Andrew Litteken [Tue, 15 Sep 2020 23:05:38 +0000 (18:05 -0500)]
[IRSim][IROutliner] Adding the extraction basics for the IROutliner.
Extracting the similar regions is the first step in the IROutliner.
Using the IRSimilarityIdentifier, we collect the SimilarityGroups and
sort them by how many instructions will be removed. Each
IRSimilarityCandidate is used to define an OutlinableRegion. Each
region is ordered by their occurrence in the Module and the regions that
are not compatible with previously outlined regions are discarded.
Each region is then extracted with the CodeExtractor into its own
function.
We test that correctly extract in:
test/Transforms/IROutliner/extraction.ll
test/Transforms/IROutliner/address-taken.ll
test/Transforms/IROutliner/outlining-same-globals.ll
test/Transforms/IROutliner/outlining-same-constants.ll
test/Transforms/IROutliner/outlining-different-structure.ll
Reviewers: paquette, jroelofs, yroux
Differential Revision: https://reviews.llvm.org/D86975
Arthur O'Dwyer [Fri, 27 Nov 2020 23:51:57 +0000 (18:51 -0500)]
[libc++] s/constpexr/constexpr/ in some comments. NFC.
Kazushi (Jam) Marukawa [Thu, 26 Nov 2020 13:27:34 +0000 (22:27 +0900)]
[VE] Optimize emitSPAdjustment function
Optimize emitSPAdjustment function to generate as small as possible
instructions to adjust SP.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D92174
Vladimir Vereschaka [Fri, 27 Nov 2020 23:01:09 +0000 (15:01 -0800)]
[libc++] Support no libc++ namespace in the iterator test.
The
5f12f4ff9078455cad9d4806da01f570553a5bf9 commit suppress printing of
inline namespace names in diagnostics by default that breaks the libc++
iterator test, which expects __1 in the namespace.
This patch fixes the test by supporting a test case without __1 in the
namespace.
Differential Revision: https://reviews.llvm.org/D92142
Arthur O'Dwyer [Fri, 27 Nov 2020 19:13:05 +0000 (14:13 -0500)]
[libc++] Consistently unparenthesize `numeric_limits<T>::max`. NFCI.
I think people were sometimes parenthesizing `(foo::max)()` out of
misplaced concern that an unparenthesized `foo::max()` would trip up
Windows' `max(a,b)` macro. However, this is not the case: `max(a,b)`
should be tripped up only by an unparenthesized call to `foo::max(a,b)`,
and in fact we already do `_VSTD::max(a,b)` all over the place anyway
without any guards.
However, in order to do it without guards, we must also
wrap the header in _LIBCPP_PUSH_MACROS, which <span> was not.
Differential Revision: https://reviews.llvm.org/D92240
Zarko Todorovski [Fri, 27 Nov 2020 20:54:37 +0000 (15:54 -0500)]
Add -fintegrated-as to clang invocation
On platforms where the integrated as isn't called by default this
test fails since the output is not what it expects. Adding this
option generates the expected output on those platforms as well.
Craig Topper [Fri, 27 Nov 2020 20:41:55 +0000 (12:41 -0800)]
[RISCV] Replace getSimpleValueType() with getValueType() in DAG combines to prevent asserts with weird types.
Nathan James [Fri, 27 Nov 2020 20:36:09 +0000 (20:36 +0000)]
[NFC] SmallVector<char...> to SmallString<...>
Nikita Popov [Fri, 27 Nov 2020 20:17:48 +0000 (21:17 +0100)]
[ValueTracking] Fix assert on shufflevector of pointers
In this case getScalarSizeInBits() is not well-defined. Use the
existing TyBits variable that handles vectors of pointers correctly.
Craig Topper [Fri, 27 Nov 2020 20:06:04 +0000 (12:06 -0800)]
[RISCV] Remove stale FIXMEs from a couple test cases. NFC
Martin Storsjö [Fri, 27 Nov 2020 19:38:39 +0000 (21:38 +0200)]
Revert "[BasicAA] Fix BatchAA results for phi-phi assumptions"
This reverts commit
8166ed1a7a26ee8ea8db9005cc8ee5d156adad9b,
as it caused some compilations to hang/loop indefinitely, see
https://reviews.llvm.org/D91936 for details.
Craig Topper [Fri, 27 Nov 2020 19:20:24 +0000 (11:20 -0800)]
[RISCV][LegalizeTypes] Teach type legalizer that it can promote UMIN/UMAX using SExtPromotedInteger if that's better for the target.
If Sext is cheaper than Zext for a target, we can use that to promote the operands of UMIN/UMAX. Using sext just makes numbers with the sign bit set even larger when treated as an unsigned number and it has no effect on number without the sign bit set. So the relative order doesn't change. This is similar to what we already do for promoting SETCC.
This is helpful on RISCV where i32 arguments are sign extended on RV64 and many instructions are able to produce results with 33 sign bits.
Differential Revision: https://reviews.llvm.org/D92128
Craig Topper [Fri, 27 Nov 2020 19:11:41 +0000 (11:11 -0800)]
[RISCV] Replace sexti32/zexti32 in isel patterns where only one part of their PatFrags can match. NFCI
We had an zexti32 after a sign_extend_inreg. The AND X, 0xffffffff
part of the zexti32 should never occur since SimplifyDemandedBits
from the sign_extend_inreg would have removed it.
We also had sexti32 as the root node of a pattern, but SelectionDAGISel
matches assertsext early before the tablegen based patterns are
evaluated.
Krzysztof Parzyszek [Fri, 27 Nov 2020 19:09:34 +0000 (13:09 -0600)]
[Hexagon] Improve check for HVX types
Allow non-simple types, like <17 x i32> to be treated as HVX vector
types.
zoecarver [Fri, 27 Nov 2020 19:05:03 +0000 (11:05 -0800)]
[libc++] Use std::move in numeric algorithms (P0616R0).
This patch updates algorithms in <numeric> to use std::move
based on p0616r0. Moving values instead of copying them
creates huge speed improvements (see the paper for details).
Differential Revision: https://reviews.llvm.org/D61170
diggerlin [Fri, 27 Nov 2020 18:37:43 +0000 (13:37 -0500)]
[AIX][XCOFF][NFC] Change geNumberOfVRSaved function name to getNumberOfVRSaved.
SUMMARY:
Change geNumberOfVRSaved function name to getNumberOfVRSaved of class TBVectorExt
Reviewers: hubert.reinterpretcast, Jason Liu
Differential Revision: https://reviews.llvm.org/D92225
Nico Weber [Fri, 27 Nov 2020 18:33:55 +0000 (13:33 -0500)]
lld-link: Let LLD_REPRODUCE control /reproduce:, like in ld.lld
Also sync help texts for the option between elf and coff ports.
Decisions:
- Do this even if /lldignoreenv is passed. /reproduce: does not affect
the main output, and this makes the env var more convenient to use.
(On the other hand, it's now possible to set this env var and forget
about it, and all future builds in the same shell will be much slower.
That's true for ld.lld, but posix shells have an easy way to set an
env var for a single command; in cmd.exe this is not possible without
contortions. Then again, lld-link runs in posix shells too.)
Original patch rebased across D68378 and D68381.
Differential Revision: https://reviews.llvm.org/D67707
Andre Vieira [Fri, 13 Nov 2020 13:56:43 +0000 (13:56 +0000)]
[AArch64] Define __ARM_FEATURE_{CRC32,ATOMICS}
This patch implements the definition of __ARM_FEATURE_ATOMICS and fixes the
missing definition of __ARM_FEATURE_CRC32 for Armv8.1-A.
Differential Revision: https://reviews.llvm.org/D91438
Sylvestre Ledru [Fri, 27 Nov 2020 17:19:35 +0000 (18:19 +0100)]
TargetProcessControl.cpp - Remove warning: extra ‘;’
Florian Hahn [Fri, 27 Nov 2020 16:59:30 +0000 (16:59 +0000)]
[VPlan] Use VPTransformState::set in widenGEP.
This patch updates widenGEP to manage the resulting vector values using
the VPValue of VPWidenGEP recipe.
Louis Dionne [Mon, 23 Nov 2020 20:52:03 +0000 (15:52 -0500)]
[libc++] Introduce an indirection to create threads in the test suite
We create threads using std::thread in various places in the test suite.
However, the usual std::thread constructor may not work on all platforms,
e.g. on platforms where passing a stack size is required to create a thread.
This commit introduces a simple indirection that makes it easier to tweak
how threads are created inside the test suite on various platforms. Note
that tests that are purposefully calling std::thread's constructor directly
(e.g. because that is what they're testing) were not modified.
Francesco Petrogalli [Fri, 27 Nov 2020 16:32:05 +0000 (16:32 +0000)]
[AllocaInst] Update `getAllocationSizeInBits` to return `TypeSize`.
Reviewed By: peterwaller-arm, sdesmalen
Differential Revision: https://reviews.llvm.org/D92020
Simon Pilgrim [Fri, 27 Nov 2020 16:11:02 +0000 (16:11 +0000)]
[X86] Add AVX2/AVX512 test coverage in sat-add.ll
Shows the failure to combine to uaddsat
Simon Pilgrim [Fri, 27 Nov 2020 15:45:58 +0000 (15:45 +0000)]
[Hexagon] Add HVX support for ISD::SMAX/SMIN/UMAX/UMIN instead of custom dag patterns
Followup to D92112 now that I've learnt about HVX type splitting.
This is some necessary cleanup work for min/max ops to eventually help us move the add/sub sat patterns into DAGCombine - D91876.
Differential Revision: https://reviews.llvm.org/D92169
Jon Chesterfield [Fri, 27 Nov 2020 15:39:20 +0000 (15:39 +0000)]
[libomptarget][cuda] Detect missing symbols in plugin at build time
[libomptarget][cuda] Detect missing symbols in plugin at build time
Passes -z,defs to the linker. Error on unresolved symbol references.
Otherwise, those unresolved symbols present as target code running on the host
as the plugin fails to load. This is significantly harder to debug than a link
time error. Flag matches that passed by amdgcn and ve plugins.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D92143
Louis Dionne [Thu, 26 Nov 2020 20:00:42 +0000 (15:00 -0500)]
[libc++] Formalize what configurations are covered by the ABI lists
By encoding ABI-affecting properties in the name of the ABI list, it
makes it clear when an ABI list test should or should not be available,
and what results we should expect.
Note that we clearly don't encode all ABI-affecting parameters in the
name right now -- I just ported over what we supported in the code that
was there previously. As we encounter configurations that we wish to
support but produce different ABI lists, we can add those to the ABI
identifier and start supporting them.
This commit also starts checking the ABI list in the CI jobs that run
a supported configuration. Eventually, all configurations should have
a generated ABI list and the test should even run implicitly as part of
the Lit test suite.
Differential Revision: https://reviews.llvm.org/D92194
Bruce Mitchener [Tue, 24 Nov 2020 17:53:53 +0000 (12:53 -0500)]
[libc++] Replace several uses of 0 by nullptr
Differential Revision: https://reviews.llvm.org/D43159
Evgeny Leviant [Fri, 27 Nov 2020 15:00:02 +0000 (18:00 +0300)]
[TableGen][SchedModels] Get rid of hasVariant. NFC
Differential revision: https://reviews.llvm.org/D92026
Zarko Todorovski [Fri, 27 Nov 2020 13:31:56 +0000 (08:31 -0500)]
[AIX] Enabling vector type arguments and return for AIX
This patch enables vector type arguments on AIX. All non-aggregate Altivec vector types are 16bytes in size and are 16byte aligned.
Reviewed By: Xiangling_L
Differential Revision: https://reviews.llvm.org/D92117
Tamas Berghammer [Thu, 26 Nov 2020 11:03:12 +0000 (11:03 +0000)]
Don't elide splat attributes during printing
A splat attribute have a single element during printing so we should
treat it as such when we decide if we elide it or not based on the flag
intended to elide large attributes.
Reviewed By: rriddle, mehdi_amini
Differential Revision: https://reviews.llvm.org/D92165
Hans Wennborg [Fri, 27 Nov 2020 13:43:14 +0000 (14:43 +0100)]
Revert "[ASTMatchers] Matcher macros with params move params instead of copying"
This broke some clang matcher tests in 32-bit MSVC builds; see PR46781.
> Summary: Use move semantics instead of copying for AST Matchers with parameters
>
> Reviewers: aaron.ballman, gribozavr2
>
> Reviewed By: gribozavr2
>
> Subscribers: cfe-commits
>
> Tags: #clang
>
> Differential Revision: https://reviews.llvm.org/D75096
This reverts commit
3e9a7b2ba470bbe9cf3de3e4b15ba09dcfd206aa.
Sjoerd Meijer [Fri, 27 Nov 2020 13:41:35 +0000 (13:41 +0000)]
[AArch64][SLP] Precommit tests which would be better not to SLP vectorize. NFC.
Erik Tomusk [Fri, 27 Nov 2020 10:14:16 +0000 (10:14 +0000)]
[OpenCL] Check for extension string extension lookup
Calling any of the OpenCLOptions::is*() functions (except isKnown())
with an unknown extension string results in a seg fault. This patch
checks that the extension exists in the map before attempting to access
it.
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D90928
Felipe de Azevedo Piovezan [Fri, 27 Nov 2020 13:07:01 +0000 (14:07 +0100)]
[mlir] Fix typos and broken links in LangRef
Many pages have had their titles renamed over time,
causing broken links to spread throughout the documentation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D92093
Georgii Rymar [Thu, 26 Nov 2020 13:05:38 +0000 (16:05 +0300)]
[obj2yaml] - Dump the `EShNum` key in some cases.
This patch starts emitting the `EShNum` key, when the `e_shnum = 0`
and the section header table exists.
`e_shnum` might be 0, when the the number of entries in the section header
table is larger than or equal to SHN_LORESERVE (0xff00).
In this case the real number of entries
in the section header table is held in the `sh_size`
member of the initial entry in section header table.
Currently, obj2yaml crashes when an object has `e_shoff != 0` and the `sh_size`
member of the initial entry in section header table is `0`.
This patch fixes it.
Differential revision: https://reviews.llvm.org/D92098
Marek Kurdej [Fri, 27 Nov 2020 12:47:20 +0000 (13:47 +0100)]
[libc++] [docs] Exclude helper files from Sphinx configuration to avoid generating empty pages.
Georgii Rymar [Thu, 26 Nov 2020 10:40:39 +0000 (13:40 +0300)]
[obj2yaml] - Don't assert when trying to calculate the expected section offset.
The following line asserts when `sh_addralign > MAX_UINT32 && (uint32_t)sh_addralign == 0`:
```
ExpectedOffset = alignTo(ExpectedOffset,
SecHdr.sh_addralign ? SecHdr.sh_addralign : 1);
```
it happens because `sh_addralign` is truncated to 32-bit value, but `alignTo`
doesn't accept `Align == 0`. We should change `1` to `1uLL`.
Differential revision: https://reviews.llvm.org/D92163
David Green [Fri, 27 Nov 2020 12:21:11 +0000 (12:21 +0000)]
[ARM] MVE qabs vectorization test. NFC
Sjoerd Meijer [Fri, 27 Nov 2020 11:30:45 +0000 (11:30 +0000)]
[SLP] Dump Tree costs. NFC.
This adds LLVM_DEBUG messages to dump the (intermediate) tree cost
calculations, which is useful to trace and see how the final cost is
calculated.
Simon Pilgrim [Fri, 27 Nov 2020 11:00:34 +0000 (11:00 +0000)]
[DAG] Legalize umin(x,y) -> sub(x,usubsat(x,y)) and umax(x,y) -> add(x,usubsat(y,x)) iff usubsat is legal
If usubsat() is legal, this is likely to result in smaller codegen expansion than the default cmp+select codegen expansion.
Allows us to move the x86-specific lowering to the generic expansion code.
Differential Revision: https://reviews.llvm.org/D92183
Simon Pilgrim [Thu, 26 Nov 2020 18:50:29 +0000 (18:50 +0000)]
[X86] Regenerate extract-store.ll tests
Rename prefix from X32 to X86 as we typically use X32 for gnux32 triples
Cheng Wang [Thu, 29 Oct 2020 11:10:44 +0000 (19:10 +0800)]
[libc] Fix typo in buildbot README.txt.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D90381
Jay Foad [Thu, 26 Nov 2020 16:48:14 +0000 (16:48 +0000)]
[LegacyPM] Avoid a redundant map lookup in setLastUser. NFC.
As a bonus this makes it (IMO) obvious that the iterator is not
invalidated, so remove the comment explaining that.
Jay Foad [Fri, 27 Nov 2020 10:03:05 +0000 (10:03 +0000)]
[LegacyPM] Remove unused undocumented parameter. NFC.
The Direction parameter to AnalysisResolver::getAnalysisIfAvailable has
never been documented or used for anything.
Raphael Isemann [Thu, 26 Nov 2020 17:02:25 +0000 (18:02 +0100)]
[ASTImporter] Import the default argument of NonTypeTemplateParmDecl
The test case isn't using the AST matchers for all checks as there doesn't seem to be support for
matching NonTypeTemplateParmDecl default arguments. Otherwise this is simply importing the
default arguments.
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D92106
Richard Sandiford [Thu, 5 Mar 2020 19:23:35 +0000 (19:23 +0000)]
[SVE] Add a couple of extra sizeless type tests
This patch adds tests for things that happened to be fixed by previous
patches, but that should continue working if we do decide to treat
sizeless types as incomplete types.
Differential Revision: https://reviews.llvm.org/D79584
Cullen Rhodes [Tue, 24 Nov 2020 18:07:13 +0000 (18:07 +0000)]
[InstSimplify] Clarify use of FixedVectorType in SimplifySelectInst
Folding a select of vector constants that include undef elements only
applies to fixed vectors, but there's no earlier check the type is not
scalable so it crashes for scalable vectors. This adds a check so this
optimization is only attempted for fixed vectors.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D92046
Roman Lebedev [Fri, 27 Nov 2020 07:21:33 +0000 (10:21 +0300)]
Reland [SimplifyCFG] FoldBranchToCommonDest: lift use-restriction on bonus instructions
This was orginally committed in
2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe.
but was immediately reverted in
f3abd54958ab90ba7c100d3fa936a3ce0dd2ad04
because of a PHI handling issue.
Original commit message:
1. It doesn't make sense to enforce that the bonus instruction
is only used once in it's basic block. What matters is
whether those user instructions fit within our budget, sure,
but that is another question.
2. It doesn't make sense to enforce that said bonus instructions
are only used within their basic block. Perhaps the branch
condition isn't using the value computed by said bonus instruction,
and said bonus instruction is simply being calculated
to be used in successors?
So iff we can clone bonus instructions, to lift these restrictions,
we just need to carefully update their external uses
to use the new cloned instructions.
Notably, this transform (even without this change) appears to be
poison-unsafe as per alive2, but is otherwise (including the patch) legal.
We don't introduce any new PHI nodes, but only "move" the instructions
around, i'm not really seeing much potential for extra cost modelling
for the transform, especially since now we allow at most one such
bonus instruction by default.
This causes the fold to fire +11.4% more (13216 -> 14725)
as of vanilla llvm test-suite + RawSpeed.
The motivational pattern is IEEE-754-2008 Binary16->Binary32
extension code:
https://github.com/darktable-org/rawspeed/blob/
ca57d77fb2ba81f21fc712cfac26e54f46406473/src/librawspeed/common/FloatingPoint.h#L115-L120
^ that should be a switch, but it is not now: https://godbolt.org/z/bvja5v
That being said, even thought this seemed like this would fix it: https://godbolt.org/z/xGq3TM
apparently that fold is happening somewhere else afterall,
so something else also has a similar 'artificial' restriction.
Roman Lebedev [Fri, 27 Nov 2020 08:47:41 +0000 (11:47 +0300)]
[NFC][SimplifyCFG] FoldBranchToCommonDest: add one more test with PHI
This is the problematic pattern i didn't think of,
that lead to revert of
2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe
in
f3abd54958ab90ba7c100d3fa936a3ce0dd2ad04.
Frederik Gossen [Fri, 27 Nov 2020 09:06:29 +0000 (10:06 +0100)]
[MLIR][SCF] Find all innermost loops for parallel loop tiling
Overcome the assumption that parallel loops are only nested in other parallel
loops.
Differential Revision: https://reviews.llvm.org/D92188
Christian Sigg [Tue, 24 Nov 2020 21:07:34 +0000 (22:07 +0100)]
[mlir] Add gpu.allocate, gpu.deallocate ops with LLVM lowering to runtime function calls.
The ops are very similar to the std variants, but support async GPU execution.
gpu.alloc does not currently support an alignment attribute, and the new ops do not have
canonicalizers/folders like their std siblings do.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D91698
Wang, Pengfei [Fri, 27 Nov 2020 08:02:09 +0000 (16:02 +0800)]
[msan] Fix bugs when instrument x86.avx512*_cvt* intrinsics.
Scalar intrinsics x86.avx512*_cvt* have an extra rounding mode operand.
We can directly ignore it to reuse the SSE/AVX math.
This fix the bug https://bugs.llvm.org/show_bug.cgi?id=48298.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D92206
Nicolas Vasilache [Thu, 26 Nov 2020 15:58:34 +0000 (15:58 +0000)]
[mlir] Let ModuleTranslate propagate LLVM triple
This adds LLVM triple propagation and updates the test that did not check it properly.
Differential Revision: https://reviews.llvm.org/D92182
Cheng Wang [Fri, 27 Nov 2020 07:38:16 +0000 (15:38 +0800)]
[libc][Obvious] Fix typo in strnlen_test.cpp.
Markus Lavin [Fri, 27 Nov 2020 07:49:20 +0000 (08:49 +0100)]
Revert "[DebugInfo] Improve dbg preservation in LSR."
This reverts commit
06758c6a6135f59deec8e73d4fcb69946ab47f54.
Bug: https://bugs.llvm.org/show_bug.cgi?id=48166
Additional discussion in: https://reviews.llvm.org/D91711
Georgii Rymar [Wed, 25 Nov 2020 11:54:35 +0000 (14:54 +0300)]
[llvm-readelf/obj] - Stop calling `reportError` in `printArchSpecificInfo()`.
This is related to MIPS. Currently we might report an error and exit,
though there is no problem to report a warning and try to continue dumping
an object. The code uses `MipsGOTParser<ELFT> Parser`, which is isolated
in this method.
Differential revision: https://reviews.llvm.org/D92090
Craig Topper [Fri, 27 Nov 2020 06:31:13 +0000 (22:31 -0800)]
[RISCV] Don't remove (and X, 0xffffffff) from inputs when matching RISCVISD::DIVUW/REMUW to 64-bit DIVU/REMU.
These patterns are using zexti32 which matches either assertzexti32
or (and X, 0xffffffff). But if we match (and X, 0xffffffff) it will
remove the AND and the inputs may no longer have the zero bits
needed to guarantee the result has enough zeros.
This commit changes the patterns to only match assertzexti32.
I'm not sure how to test the broken case since the DIVUW/REMUW nodes
are created during type legalization, but type legalization won't
create an (and X, 0xfffffffff) directly on the inputs.
I've also changed the zexti32 on the root of the pattern to just
checking for AND. We were previously also matching assertzexti32,
but I doubt that pattern would ever occur.
Max Kazantsev [Fri, 27 Nov 2020 07:03:05 +0000 (14:03 +0700)]
[Test] Add some more tests showing how we fail to widen IV
Kazu Hirata [Fri, 27 Nov 2020 06:02:04 +0000 (22:02 -0800)]
[IR] Use llvm::is_contained and pred_size (NFC)
Max Kazantsev [Fri, 27 Nov 2020 04:19:54 +0000 (11:19 +0700)]
[IndVars] LCSSA Phi users should not prevent widening
When widening an IndVar that has LCSSA Phi users outside
the loop, we can safely widen it as usual and then truncate
the result outside the loop without hurting the performance.
Differential Revision: https://reviews.llvm.org/D91593
Reviewed By: skatkov
Kirill Bobyrev [Fri, 27 Nov 2020 02:59:24 +0000 (03:59 +0100)]
[clangd] Add support for within-file rename of complicated fields
This was originally a part of D71880 but is separated for simplicity and ease
of reviewing.
Fixes: https://github.com/clangd/clangd/issues/582
Reviewed By: hokein
Differential Revision: https://reviews.llvm.org/D91952
QingShan Zhang [Fri, 27 Nov 2020 02:10:55 +0000 (02:10 +0000)]
[DAGCombine] Adding a hook to improve the precision of fsqrt if the input is denormal
For now, we will hardcode the result as 0.0 if the input is denormal or 0. That will
have the impact the precision. As the fsqrt added belong to the cold path of the
cmp+branch, it won't impact the performance for normal inputs for PowerPC, but improve
the precision if the input is denormal.
Reviewed By: Spatel
Differential Revision: https://reviews.llvm.org/D80974
Kazu Hirata [Fri, 27 Nov 2020 02:00:54 +0000 (18:00 -0800)]
[InlineCost] Fix indentation (NFC)
Juneyoung Lee [Thu, 26 Nov 2020 10:33:42 +0000 (19:33 +0900)]
[LangRef] Add poison constant
This patch adds a description about the newly added poison constant to LangRef.
Differential Revision: https://reviews.llvm.org/D92162
Craig Topper [Thu, 26 Nov 2020 23:08:54 +0000 (15:08 -0800)]
[RISCV] Add isel patterns to use SBSET for (1 << X) by using X0 as the input.
Sam McCall [Wed, 25 Nov 2020 16:18:00 +0000 (17:18 +0100)]
[clangd] Fix use of system-installed GRPC after
f726101b6240a6740b3c0926af759da5e7336f8a
We need a real target now, and it was only being created if grpc was
built from source or imported from homebrew.
Differential Revision: https://reviews.llvm.org/D92107
Nikita Popov [Sun, 22 Nov 2020 17:23:53 +0000 (18:23 +0100)]
[BasicAA] Fix BatchAA results for phi-phi assumptions
Add a flag that disables caching when computing aliasing results
potentially based on a phi-phi NoAlias assumption. We'll still
insert cache entries temporarily to catch infinite recursion,
but will drop them afterwards, so they won't persist in BatchAA.
Differential Revision: https://reviews.llvm.org/D91936
Arthur Eubanks [Thu, 26 Nov 2020 20:33:24 +0000 (12:33 -0800)]
[test] Fix runtime-pointer-checking-insert-typesize.ll under NPM
Also clean it up a bit.
Louis Dionne [Thu, 26 Nov 2020 20:14:02 +0000 (15:14 -0500)]
[libc++] Install missing package in the Dockerfile
python3-distutils is required to use `import distutils.spawn`, which is
required by the ABI list targets.
Roman Lebedev [Thu, 26 Nov 2020 20:11:26 +0000 (23:11 +0300)]
Revert "[SimplifyCFG] FoldBranchToCommonDest: lift use-restriction on bonus instructions"
Many bots are unhappy, at the very least missed a few codegen tests,
and possibly this has a logic hole inducing a miscompile
(will be really awesome to have ready reproducer..)
Need to investigate.
This reverts commit
2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe.
Mariusz Ceier [Thu, 26 Nov 2020 20:03:22 +0000 (21:03 +0100)]
[llvmbuildectomy] Also include component groups in llvm-config --components list
Without this change mesa fails while looking for llvm components like amdgpu, engine or native:
Run-time dependency LLVM (modules: amdgpu(missing), bitreader, bitwriter, core, engine(missing), executionengine, instcombine, ipo, mcdisassembler, mcjit, native(missing), scalaropts, transformutils, coroutines)
Looking for a fallback subproject for the dependency llvm (modules: bitwriter, engine, mcdisassembler, mcjit, core, executionengine, scalaropts, transformutils, instcombine, amdgpu, native, bitreader, ipo)
This change adds component groups (like all-targets, engine, native, amdgpu) to the "all" component.
Differential Revision: https://reviews.llvm.org/D92158
Roman Lebedev [Thu, 26 Nov 2020 16:41:31 +0000 (19:41 +0300)]
[SimplifyCFG] FoldBranchToCommonDest: lift use-restriction on bonus instructions
1. It doesn't make sense to enforce that the bonus instruction
is only used once in it's basic block. What matters is
whether those user instructions fit within our budget, sure,
but that is another question.
2. It doesn't make sense to enforce that said bonus instructions
are only used within their basic block. Perhaps the branch
condition isn't using the value computed by said bonus instruction,
and said bonus instruction is simply being calculated
to be used in successors?
So iff we can clone bonus instructions, to lift these restrictions,
we just need to carefully update their external uses
to use the new cloned instructions.
Notably, this transform (even without this change) appears to be
poison-unsafe as per alive2, but is otherwise (including the patch) legal.
We don't introduce any new PHI nodes, but only "move" the instructions
around, i'm not really seeing much potential for extra cost modelling
for the transform, especially since now we allow at most one such
bonus instruction by default.
This causes the fold to fire +11.4% more (13216 -> 14725)
as of vanilla llvm test-suite + RawSpeed.
The motivational pattern is IEEE-754-2008 Binary16->Binary32
extension code:
https://github.com/darktable-org/rawspeed/blob/
ca57d77fb2ba81f21fc712cfac26e54f46406473/src/librawspeed/common/FloatingPoint.h#L115-L120
^ that should be a switch, but it is not now: https://godbolt.org/z/bvja5v
That being said, even thought this seemed like this would fix it: https://godbolt.org/z/xGq3TM
apparently that fold is happening somewhere else afterall,
so something else also has a similar 'artificial' restriction.
Roman Lebedev [Thu, 26 Nov 2020 12:21:56 +0000 (15:21 +0300)]
[NFC][SimplifyCFG] Add test coverage for FoldBranchToCommonDest xform with live-out bonus instuctions
The uses of the bonus instructions should not be preventing the transformation.
Roman Lebedev [Thu, 26 Nov 2020 11:58:28 +0000 (14:58 +0300)]
[NFC][SimplifyCFG] Add statistic to `FoldBranchToCommonDest()` fold
Louis Dionne [Thu, 26 Nov 2020 19:45:07 +0000 (14:45 -0500)]
[libc++] Remove ABI lists for previous releases
We don't actually update the ABI lists at every release -- it's too much
work, since we'd technically have to do it even for minor releases.
Furthermore, I don't think anybody uses those (I certainly don't rely
on them for anything).
Instead, it is better to rely on the ABI list changelog and the canonical
ABI list that we always keep up to date. If one wants to know what symbols
were shipped in a specific release, that can be discovered easily using
Git, which is a superior tool than keeping textual copies of old versions.
Louis Dionne [Thu, 26 Nov 2020 19:40:38 +0000 (14:40 -0500)]
[libc++] Fix the Homebrew tap to install Buildkite on macOS hosts
Adam Czachorowski [Thu, 26 Nov 2020 16:51:04 +0000 (17:51 +0100)]
[clangd] AddUsing: do not crash on non-namespace using decls.
Differential Revision: https://reviews.llvm.org/D92186
Simon Pilgrim [Thu, 26 Nov 2020 18:29:37 +0000 (18:29 +0000)]
[X86] Regenerate i64-to-float.ll tests with avx512 coverage
Rename prefix from X32 to X86 as we typically use X32 for gnux32 triples
Arthur Eubanks [Thu, 19 Nov 2020 17:37:40 +0000 (09:37 -0800)]
[BPF][NewPM] Port bpf-adjust-opt to NPM and add it to pipeline
Reviewed By: yonghong-song
Differential Revision: https://reviews.llvm.org/D91990
Nikita Popov [Tue, 17 Nov 2020 19:11:09 +0000 (20:11 +0100)]
[AA] Split up LocationSize::unknown()
Currently, we have some confusion in the codebase regarding the
meaning of LocationSize::unknown(): Some parts (including most of
BasicAA) assume that LocationSize::unknown() only allows accesses
after the base pointer. Some parts (various callers of AA) assume
that LocationSize::unknown() allows accesses both before and after
the base pointer (but within the underlying object).
This patch splits up LocationSize::unknown() into
LocationSize::afterPointer() and LocationSize::beforeOrAfterPointer()
to make this completely unambiguous. I tried my best to determine
which one is appropriate for all the existing uses.
The test changes in cs-cs.ll in particular illustrate a previously
clearly incorrect AA result: We were effectively assuming that
argmemonly functions were only allowed to access their arguments
after the passed pointer, but not before it. I'm pretty sure that
this was not intentional, and it's certainly not specified by
LangRef that way.
Differential Revision: https://reviews.llvm.org/D91649
Fangrui Song [Thu, 26 Nov 2020 17:13:08 +0000 (09:13 -0800)]
[MC] Set the unique id of .stack_sizes to the associated .text section's
Similar to D92113. Currently `clang -fstack-size-section -fno-unique-section-names`
sets the linked-to symbol to the first `.text`, which is:
* incorrect for COMDAT sections
* inferior for non-COMDAT sections in -ffunction-sections mode (poor --gc-sections: .stack_sizes cannot be separately discarded)
Note, if the section symbol can be referenced in more places (if the
function begin symbol does not apply), we probably should consider
defining a different BeginSymbol for sections with ",unique" linkage.
Reviewed By: grimar, jhenderson
Differential Revision: https://reviews.llvm.org/D92151
Zarko Todorovski [Thu, 26 Nov 2020 16:15:00 +0000 (11:15 -0500)]
[AIX] Add support for non var_arg extended vector ABI calling convention on AIX
This patch enables passing non variadic vector type parameters on the caller and callee side and vector return on AIX that are passed in vector registers only.
So far, support is enabled for only the AIX extended Altivec ABI Calling convention.
Reviewed By: sfertile, DiggerLin
Differential Revision: https://reviews.llvm.org/D86476
Raphael Isemann [Thu, 26 Nov 2020 16:02:31 +0000 (17:02 +0100)]
[ASTImporter] Import the default argument of TemplateTypeParmDecl
The test case isn't using the AST matchers for all checks as there doesn't seem to be support for
matching TemplateTypeParmDecl default arguments. Otherwise this is simply importing the
default arguments.
Also updates several LLDB tests that now as intended omit the default template
arguments of several std templates.
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D92103
Louis Dionne [Wed, 25 Nov 2020 21:42:42 +0000 (16:42 -0500)]
[libc++] Remove sysctl-based implementation of thread::hardware_concurrency()
Using sysctl requires including headers that are considered internal on
Linux, like <sys/sysctl.h> & friends. Instead, sysconf is defined by POSIX
(and we have a fallback for Windows), so all the systems we support should
be happy with just sysconf.
Differential Revision: https://reviews.llvm.org/D92135
Louis Dionne [Wed, 25 Nov 2020 21:17:23 +0000 (16:17 -0500)]
[libc++] Attempt to fix spurious modules-related failures in the CI
I'm not 100% sure what the issue actually is since I can't reproduce it
locally, however what I explain in the comment is my best attempt to
explain what's going on.
Differential Revision: https://reviews.llvm.org/D92131
Stephan Herhut [Thu, 26 Nov 2020 16:08:49 +0000 (17:08 +0100)]
[mlir][DialectConversion] Do not prematurely drop unused cast operations
The rewrite logic has an optimization to drop a cast operation after
rewriting block arguments if the cast operation has no users. This is
unsafe as there might be a pending rewrite that replaced the cast operation
itself and hence would trigger a second free.
Instead, do not remove the casts and leave it up to a later canonicalization
to do so.
Differential Revision: https://reviews.llvm.org/D92184
Alex Richardson [Thu, 26 Nov 2020 14:47:31 +0000 (14:47 +0000)]
[libc++] Fix two fr_FR locale tests on FreeBSD
FreeBSD's locale data uses the same U+2027 separator as Glibc 2.27 and newer.
Reviewed By: #libc, emaste, ldionne
Differential Revision: https://reviews.llvm.org/D91165
Simon Pilgrim [Thu, 26 Nov 2020 15:33:27 +0000 (15:33 +0000)]
[AArch64] Regenerate min/max tests and add vXi64 umin/umax test coverage
David Green [Thu, 26 Nov 2020 15:10:44 +0000 (15:10 +0000)]
[ARM] Cleanup for the MVETailPrediction pass
This strips out a lot of the code that should no longer be needed from
the MVETailPredictionPass, leaving the important part - find active lane
mask instructions and convert them to VCTP operations.
Differential Revision: https://reviews.llvm.org/D91866
Simon Pilgrim [Thu, 26 Nov 2020 15:07:45 +0000 (15:07 +0000)]
Revert rG12d59b696b330 "[DAG] Legalize umin(x,y) -> sub(x,usubsat(x,y)) and umax(x,y) -> add(x,usubsat(y,x)) iff usubsat is legal"
This reverts commit
12d59b696b33065e070d6ee7a55d2e8c019d138b.
Prematurely pushed this to trunk
Simon Pilgrim [Thu, 26 Nov 2020 14:47:06 +0000 (14:47 +0000)]
[DAG] Legalize umin(x,y) -> sub(x,usubsat(x,y)) and umax(x,y) -> add(x,usubsat(y,x)) iff usubsat is legal
If usubsat() is legal, this is likely to result in smaller codegen expansion than the default cmp+select codegen expansion.
Allows us to move the x86-specific lowering to the generic expansion code.
Benjamin Kramer [Thu, 26 Nov 2020 14:37:18 +0000 (15:37 +0100)]
Remove stray debug-only from test
Jan Svoboda [Tue, 24 Nov 2020 11:48:11 +0000 (12:48 +0100)]
[clang][cli] Port Target option flags to new option parsing system
Depends on D83697
Original patch by Daniel Grumberg.
Differential Revision: https://reviews.llvm.org/D83698
Robert Lougher [Wed, 25 Nov 2020 19:52:20 +0000 (19:52 +0000)]
[LiveDebugVariables] Strip all debug instructions from nodebug functions
A crash/assertion failure in the greedy register allocator was tracked
down to a debug instr being passed to LiveIntervals::getInstructionIndex.
Normally this should not occur as debug instructions are collected and
removed by LiveDebugVariables before RA, and reinserted afterwards.
However, when a function has no debug info, LiveDebugVariables simply
strips any debug values that are present as they're not needed (this
situation will occur when a function with debug info is inlined into a
nodebug function). The problem is, it only removes DBG_VALUE instructions,
leaving DBG_LABELs (the cause of the crash).
This patch updates the LiveDebugVariables nodebug path to remove all debug
instructions. The test case verifies that DBG_VALUE/DBG_LABEL instructions
are present, and that they are stripped.
When -experimental-debug-variable-locations is enabled, certain variable
locations are represented by DBG_INSTR_REF instead of DBG_VALUE. The test
case verifies that a DBG_INSTR_REF is emitted by the option, and that it
is also stripped.
Differential Revision: https://reviews.llvm.org/D92127
Raphael Isemann [Thu, 26 Nov 2020 12:38:49 +0000 (13:38 +0100)]
[ASTImporter] Import the default argument of TemplateTemplateParmDecl
Same idea as in D92103 and D92106, but I realised after creating those reviews that there are
also TemplateTemplateParmDecls that can have default arguments, so here's hopefully the
last patch for default template arguments.
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D92119
Florian Hahn [Thu, 26 Nov 2020 13:50:24 +0000 (13:50 +0000)]
[VPlan] Turn VPReplicateRecipe into a VPValue.
Update VPReplicateRecipe to inherit from VPValue. This still does not
update scalarizeInstruction to set the result for the VPValue of
VPReplicateRecipe, because this first requires tracking scalar values in
VPTransformState.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D91500
Stephan Herhut [Thu, 26 Nov 2020 13:40:21 +0000 (14:40 +0100)]
[mlir][linalg] Mark linalg.yield as ReturnLike
This change is required so that bufferization can properly identify
the linalg.yield as a terminator with an associated parent op.
Differential Revision: https://reviews.llvm.org/D92173
David Stenberg [Thu, 26 Nov 2020 09:15:59 +0000 (10:15 +0100)]
[IndVarSimplify] Fix Modified status when handling dead PHI nodes
When bailing out in rewriteLoopExitValues() you could be left with PHI
nodes in the DeadInsts vector. Those would be not handled by the use of
RecursivelyDeleteTriviallyDeadInstructions() in IndVarSimplify. This
resulted in the IndVarSimplify pass returning an incorrect modified
status. This was caught by the expensive check introduced in D86589.
This patches changes IndVarSimplify so that it deletes those PHI nodes,
using RecursivelyDeleteDeadPHINode().
This fixes PR47486.
Reviewed By: mkazantsev
Differential Revision: https://reviews.llvm.org/D91153
Sjoerd Meijer [Thu, 26 Nov 2020 13:03:37 +0000 (13:03 +0000)]
[AArch64][CostModel] Precommit some vector mul tests. NFC.
The cost-model is not getting the cost right for a mul with <2 x i64>
operands, i.e. we don't have a MUL.2d, and this is precommitting some
tests before adjusting this.
Kazushi (Jam) Marukawa [Mon, 23 Nov 2020 06:37:46 +0000 (15:37 +0900)]
[VE] Add comprehensive stackframe tests
Add comprehensive stackframe regression tests as a preparation of
VEFrameLowering.cpp optimizations.
Reviewed By: simoll
Differential Revision: https://reviews.llvm.org/D92170