Craig Topper [Fri, 9 Aug 2019 05:53:37 +0000 (05:53 +0000)]
[X86] Remove DAG combine expansion of extending masked load and truncating masked store.
The only way to generate these was through promoting legalization
of narrow vectors, but we widen those types now. So we shouldn't
produce these nodes.
Craig Topper [Fri, 9 Aug 2019 05:17:52 +0000 (05:17 +0000)]
[X86] Remove handler for (U/S)(ADD/SUB)SAT from ReplaceNodeResults. Remove TypeWidenVector check from code that handles X86ISD::VPMADDWD and X86ISD::AVG.
More unneeded code since we now legalize narrow vectors by widening.
David Blaikie [Fri, 9 Aug 2019 01:14:33 +0000 (01:14 +0000)]
DebugInfo/DWARF: Provide some (pretty half-hearted) error handling access when parsing units
This isn't the most robust error handling API, but does allow clients to
opt-in to getting Errors they can handle. I suspect the long-term
solution would be to move away from the lazy unit parsing and have an
explicit step that parses the unit and then allows access to the other
APIs that require a parsed unit.
llvm-dwarfdump could be expanded to use this (or newer/better API) to
demonstrate the benefit of it - but for now lld will use this in a
follow-up cl which ensures lld can exit non-zero on errors like this (&
provide more descriptive diagnostics including which object file the
error came from).
(error access to later errors when parsing nested DIEs would be good too
- but, again, exposing that without it being a hassle for every consumer
may be tricky)
GlobalAlias and GlobalIFunc ought to be treated the same by the IR
linker, so we can generalize the code to be in terms of their common
base class GlobalIndirectSymbol.
Craig Topper [Thu, 8 Aug 2019 21:36:47 +0000 (21:36 +0000)]
[X86] Improve codegen of v8i64->v8i16 and v16i32->v16i8 truncate with avx512vl, avx512bw, min-legal-vector-width<=256 and prefer-vector-width=256
Under this configuration we'll want to split the v8i64 or v16i32 into two vectors. The default legalization will try to truncate each of those 256-bit pieces one step to 128-bit, concatenate those, then truncate one more time from the new 256 to 128 bits.
With this patch we now truncate the two splits to 64-bits then concatenate those. We have to do this two different ways depending on whether have widening legalization enabled. Without widening legalization we have to manually construct X86ISD::VTRUNC to prevent the ISD::TRUNCATE with a narrow result being promoted to 128 bits with a larger element type than what we want followed by something like a pshufb to grab the lower half of each element to finish the job. With widening legalization we just get the right thing. When we switch to widening by default we can just delete the other code path.
Guozhi Wei [Thu, 8 Aug 2019 20:25:23 +0000 (20:25 +0000)]
[MBP] Disable aggressive loop rotate in plain mode
Patch https://reviews.llvm.org/D43256 introduced more aggressive loop layout optimization which depends on profile information. If profile information is not available, the statically estimated profile information(generated by BranchProbabilityInfo.cpp) is used. If user program doesn't behave as BranchProbabilityInfo.cpp expected, the layout may be worse.
To be conservative this patch restores the original layout algorithm in plain mode. But user can still try the aggressive layout optimization with -force-precise-rotation-cost=true.
Craig Topper [Thu, 8 Aug 2019 18:11:17 +0000 (18:11 +0000)]
[X86] Make CMPXCHG16B feature imply CMPXCHG8B feature.
This fixes znver1 so that it properly enables CMPXHG8B. We can
probably remove explicit CMPXCHG8B from CPUs that also have
CMPXCHG16B, but keeping this simple to allow cherry pick to 9.0.
[AArch64] Do not emit '#' before immediates in inline asm
Summary:
The A64 assembly language does not require the '#' character to
introduce constant immediate operands. Avoid the '#' since the AArch64
asm parser does not accept '#' before the lane specifier and rejects the
following:
__asm__ ("fmla v2.4s, v0.4s, v1.s[%0]" :: "I"(0x1))
Fix a test to not expect the '#' and add a new test case with the above
asm.
Tom Stellard [Thu, 8 Aug 2019 17:23:33 +0000 (17:23 +0000)]
lit: Use a License classifier that pypi will accept
Summary:
'OSI Approved :: Apache-2.0 with LLVM exception' is not a valid
classifier. 'OSI Approved :: Apache Software License' is the closest
fit for the new license, so we've decided to use this one.
The classifiers seem to only be used for searching on the pypi website,
so this does not actually change the license of the code.
We still pass 'Apache-2.0 with LLVM exception' as the license to setup(),
and this appears alongside the classifier on the pypi webpage for lit.
Akira Hatanaka [Thu, 8 Aug 2019 16:59:31 +0000 (16:59 +0000)]
[ObjC][ARC] Upgrade calls to ARC runtime functions to intrinsic calls if
the bitcode has the arm64 retainAutoreleasedReturnValue marker
The ARC middle-end passes stopped optimizing or transforming bitcode
that has been compiled with old compilers after we started emitting
calls to ARC runtime functions as intrinsic calls instead of normal
function calls in the front-end and made changes to teach the ARC
middle-end passes about those intrinsics (see r349534). This patch
converts calls to ARC runtime functions that are not intrinsic functions
to intrinsic function calls if the bitcode has the arm64
retainAutoreleasedReturnValue marker. Checking for the presence of the
marker is necessary to make sure we aren't changing ARC function calls
that were originally MRR message sends (see r349952).
Simon Pilgrim [Thu, 8 Aug 2019 15:54:20 +0000 (15:54 +0000)]
[X86] SimplifyDemandedVectorElts - attempt to recombine target shuffle using DemandedElts mask
If we don't demand all elements, then attempt to combine to a simpler shuffle.
At the moment we can only do this if Depth == 0 as combineX86ShufflesRecursively uses Depth to track whether the shuffle has really changed or not - we'll need to change this before we can properly start merging combineX86ShufflesRecursively into SimplifyDemandedVectorElts.
The insertps-combine.ll regression is because XFormVExtractWithShuffleIntoLoad can't see through shuffles of different widths - this will be fixed in a follow-up commit.
David Tenty [Thu, 8 Aug 2019 15:40:35 +0000 (15:40 +0000)]
Enable assembly output of local commons for AIX
Summary:
This patch enable assembly output of local commons for AIX using .lcomm
directives. Adds a EmitXCOFFLocalCommonSymbol to MCStreamer so we can emit the
AIX version of .lcomm assembly directives which include a csect name. Handle the
case of BSS locals in PPCAIXAsmPrinter by using EmitXCOFFLocalCommonSymbol. Adds
a test for generating .lcomm on AIX Targets.
David Green [Thu, 8 Aug 2019 15:27:58 +0000 (15:27 +0000)]
[ARM] Add support for MVE pre and post inc loads and stores
This adds pre- and post- increment and decrements for MVE loads and stores. It
uses the builtin pre and post load/store detection, unlike Neon. Loads are
selected with the code in tryT2IndexedLoad, stores are selected with tablegen
patterns. The immediates have a +/-7bit range, multiplied by the size of the
element.
David Green [Thu, 8 Aug 2019 15:15:19 +0000 (15:15 +0000)]
[ARM] MVE big endian loads/stores
This adds some missing patterns for big endian loads/stores, allowing unaligned
loads/stores to also be selected with an extra VREV, which produces better code
than aligning through a stack. Also moves VLDR_P0 to not be LE only, and
adjusts some of the tests to show all that working.
Sam Elliott [Thu, 8 Aug 2019 14:59:16 +0000 (14:59 +0000)]
[RISCV] Allow ABI Names in Inline Assembly Constraints
Summary:
Clang will replace references to registers using ABI names in inline
assembly constraints with references to architecture names, but other
frontends do not. LLVM uses the regular assembly parser to parse inline asm,
so inline assembly strings can contain references to registers using their
ABI names.
This patch adds support for parsing constraints using either the ABI name or
the architectural register name. This means we do not need to implement the
ABI name replacement code in every single frontend, especially those like
Rust which are a very thin shim on top of LLVM IR's inline asm, and that
constraints can more closely match the assembly strings they refer to.
Sam Elliott [Thu, 8 Aug 2019 14:40:54 +0000 (14:40 +0000)]
[RISCV] Minimal stack realignment support
Summary:
Currently the RISC-V backend does not realign the stack. This can be an issue even for the RV32I/RV64I ABIs (where the stack is 16-byte aligned), though is rare. It will be much more comment with RV32E (though the alignment requirements for common data types remain under-documented...).
This patch adds minimal support for stack realignment. It should cope with large realignments. It will error out if the stack needs realignment and variable sized objects are present.
It feels like a lot of the code like getFrameIndexReference and determineFrameLayout could be refactored somehow, as right now it feels fiddly and brittle. We also seem to allocate a lot more memory than GCC does for equivalent C code.
Tim Corringham [Thu, 8 Aug 2019 13:46:17 +0000 (13:46 +0000)]
Add llvm.licm.disable metadata
For some targets the LICM pass can result in sub-optimal code in some
cases where it would be better not to run the pass, but it isn't
always possible to suppress the transformations heuristically.
Where the front-end has insight into such cases it is beneficial
to attach loop metadata to disable the pass - this change adds the
llvm.licm.disable metadata to enable that.
Fix check in tools/gold/X86/strip_names.ll regarding unnamed args
After r367755 value numbers are printed for unnamed
function arguments. The tools/gold/X86/strip_names.ll
was not updated in that commit, so this patch can be
seen as a follow up to r367755.
Simon Pilgrim [Thu, 8 Aug 2019 10:37:03 +0000 (10:37 +0000)]
[TargetLowering] SimplifyDemandedBits - call SimplifyMultipleUseDemandedBits for ISD::EXTRACT_VECTOR_ELT
This patch attempts to peek through vectors based on the demanded bits/elt of a particular ISD::EXTRACT_VECTOR_ELT node, allowing us to avoid dependencies on ops that have no impact on the extract.
In particular this helps remove some unnecessary scalar->vector->scalar patterns.
The wasm shift patterns are annoying - @tlively has indicated that the wasm vector shift codegen are to be refactored in the near-term and isn't considered a major issue.
George Rimar [Thu, 8 Aug 2019 09:49:05 +0000 (09:49 +0000)]
[yaml2obj/obj2yaml] - Add a basic support for extended section indexes.
In some cases a symbol might have section index == SHN_XINDEX.
This is an escape value indicating that the actual section header index
is too large to fit in the containing field.
Then the SHT_SYMTAB_SHNDX section is used. It contains the 32bit values
that stores section indexes.
ELF gABI says that there can be multiple SHT_SYMTAB_SHNDX sections,
i.e. for example one for .symtab and one for .dynsym
(1) https://groups.google.com/forum/#!topic/generic-abi/-XJAV5d8PRg
(2) DT_SYMTAB_SHNDX: http://www.sco.com/developers/gabi/latest/ch5.dynamic.html
In this patch I am only supporting a single SHT_SYMTAB_SHNDX associated
with a .symtab. This is a more or less common case which is used a few tests I saw in LLVM.
I decided not to create the SHT_SYMTAB_SHNDX section as "implicit",
but implement is like a kind of regular section for now.
i.e. tools do not recreate this section or its content, like they do for
symbol table sections, for example. That should allow to write all kind of
possible broken test cases for our needs and keep the output closer to requested.
The error is:
/home/buildslave/slave_as-bldslv8/lld-perf-testsuite/llvm/include/llvm/Object/ELF.h:67:14:
error: 'static' function 'defaultWarningHandler' declared in header file should be declared 'static inline' [-Werror,-Wunneeded-internal-declaration]
static Error defaultWarningHandler(const Twine &Msg) { return createError(Msg); }
^
George Rimar [Thu, 8 Aug 2019 07:17:35 +0000 (07:17 +0000)]
[llvm-readobj/libObject] - Introduce a custom warning handler for `ELFFile<ELFT>` methods.
Currently, we have a code duplication in llvm-readobj which was introduced in D63266.
The duplication was introduced to allow llvm-readobj to dump the partially
broken object. Methods in ELFFile<ELFT> perform a strict validation of the inputs,
what is itself good, but not for dumper tools, that might want to dump the information,
even if some pieces are broken/unexpected.
This patch introduces a warning handler which can be passed to ELFFile<ELFT> methods
and can allow skipping the non-critical errors when needed/possible.
For demonstration, I removed the duplication from llvm-readobj and implemented a warning using
the new custom warning handler. It also deduplicates the strings printed, making the output less verbose.
David Green [Thu, 8 Aug 2019 06:22:03 +0000 (06:22 +0000)]
[ARM] Tighten up VLDRH.32 with low alignments
VLDRH needs to have an alignment of at least 2, including the
widening/narrowing versions. This tightens up the ISel patterns for it and
alters allowsMisalignedMemoryAccesses so that unaligned accesses are expanded
through the stack. It also fixed some incorrect shift amounts, which seemed to
be passing a multiple not a shift.
It's been in for more than 30 min and no bots have complained. Let's see if some
slow ones catch up. I'll do another manual pass on bots later (in case some that
were down are back up), and then turn this on permanently through a regular
review.
JF Bastien [Thu, 8 Aug 2019 05:12:20 +0000 (05:12 +0000)]
Temporarily bump minimum compiler version
It's pretty hard to find a reliable list of which bots use which compiler version... so I'm going to commit this change which allows us to mandate the compilers required for C++14. This bump is what we've already agreed to do, so I'll use the list of failures to figure out which bots need to bump their compiler version. I'll revert the change in a few minutes.
The last discussion of this is here: http://lists.llvm.org/pipermail/llvm-dev/2019-August/134360.html
[Attributor][Stats] Locate statistics tracking with the attributes
Summary:
The ever growing switch required Attribute::AttrKind values but they
might not be available for all abstract attributes we deduce. With the
new method we track statistics at the abstract attribute level. The
provided macros simplify the usage and make the messages uniform.
Summary:
The wrapper reduces boilerplate code and also provide a nice way to
determine the state type used by an abstract attributes statically via
AAType::StateType.
This was already discussed as part of the review of D65711.
If we know everything is live there is no need to query for liveness.
Indicating a pessimistic fixpoint will cause the state to be "invalid"
which will cause the Attributor to not return the AAIsDead on request,
which will prevent us from querying isAssumedDead().
[Attributor] Provide easier checkForallReturnedValues functionality
Summary:
So far, whenever one wants to look at returned values, one had to deal
with the AAReturnedValues and potentially with the AAIsDead attribute.
In the same spirit as other checkForAllXXX methods, we add this
functionality now to the Attributor. By adopting the use sites we got
better results when return instructions were dead.
Bob Haarman [Wed, 7 Aug 2019 20:45:23 +0000 (20:45 +0000)]
Revert r367501 "Create unique, but identically-named ELF sections..."
This reverts commit fbc563e2cb6c5f1d0200b390513506b6aca4d2e9 "Create
unique, but identically-named ELF sections for explicitly-sectioned
functions and globals when using -function-sections and
-data-sections."
Reason for revert: sections are created with potentially wrong
attributes.
Craig Topper [Wed, 7 Aug 2019 18:28:16 +0000 (18:28 +0000)]
[ValueTracking] When calculating known bits for integer abs, make sure we're looking at a negate and not just any instruction with the nsw flag set.
The matchSelectPattern code can match patterns like (x >= 0) ? x : -x
for absolute value. But it can also match ((x-y) >= 0) ? (x-y) : (y-x).
If the latter form was matched we can only use the nsw flag if its
set on both subtracts.
This match makes sure we're looking at the former case only.
David Blaikie [Wed, 7 Aug 2019 17:18:11 +0000 (17:18 +0000)]
DebugInfo/DWARF: Normalize DWARFObject members on the DWARF spec section names
Some of these names were abbreviated, some were not, some pluralised,
some not. Made the API difficult to use - since it's an exact 1:1
mapping to the DWARF sections - use those names (changing underscore
separation for camel casing).
Nico Weber [Wed, 7 Aug 2019 17:00:19 +0000 (17:00 +0000)]
Support: Remove needless allocation when getMainExecutable() calls readlink()
We built a StringRef from a string literal which we then converted to a
std::string to call c_str(). Just use a pointer to the string literal
instead of a StringRef.
Craig Topper [Wed, 7 Aug 2019 16:24:26 +0000 (16:24 +0000)]
Recommit r367901 "[X86] Enable -x86-experimental-vector-widening-legalization by default."
The assert that caused this to be reverted should be fixed now.
Original commit message:
This patch changes our defualt legalization behavior for 16, 32, and
64 bit vectors with i8/i16/i32/i64 scalar types from promotion to
widening. For example, v8i8 will now be widened to v16i8 instead of
promoted to v8i16. This keeps the elements widths the same and pads
with undef elements. We believe this is a better legalization strategy.
But it carries some issues due to the fragmented vector ISA. For
example, i8 shifts and multiplies get widened and then later have
to be promoted/split into vXi16 vectors.
This has the potential to cause regressions so we wanted to get
it in early in the 10.0 cycle so we have plenty of time to
address them.
Next steps will be to merge tests that explicitly test the command
line option. And then we can remove the option and its associated
code.
Jay Foad [Wed, 7 Aug 2019 15:16:28 +0000 (15:16 +0000)]
[InstCombine] Propagate fast math flags through selects
Summary:
In SimplifySelectsFeedingBinaryOp, propagate fast math flags from the
outer op into both arms of the new select, to take advantage of
simplifications that require fast math flags.
David Bolvansky [Wed, 7 Aug 2019 14:44:50 +0000 (14:44 +0000)]
[UpdateTestChecks] Update tests option
Summary:
Port of new feature introduced https://reviews.llvm.org/D65610 to other update scripts.
- update_*_checks.py: add an alias -u for --update-only
- port --update-only to other update_*_test_checks.py scripts
- update script aborts if the test file was generated by another update_*_test_checks.py utility
projects / llvm / log