Alexander Shaposhnikov [Wed, 9 Nov 2022 21:36:18 +0000 (21:36 +0000)]
[CodeGen][AArch64] Enable LDAPR under +RCPC
This is a follow-up to D126250 and enables LDAPR
if the RCPC extensions are enabled.
Test plan: ninja check-all
Differential revision: https://reviews.llvm.org/D137590
Nikolas Klauser [Wed, 9 Nov 2022 20:01:42 +0000 (21:01 +0100)]
[libc++] Add FTM for constexpr vector
It looks like we forgot to set the FTM when adding constexpr vector support.
Reviewed By: ldionne, #libc
Spies: libcxx-commits, arichardson
Differential Revision: https://reviews.llvm.org/D137729
Aart Bik [Wed, 9 Nov 2022 21:05:43 +0000 (13:05 -0800)]
[mlir][sparse] first end-to-end matmul with codegen
(1) also fixes memory leak in sparse2dense rewriting
(2) still needs fix in dense2sparse by skipping zeros
Reviewed By: wrengr
Differential Revision: https://reviews.llvm.org/D137736
Sander de Smalen [Wed, 9 Nov 2022 21:14:25 +0000 (21:14 +0000)]
Revert "[AArch64][SME] Disable GlobalISel/FastISel for SME functions."
Reverting the patch due to a buildbot failure.
This reverts commit
e1e260cc64bd900d5f3f88187c60cb02d3a805f5.
Michael Francis [Tue, 8 Nov 2022 19:37:29 +0000 (19:37 +0000)]
[Test][AIX][pg] Add 32-bit linker invocation tests
Differential Review: https://reviews.llvm.org/D137372
stanley-nod [Wed, 9 Nov 2022 19:26:20 +0000 (11:26 -0800)]
[mlir][vector] Modify constraint and interface for warp reduce on f16 and i8
Quantization method is crucial and ubiqutous in accelerating machine
learning workloads. Most of these methods uses f16 and i8 types.
This patch relaxes the type contraints on warp reduce distribution to
allow these types. Furthermore, this patch also changed the interface
and moved the initial reduction of data to a single thread into the
distributedReductionFn, this gives flexibility for developers to control
how they are obtaining the initial lane value, which might differ based
on the input types. (i.e to shuffle 32-width type, we need to reduce f16
to 2xf16 types rather than a single element).
Reviewed By: ThomasRaoux
Differential Revision: https://reviews.llvm.org/D137691
Michael Francis [Tue, 8 Nov 2022 19:32:44 +0000 (19:32 +0000)]
[Test][AIX][p] Add 64-bit linker invocation tests
Differential Review: https://reviews.llvm.org/D137373
Leonard Chan [Tue, 8 Nov 2022 23:11:26 +0000 (23:11 +0000)]
[compiler-rt][hwasan] Call __hwasan_library_loaded via
__sanitizer_library_loaded on Fuchsia
Differential Revision: https://reviews.llvm.org/D133806
Leonard Chan [Tue, 8 Nov 2022 23:06:19 +0000 (23:06 +0000)]
[compiler-rt][hwasan] Do not call InitLoadedGlobals in __hwasan_init
Fuchsia's libc provides a new hook (__sanitizer_module_loaded) which calls
hwasan_library_loaded in the startup path which will register globals in
loaded modules.
Differential Revision: https://reviews.llvm.org/D137676
Jannik Silvanus [Thu, 27 Oct 2022 07:14:40 +0000 (09:14 +0200)]
[llvm-diff] Fix false-positive diffs on forward-referencing phi nodes
When a phi node references a variable defined in a basic block dominated
by the the basic block containing the phi node, llvm-diff currently cannot determine
whether the variable is equivalent, and thus treats the phi node as different
and reports a difference. This leads to false positive differences as demonstrated
by the loop.ll diff, for which llvm-diff reports a diff when comparing the file
with itself.
Fix that issue by adding the concept of *equivalence assumptions*.
When encountering a pair of values which can neither be proven to be equivalent
nor to be non-equivalent, instead optimistically assume equivalence, and store
somewhere that the equivalence of the currently compared basic blocks depends
on this assumption.
Later, once all BBs have been processed, check all made assumptions and report
blocks as different whose equivalence was depending on an incorrect assumption,
or an assumption we could not prove to be correct.
In order to preserve the original diff report order, also schedule diffs
of blocks already known to be different using the same mechanism, so all block
diffs are now generated at the very end of function diffing.
In case an incorrect assumption was made, all further shown equivalences between
old and new values implictly depend on the incorrect assumption. Some of these
may in fact be not equivalent, but these are neither reverted nor reported,
because they are considered indirect diffs caused by an earlier direct diff.
See inline comments for an argument why we do not run into issues caused by circular
proof dependencies.
Differential Revision: https://reviews.llvm.org/D137318
Tom Stellard [Wed, 9 Nov 2022 19:32:26 +0000 (11:32 -0800)]
Revert "Move googletest to the third-party directory"
This reverts commit
59052468c3e38cab15582cefbb5133fd4c2ffce5.
It looks like this patch breaks the build when compiler-rt is passed to
LLVM_ENABLE_PROJECTS instead of LLVM_ENABLE_RUNTIMES.
Craig Topper [Wed, 9 Nov 2022 19:16:00 +0000 (11:16 -0800)]
[RISCV] Support BSET/BCLR/BINV in hasAllWUsers.
These instructions set, clear, or invert a single bit. If their
users don't use the upper 32 bits, they don't use the upper 32 bits
of their input.
Additionally the second operand BSET/BCLR/BINV is a shift amount
which only uses 6 bits of the source.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D137452
Craig Topper [Wed, 9 Nov 2022 19:15:30 +0000 (11:15 -0800)]
[RISCV] Support shift/rotate amount operands in isAllUsesReadW.
These operands use 6 bits.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D137450
Craig Topper [Wed, 9 Nov 2022 19:15:03 +0000 (11:15 -0800)]
[RISCV] Improve support for ADD_UW/SHXADD_UW in hasAllWUsers.
The first use operand of these is implicitly zero extended. We
can consider that a W read. If the use is the other operand, we
need to look through the instruction.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D137449
Craig Topper [Wed, 9 Nov 2022 19:14:21 +0000 (11:14 -0800)]
[RISCV] Support SB/SH/SW in hasAllWUsers in RISCVSExtWRemoval.
After D137446 we can see which operand is the user. If the user
is the value operand of a SB/SH/SW then the upper 32 bits aren't
used.
Reviewed By: asb
Differential Revision: https://reviews.llvm.org/D137448
Craig Topper [Wed, 9 Nov 2022 07:40:15 +0000 (23:40 -0800)]
[RISCV] Rework hasAllWUsers in RISCVSExtWRemoval. NFCI
Instead of storing the uses to check in the worklist, store the
instruction we want to check uses for.
Now we pop and instruction from the worklist, loop over its uses
and check them. If it's something we need to look across, we'll push
it to the worklist.
By doing it this way, we can have access to which operand
of the user is using the instruction. This will allow supporting
store instructions since we'll be able to disambiguate the the value
operand and the pointer operand. We can also improve support for
*add.uw instructions and shift amount uses.
Reviewed By: mohammed-nurulhoque, asb
Differential Revision: https://reviews.llvm.org/D137446
Craig Topper [Wed, 9 Nov 2022 07:24:01 +0000 (23:24 -0800)]
[clang] Mention vector in the description for -mno-implict-float.
As far as I understand, this disables all integer vectors too.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D134089
Lang Hames [Wed, 9 Nov 2022 19:13:18 +0000 (11:13 -0800)]
[ORC] Add a unit test to verify that bound weak symbols can't be overridden.
Weak symbols can be overridden while they're in the NeverSearched state, but
should not be able to be overridden once they've been bound by some lookup.
Historically we guaranteed this by stripping the weak flag once a symbol as
bound, causing it to appear as if it were strong. In
ffe2dda29f3 we changed
that behavior to retain weak flags on symbols (to facilitate tracking for
dynamic re-binding during dlopen). This test checks that we still fail as
required after
ffe2dda29f3.
Tom Stellard [Wed, 9 Nov 2022 16:51:34 +0000 (08:51 -0800)]
Move googletest to the third-party directory
This will help improve the project's layering, so that sub-projects
that don't actually need any llvm code can still use googletest
without having to reference code in the llvm directory.
This will also make it easier to consolidate and simplify the standalone
build configurations.
Reviewed By: stellaraccident, lattner, probinson, phosek
Differential Revision: https://reviews.llvm.org/D131919
Renaud-K [Wed, 9 Nov 2022 19:04:22 +0000 (11:04 -0800)]
Using higher level interface to insert new arguments so arguments
and their corresponding attributes are moved together
Differential revision: https://reviews.llvm.org/D13767
v1nh1shungry [Wed, 9 Nov 2022 20:04:20 +0000 (21:04 +0100)]
[clangd] Fix the code action `RemoveUsingNamespace`
Avoid adding qualifiers before user-defined literals
Reviewed By: tom-anders
Differential Revision: https://reviews.llvm.org/D137550
Jim Ingham [Wed, 9 Nov 2022 18:11:16 +0000 (10:11 -0800)]
Handle aliasing a non-top-level command.
This didn't work previously because we had to check whether the incoming
command was an alias command, but if it wasn't we still used the result
of that lookup - which was by the command's node name. That fails for
non-top-level commands. In this case, the resolution is pretty simple since
we already have the node's CommandObject, so all we needed to do was turn
it into a shared pointer, for which I added enable_shared_from_this to the
CommandObject.
Differential Revision: https://reviews.llvm.org/D137662
Jim Ingham [Wed, 9 Nov 2022 18:03:30 +0000 (10:03 -0800)]
Don't try to create Expressions when the process is running.
We generally prohibit this at a higher level - for instance requiring
the process to be stopped for "expr". But when we trigger an expression
for internal purposes (e.g. to fetch types from the ObjC runtime) we weren't
checking the process state. Now we explicitly check this at the very start
of the job so we don't get into bad states.
Differential Revision: https://reviews.llvm.org/D137684
Ilya Biryukov [Wed, 9 Nov 2022 17:31:14 +0000 (18:31 +0100)]
[include-cleaner] Add dependency to fix build with shared libraries
Jason Molenda [Wed, 9 Nov 2022 16:38:57 +0000 (08:38 -0800)]
Rename unique-types2/TestUniqueTypes.py to TestUniqueTypes2.py
lang/cpp/unique-types/TestUniqueTypes.py already exists.
Simon Pilgrim [Wed, 9 Nov 2022 17:08:36 +0000 (17:08 +0000)]
[X86] Replace unnecessary CVTPS2PI/CVTPS2DQ overrides with better base class defs
Broadwell/Haswell were completely overriding the WriteCvtPD2I class defs - we can remove those overrides entirely by just choosing better class defs.
Also fixes the scheduler for a missing YMM folded case - confirmed with Agner + uops.info that the port usage is correct
Jon Chesterfield [Wed, 9 Nov 2022 16:54:20 +0000 (16:54 +0000)]
Revert "[amdgpu][lds] Use the same isKernel predicate consistently"
Looks like this composed poorly with a nominally independent patch, will fix
This reverts commit
0ba0398517778514eb44cb7ba9bf9d4d20a856e0.
Jon Chesterfield [Wed, 9 Nov 2022 16:45:02 +0000 (16:45 +0000)]
[amdgpu][lds] Use the same isKernel predicate consistently
isKernelCC != isKernel(F->getCallingConv())
There's a test case (lower-kernel-lds.ll) that explicitly skips amdgpu_ps
so this change picks the isKernel predicate that continues to skip that
calling convention.
isKernel returns true for AMDGPU_KERNEL and SPIR_KERNEL. isKernelCC also
returns true for other calling conventions.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D136599
Jon Chesterfield [Wed, 9 Nov 2022 16:17:33 +0000 (16:17 +0000)]
Revert "[LLD] Enable --no-undefined-version by default."
This reverts commit
eedbe44b8755f7d162eee43cb3e8c9da1e61ebad.
Sander de Smalen [Wed, 9 Nov 2022 15:53:53 +0000 (15:53 +0000)]
[AArch64][SME] Disable GlobalISel/FastISel for SME functions.
This patch ensures that GlobalISel and FastISel fall back to regular DAG ISel when:
* A function requires streaming-mode to be enabled at the start/end of the function.
This happens when the function has no streaming interface, but does have a streaming body.
* A function requires a lazy-save to be committed at the start of the function.
This happens if the function has the `aarch64_pstate_za_new` attribute.
* A call to a function requires a change in streaming-mode.
* A call to a function requires a lazy-save buffer to be set up.
Patch by @CarolineConcatto
Reviewed By: david-arm
Differential Revision: https://reviews.llvm.org/D136361
Jon Chesterfield [Wed, 9 Nov 2022 15:57:38 +0000 (15:57 +0000)]
[amdgpu][lds] Use a consistent order of fields in generated structs
Avoids spurious and confusing test failures on changing implementation.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D136598
Sander de Smalen [Tue, 8 Nov 2022 13:22:48 +0000 (13:22 +0000)]
[AArch64] NFC: Remove unused parameter from allocateLazySaveBuffer
Oleg Shyshkov [Wed, 9 Nov 2022 14:59:54 +0000 (15:59 +0100)]
Revert "[mlir][linalg] Replace "string" iterator_types attr with enums in LinalgInterface."
Breaks linalg python tests. Would need to also update python/mlir/dialects/linalg/opdsl.
This reverts commit
b809d73973bb5aeedeb6a18cac2a7b3111d0c8d2.
Simon Pilgrim [Wed, 9 Nov 2022 14:54:03 +0000 (14:54 +0000)]
[X86] Replace unnecessary CVTSD2SI/CVTSS2SI overrides with better base class defs
Broadwell/Haswell were completely overriding the WriteCvtSD2I/WriteCvtSS2I class defs - we can remove those overrides entirely by just choosing better class defs.
Oleg Shyshkov [Wed, 9 Nov 2022 14:42:40 +0000 (15:42 +0100)]
[mlir][linalg] Replace "string" iterator_types attr with enums in LinalgInterface.
[RFC: EnumAttr for iterator types in Linalg](https://discourse.llvm.org/t/rfc-enumattr-for-iterator-types-in-linalg/64535)
This affect touches and probably breaks most of the code that creates `linalg.generic`. A fix would be to replace calls to `getParallelIteratorTypeName/getReductionIteratorTypeName` with `mlir::utils::IteratorType::parallel/reduction` and types from `StringRef` to `mlir::utils::IteratorType`.
Due to limitations of tablegen, shared C++ definition of IteratorType enum lives in StructuredOpsUtils.td, but each dialect should have it's own EnumAttr wrapper. To avoid conflict, all enums in a dialect are put into a separate file with a separate tablegen rule.
Test dialect td files are refactored a bit.
Printed format of `linalg.generic` temporarily remains unchanged to avoid breaking code and tests in the same change.
Differential Revision: https://reviews.llvm.org/D137658
David Sherwood [Thu, 3 Nov 2022 11:04:43 +0000 (11:04 +0000)]
[AArch64][SVE2] Add the SVE2.1 integer quadword reduction instructions
This patch adds the assembly/disassembly for the following instructions:
addqv : Unsigned add reduction of quadword vector segments
andqv : Bitwise AND reduction of quadword vector segments
eorqv : Bitwise exclusive OR reduction of quadword vector segments
orqv : Bitwise inclusive OR reduction of quadword vector segments
smaxqv : Signed maximum reduction of quadword vector segments
sminqv : Signed minimum reduction of quadword vector segments
umaxqv : Unsigned maximum reduction of quadword vector segments
uminqv : Unsigned minimum reduction of quadword vector segments
The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09
Differential Revision: https://reviews.llvm.org/D137411
Nikita Popov [Wed, 9 Nov 2022 14:25:38 +0000 (15:25 +0100)]
[SCCP] Add tests for with.overflow intrinsics (NFC)
Aaron Ballman [Wed, 9 Nov 2022 14:19:24 +0000 (09:19 -0500)]
Fix Clang sphinx build
This addresses the issue found by:
https://lab.llvm.org/buildbot/#/builders/92/builds/35350
Tomasz Kamiński [Wed, 9 Nov 2022 13:31:49 +0000 (14:31 +0100)]
[analyzer] Fix crash for array-delete of UnknownVal values.
We now skip the destruction of array elements for `delete[] p`,
if the value of `p` is UnknownVal and does not have corresponding region.
This eliminate the crash in `getDynamicElementCount` on that
region and matches the behavior for deleting the array of
non-constant range.
Reviewed By: isuckatcs
Differential Revision: https://reviews.llvm.org/D136671
Vladislav Vinogradov [Wed, 2 Nov 2022 11:22:14 +0000 (14:22 +0300)]
[Support] Format provider improvements
Remove `std::forward` call for `iterator_range` iterator de-reference.
It fixes formatting usage for some tricky cases, like special ranges,
which de-reference to value type.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D94769
Nicolas Vasilache [Wed, 9 Nov 2022 13:09:39 +0000 (05:09 -0800)]
[mlir][Transform] Fix transform.sequence crash in the presence of propagated silenceable errors and yield operations
Differential Revision: https://reviews.llvm.org/D137708
Sven van Haastregt [Wed, 9 Nov 2022 13:17:23 +0000 (13:17 +0000)]
[OpenCL] Guard read_write image3d with cl_khr_3d_image_writes
Not all `read_write image3d_t` occurrences in opencl-c.h were guarded
with `cl_khr_3d_image_writes`. Align with `-fdeclare-opencl-builtins`.
Vladislav Vinogradov [Wed, 9 Nov 2022 13:14:39 +0000 (16:14 +0300)]
Revert "[Support] Format provider improvements"
This reverts commit
791bdba0b183371ca3e16bb8411a059b4f8636ac.
Bot failure: https://lab.llvm.org/buildbot/#/builders/223/builds/9914
Vladislav Vinogradov [Wed, 2 Nov 2022 11:22:14 +0000 (14:22 +0300)]
[Support] Format provider improvements
Remove `std::forward` call for `iterator_range` iterator de-reference.
It fixes formatting usage for some tricky cases, like special ranges,
which de-reference to value type.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D94769
Haojian Wu [Wed, 9 Nov 2022 12:03:12 +0000 (13:03 +0100)]
Fix an unused-variable warning in release build, NFC.
Tomas Matheson [Sun, 23 Oct 2022 20:44:21 +0000 (21:44 +0100)]
[ARM] Move Triple::getARMCPUForArch into ARMTargetParser
This is very backend specific so either belongs in Toolchains/ARM or in
ARMTargetParser. Since it is used in lldb, ARMTargetParser made more sense.
This is part of an effort to move information about ARM/AArch64 architecture
versions, extensions and CPUs into their respective TargetParsers.
Differential Revision: https://reviews.llvm.org/D137564
OCHyams [Wed, 9 Nov 2022 11:39:29 +0000 (11:39 +0000)]
[Assignment Tracking][10/*] salvageDebugInfo for dbg.assign intrinsics
The Assignment Tracking debug-info feature is outlined in this RFC:
https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir
Plumb in salvaging for the address part of dbg.assign intrinsics.
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D133293
Nikita Popov [Wed, 9 Nov 2022 11:18:18 +0000 (12:18 +0100)]
[SCCP] Add helper for getting constant range (NFC)
Add a helper for the recurring pattern of getting a constant range
if the value lattice element is one, or a full range otherwise.
Victor Campos [Wed, 19 Oct 2022 13:03:57 +0000 (14:03 +0100)]
[AArch64] Add support for the Cortex-X3 CPU
Cortex-X3 is an Armv9-A AArch64 CPU.
This patch introduces support for Cortex-X3.
Technical Reference Manual: https://developer.arm.com/documentation/101593/latest
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D136589
Kai Sasaki [Wed, 9 Nov 2022 11:11:55 +0000 (12:11 +0100)]
[mlir][complex] Canonicalize complex.sub zero
Subtracting zero constant can be fold in no complex.sub operation.
Reviewed By: pifon2a
Differential Revision: https://reviews.llvm.org/D137523
OCHyams [Wed, 9 Nov 2022 10:49:55 +0000 (10:49 +0000)]
[Assignment Tracking][9/*] Don't drop DIAssignID in dropUnknownNonDebugMetadata
The Assignment Tracking debug-info feature is outlined in this RFC:
https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir
A DIAssignID attachment is debug metadata, so don't drop it.
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D133292
OCHyams [Wed, 9 Nov 2022 10:41:28 +0000 (10:41 +0000)]
[Assignment Tracking][8/*] Add DIAssignID merging utilities
The Assignment Tracking debug-info feature is outlined in this RFC:
https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir
Add method:
Instruction::mergeDIAssignID(
ArrayRef<const Instruction* > SourceInstructions)
which merges the DIAssignID metadata attachments on `SourceInstructions` and
`this` and replaces uses of the original IDs with the new shared one.
This is used when stores are merged, for example sinking stores out of a
if-diamond CFG or vectorizing contiguous stores.
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D133291
Jean Perier [Wed, 9 Nov 2022 10:40:56 +0000 (11:40 +0100)]
[flang][NFC] move fir namespace into FortranVariableOpInterface ODS declaration
This will be needed to be able to use this interface in HLFIR ODS operation definition
and get the right namespaces.
Differential Revision: https://reviews.llvm.org/D137702
Jean Perier [Wed, 9 Nov 2022 10:37:53 +0000 (11:37 +0100)]
[flang][NFC] move HLFIR dialect definition in IR
Transforms will need to be put in a different library to avoid circular
dependencies with other parts of the compiler. So create an IR
directory, this (flat include directory, IR and Transform directories in
lib) matches how dialects like LLVM dialect are implemented.
Differential Revision: https://reviews.llvm.org/D137701
Sheng [Wed, 9 Nov 2022 10:37:03 +0000 (18:37 +0800)]
[M68k] Add support for atomic instructions
This adds support for atomic_load, atomic_store, atomic_cmpxchg
and atomic_rmw
Fixes #48236
Reviewed by: myhsu, efriedma
Differential Revision: https://reviews.llvm.org/D136525
Sheng [Wed, 9 Nov 2022 10:25:09 +0000 (18:25 +0800)]
[M68k] Replace `IsM680x0` with predicates `AtLeastM680x0`
The former is incorrect.
Reviewed By: myhsu
Differential Revision: https://reviews.llvm.org/D137425
Thomas Symalla [Wed, 9 Nov 2022 10:23:24 +0000 (11:23 +0100)]
[NFC][AMDGPU] Pre-commit tests for D136432.
chenglin.bi [Wed, 9 Nov 2022 10:06:21 +0000 (18:06 +0800)]
[AArch64] Support all extend/shift op for pattern: (ExtendOrShfitNode - Y) + Z --> (Z - Y) + ExtendOrShfitNode
Followup rG325a3083b5b24
Add SRL/SRA/xxx_EXTEND/AND support
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D137069
OCHyams [Tue, 8 Nov 2022 16:58:53 +0000 (16:58 +0000)]
Reapply: [Assignment Tracking][7/*] Add assignment tracking functionality to clang
Reverted in
98fa95492f3bbd5befdeb36c88a3ac5ef2740b4e.
The Assignment Tracking debug-info feature is outlined in this RFC:
https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir
This patch plumbs the AssignmentTrackingPass (AKA declare-to-assign), added in
the previous patch in this set, into the optimisation pipeline from
clang. clang/test/CodeGen/assignment-tracking/assignment-tracking.cpp is the
main test for this patch.
Note: while clang (with the help of the declare-to-assign pass) can now emit
Assignment Tracking metadata, the llvm middle and back ends don't yet
understand it.
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D132226
OCHyams [Wed, 9 Nov 2022 09:11:02 +0000 (09:11 +0000)]
Fix D132221 rebase: Add flags to enable Assignment Tracking
The flags were not set up properly after rebasing the patch, causing bot
failures for D132226, for example:
https://lab.llvm.org/buildbot/#/builders/109/builds/50251
David Sherwood [Mon, 7 Nov 2022 10:18:34 +0000 (10:18 +0000)]
[AArch64][SVE2] Add the SVE2.1 quadword structured load/store instructions
This patch adds the assembly/disassembly for the following instructions:
ld2q : Contiguous load two-quadword structures to two vectors
ld3q : Contiguous load three-quadword structures to three vectors
ld4q : Contiguous load four-quadword structures to four vectors
st2q : Contiguous store two-quadword structures from two vectors
st3q : Contiguous store three-quadword structures to three vectors
st4q : Contiguous store four-quadword structures to four vectors
The reference can be found here:
https://developer.arm.com/documentation/ddi0602/2022-09
Differential Revision: https://reviews.llvm.org/D137554
Bjorn Pettersson [Tue, 8 Nov 2022 07:04:39 +0000 (08:04 +0100)]
[test] Avoid legacy PM default pipelines (O0,O1 etc) when running opt
Two lit tests were found running something like this:
opt -O<n> -pass-locked-to-legacy-PM ...
The expand-atomicrmw-xchg-fp.ll seem to have used -O1 just to ensure
that the -atomic-expand pass were thinking that it wasn't running at
O0 level. Same thing can be ensured by using the -codegen-opt-level=1
option, making it possible to avoid using O1 in that test case.
In the vector-reductions-expanded.ll test case it was possible to
split the RUN line into using two opt invocations. First running
"opt -O2" using the new PM, and then running "opt -expand-reductions"
using the legacy PM.
I think that given this patch we get closer to removing code related
to 'AddOptimizationPasses' in opt.cpp.
Differential Revision: https://reviews.llvm.org/D137626
Freddy Ye [Wed, 9 Nov 2022 08:12:25 +0000 (16:12 +0800)]
[X86] Support -march=sierraforest, grandridge, graniterapids.
Reviewed By: skan, pengfei, MaskRay
Differential Revision: https://reviews.llvm.org/D137153
melonedo [Tue, 16 Aug 2022 04:56:59 +0000 (12:56 +0800)]
[RISCV] Add support for static chain
The static chain parameter is a special parameter that is not passed in the usual argument registers or stack space. For example, in x64 System V ABI it is always passed in R10. Although the ABI of RISCV does not assign a register for this purpose, GCC had support for it on RISC-V a long time ago, and it is exposed via `__builtin_call_with_static_chain` intrinsic, and assign t2 for static chain parameters. This patch also chose t2 for compatibility.
In LLVM, static chain parameters are handled by the `nest` attribute of an argument to a function ([D6332](https://reviews.llvm.org/D6332)), so tests are added to ensure `nest` arguments are handled correctly.
Reviewed By: kito-cheng, MaskRay
Differential Revision: https://reviews.llvm.org/D129106
chenglin.bi [Wed, 9 Nov 2022 07:36:47 +0000 (15:36 +0800)]
[AArch64] Precommit test for D137069; NFC
Vitaly Buka [Wed, 9 Nov 2022 06:59:03 +0000 (22:59 -0800)]
[compiler-rt] Set CMP0114 policy for standalone build
Fixes sanitizer-x86_64-linux-qemu bot.
Animesh Kumar [Fri, 30 Sep 2022 17:14:42 +0000 (22:44 +0530)]
[OpenMP] Add map clause to the LIT test on use_device_addr clause
As per the OpenMP Spec, "A list item in a use_device_addr clause
must have a corresponding list item in the device data environment"
. Therefore a `map` clause is added which will make sure that the
respective list items are mapped to the device data environment
before the `use_device_addr` clause is specified. The CHECK lines
are also modified based on this change.
Differential Revision: https://reviews.llvm.org/D134974
jinge90 [Wed, 9 Nov 2022 06:24:39 +0000 (14:24 +0800)]
[libunwind] Check corrupted return address in unwind_phase2 when CET is enabled.
If CET shadow stack is enabled, we count the number of stack frames skipped
and adjust CET shadow stack based on the number in libunwind unwind_phase2.
At the same time, we can enhance security via comparing the return address in
normal stack against counterpart in CET shadow stack, if they don't match,
it means the return address stored in normal stack has been corrupted and we
will return _URC_FATAL_PHASE2_ERROR in that case.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D136667
Signed-off-by: jinge90 <ge.jin@intel.com>
Haohai Wen [Wed, 9 Nov 2022 06:08:01 +0000 (14:08 +0800)]
[X86] Add In64BitMode requirement for MMXRI
REX_W prefix is only encodable in 64bit.
Reviewed By: skan
Differential Revision: https://reviews.llvm.org/D137686
chenglin.bi [Tue, 8 Nov 2022 21:07:44 +0000 (05:07 +0800)]
[TypePromotion] Replace Zext to Truncate for the case src bitwidth is larger
Fix: https://github.com/llvm/llvm-project/issues/58843
Reviewed By: samtebbs
Differential Revision: https://reviews.llvm.org/D137613
River Riddle [Tue, 8 Nov 2022 20:46:41 +0000 (12:46 -0800)]
[mlir][LLVM] Add base type and flag support to composite type
This requires defining the DIFlag enum from LLVM.
Differential Revision: https://reviews.llvm.org/D137667
Peiming Liu [Wed, 9 Nov 2022 01:42:26 +0000 (01:42 +0000)]
[mlir][sparse] fix SSA chain issue in dense2sparse conversion.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D137683
Peiming Liu [Wed, 9 Nov 2022 00:43:27 +0000 (00:43 +0000)]
[mlir][sparse] extend foreach operation to iterator over sparse constant.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D137679
Jakub Kuderski [Wed, 9 Nov 2022 01:34:31 +0000 (20:34 -0500)]
[mlir][arith] Add `select` support to WIE
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D137589
Fangrui Song [Wed, 9 Nov 2022 01:28:04 +0000 (17:28 -0800)]
[lld] Fix duplicate word typos. NFC
Based on lld/ part of D137338 but reflowed comments.
Peixin-Qiao [Wed, 9 Nov 2022 01:24:38 +0000 (09:24 +0800)]
[flang] Fix function result rewrite for CPTR type
Not all derived type can be taken as abstract result. The CPTR type
should be treated as return by value so to interoperable with C
functions. Fix the function result rewrite for CPTR type, but it
should be generalized for all derived types. The ABI of
interoperability with C for derived type is architecture dependent,
which should be supported later.
Reviewed By: PeteSteinfeld, jeanPerier
Differential Revision: https://reviews.llvm.org/D137548
wren romano [Wed, 9 Nov 2022 00:43:44 +0000 (16:43 -0800)]
[mlir][sparse] Factoring out NewCallParams class in SparseTensorConversion.cpp
The new class helps encapsulate the arguments to `_mlir_ciface_newSparseTensor` so that client code doesn't depend on the details of the API. (This makes way for the next differential which significantly alters the API.)
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D137680
Jim Ingham [Wed, 9 Nov 2022 00:57:57 +0000 (16:57 -0800)]
Move the second instance of TestUniqueTypes.py to a unique file
name. lldb-dotest.py errors out if two tests have the same filename.
Mehdi Amini [Wed, 9 Nov 2022 00:42:41 +0000 (00:42 +0000)]
Do not build with Werror by default (Bazel build)
This seems like an anti-pattern to have -Werror on by default:
it is hostile to user as we can't ensure that all of the supported
platforms are warning-free, and any newer compiler could break the build
for a user who does not have a clear actionable way around it.
Reviewed By: GMNGeoffrey, kuhar
Differential Revision: https://reviews.llvm.org/D123481
Mehdi Amini [Thu, 3 Nov 2022 20:29:28 +0000 (20:29 +0000)]
Apply clang-tidy fixes for readability-redundant-smartptr-get in SCF.cpp (NFC)
Mehdi Amini [Thu, 3 Nov 2022 20:28:33 +0000 (20:28 +0000)]
Apply clang-tidy fixes for llvm-qualified-auto in SCF.cpp (NFC)
Rob Suderman [Tue, 8 Nov 2022 00:18:39 +0000 (16:18 -0800)]
[mlir][linalg] Fix vectorization of linalg depthwise conv for int types
Vectorization of Linalg's depthwise convolution only supports floating
point types. Previous version assumed floating point operations would
work. This version checks whether the computation is integer or floating
point and adjust the inner loop computation.
Reviewed By: hanchung
Differential Revision: https://reviews.llvm.org/D137595
Aart Bik [Tue, 8 Nov 2022 22:58:45 +0000 (14:58 -0800)]
[mlir][sparse] add sparse matmul codegen check tests
This found an insertion point bug in access pattern expansion
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D137675
Akira Hatanaka [Tue, 8 Nov 2022 22:26:51 +0000 (14:26 -0800)]
[ObjC][ARC] Fix non-deterministic behavior in ProvenanceAnalysis
ProvenanceAnalysis::relatedCheck was giving different answers depending
on the order in which the pointers were passed.
Specifically, it was returning different values when A and B were both
loads and were both referring to identifiable objects, but only one was
used by a store instruction.
Aart Bik [Tue, 8 Nov 2022 21:10:54 +0000 (13:10 -0800)]
[mlir][sparse] 3-dimensional sparse tensor insertion test
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D137668
Sam McCall [Tue, 8 Nov 2022 22:48:15 +0000 (23:48 +0100)]
[clangd] Avoid escaping issues in system-include-extractor.test
Fangrui Song [Tue, 8 Nov 2022 22:39:09 +0000 (14:39 -0800)]
[Driver] Refactor err_drv_unsupported_option_argument call sites to use llvm::opt::Arg::getSpelling
For `-foo=bar`, getSpelling return `-foo=` which is exactly what we need from
the diagnostic. Drop `-` from the err_drv_unsupported_option_argument template.
This change makes `--` long option diagnostics more convenient.
Reviewed By: lenary
Differential Revision: https://reviews.llvm.org/D137659
Jez Ng [Tue, 8 Nov 2022 21:33:32 +0000 (16:33 -0500)]
[lld-macho] Fix bugs around EH_Frame symbols
While extending the map file to cover unwind info, I realized we had two
issues with our EH_Frame symbols:
1. Their size was not set
2. We would create two EH_Frame symbols per frame when we only needed
one. This was because the Defined constructor would add the symbol
itself to InputSection::symbols, but we were also manually appending
the symbol to that same vector.
Note that ld64 prints "CIE" and "FDE for: <function>" instead of just
"EH_Frame", but I'm punting on that for now unless we discover that
users really depend upon it.
Reviewed By: #lld-macho, Roger
Differential Revision: https://reviews.llvm.org/D137370
Jez Ng [Tue, 8 Nov 2022 21:33:28 +0000 (16:33 -0500)]
[lld-macho] Emit map file entry for compact unwind info
Just like ld64 does.
Reviewed By: #lld-macho, Roger
Differential Revision: https://reviews.llvm.org/D137369
Jez Ng [Tue, 8 Nov 2022 21:33:22 +0000 (16:33 -0500)]
[lld-macho] Overhaul map file code
The previous map file code left out was modeled after LLD-ELF's
implementation. However, ld64's map file differs quite a bit from
LLD-ELF's. I've revamped our map file implementation so it is better
able to emit ld64-style map files.
Notable differences:
* ld64 doesn't demangle symbols in map files, regardless of whether
`-demangle` is passed. So we don't have to bother with
`getSymbolStrings()`.
* ld64 doesn't emit symbols in cstring sections; it emits just the
literal values. Moreover, it emits these literal values regardless of
whether they are labeled with a symbol.
* ld64 emits map file entries for things that are not strictly symbols,
such as unwind info, GOT entries, etc. That isn't handled in this
diff, but this redesign makes them easy to implement.
Additionally, the previous implementation sorted the symbols so as to
emit them in address order. This was slow and unnecessary -- the symbols
can already be traversed in address order by walking the list of
OutputSections. This brings significant speedups. Here's the numbers
from the chromium_framework_less_dwarf benchmark on my Mac Pro, with the
`-map` argument added to the response file:
base diff difference (95% CI)
sys_time 2.922 ± 0.059 2.950 ± 0.085 [ -0.7% .. +2.5%]
user_time 11.464 ± 0.191 8.290 ± 0.123 [ -28.7% .. -26.7%]
wall_time 11.235 ± 0.175 9.184 ± 0.169 [ -19.3% .. -17.2%]
samples 16 23
(It's worth noting that map files are written in parallel with the
output binary, but they often took longer to write than the binary
itself.)
Finally, I did further cleanups to the map-file.s test -- there was no
real need to have a custom-named section. There were also alt_entry
symbol declarations that had no corresponding definition. Either way,
neither custom-named sections nor alt_entry symbols trigger special code
paths in our map file implementation.
Reviewed By: #lld-macho, Roger
Differential Revision: https://reviews.llvm.org/D137368
Aart Bik [Fri, 4 Nov 2022 19:23:03 +0000 (12:23 -0700)]
[mlir][sparse] first general insertion implementation with pure codegen
This revision generalizes lowering the sparse_tensor.insert op into actual code that directly operates on the memrefs of a sparse storage scheme. The current insertion strategy does *not* rely on a cursor anymore, with introduces some testing overhead for each insertion (but still proportional to the rank, as before). Over time, we can optimize the code generation, but this version enables us to finish the effort to migrate from library to actual codegen.
Things to do:
(1) carefully deal with (un)ordered and (not)unique
(2) omit overhead when not needed
(3) optimize and specialize
(4) try to avoid the pointer "cleanup" (at HasInserts), and make sure the storage scheme is consistent at every insertion point (so that it can "escape" without concerns).
Reviewed By: Peiming
Differential Revision: https://reviews.llvm.org/D137457
River Riddle [Tue, 8 Nov 2022 03:02:18 +0000 (19:02 -0800)]
[mlir][LLVMDebugTranslation] Move the check for FusedLoc scope
Shift it to where we already check for FusedLoc, which removes the
need to cast multiple times.
River Riddle [Tue, 8 Nov 2022 03:01:14 +0000 (19:01 -0800)]
[mlir][LLVMDebugTranslation] Add inlinedAt to the key when caching locations
This avoids accidentally using the wrong inlined at scope when translating
locations.
rkayaith [Tue, 8 Nov 2022 20:42:50 +0000 (15:42 -0500)]
[mlir][NVGPU] Fix -Wunsequenced warning
llvm-project/mlir/lib/Conversion/NVGPUToNVVM/NVGPUToNVVM.cpp:441:25:
warning: multiple unsequenced modifications to 'asmArgIdx'
[-Wunsequenced]
ss << "$" << asmArgIdx++ << ",$" << asmArgIdx++ << ";";
^ ~~
Björn Schäpers [Fri, 4 Nov 2022 11:38:06 +0000 (12:38 +0100)]
[clang-format][NFCish] Alphabetical sort Format.(h|pp)
I've:
- Sorted the members of FormatStyle alphabetical. The enums and structs
are kept close to the member.
- Sorted the yaml io functions, based on the type they operate on.
- Sorted the initializers in getLLVMStyle(), except that penalities are
kept at the end.
- Sorted the io of FormatStyle, this changes the --dump-config behavior.
- Moved the deprecated options into the only input case, this also
changes --dump-config, it does not put the not directly used options
in the .clang-format anymore.
- Sorted the comparisons in operator==.
- Added WhiteSpaceMacros in operator==, I've not actively looked if all
other members are compared.
- This showed flawed tests (or in my opinion a flawed io operation, but
that is another discussion and change).
Differential Revision: https://reviews.llvm.org/D137409
Trass3r [Tue, 8 Nov 2022 20:44:46 +0000 (20:44 +0000)]
[Visualizers] Add natvis visualizers for various internal llvm classes
Also improve DisplayStrings for array and string types.
Differential Revision: https://reviews.llvm.org/D135685
rkayaith [Thu, 3 Nov 2022 18:29:44 +0000 (14:29 -0400)]
[mlir-opt] Add '-p' as an alias for '-pass-pipeline'
The pipeline strings have been getting more verbose over time, adding an
alias for the option should help improve the ergonomics a bit.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D137347
rkayaith [Wed, 12 Oct 2022 21:32:32 +0000 (17:32 -0400)]
[mlir][spirv] Change translation to use spirv.module
Update the SPIRV `mlir-translate` translations to translate to/from
`spirv.module` instead of `builtin.module`. This simplifies the
translation since the code no longer needs to walk the module looking
for a SPIRV module, however it requires passing `-no-implicit-module` to
all the tests.
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D135819
David Green [Tue, 8 Nov 2022 19:30:26 +0000 (19:30 +0000)]
[AArch64] Allow users-facing feature names in clang target attributes
D133848 added support for the GCC format of target("..") attributes. The
supported formats to match gcc are:
// "arch=<arch>" - parsed to features as per -march=..
// "cpu=<cpu>" - parsed to features as per -mcpu=.., with CPU set to <cpu>
// "tune=<cpu>" - TuneCPU set to <cpu>
// "+feature", "+nofeature" - Add (or remove) feature.
We also support the existing formats, previously accepted by clang, for
compatibility with the existing code and intrinsics code:
// "feature", "no-feature" - Add (or remove) feature.
The clang formats would accept and use internal feature names
("fullfp16"/"neon"/"sve") as opposed to the user facing names
("fp16"/"simd"/"sve"). Usually they use the same names, but can be
different for cases like fp, fullfp16 and mte (among others).
This patch makes the clang format also except the user facing names, by
parsing the features through getArchExtFeature. There is a fallback if
the name is not recognized (like "fullfp16"), where we add the existing
string which should then be checked later for consistency. This allows
the internal names to be used as before, so long as they are recognized
as internal names. (Note that we currently don't have an implementation
of isValidFeatureName. The backend will currently give an error like
"'-sid' is not a recognized feature for this target (ignoring feature)."
This should be improved in a later patch once an implementation of
isValidFeatureName in clang is present).
Differential Revision: https://reviews.llvm.org/D137617
Adhemerval Zanella [Wed, 10 Aug 2022 14:47:07 +0000 (11:47 -0300)]
[clang][Headers] Do not define varargs macros for __need___va_list
The glibc uses the define to avoid namespace polution on headers
that requires variadic argument, where the inclusion of stdarg.h is
required to obtain the va_list definition.
For such cases only __gnuc_va_list is required.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D137268