Scott Linder [Mon, 5 Oct 2020 22:45:35 +0000 (22:45 +0000)]
[AMDGPU] Emit correct kernel descriptor on big-endian hosts
Previously we wrote multi-byte values out as-is from host memory. Use
the `emitIntN` helpers in `MCStreamer` to produce a valid descriptor
irrespective of the host endianness.
Reviewed By: arsenm, rochauha
Differential Revision: https://reviews.llvm.org/D88858
Thomas Raoux [Tue, 6 Oct 2020 16:56:35 +0000 (09:56 -0700)]
[mlir][vector] Fold extractOp coming from broadcastOp
Combine ExtractOp with scalar result with BroadcastOp source. This is useful to
be able to incrementally convert degenerated vector of one element into scalar.
Differential Revision: https://reviews.llvm.org/D88751
Stanislav Mekhanoshin [Tue, 6 Oct 2020 17:05:43 +0000 (10:05 -0700)]
[AMDGPU] Create isGFX9Plus utility function
Introduce a utility function to make it more
convenient to write code that is the same on
the GFX9 and GFX10 subtargets.
Use isGFX9Plus in the AsmParser for AMDGPU.
Authored By: Joe_Nash
Differential Revision: https://reviews.llvm.org/D88908
Fanbo Meng [Tue, 6 Oct 2020 17:13:01 +0000 (13:13 -0400)]
[SystemZ][z/OS] Set default alignment rules for z/OS target
Set the default alignment control variables for z/OS target and add test case for alignment rules on z/OS.
Reviewed By: abhina.sreeskantharajan
Differential Revision: https://reviews.llvm.org/D88845
Simon Pilgrim [Tue, 6 Oct 2020 16:32:35 +0000 (17:32 +0100)]
[X86][SSE] combineX86ShuffleChain add 'CanonicalizeShuffleInput' helper. NFCI.
As part of PR45974, we're getting closer to not creating 'padded' vectors on-the-fly in combineX86ShufflesRecursively, and only pad the source inputs if we have a definite match inside combineX86ShuffleChain.
At the moment combineX86ShuffleChain just has to bitcast an input to the correct shuffle type, but eventually we'll need to pad them as well. So, move the bitcast into a 'CanonicalizeShuffleInput helper for now, making the diff for future padding support a lot smaller.
Sebastian Neubauer [Tue, 6 Oct 2020 15:08:34 +0000 (17:08 +0200)]
[AMDGPU] Remove SIInstrInfo::calculateLDSSpillAddress
This function does not seem to be used anymore.
Differential Revision: https://reviews.llvm.org/D88904
Nikita Popov [Sun, 4 Oct 2020 19:03:02 +0000 (21:03 +0200)]
[MemCpyOpt] Use dereferenceable pointer helper
The call slot optimization has some home-grown code for checking
whether the destination is dereferenceable. Replace this with the
generic isDereferenceableAndAlignedPointer() helper.
I'm not checking alignment here, because that is currently handled
separately and may be an enforced alignment for allocas. The clean
way of integrating that part would probably be to accept a callback
in isDereferenceableAndAlignedPointer() for the actual isAligned check,
which would then have a chance to use an enforced alignment instead.
This allows the destination to be a GEP (among other things), though
the two open TODOs may prevent it from working in practice.
Differential Revision: https://reviews.llvm.org/D88805
Nikita Popov [Sun, 4 Oct 2020 16:04:28 +0000 (18:04 +0200)]
[MemCpyOpt] Check for throwing calls during call slot optimization
When performing call slot optimization for a non-local destination,
we need to check whether there may be throwing calls between the
call and the copy. Otherwise, the early write to the destination
may be observable by the caller.
This was already done for call slot optimization of load/store,
but not for memcpys. For the sake of clarity, I'm moving this check
into the common optimization function, even if that does need an
additional instruction scan for the load/store case.
As efriedma pointed out, this check is not sufficient due to
potential accesses from another thread. This case is left as a TODO.
Differential Revision: https://reviews.llvm.org/D88799
Nikita Popov [Tue, 6 Oct 2020 16:14:10 +0000 (18:14 +0200)]
[MemCpyOpt] Add separate statistic for call slot optimization (NFC)
Hafiz Abid Qadeer [Mon, 5 Oct 2020 21:28:25 +0000 (17:28 -0400)]
[libc++] Check _LIBCPP_USE_CLOCK_GETTIME before using clock_gettime
The clock_gettime function is available when _POSIX_TIMERS is defined.
We check for this and set _LIBCPP_USE_CLOCK_GETTIME accordingly since
59b3102739c. But check for _LIBCPP_USE_CLOCK_GETTIME was removed in
babd3aefc91. As a result, code is now trying to use clock_gettime even
on platforms where it is not available and it is causing build failure
with newlib.
This patch restores the checks to fix this.
Differential Revision: https://reviews.llvm.org/D88825
peter klausler [Thu, 1 Oct 2020 18:46:24 +0000 (11:46 -0700)]
[flang] Track CHARACTER length better in TypeAndShape
CHARACTER length expressions were not always being
captured or computed as part of procedure "characteristics",
leading to test failures due to an inability to compute
memory size expressions accurately.
Differential revision: https://reviews.llvm.org/D88689
Simon Pilgrim [Tue, 6 Oct 2020 15:30:09 +0000 (16:30 +0100)]
[APIntTest] Extend extractBits to check 'lshr+trunc' pattern for each case as well.
Noticed while triaging PR47731 that we don't have great coverage for such patterns.
Louis Dionne [Tue, 6 Oct 2020 15:27:57 +0000 (11:27 -0400)]
[libc++] Allow retries in two flaky tests
Fangrui Song [Tue, 6 Oct 2020 15:26:12 +0000 (08:26 -0700)]
[X86] .code16: temporarily set Mode32Bit when matching an instruction with the data32 prefix
PR47632
This allows MC to match `data32 ...` as one instruction instead of two (data32 without insn + insn).
The compatibility with GNU as improves: `data32 ljmp` will be matched as ljmpl.
`data32 lgdt 4(%eax)` will be matched as `lgdtl` (prefixes: 0x67 0x66, instead
of 0x66 0x67).
GNU as supports many other `data32 *w` as `*l`. We currently just hard code
`data32 callw` and `data32 ljmpw`. Generalizing the suffix replacement is
tricky and requires a think about the "bwlq" appending suffix rules in MatchAndEmitATTInstruction.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D88772
Aaron En Ye Shi [Tue, 6 Oct 2020 15:26:56 +0000 (15:26 +0000)]
[HIP] NFC Add comments to cmath functions
Add missing comments to cmath functions.
Differential Revision: https://reviews.llvm.org/D88837
Aaron En Ye Shi [Tue, 6 Oct 2020 15:19:00 +0000 (15:19 +0000)]
[HIP] NFC properly reference Differential Revision
Committed [HIP] Restructure hip headers to add cmath
with typo in commit message. Should be Differential
Revision instead of Review. Using this to close the
diff.
Differential Revision: https://reviews.llvm.org/D88837
Dávid Bolvanský [Mon, 5 Oct 2020 20:16:59 +0000 (22:16 +0200)]
[SimplifyLibCalls] Optimize mempcpy_chk to mempcpy
LLVM GN Syncbot [Tue, 6 Oct 2020 14:49:44 +0000 (14:49 +0000)]
[gn build] Port
aa2b593f149
Aaron En Ye Shi [Thu, 1 Oct 2020 21:43:52 +0000 (21:43 +0000)]
[HIP] Restructure hip headers to add cmath
Separate __clang_hip_math.h header into __clang_hip_cmath.h
and __clang_hip_math.h. Improve the math function definition,
and add missing definitions or declarations. Add missing
overloads.
Reviewed By: tra, JonChesterfield
Differential Review: https://reviews.llvm.org/D88837
Arthur Eubanks [Mon, 5 Oct 2020 22:17:12 +0000 (15:17 -0700)]
[BPF][NewPM] Make BPFTargetMachine properly adjust NPM optimizer pipeline
This involves porting BPFAbstractMemberAccess and BPFPreserveDIType to
NPM, then adding them BPFTargetMachine::registerPassBuilderCallbacks
(the NPM equivalent of adjustPassManager()).
Reviewed By: yonghong-song, asbirlea
Differential Revision: https://reviews.llvm.org/D88855
Arthur Eubanks [Wed, 23 Sep 2020 22:11:40 +0000 (15:11 -0700)]
[test][InstCombine][NewPM] Fix InstCombine tests under NPM
Some of these depended on analyses being present that aren't provided
automatically in NPM.
early_dce_clobbers_callgraph.ll was previously inlining a noinline function?
cast-call-combine.ll relied on the legacy always-inline pass being a
CGSCC pass and getting rerun.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D88187
Arthur Eubanks [Tue, 22 Sep 2020 22:18:55 +0000 (15:18 -0700)]
[test][NewPM] Make dead-uses.ll work under NPM
This one is weird...
globals-aa needs to be already computed at licm, or else a function pass
can't run a module analysis and won't have access to globals-aa.
But the globals-aa result is impacted by instcombine in a way that
affects what the test is expecting. If globals-aa is computed before
instcombine, it is cached and globals-aa used in licm won't contain the
necessary info provided by instcombine.
Another catch is that if we don't invalidate AAManager, it will use the
cached AAManager that instcombine requested, which may not contain
globals-aa. So we have to invalidate<aa> so that licm can recompute
an AAManager with the globals-aa created by the require<globals-aa>.
This is essentially the problem described in https://reviews.llvm.org/D84259.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D88118
Johannes Doerfert [Wed, 9 Sep 2020 21:08:20 +0000 (16:08 -0500)]
[Attributor][FIX] Move assertion to make it not trivially fail
The idea of this assertion was to check the simplified value before we
assign it, not after, which caused this to trivially fail all the time.
Johannes Doerfert [Wed, 9 Sep 2020 01:18:54 +0000 (20:18 -0500)]
[Attributor][FIX] Dead return values are not `noundef`
When we assume a return value is dead we might still visit return
instructions via `Attributor::checkForAllReturnedValuesAndReturnInsts(..)`.
When we do so the "returned value" is potentially simplified to `undef`
as it is the assumed "returned value". This is a problem if there was a
preexisting `noundef` attribute that will only be removed as we manifest
the `undef` return value. We should not use this combination to derive
`unreachable` though. Two test cases fixed.
Johannes Doerfert [Mon, 7 Sep 2020 22:55:05 +0000 (17:55 -0500)]
[Attributor][NFC] Ignore benign uses in AAMemoryBehaviorFloating
In AAMemoryBehaviorFloating we used to track benign uses in a SetVector.
With this change we look through benign uses eagerly to reduce the
number of elements (=Uses) we look at during an update.
The test does actually not fail prior to this commit but I already wrote
it so I kept it.
Shivanshu Goyal [Tue, 6 Oct 2020 14:12:48 +0000 (16:12 +0200)]
Add ability to turn off -fpch-instantiate-templates in clang-cl
A lot of our code building with clang-cl.exe using Clang 11 was failing with
the following 2 type of errors:
1. explicit specialization of 'foo' after instantiation
2. no matching function for call to 'bar'
Note that we also use -fdelayed-template-parsing in our builds.
I tried pretty hard to get a small repro for these failures, but couldn't. So
there is some subtle edge case in the -fpch-instantiate-templates feature
introduced by this change: https://reviews.llvm.org/D69585
When I tried turning this off using -fno-pch-instantiate-templates, builds
would silently fail with the same error without any indication that
-fno-pch-instantiate-templates was being ignored by the compiler. Then I
realized this "no" option wasn't actually working when I ran Clang under a
debugger.
Differential revision: https://reviews.llvm.org/D88680
Dmitri Gribenko [Tue, 6 Oct 2020 14:02:17 +0000 (16:02 +0200)]
Silence -Wunused-variable in NDEBUG mode
Dmitri Gribenko [Tue, 6 Oct 2020 13:28:19 +0000 (15:28 +0200)]
Revert "[c++17] Implement P0145R3 during constant evaluation."
This reverts commit
ded79be63555f4e5bfdb0db27ef22b71fe568474. It causes
a crash (I sent the crash reproducer directly to the author).
Simon Pilgrim [Tue, 6 Oct 2020 12:52:25 +0000 (13:52 +0100)]
[InstCombine] canRewriteGEPAsOffset - don't dereference a dyn_cast<>. NFCI.
We know V is a IntToPtrInst or PtrToIntInst type so we know its a CastInst - so use cast<> directly.
Prevents clang static analyzer warning that we could deference a null pointer.
Simon Pilgrim [Tue, 6 Oct 2020 12:44:17 +0000 (13:44 +0100)]
[InstCombine] FoldShiftByConstant - consistently use ConstantExpr in logicalshift(trunc(shift(x,c1)),c2) fold. NFCI.
This still only gets used for scalar types but now always uses ConstantExpr in preparation for vector support - it was using APInt methods in some places.
Haojian Wu [Tue, 6 Oct 2020 13:46:40 +0000 (15:46 +0200)]
[clangd] Add basic keyword-name-validation in rename.
Differential Revision: https://reviews.llvm.org/D88875
Sam Tebbs [Wed, 23 Sep 2020 10:43:27 +0000 (11:43 +0100)]
[ARM] Fold select_cc(vecreduce_[u|s][min|max], x) into VMINV or VMAXV
This folds a select_cc or select(set_cc) of a max or min vector reduction with a scalar value into a VMAXV or VMINV.
Differential Revision: https://reviews.llvm.org/D87836
Dmitry Preobrazhensky [Tue, 6 Oct 2020 12:30:20 +0000 (15:30 +0300)]
[AMDGPU][MC] Added detection of unsupported instructions
Implemented identification of unsupported instructions; improved errors reporting.
See bug 42590.
Reviewers: rampitec
Differential Revision: https://reviews.llvm.org/D88211
Nicolas Vasilache [Tue, 6 Oct 2020 11:40:52 +0000 (11:40 +0000)]
[mlir][Linalg] Extend buffer allocation to support Linalg init tensors
This revision adds init_tensors support to buffer allocation for Linalg on tensors.
Currently makes the assumption that the init_tensors fold onto the first output tensors.
This assumption is not currently enforced or cast in stone and requires experimenting with tiling linalg on tensors for ops **without reductions**.
Still this allows progress towards the end-to-end goal.
Chuyang Chen [Tue, 6 Oct 2020 12:45:26 +0000 (08:45 -0400)]
Convert diagnostics about multi-character literals from extension to warning
This addresses PR46797.
Jonas Paulsson [Tue, 6 Oct 2020 11:20:15 +0000 (13:20 +0200)]
[SystemZAsmParser] Treat VR128 separately in ParseDirectiveInsn().
This patch makes the parser
- reject higher vector registers (>=16) in operands where they should not
be accepted.
- accept higher integers (>=16) in vector register operands.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D88888
Michał Górny [Mon, 5 Oct 2020 21:07:00 +0000 (23:07 +0200)]
[lldb] [Platform] Move common ::DebugProcess() to PlatformPOSIX
Move common ::DebugProcess() implementation shared by Linux and NetBSD
(and to be shared by FreeBSD shortly) into PlatformPOSIX, and move
the old base implementation used only by Darwin to PlatformDarwin.
Differential Revision: https://reviews.llvm.org/D88852
Simon Pilgrim [Tue, 6 Oct 2020 12:12:47 +0000 (13:12 +0100)]
[InstCombine] FoldShiftByConstant - use PatternMatch for logicalshift(trunc(shift(x,c1)),c2) fold. NFCI.
Simon Pilgrim [Tue, 6 Oct 2020 11:50:04 +0000 (12:50 +0100)]
[InstCombine] FoldShiftByConstant - remove unnecessary cast<>. NFC.
Op1 is already a Constant*
Alexey Lapshin [Tue, 6 Oct 2020 11:55:45 +0000 (14:55 +0300)]
[llvm-objcopy][NFC] fix style issues reported by clang-format.
LLVM GN Syncbot [Tue, 6 Oct 2020 12:02:07 +0000 (12:02 +0000)]
[gn build] Port
d6c9dc3c17e
Adam Balogh [Tue, 6 Oct 2020 07:25:58 +0000 (09:25 +0200)]
[clang-tidy] Remove obsolete checker google-runtime-references
The rules which is the base of this checker is removed from the
//Google C++ Style Guide// in May:
[[ https://github.com/google/styleguide/pull/553 | Update C++ styleguide ]].
Now this checker became obsolete.
Differential Revision: https://reviews.llvm.org/D88831
Alexander Shaposhnikov [Tue, 6 Oct 2020 10:41:19 +0000 (03:41 -0700)]
[llvm-objcopy][MachO] Add support for universal binaries
This diff adds support for universal binaries to llvm-objcopy.
This is a recommit of
32c8435ef70031 with the asan issue fixed.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D88400
Denis Antrushin [Fri, 4 Sep 2020 17:45:41 +0000 (00:45 +0700)]
[Statepoints] Change statepoint machine instr format to better suit VReg lowering.
Current Statepoint MI format is this:
STATEPOINT
<id>, <num patch bytes >, <num call arguments>, <call target>,
[call arguments...],
<StackMaps::ConstantOp>, <calling convention>,
<StackMaps::ConstantOp>, <statepoint flags>,
<StackMaps::ConstantOp>, <num deopt args>, [deopt args...],
<gc base/derived pairs...> <gc allocas...>
Note that GC pointers are listed in pairs <base,derived>.
This causes base pointers to appear many times (at least twice) in
instruction, which is bad for us when VReg lowering is ON.
The problem is that machine operand tiedness is 1-1 relation, so
it might look like this:
%vr2 = STATEPOINT ... %vr1, %vr1(tied-def0)
Since only one instance of %vr1 is tied, that may lead to incorrect
codegen (see PR46917 for more details), so we have to always spill
base pointers. This mostly defeats new VReg lowering scheme.
This patch changes statepoint instruction format so that every
gc pointer appears only once in operand list. That way they all can
be tied. Additional set of operands is added to preserve base-derived
relation required to build stackmap.
New statepoint has following format:
STATEPOINT
<id>, <num patch bytes>, <num call arguments>, <call target>,
[call arguments...],
<StackMaps::ConstantOp>, <calling convention>,
<StackMaps::ConstantOp>, <statepoint flags>,
<StackMaps::ConstantOp>, <num deopt args>, [deopt args...],
<StackMaps::ConstantOp>, <num gc pointers>, [gc pointers...],
<StackMaps::ConstantOp>, <num gc allocas>, [gc allocas...]
<StackMaps::ConstantOp>, <num entries in gc map>, [base/derived indices...]
Changes are:
- every gc pointer is listed only once in a flat length-prefixed list;
- alloca list is prefixed with its length too;
- following alloca list is length-prefixed list of base-derived
indices of pointers from gc pointer list. Note that indices are
logical (number of pointer), not absolute (index of machine operand).
Differential Revision: https://reviews.llvm.org/D87154
Alex Richardson [Tue, 6 Oct 2020 10:38:52 +0000 (11:38 +0100)]
[libcxx][lit] Add support for custom ssh/scp flags in ssh.py
In our CHERI Jenkins CI we need to pass `-F <custom_config_file>` to each
ssh/scp command to set various arguments such as the localhost port, usage
of controlmaster, etc. to speed up connections to our emulated QEMU systems.
For our specific use-case I could have also added a single --ssh-config-file
argument that can be used for both the scp and ssh commands, but being able
to pass arbitrary extra flags for both commands seems more flexible.
Reviewed By: #libc, ldionne
Differential Revision: https://reviews.llvm.org/D84097
David Spickett [Mon, 21 Sep 2020 10:29:06 +0000 (11:29 +0100)]
[AArch64] Correct parameter type for unsigned Neon scalar shift intrinsics
In the following intrinsics the shift amount
(parameter 2) should be signed.
vqshlb_u8 vqshlh_u16 vqshls_u32 vqshld_u64
vqrshlb_u8 vqrshlh_u16 vqrshls_u32 vqrshld_u64
vshld_u64
vrshld_u64
See https://developer.arm.com/documentation/ihi0073/latest
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D88013
Sam McCall [Mon, 5 Oct 2020 19:41:35 +0000 (21:41 +0200)]
[clangd] Add `score` extension to workspace/symbol response.
The protocol doesn't really incorporate ranking.
As with code completion, most clients respect what the server sends, but
VSCode re-ranks items, with predictable results.
See https://github.com/clangd/vscode-clangd/issues/81
There's no filterText field so we may be unable to construct a good workaround.
But expose the score so we may be able to do this on the client in future.
Differential Revision: https://reviews.llvm.org/D88844
Paul Walker [Thu, 1 Oct 2020 17:08:33 +0000 (18:08 +0100)]
[SVE] Lower fixed length vector fneg and fsqrt operations.
Also updates sve-fp.ll to use fneg directly.
Differential Revision: https://reviews.llvm.org/D88683
Paul Walker [Thu, 1 Oct 2020 15:37:55 +0000 (16:37 +0100)]
[SVE] Lower fixed length vector floating point rounding operations.
Adds lowering for:
llvm.ceil
llvm.floor
llvm.nearbyint
llvm.rint
llvm.round
llvm.trunc
Differential Revision: https://reviews.llvm.org/D88671
Pushpinder Singh [Mon, 5 Oct 2020 12:59:26 +0000 (08:59 -0400)]
[OpenMP][RTL] Remove dead code
RequiresDataSharing was always 0, resulting dead code in device runtime library.
Reviewed By: jdoerfert, JonChesterfield
Differential Revision: https://reviews.llvm.org/D88829
Tres Popp [Tue, 22 Sep 2020 10:23:43 +0000 (12:23 +0200)]
[mlir] Add file to implement bufferization for shape ops.
This adds a shape-bufferize pass and implements the pattern for
shape.assuming.
Differential Revision: https://reviews.llvm.org/D88083
Dmitri Gribenko [Tue, 6 Oct 2020 09:29:24 +0000 (11:29 +0200)]
Revert "[llvm-objcopy][MachO] Add support for universal binaries"
This reverts commit
32c8435ef70031d7bd3dce48e41bdce65747e123. It fails
ASan, details in https://reviews.llvm.org/D88400.
Dmitri Gribenko [Tue, 6 Oct 2020 09:28:55 +0000 (11:28 +0200)]
Revert "[llvm-objcopy][MachO] Add missing std::move."
This reverts commit
6e25586990b93e2c9eaaa4f473b6720ccd646c46. It depends
on
32c8435ef70031d7bd3dce48e41bdce65747e123, which I'm reverting due to
ASan failures. Details in https://reviews.llvm.org/D88400.
Mauri Mustonen [Tue, 6 Oct 2020 09:11:20 +0000 (10:11 +0100)]
[VPlan] Add vplan native path vectorization test case for inner loop reduction
Regarding this bug I posted earlier: https://bugs.llvm.org/show_bug.cgi?id=47035
After reading through LLVM source code and getting familiar with VPlan I was able to vectorize the code using by enabling VPlan native path. After talking with @fhahn he suggested that I contribute this as a test case. So here it is. I tried to follow the available guides how to do this best I could. I modified IR code by hand to have more clear variable names instead of numbers.
One thing what I'd like to get input from someone is that is current CHECK lines sufficient enough to verify that the inner loop has been vectorized properly?
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D87564
Georgii Rymar [Wed, 30 Sep 2020 13:23:01 +0000 (16:23 +0300)]
[llvm-readobj/elf][test] - Stop using precompiled binaries in mips-got.test
This removed 2 last precompiled binaries from the mips-got.test.
YAML descriptions are used instead.
Differential revision: https://reviews.llvm.org/D88565
Haojian Wu [Mon, 5 Oct 2020 12:02:59 +0000 (14:02 +0200)]
[clangd] Verify the diagnostic code in include-fixer diagnostic tests, NFC.
Make it easier to spot which diagnostics in the include-fixer list are tested.
Differential Revision: https://reviews.llvm.org/D88828
Sebastian Neubauer [Tue, 6 Oct 2020 08:45:31 +0000 (10:45 +0200)]
[AMDGPU] Fix gcc warnings
uint8_t types are implicitly promoted to int, leading to a
unsigned-signed comparison.
Thanks for the heads-up @uabelho.
Differential Revision: https://reviews.llvm.org/D88876
George Mitenkov [Tue, 6 Oct 2020 08:20:10 +0000 (11:20 +0300)]
[MLIR][SPIRVToLLVM] Conversion for composite extract and insert
A pattern to convert `spv.CompositeInsert` and `spv.CompositeExtract`.
In LLVM, there are 2 ops that correspond to each instruction depending
on the container type. If the container type is a vector type, then
the result of conversion is `llvm.insertelement` or `llvm.extractelement`.
If the container type is an aggregate type (i.e. struct, array), the
result of conversion is `llvm.insertvalue` or `llvm.extractvalue`.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D88205
Haojian Wu [Tue, 6 Oct 2020 08:17:47 +0000 (10:17 +0200)]
[clangd] Fix an inconsistent ReasonToReject enum usage, NFC.
Nicolas Vasilache [Tue, 6 Oct 2020 07:59:02 +0000 (07:59 +0000)]
[mlir][Linalg] Reintroduced missing verification check
A verification check on the number of indexing maps seems to have dropped inadvertently. Also update the relevant roundtrip tests.
Andrzej Warzynski [Tue, 6 Oct 2020 07:45:53 +0000 (08:45 +0100)]
[flang][NFC] Remove redundant `;`
Sadly this has been causing gcc-10 builds to fail.
Muhammad Omair Javaid [Tue, 6 Oct 2020 07:27:27 +0000 (12:27 +0500)]
[LLDB] Add QEMU testing environment setup guide for SVE testing
This patch adds a HowTo document to lldb docs which gives instruction for
setting up a virtual environment based on QEMU emulator for LLDB testing.
Instruction in this document are tested on Arm and AArch64 targets but
can easily be duplicated for other targets supported by QEMU.
This helps test LLDB in absence for modern AArch64 features not released
in publicly available hardware till date.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D82064
Raphael Isemann [Tue, 6 Oct 2020 07:27:24 +0000 (09:27 +0200)]
[lldb] Symlink the Clang resource directory to the LLDB build directory in standalone builds
When doing a standalone build (i.e., building just LLDB against an existing
LLVM/Clang installation), LLDB is currently unable to find any Clang resource
directory that contains all the builtin headers we need to parse real source
code. This causes several tests that actually parse source code on disk within
the expression parser to fail (most notably nearly all the import-std-module
tests).
The reason why LLDB can't find the resource directory is that we search based on
the path of the LLDB shared library path. We assumed that the Clang resource
directory is in the same prefix and has the same relative path to the LLDB
shared library (e.g., `../clang/10.0.0/include`). However for a standalone build
where the existing Clang can be anywhere on the disk, so we can't just rely on
the hardcoded relative paths to the LLDB shared library.
It seems we can either solve this by copying the resource directory to the LLDB
installation, symlinking it there or we pass the path to the Clang installation
to the code that is trying to find the resource directory. When building the
LLDB framework we currently copy the resource directory over to the framework
folder (this is why the import-std-module are not failing on the Green Dragon
standalone bot).
This patch symlinks the resource directory of Clang into the LLDB build
directory. The reason for that is simply that this is only needed when running
LLDB from the build directory. Once LLDB and Clang/LLVM are installed the
already existing logic can find the Clang resource directory by searching
relative to the LLDB shared library.
Reviewed By: kastiglione, JDevlieghere
Differential Revision: https://reviews.llvm.org/D88581
David Sherwood [Thu, 3 Sep 2020 07:28:57 +0000 (08:28 +0100)]
[SVE][CodeGen] Fix DAGCombiner::ForwardStoreValueToDirectLoad for scalable vectors
In DAGCombiner::ForwardStoreValueToDirectLoad I have fixed up some
implicit casts from TypeSize -> uint64_t and replaced calls to
getVectorNumElements() with getVectorElementCount(). There are some
simple cases of forwarding that we can definitely support for
scalable vectors, i.e. when the store and load are both scalable
vectors and have the same size. I have added tests for the new
code paths here:
CodeGen/AArch64/sve-forward-st-to-ld.ll
Differential Revision: https://reviews.llvm.org/D87098
Haojian Wu [Tue, 6 Oct 2020 06:51:49 +0000 (08:51 +0200)]
[AST][RecoveryExpr] Support dependent binary operator in C for error recovery.
see the whole context in: https://reviews.llvm.org/D85025
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D84226
Johannes Doerfert [Tue, 6 Oct 2020 04:53:41 +0000 (23:53 -0500)]
[AttributeFuncs][FIX] Update new tests (D87304, D87306) after sret changes
Hopefully the last of these, apologies for the noise.
Max Kazantsev [Tue, 6 Oct 2020 04:18:13 +0000 (11:18 +0700)]
Revert "[SCEV] Prove implicaitons via AddRec start"
This reverts commit
69acdfe075fa8eb18781f88f4d0cd1ea40fa6e48.
Need to investigate reported miscompiles.
Johannes Doerfert [Tue, 6 Oct 2020 04:35:49 +0000 (23:35 -0500)]
[AttributeFuncs][FIX] Update new tests (D87304) after sret changes
Lang Hames [Tue, 6 Oct 2020 04:28:07 +0000 (21:28 -0700)]
[JITLink][ELF] Handle BSS sections, improve some error messages.
This patch enables basic BSS section handling, and improves a couple of error
messages in the ELF section parsing code.
Patch by Christian Schafmeister. Thanks Christian!
Differential Revision: https://reviews.llvm.org/D88867
Johannes Doerfert [Tue, 8 Sep 2020 15:13:11 +0000 (10:13 -0500)]
[AttributeFuncs] Consider `noundef` in `typeIncompatible`
Drop `noundef` for return values that are replaced by void and make it
illegal to put `noundef` on a void value.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D87306
Johannes Doerfert [Tue, 8 Sep 2020 15:10:36 +0000 (10:10 -0500)]
[AttributeFuncs] Consider `align` in `typeIncompatible`
Alignment attributes need to be dropped for non-pointer values.
This also introduces a check into the verifier to ensure you don't use
`align` on anything but a pointer. Test needed to be adjusted
accordingly.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D87304
Serguei Katkov [Fri, 2 Oct 2020 03:40:52 +0000 (10:40 +0700)]
[GVN LoadPRE] Extend the scope of optimization by using context to prove safety of speculation
Use context to prove that load can be safely executed at a point where load is being hoisted.
Postpone the decision about safety of speculative load execution till the moment we know
where we hoist load and check safety at that context.
Reviewers: nikic, fhahn, mkazantsev, lebedev.ri, efriedma, reames
Reviewed By: reames, mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D88725
Jonas Devlieghere [Mon, 5 Oct 2020 23:35:23 +0000 (16:35 -0700)]
[dotest] Simplify logic to find the Python path
Simplify the logic of parsing the lldb -P output to find the python
path. This removes the special handling for the LLDB.framework case and
instead of pattern matching known errors focus on finding a directory
path that contains an __init__.py.
Differential revision: https://reviews.llvm.org/D88840
Richard Smith [Tue, 6 Oct 2020 00:52:23 +0000 (17:52 -0700)]
[c++17] Implement P0145R3 during constant evaluation.
Ensure that we evaluate assignment and compound-assignment
right-to-left, and array subscripting left-to-right.
Fixes PR47724.
Richard Smith [Tue, 6 Oct 2020 01:39:33 +0000 (18:39 -0700)]
Make OpenMP tests less brittle in the face of changes in constant
evaluation diagnostics.
Arthur Eubanks [Thu, 17 Sep 2020 23:01:30 +0000 (16:01 -0700)]
[NewPM] Set -enable-npm-optnone to true by default
This makes the NPM skip not required passes on functions marked optnone.
If this causes a pass that should be required but has not been marked
required to be skipped, add
`static bool isRequired() { return true; }`
to the pass class. AlwaysInlinerPass is an example.
clang/test/CodeGen/O0-no-skipped-passes.c is useful for checking that
no passes are skipped under -O0.
The -enable-npm-optnone option will be removed once this has been stable
for long enough without issues.
Reviewed By: ychen, asbirlea
Differential Revision: https://reviews.llvm.org/D87869
Mircea Trofin [Sat, 3 Oct 2020 03:28:49 +0000 (20:28 -0700)]
[MLInliner] Factor out logging
Factored out the logging facility, to allow its reuse outside the
inliner.
Differential Revision: https://reviews.llvm.org/D88770
Carl Ritson [Tue, 6 Oct 2020 00:44:50 +0000 (09:44 +0900)]
[AMDGPU] SIInsertSkips: Refactor early exit block creation
Refactor exit block creation to a single call ensureEarlyExitBlock.
Add support for generating an early exit block which clears the
exec mask, but only add this instruction when required.
These changes are to facilitate adding more forms of early
termination for PS shaders in the near future.
Reviewed By: nhaehnle
Differential Revision: https://reviews.llvm.org/D88775
Carl Ritson [Tue, 6 Oct 2020 00:31:30 +0000 (09:31 +0900)]
Fix reordering of instructions during VirtRegRewriter unbundling
When unbundling COPY bundles in VirtRegRewriter the start of the
bundle is not correctly referenced in the unbundling loop.
The effect of this is that unbundled instructions are sometimes
inserted out-of-order, particular in cases where multiple
reordering have been applied to avoid clobbering dependencies.
The resulting instruction sequence clobbers dependencies.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D88821
Evandro Menezes [Tue, 6 Oct 2020 00:20:59 +0000 (19:20 -0500)]
[RISCV] Fix broken test
Fix test for the SiFive E76 core.
This patch fixes the issue introduced by the commit
5d6d8a2769.
Mircea Trofin [Thu, 24 Sep 2020 04:58:45 +0000 (21:58 -0700)]
[NFC][regalloc] Separate iteration from AllocationOrder
This separates the two concerns - encapsulation of traversal order; and
iteration.
Differential Revision: https://reviews.llvm.org/D88256
Louis Dionne [Mon, 5 Oct 2020 22:50:37 +0000 (18:50 -0400)]
[libc++] Improve tests for std::quoted
Instead of using ad-hoc mechanisms to disable the tests in C++ < 14, use
UNSUPPORTED markup.
Jim Ingham [Fri, 2 Oct 2020 19:43:24 +0000 (12:43 -0700)]
Fix raciness in the StopHook check for "has the target run".
This was looking at the privateState, but it's possible that
the actual process has started up and then stopped again by the
time we get to the check, which would lead us to get out of running
the stop hooks too early.
Instead we need to track the intention of the stop hooks directly.
Differential Revision: https://reviews.llvm.org/D88753
Greg Clayton [Sat, 3 Oct 2020 01:05:10 +0000 (18:05 -0700)]
Show register names in DWARF unwind info.
Register context information was already being passed into the DWARFDebugFrame code that dumps unwind information but it wasn't being used. This change adds the ability to dump registers names of a valid MC register context was passed in and if it knows about the register. Updated the tests to use the newly returned register names.
Differential Revision: https://reviews.llvm.org/D88767
Joachim Protze [Mon, 5 Oct 2020 22:06:40 +0000 (00:06 +0200)]
[OpenMP][Archer][Tests] NFC: fix spurious test failure
The test disables suppression and therefore sometimes triggers a know false
positive in the openmp runtime. The test should only verify that the env
var is handles as expected.
Alex Richardson [Mon, 5 Oct 2020 21:54:22 +0000 (22:54 +0100)]
[libcxx][dsl] Fix mutable default argument warning
This is flagged by PyCharm and can cause subtle bugs. While changing this
also re-sort the imports and add missing ones.
Reviewed By: #libc, ldionne
Differential Revision: https://reviews.llvm.org/D88816
Craig Topper [Mon, 5 Oct 2020 22:03:07 +0000 (15:03 -0700)]
[X86] Remove X86ISD::LCMPXCHG8_SAVE_EBX_DAG and LCMPXCHG8B_SAVE_EBX pseudo instruction
This and its friend X86ISD::LCMPXCHG8_SAVE_RBX_DAG are used if we need to avoid clobbering the frame pointer in EBX/RBX. EBX/RBX are only used a frame pointer in 64-bit mode. In 64-bit mode we don't use CMPXCHG8B since we have a GR64 cmpxchg available. So we don't need special handling for LCMPXCHG8B.
Split from D88808
Differential Revision: https://reviews.llvm.org/D88853
Craig Topper [Mon, 5 Oct 2020 21:27:14 +0000 (14:27 -0700)]
[SelectionDAG] Make sure FMF are propagated when getSetcc canonicalizes FP constants to RHS.
getNode handling for ISD:SETCC calls FoldSETCC which can canonicalize
FP constants to the RHS. When this happens we should create the node
with the FMF that was requested. By using FlagInserter when can ensure
any calls to getNode/getSetcc during canonicalization will also get the flags.
Differential Revision: https://reviews.llvm.org/D88063
Fangrui Song [Mon, 5 Oct 2020 21:47:46 +0000 (14:47 -0700)]
Cleanup CodeGen/CallingConvLower.cpp
Patch by pi1024e (email unavailable)
Differential Revision: https://reviews.llvm.org/D82593
Vedant Kumar [Mon, 5 Oct 2020 20:41:12 +0000 (13:41 -0700)]
Revert "Outline non returning functions unless a longjmp"
This reverts commit
20797989ea190f2ef22d13c5a7a0535fe9afa58b.
This patch (https://reviews.llvm.org/D69257) cannot complete a stage2
build due to the change:
```
CI->getCalledFunction()->getName().contains("longjmp")
```
There are several concrete issues here:
- The callee may not be a function, so `getCalledFunction` can assert.
- The called value may not have a name, so `getName` can assert.
- There's no distinction made between "my_longjmp_test_helper" and the
actual longjmp libcall.
At a higher level, there's a serious layering problem here. The
splitting pass makes policy decisions in a general way (e.g. based on
attributes or profile data). Special-casing certain names breaks the
layering. It subverts the work of library maintainers (who may now need
to opt-out of unexpected optimization behavior for any affected
functions) and can lead to inconsistent optimization behavior (as not
all llvm passes special-case ".*longjmp.*" in the same way).
The patch may need significant revision to address these issues.
But the immediate issue is that this crashes while compiling llvm's unit
tests in a stage2 build (due to the `getName` problem).
Cameron McInally [Mon, 5 Oct 2020 21:05:14 +0000 (16:05 -0500)]
[SVE] Add legalisation tests to sve-fixed-length-int-reduce.ll
ergawy [Mon, 5 Oct 2020 20:39:39 +0000 (16:39 -0400)]
[MLIR][SPIRV] Extend _reference_of to support SpecConstantCompositeOp.
Adds support for SPIR-V composite speciailization constants to spv._reference_of.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D88732
Roman Lebedev [Mon, 5 Oct 2020 19:35:59 +0000 (22:35 +0300)]
[InstCombine] Revert rL226781 "Teach InstCombine to canonicalize loads which are only ever stored to always use a legal integer type if one is available." (PR47592)
(it was introduced in https://lists.llvm.org/pipermail/llvm-dev/2015-January/080956.html)
This canonicalization seems dubious.
Most importantly, while it does not create `inttoptr` casts by itself,
it may cause them to appear later, see e.g. D88788.
I think it's pretty obvious that it is an undesirable outcome,
by now we've established that seemingly no-op `inttoptr`/`ptrtoint` casts
are not no-op, and are no longer eager to look past them.
Which e.g. means that given
```
%a = load i32
%b = inttoptr %a
%c = inttoptr %a
```
we likely won't be able to tell that `%b` and `%c` is the same thing.
As we can see in D88789 / D88788 / D88806 / D75505,
we can't really teach SCEV about this (not without the https://bugs.llvm.org/show_bug.cgi?id=47592 at least)
And we can't recover the situation post-inlining in instcombine.
So it really does look like this fold is actively breaking
otherwise-good IR, in a way that is not recoverable.
And that means, this fold isn't helpful in exposing the passes
that are otherwise unaware of these patterns it produces.
Thusly, i propose to simply not perform such a canonicalization.
The original motivational RFC does not state what larger problem
that canonicalization was trying to solve, so i'm not sure
how this plays out in the larger picture.
On vanilla llvm test-suite + RawSpeed, this results in
increase of asm instructions and final object size by ~+0.05%
decreases final count of bitcasts by -4.79% (-28990),
ptrtoint casts by -15.41% (-3423),
and of inttoptr casts by -25.59% (-6919, *sic*).
Overall, there's -0.04% less IR blocks, -0.39% instructions.
See https://bugs.llvm.org/show_bug.cgi?id=47592
Differential Revision: https://reviews.llvm.org/D88789
peter klausler [Wed, 30 Sep 2020 20:04:44 +0000 (13:04 -0700)]
[flang] Correct kP scaling on F output
The sign of the scaling factor was misinterpreted for output
as meaning what it does for input. To be correct, they
should cancel each other out.
print '(1P,F4.3)', 1. ! printed 0.1 but should print 10.0
Differential revision: https://reviews.llvm.org/D88610
Evandro Menezes [Fri, 2 Oct 2020 21:44:32 +0000 (16:44 -0500)]
[RISCV] Add SiFive cores to the CPU option
Add the SiFive cores E76 and U74 using the SiFive 7 series microarchitecture.
Differential Revision: https://reviews.llvm.org/D88759
Evandro Menezes [Fri, 2 Oct 2020 21:26:49 +0000 (16:26 -0500)]
[RISCV] Use the extensions in the canonical order (NFC)
Fix a mistake in the ordering.
Louis Dionne [Mon, 5 Oct 2020 20:39:33 +0000 (16:39 -0400)]
[libc++] Use __has_include instead of complex logic in thread.cpp
We might end up including more headers than strictly necessary this way,
but it's much simpler and it makes it easier to port thread.cpp to systems
not handled by the existing conditionals.
Louis Dionne [Mon, 5 Oct 2020 20:30:23 +0000 (16:30 -0400)]
[libc++] NFC: Rename variant helpers to avoid name clashes
Some system headers define __constructor and __destructor macros (for
Clang attributes constructor and destructor). While this is badly
behaved, it is easy for libc++ to work around this issue.
Louis Dionne [Mon, 5 Oct 2020 20:24:41 +0000 (16:24 -0400)]
[libc++] NFC: Remove unused <iostream> include in atomic.cpp
Dávid Bolvanský [Mon, 5 Oct 2020 20:27:14 +0000 (22:27 +0200)]
Revert "[SLC] Optimize mempcpy_chk to mempcpy"
This reverts commit
3f1fd59de3002e3d5a4eca98cd49c45755ab0110.