 2023 18:13:25 +0000 (11:13 -0700)]
[lldb] Add 'CHECK' to class-type-nullptr-deref.s test.
This test previously relied on just segfaulting or not. This commit adds
a CHECK statement to the test.
Differential Revision: https://reviews.llvm.org/D148151
Chris Jones [Thu, 13 Apr 2023 15:05:26 +0000 (17:05 +0200)]
Use `bytes`, not `str`, to return C++ strings to Python.
`str` must be valid UTF-8, which is not guaranteed for C++ strings.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D147818
Nicole Rabjohn [Thu, 13 Apr 2023 15:01:47 +0000 (11:01 -0400)]
[libcxx][AIX] Reverting XFAILs for test cases after OS update
These test cases were fixed with AIX 73TL1, and are currently passing on AIX machines with that fix. This fix has also been backported to the 7.2 service line. These were tested on a machine with AIX 7.2 TL 5 SP4 installed.
Differential Revision: https://reviews.llvm.org/D148040
Zequan Wu [Tue, 28 Mar 2023 16:55:47 +0000 (12:55 -0400)]
[Coverage] Handle invalid end location of an expression/statement.
Fix a crash when an expression/statement can have valid start location but invalid end location in some situations. For example: https://github.com/llvm/llvm-project/blob/llvmorg-16.0.1/clang/lib/Sema/SemaExprCXX.cpp#L1536
This confuses `CounterCoverageMappingBuilder` when popping a region from region
stack as if the end location is a macro or include location.
Reviewed By: hans, aaron.ballman
Differential Revision: https://reviews.llvm.org/D147073
Utkarsh Saxena [Thu, 13 Apr 2023 10:47:52 +0000 (12:47 +0200)]
[clangd] Use FileEntryRef for canonicalizing filepaths.
Using FileEntry for retrieving filenames give the name used for the last access. FileEntryRef gives the first access name.
Last access name is suspected to change with unrelated changes in clang or the underlying filesystem. First access name gives more stability to the name and makes it easier to track.
Differential Revision: https://reviews.llvm.org/D148213
Vlad Serebrennikov [Thu, 13 Apr 2023 14:48:24 +0000 (17:48 +0300)]
[clang][NFC] Use a different service for CWG issue links
We've been using https://wg21.link for C++ DR status page, but it forwards non-resolved issues to EDG wiki, which is not useful for general public. This patch replace it with https://cplusplus.github.io/CWG/issues/ .
Job Noorman [Thu, 13 Apr 2023 14:38:45 +0000 (16:38 +0200)]
[BOLT][NFC] Fix UB due to unaligned load in DebugStrOffsetsWriter
The following tests fail when enabling UBSan due to an unaligned memory
load:
> runtime error: load of misaligned address 0x620000000643 for type
> 'const uint32_t' (aka 'const unsigned int'), which requires 4 byte
> alignment
BOLT :: AArch64/asm-func-debug.test
BOLT :: AArch64/update-debug-reloc.test
BOLT :: X86/asm-func-debug.test
BOLT :: X86/dwarf5-df-dualcu.test
BOLT :: X86/dwarf5-df-mono-dualcu.test
BOLT :: X86/dwarf5-ftypes-dwp-input-dwo-output.test
BOLT :: X86/dwarf5-locaddrx.test
BOLT :: X86/dwarf5-split-dwarf4-monolithic.test
BOLT :: X86/inlined-function-mixed.test
BOLT :: non-empty-debug-line.test
This patch fixes this by using read32le for the load.
Reviewed By: ayermolo
Differential Revision: https://reviews.llvm.org/D148217
Nikita Popov [Tue, 11 Apr 2023 14:50:59 +0000 (16:50 +0200)]
[Attributor] Convert some tests to opaque pointers (NFC)
Archibald Elliott [Thu, 13 Apr 2023 10:51:59 +0000 (11:51 +0100)]
[ARM] Fix Crashes in fp16/bf16 Inline Asm
We were still seeing occasional crashes with inline assembly blocks
using fp16/bf16 after my previous patches:
- https://reviews.llvm.org/rGff4027d152d0
- https://reviews.llvm.org/rG7d15212b8c0c
- https://reviews.llvm.org/rG20b2d11896d9
It turns out:
- The original two commits were wrong, and we should have always been
choosing the SPR register class, not the HPR register class, so that
LLVM's SelectionDAGBuilder correctly did the right splits/joins.
- The `splitValueIntoRegisterParts`/`joinRegisterPartsIntoValue` changes
from rG20b2d11896d9 are still correct, even though they sometimes
result in inefficient codegen of casts between fp16/bf16 and i32/f32
(which is visible in these tests).
This patch fixes crashes in `getCopyToParts` and when trying to select
`(bf16 (bitconvert (fp16 ...)))` dags when Neon is enabled.
This patch also adds support for passing fp16/bf16 values using the 'x'
constraint that is LLVM-specific. This should broadly match how we pass
with 't' and 'w', but with a different set of valid S registers.
Differential Revision: https://reviews.llvm.org/D147715
Timm Bäder [Wed, 1 Feb 2023 06:53:14 +0000 (07:53 +0100)]
[clang][Interp] Add a failing test case
Timm Bäder [Fri, 27 Jan 2023 08:32:20 +0000 (09:32 +0100)]
[clang][Interp] Don't create global variables more than once
just because we're being told to evaluate it twice. This sometimes
happens when a variable is evaluated again during codegen.
Differential Revision: https://reviews.llvm.org/D147535
Jun Zhang [Thu, 13 Apr 2023 13:33:45 +0000 (21:33 +0800)]
[Docs] Point to the correct bug tracker
Signed-off-by: Jun Zhang <jun@junz.org>
Timm Bäder [Mon, 30 Jan 2023 09:24:07 +0000 (10:24 +0100)]
[clang][Interp] Make sure we have a variable scope for initializers
Otherwise, we run into an assertion when trying to use the current
variable scope while creating temporaries for constructor initializers.
Differential Revision: https://reviews.llvm.org/D147534
David Spickett [Fri, 13 Jan 2023 15:50:37 +0000 (15:50 +0000)]
[lldb] Show register fields using bitfield struct types
This change uses the information from target.xml sent by
the GDB stub to produce C types that we can use to print
register fields.
lldb-server *does not* produce this information yet. This will
only work with GDB stubs that do. gdbserver or qemu
are 2 I know of. Testing is added that uses a mocked lldb-server.
```
(lldb) register read cpsr x0 fpcr fpsr x1
cpsr = 0x60001000
= (N = 0, Z = 1, C = 1, V = 0, TCO = 0, DIT = 0, UAO = 0, PAN = 0, SS = 0, IL = 0, SSBS = 1, BTYPE = 0, D = 0, A = 0, I = 0, F = 0, nRW = 0, EL = 0, SP = 0)
```
Only "register read" will display fields, and only when
we are not printing a register block.
For example, cpsr is a 32 bit register. Using the target's scratch type
system we construct a type:
```
struct __attribute__((__packed__)) cpsr {
uint32_t N : 1;
uint32_t Z : 1;
...
uint32_t EL : 2;
uint32_t SP : 1;
};
```
If this register had unallocated bits in it, those would
have been filled in by RegisterFlags as anonymous fields.
A new option "SetChildPrintingDecider" is added so we
can disable printing those.
Important things about this type:
* It is packed so that sizeof(struct cpsr) == sizeof(the real register).
(this will hold for all flags types we create)
* Each field has the same storage type, which is the same as the type
of the raw register value. This prevents fields being spilt over
into more storage units, as is allowed by most ABIs.
* Each bitfield size matches that of its register field.
* The most significant field is first.
The last point is required because the most significant bit (MSB)
being on the left/top of a print out matches what you'd expect to
see in an architecture manual. In addition, having lldb print a
different field order on big/little endian hosts is not acceptable.
As a consequence, if the target is little endian we have to
reverse the order of the fields in the value. The value of each field
remains the same. For example 0b01 doesn't become 0b10, it just shifts
up or down.
This is needed because clang's type system assumes that for a struct
like the one above, the least significant bit (LSB) will be first
for a little endian target. We need the MSB to be first.
Finally, if lldb's host is a different endian to the target we have
to byte swap the host endian value to match the endian of the target's
typesystem.
| Host Endian | Target Endian | Field Order Swap | Byte Order Swap |
|-------------|---------------|------------------|-----------------|
| Little | Little | Yes | No |
| Big | Little | Yes | Yes |
| Little | Big | No | Yes |
| Big | Big | No | No |
Testing was done as follows:
* Little -> Little
* LE AArch64 native debug.
* Big -> Little
* s390x lldb running under QEMU, connected to LE AArch64 target.
* Little -> Big
* LE AArch64 lldb connected to QEMU's GDB stub, which is running
an s390x program.
* Big -> Big
* s390x lldb running under QEMU, connected to another QEMU's GDB
stub, which is running an s390x program.
As we are not allowed to link core code to plugins directly,
I have added a new plugin RegisterTypeBuilder. There is one implementation
of this, RegisterTypeBuilderClang, which uses TypeSystemClang to build
the CompilerType from the register fields.
Reviewed By: jasonmolenda
Differential Revision: https://reviews.llvm.org/D145580
Alex Zinenko [Thu, 13 Apr 2023 12:58:11 +0000 (14:58 +0200)]
[mlir] fix mismerge in bazel
Timm Bäder [Thu, 13 Apr 2023 12:46:30 +0000 (14:46 +0200)]
[clang][github] Try to fix multi-line py comment
Introduced in
09effa706a026c7ebcc01bf14f9f710cb1a8fa87
Archibald Elliott [Thu, 13 Apr 2023 12:40:18 +0000 (13:40 +0100)]
[NFC][ARM] Fix Type in Test
I landed this test with a typo, the callsites all show `fp16_inner`
returning `half`, so the declaration should too.
Timm Bäder [Fri, 24 Mar 2023 16:05:14 +0000 (17:05 +0100)]
[llvm][github] Add good-first-issue comment to issues
Differential Revision: https://reviews.llvm.org/D146819
Archibald Elliott [Wed, 5 Apr 2023 11:33:46 +0000 (12:33 +0100)]
[AArch64][GISel] Legalize non-power-of-two G_CTTZ
The main change here is to add a `widenScalarToNextPow2` before the
`clampScalar` so that non-power-of-two sizes between 32 and 64 are
turned into s64 count trailing zeroes.
However, if you make the legalisation rules depend on TypeIdx 0 (the
output), then you still get crashes for the s65 testcase, which I solved
by instead flipping the rules around to be about TypeIdx 1 (the input),
with a `scalarSameSizeAs` at the end to tie index 0 to index 1. This,
incidentally, is how things are written for `G_CTLZ`.
Differential Revision: https://reviews.llvm.org/D147602
Alex Zinenko [Tue, 11 Apr 2023 14:06:06 +0000 (14:06 +0000)]
[mlir] add structured (Linalg) transform op matchers
Add a set of transform operations into the "structured" extension of the
Transform dialect that allow one to select transformation targets more
specifically than the currently available matching. In particular, add
the mechanism for identifying the producers of operands (input and init
in destination-passing style) and users of results, as well as
mechanisms for reasoning about the shape of the iteration space.
Additionally, add several transform operations to manipulate parameters
that could be useful to implement more advanced selectors. Specifically,
new operations let one produce and compare parameter values to implement
shape-driven transformations.
New operations are placed in separate files to decrease compilation
time. Some relayering of the extension is necessary to avoid repeated
generation of enums.
Depends on D148013
Depends on D148014
Depends on D148015
Reviewed By: chelini
Differential Revision: https://reviews.llvm.org/D148017
David Spickett [Fri, 13 Jan 2023 15:50:08 +0000 (15:50 +0000)]
[lldb] Read register fields from target XML
This teaches ProcessGDBRemote to look for "flags" nodes
in the target XML that tell you what fields a register has.
https://sourceware.org/gdb/onlinedocs/gdb/Target-Description-Format.html
It will check for various invalid inputs like:
* Flags nodes with 0 fields in them.
* Start or end being > the register size.
* Fields that overlap.
* Required properties not being present (e.g. no name).
* Flag sets being redefined.
If anything untoward is found, we'll just drop the field or the
flag set altogether. Register fields are a "nice to have" so LLDB
shouldn't be crashing because of them, instead just log anything
we throw away. So the user can fix their XML/file a bug with their
vendor.
Once that is done it will sort the fields and pass them to
the RegisterFields class I added previously.
There is no way to see these fields yet, so tests for this code
will come later when the formatting code is added.
The fields are stored in a map of unique pointers on the
ProcessGDBRemote class. It will give out raw pointers on the
assumption that the GDB process lives longer than the users
of those pointers do. Which means RegisterInfo is still a trivial struct
but we are properly destroying the fields when the GDB process ends.
We can't store the fields directly in the map because adding new
items may cause its storage to be reallocated, which would invalidate
pointers we've already given out.
Reviewed By: jasonmolenda, JDevlieghere
Differential Revision: https://reviews.llvm.org/D145574
Stefan Gränitz [Thu, 13 Apr 2023 12:21:20 +0000 (14:21 +0200)]
[lldb][ObjectFileELF] Improve error output for unsupported arch/relocations
ObjectFileELF::ApplyRelocations() considered all 32-bit input objects to be i386 and didn't provide good error messages for AArch32 objects. Please find an example in https://github.com/llvm/llvm-project/issues/61948
While we are here, let' improve the situation for unsupported architectures as well. I think we should report the error here too and not silently fail (or crash with assertions enabled).
Reviewed By: SixWeining
Differential Revision: https://reviews.llvm.org/D147627
Job Noorman [Thu, 13 Apr 2023 12:29:18 +0000 (14:29 +0200)]
[BOLT][NFC] Fix UB due to left shift of negative value
The following test fails when enabling UBSan due to a left shift of a
negative value:
> runtime error: left shift of negative value -2
BOLT :: AArch64/ext-island-ref.s
This patch fixes this by using a multiplication instead of a shift.
Reviewed By: yota9
Differential Revision: https://reviews.llvm.org/D148218
Nicolas Vasilache [Tue, 4 Apr 2023 15:17:19 +0000 (08:17 -0700)]
[mlir][Linalg] Fix hoist padding through scf.for iter_arg
Previously, hoisting through an iter_arg would mistakenly yield the unpadded value and
cast it to the padded value.
This was incorrect and resulted in out-of-bounds accesses.
The correct formulation is to yield the padded value and extract a smaller dynamic slice
out of it.
Differential Revision: https://reviews.llvm.org/D148173
David Spickett [Wed, 12 Apr 2023 13:02:45 +0000 (14:02 +0100)]
[llvm][readobj] Add AArch64 SME and SME2 note types
These are used to store new state added by the Scalable Matrix
Extension which is documented in
https://developer.arm.com/documentation/ddi0616/aa/.
The values match those defined by Linux, see:
https://github.com/torvalds/linux/blob/
e62252bc55b6d4eddc6c2bdbf95a448180d6a08d/include/uapi/linux/elf.h#L435
The ZT register(s) are added by SME2 which is not yet publicly
documented but has support in LLVM and Linux already.
Also added descriptions for SVE and PAC_MASK notes since those
were missing.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D148126
Jorge Pinto Sousa [Thu, 13 Apr 2023 12:14:14 +0000 (08:14 -0400)]
Make 'static assertion failed' diagnostics point to the static assertion expression
"static assertion failed" pointed to the static_assert token and
then underlined the static assertion expression:
<source>:3:1: error: static assertion failed
static_assert(false);
^ ~~~~~
1 error generated.
See Godbolt: https://godbolt.org/z/r38booz59
Now it points to and highlights the assertion expression.
Fixes https://github.com/llvm/llvm-project/issues/61951
Differential Revision: https://reviews.llvm.org/D147745
Timm Bäder [Thu, 13 Apr 2023 10:59:38 +0000 (12:59 +0200)]
[clang][NFC] More range for loops in TextDiagnostic.cpp
Timm Bäder [Wed, 12 Apr 2023 12:37:39 +0000 (14:37 +0200)]
[clang][NFC] Fix comment typo
Kadir Cetinkaya [Wed, 12 Apr 2023 19:54:05 +0000 (21:54 +0200)]
[include-cleaner] Handle incomplete template specializations
Instantiation pattern is null for incomplete template types and using
specializaiton decl results in not seeing re-declarations.
Differential Revision: https://reviews.llvm.org/D148158
David Spickett [Mon, 19 Dec 2022 11:58:15 +0000 (11:58 +0000)]
[lldb] Add dummy field to RegisterInfo for register flags use later
This structure is supposed to be trivial, so we cannot simply do
"= nullptr;" on the new member. Doing that means you are non trivial,
regardless of whether you emulate the previously implied constructor somehow.
The next option is to update every use of brace initialisation.
Given that this is some hundreds of lines, this change just adds a dummy
pointer that is set to nullptr. Subsequent changes will actually use that
to point to register flags information.
Note: This change is not clang-format-ted because it changes a bunch of
areas that are not themselves formatted. It would just add noise.
Reviewed By: jasonmolenda, JDevlieghere
Differential Revision: https://reviews.llvm.org/D145568
Simon Pilgrim [Thu, 13 Apr 2023 10:54:44 +0000 (11:54 +0100)]
[TTI][X86] getMinMaxCost - use existing float minnum/maxnum intrinsic cost values instead of maintaining a duplicate cost table
Without fastmath (nnan) flags, minnum/maxnum must perform isnan handling as well as fmin/fmax - meaning the costs are notably higher, this is correctly handled in getIntrinsicInstrCost but was missing from the getMinMaxCost cost tables (which assumed fastmath).
Followup to
63c3895327839ba5b57f5b99ec9e888abf976ac6 which handled the integer cases
David Green [Thu, 13 Apr 2023 10:57:32 +0000 (11:57 +0100)]
[ARM] Replace some uses of -mcpu=cortex-m33 with architectures features. NFC
This adjusts some of the tests to use the architecture features directly as
opposed to -mcpu=cortex-m33 names.
Alex Zinenko [Wed, 12 Apr 2023 09:18:25 +0000 (09:18 +0000)]
[mlir] LLVM lowering: don't use calling convention in op returns
Conversions to the LLVM dialect have an option to use the "bare pointer"
calling convention that converts memref types differently than the
default convention. It has crept into the conversion of operations that
are not related to calls but do require multiresult-to-struct packing.
Use a similar mechanism for the latter without using the calling
convention.
Reviewed By: gysit
Differential Revision: https://reviews.llvm.org/D148086
LLVM GN Syncbot [Thu, 13 Apr 2023 10:55:22 +0000 (10:55 +0000)]
[gn build] Port
6ea45e3007b8
David Spickett [Fri, 5 Aug 2022 14:40:31 +0000 (14:40 +0000)]
[lldb] Add RegisterFlags class
This models the "flags" node from GDB's target XML:
https://sourceware.org/gdb/onlinedocs/gdb/Target-Description-Format.html
This node is used to describe the fields of registers like cpsr on AArch64.
RegisterFlags is a class that contains a list of register fields.
These fields will be extracted from the XML sent by the remote.
We assume that there is at least one field, that the fields are
sorted in descending order and do not overlap. That will be
enforced by the XML processor (the GDB client code in our case).
The fields may not cover the whole register. To account for this
RegisterFields will add anonymous padding fields so that
sizeof(all fields) == sizeof(register). This will save a lot
of hasssle later.
Reviewed By: jasonmolenda, JDevlieghere
Differential Revision: https://reviews.llvm.org/D145566
Quentin Colombet [Wed, 5 Apr 2023 14:14:18 +0000 (16:14 +0200)]
[mlir][TransformDialect] Simplify the lowering of pack/unpack when these are just pad/unpad
This patch recognizes when tensor.pack/unpack operations are simple
tensor.pad/unpad (a.k.a. tensor.extract_slice) and lowers them in a simpler
sequence of instruction.
For pack, instead of doing:
```
pad
expand_shape
transpose
```
we do
```
pad
insert_slice
```
For unpack, instead of doing:
```
transpose
collapse_shape
extract_slice
```
we do
```
extract_slice
```
Note: returning nullptr for the transform dialect is fine. The related
handles are just ignored by the following transformation.
Differential Revision: https://reviews.llvm.org/D148159
Nicolas Vasilache [Thu, 13 Apr 2023 10:04:26 +0000 (03:04 -0700)]
[mlir][Tensor] Drop SplitPaddingPatterns.
These old patterns are not in use in either MLIR or downstream projects except for one test.
Additionally this is redundant with logic in the tensor.pad tiling implementation.
Drop SplitPaddingPatterns to reduce entropy.
Differential Revision: https://reviews.llvm.org/D148207
Simon Pilgrim [Thu, 13 Apr 2023 10:17:37 +0000 (11:17 +0100)]
[TTI] Remove unnecessary default CostKind args from getExtendedReductionCost/getMulAccReductionCost wrappers. NFC.
We should only ever call these from the TargetTransformInfo interface, which provides all the args.
Simon Pilgrim [Thu, 13 Apr 2023 10:07:55 +0000 (11:07 +0100)]
[TTI] getExtendedReductionCost - replace std::optional<FastMathFlags> args with FastMathFlags
Followup to D148149 where it was noticed that the std::optional wrapper wasn't helping with anything (we can just use an empty FastMathFlags()).
Haojian Wu [Thu, 6 Apr 2023 08:11:46 +0000 (10:11 +0200)]
[clangd] Fix a nullptr-dereference crash in computeIncludeCleanerFindings.
Be more robust, we shuold not crash when we cannot find the corresponding token from the
tokenbuffer.
Differential Revision: https://reviews.llvm.org/D147686
Ingo Müller [Tue, 11 Apr 2023 13:34:37 +0000 (13:34 +0000)]
[mlir] Make 1:N type conversion tests less brittle.
They had previously not followed the best practices; I must have
commited these in a rush. In particular, they used hard-coded SSA names
instead of patterns. This patch changes that, uses CHECK-DAG more
instead of CHECK-NEXT, and makes all function names unique.
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D148012
Haojian Wu [Thu, 13 Apr 2023 10:01:51 +0000 (12:01 +0200)]
[include-fixer] Add the missing variant header to the STL header list.
Krasimir Georgiev [Thu, 13 Apr 2023 10:03:37 +0000 (10:03 +0000)]
[clang] fix an unused variable warning after
9d0b55f0e4ca55d04ee8abfdf021913ea3c30082
Congcong Cai [Thu, 13 Apr 2023 09:56:09 +0000 (11:56 +0200)]
[WebAssembly] `AsmTypeCheck` support to br instr
This PR introduces the `BrStack` member to store the info about
`loop`, `block`, `if` and `try`. It can check whether `br` immediate number
out of range.
Reviewed By: aheejin
Differential Revision: https://reviews.llvm.org/D148054
Simon Pilgrim [Thu, 13 Apr 2023 09:42:36 +0000 (10:42 +0100)]
[TTI] getMinMaxReductionCost - add FastMathFlag argument
Similar to the getArithmeticReductionCost / getExtendedReductionCost calls (which really don't need to use std::optional<>).
This will be necessary to correct recognize fast/nnan fmax/fmul reductions which can avoid nan handling - which will allow us to remove the fmax/fmin special case in X86TTIImpl::getMinMaxCost and use getIntrinsicInstrCost like we do for integer reductions (
63c3895327839ba5b57f5b99ec9e888abf976ac6).
Differential Revision: https://reviews.llvm.org/D148149
Jun Zhang [Thu, 13 Apr 2023 08:42:10 +0000 (16:42 +0800)]
[InstCombine] icmp(X | OrC, C) --> icmp(X, 0)
We can eliminate the or operation based on the predicate and the
relation between OrC and C.
sge: X | OrC s>= C --> X s>= 0 iff OrC s>= C s>= 0
sgt: X | OrC s> C --> X s>= 0 iff OrC s> C s>= 0
sle: X | OrC s<= C --> X s< 0 iff OrC s> C s>= 0
slt: X | OrC s< C --> X s< 0 iff OrC s>= C s>= 0
Alive2 links:
sge: https://alive2.llvm.org/ce/z/W-6FHE
sgt: https://alive2.llvm.org/ce/z/TKK2yJ
sle: https://alive2.llvm.org/ce/z/vURQGM
slt: https://alive2.llvm.org/ce/z/JAsVfw
Related issue: https://github.com/llvm/llvm-project/issues/61538
Signed-off-by: Jun Zhang <jun@junz.org>
Differential Revision: https://reviews.llvm.org/D147597
Jun Zhang [Thu, 13 Apr 2023 08:42:01 +0000 (16:42 +0800)]
Add baseline tests for D147597
Signed-off-by: Jun Zhang <jun@junz.org>