v1nh1shungry [Sun, 6 Nov 2022 17:30:41 +0000 (18:30 +0100)]
[clangd] Fix the code action `RemoveUsingNamespace`
Avoid adding qualifiers before C++ operators declared in a non-class context
Reviewed By: tom-anders
Differential Revision: https://reviews.llvm.org/D137494
Alexey Bader [Sun, 6 Nov 2022 16:16:36 +0000 (08:16 -0800)]
Online sync-ups: add SYCL working group sync details.
luxufan [Sun, 6 Nov 2022 02:44:50 +0000 (10:44 +0800)]
[IndVars] Forget the SCEV when the instruction has been sunk.
In the past, the SCEV expression of the sunk instruction was not
forgetted. This led to the incorrect block dispositions after the
instruction be sunk.
Fixes https://github.com/llvm/llvm-project/issues/58662
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D137060
Simon Pilgrim [Sun, 6 Nov 2022 14:32:00 +0000 (14:32 +0000)]
[X86] Schedule scalar movsx/movzx load+extend ops as WriteLoad instead of WriteALULd
Although some very old x86 hardware would perform the extension as a later stage, every target we have a scheduler for always performs this as part of the load-op (avoid ALU pipes etc.). If anyone wants to model very old hardware they can always override this.
This patch just tags these as WriteLoad directly and removes unnecessary overrides - this cleans up some latency/throughput tests as they aren't being badly modelled as folded ALU ops
Sanjay Patel [Sun, 6 Nov 2022 14:00:04 +0000 (09:00 -0500)]
[SimplifyLibCalls] improve code readability for AttributeList propagation; NFC
It is possible that we can do better on some of these transforms
by passing some subset of attributes, but we were not doing that
in any of the changed code. So it's better to give that a name
to indicate we're clearing attributes or make that more obvious
by using the default-constructed empty list.
Sanjay Patel [Sun, 6 Nov 2022 13:26:39 +0000 (08:26 -0500)]
[InstCombine] reduce multi-use casts+masks
As noted in the code comment, we could generalize this:
https://alive2.llvm.org/ce/z/N5m-eZ
It saves an instruction even without a constant operand,
but the 'and' is wider. We can do that as another step
if it doesn't harm anything.
I noticed that this missing pattern with a constant operand
inhibited other transforms in a recent bug report, so this
is enough to solve that case.
Sanjay Patel [Fri, 4 Nov 2022 15:17:44 +0000 (11:17 -0400)]
[InstCombine] add tests for zext of and of trunc; NFC
The basic one-use version of this sequence is reduced,
but we don't transform these currently.
Simon Pilgrim [Sun, 6 Nov 2022 14:03:59 +0000 (14:03 +0000)]
[X86] Fix scalar load latencies for WriteLoad scheduler class
Znver1/Znver2 were using vector load latency values (which is what WriteFLoad*/WriteVecLoad* are for) instead of the scalar load latency value
TBH I'm not sure clflush/clzero/prefetch ops should be tagged as WriteLoad but at least this makes us more consistent
Simon Pilgrim [Sun, 6 Nov 2022 13:35:46 +0000 (13:35 +0000)]
[X86] Regenerate atom-pad-short-functions.ll test checks
Simon Pilgrim [Sun, 6 Nov 2022 13:16:23 +0000 (13:16 +0000)]
Revert rG244331ae833aaf33503bbd36890e704afb66a237 "[X86] Fix scalar load latencies for WriteLoad scheduler class"
Forgot to update tests outside the llvm-mca test folder :-(
Simon Pilgrim [Sun, 6 Nov 2022 12:22:05 +0000 (12:22 +0000)]
[X86] Fix scalar load latencies for WriteLoad scheduler class
Atom was missing a load latency value (so was defaulting to 1cy)
Znver1/Znver2 were using vector load latency values (which is what WriteFLoad*/WriteVecLoad* are for) instead of the scalar load latency value
TBH I'm not sure clflush/clzero/prefetch ops should be tagged as WriteLoad but at least this makes us more consistent
David Green [Sun, 6 Nov 2022 12:06:12 +0000 (12:06 +0000)]
[SLP][AArch64] Add a test case for SLP phi ordering of scalable vectors. NFC
David Green [Sun, 6 Nov 2022 11:40:08 +0000 (11:40 +0000)]
Revert "[SLP] Extend reordering data of tree entry to support PHI nodes"
This reverts commit
87a20868eb2043420d48f591c3437472f7137834 as it has
problems with scalable vectors and use-list orders. Test to follow.
Valentin Clement [Sun, 6 Nov 2022 08:56:19 +0000 (09:56 +0100)]
[flang] Fix controlSuccessor chain for select rank construct
Represent the select rank statement + select rank case statement
the same way the select case statement and case statement are represented.
controlSuccessor was not correctly attributed to the next type guard stmt.
Similar to D137460 for select type construct.
Reviewed By: vdonaldson
Differential Revision: https://reviews.llvm.org/D137490
Kazu Hirata [Sun, 6 Nov 2022 08:10:02 +0000 (00:10 -0800)]
[llvm] Remove redundaunt typename (NFC)
Michał Górny [Sun, 6 Nov 2022 08:06:16 +0000 (09:06 +0100)]
Revert "[llvm] [cmake] Set EXCLUDE_FROM_ALL on gtest and TestingSupport"
This reverts commit
b6be37b0be5e1b524456c375700c76b7dae5171c. Still
breaks flang.
Michał Górny [Sun, 30 Oct 2022 06:37:15 +0000 (07:37 +0100)]
[llvm] [cmake] Set EXCLUDE_FROM_ALL on gtest and TestingSupport
Exclude building googletest and LLVMTestingSupport libraries from
the `all` target. If unittests are being built, these libraries will
be built as a dependency anyway. If they are not being built, building
them makes little sense as they are not installed or used otherwise.
This will also make standalone builds of other projects easier, as it
makes it possible to include these directories without having to cover
them with additional conditions to prevent them from being built
unconditionally.
Differential Revision: https://reviews.llvm.org/D137035
Kazu Hirata [Sun, 6 Nov 2022 07:52:42 +0000 (23:52 -0800)]
[mlir] Fix a warning
This patch fixes:
llvm-project/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp:1855:17:
error: comparison of integers of different signs: 'int64_t' (aka
'long') and 'size_t' (aka 'unsigned long') [-Werror,-Wsign-compare]
Arnab Dutta [Sun, 6 Nov 2022 06:59:10 +0000 (12:29 +0530)]
[MLIR] Fix bug in the method constructing semi affine expression from flattened form
Set proper offset to the second element of the index pair when either
lhs or rhs of a local expression is a dimensional identifier, so that
we do not have same index values for more than one local expression.
Reviewed By: springerm, hanchung
Differential Revision: https://reviews.llvm.org/D137389
Argyrios Kyrtzidis [Sun, 6 Nov 2022 03:58:41 +0000 (20:58 -0700)]
[llvm/cmake/config-ix.cmake] If `LD64_EXECUTABLE` is already set, avoid the need to look up `ld64`
This provides option to set the `LD64_EXECUTABLE` variable to a path at CMake configure time directly.
River Riddle [Sat, 5 Nov 2022 23:36:17 +0000 (16:36 -0700)]
[mlir][SubElements] Re-add null guards to better enable downstream adoption
We used to allow this, and it can break clients that still rely on it.
River Riddle [Sat, 5 Nov 2022 23:35:25 +0000 (16:35 -0700)]
[mlir][SubElementInterfaces] Prefer calling the derived get if possible
This allows for better supporting attributes/types that override the
default builders.
Jeff Niu [Fri, 4 Nov 2022 01:40:28 +0000 (18:40 -0700)]
[mlir] Drop `const` from `SymbolTableCollection::lookupSymbolIn`
This function didn't work because it was marked const but both functions
it could dispatch to are not.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D137380
Paul Robinson [Sat, 5 Nov 2022 21:36:40 +0000 (14:36 -0700)]
Change a 'default_triple' to 'object-emission'
as pointed out in post-commit review on D136612 by Igor Kudrin.
Sylvestre Ledru [Sat, 5 Nov 2022 21:35:54 +0000 (22:35 +0100)]
Add support of the next Ubuntu (Ubuntu 23.04 - Lunar Lobster)
LLVM GN Syncbot [Sat, 5 Nov 2022 20:26:22 +0000 (20:26 +0000)]
[gn build] Port
59ef4b3686e4
Nikolas Klauser [Thu, 22 Sep 2022 19:53:13 +0000 (21:53 +0200)]
[libc++] Split __allocator_destructor out of shared_ptr.h
Reviewed By: ldionne, huixie90, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D134479
Nikolas Klauser [Wed, 2 Nov 2022 19:27:42 +0000 (20:27 +0100)]
[libc++] Granularize <concept> includes
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D137283
Simon Pilgrim [Sat, 5 Nov 2022 19:06:55 +0000 (19:06 +0000)]
[X86] Replace unnecessary int2float and float2double overrides with better base class defs
Broadwell/Haswell were completely overriding the class defs - we can remove those overrides entirely by just choosing better class defs (plus a fix for missing mmx folded load).
Michał Górny [Mon, 31 Oct 2022 14:28:32 +0000 (15:28 +0100)]
[flang] Add one missing "llvm_" prefix to gtest_main
The `target_link_libraries()` call operates on `gtest_main` rather than
`llvm_gtest_main`. I think it was missed while prefixing all the tools
in
38151a08c21e0cdacd52af03e4bdff258f0bac6f.
Differential Revision: https://reviews.llvm.org/D137080
Simon Pilgrim [Sat, 5 Nov 2022 17:28:23 +0000 (17:28 +0000)]
[MCA][X86] Add more avx512 cvt instructions test coverage
Amaury Séchet [Sat, 5 Nov 2022 16:44:42 +0000 (16:44 +0000)]
[NFC] Refactor DAGCombiner::foldSelectOfConstants to reduce nesting 2.0
Florian Hahn [Sat, 5 Nov 2022 16:55:09 +0000 (16:55 +0000)]
[SimpleLoopUnswitch] Forget block & loop dispos during trivial unswitch.
Unswitching adjusts the CFG in ways that may invalidate cached loop
dispositions. Clear all cached block and loop dispositions during
trivial unswitching. The same is already done for non-trivial
unswitching.
Fixes #58751.
Amaury Séchet [Sat, 5 Nov 2022 16:11:13 +0000 (16:11 +0000)]
[NFC] Refactor DAGCombiner::foldSelectOfConstants to reduce nesting
Simon Pilgrim [Sat, 5 Nov 2022 15:45:45 +0000 (15:45 +0000)]
[X86] Cleanup WriteCvtSD2SS/WriteCvtPD2PS overrides
The WriteCvtSD2SS/WriteCvtPD2PS* classes were mostly unused as the models were needlessly overriding all instructions - in some cases the folded pattern overrides were entirely missing (but I've confirmed they just have an additional Port23 use)
There were a couple of typos (confirmed with Agner/uops.info) - Skylake/Icelake uses Port5+Port01 for XMM/YMM, Skylake uses Port5+Port05 for ZMM but Icelake uses Port5+Port0
Nikolas Klauser [Tue, 1 Nov 2022 19:06:11 +0000 (20:06 +0100)]
[libc++] Add [[nodiscard]] extensions to ranges algorithms
This mirrors what we have done in the classic algorithms
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D137186
Simon Pilgrim [Sat, 5 Nov 2022 14:35:41 +0000 (14:35 +0000)]
[X86] Fix cvtss2si64/cvttss2si64 typo in SkylakeClient
SS2SI64 conversions use Port0/Port01/Port5 (with/without truncation), but SS2SI32 only uses Port0/Port01 like SD2SI32/SD2SI64
Valentin Clement [Sat, 5 Nov 2022 14:10:40 +0000 (15:10 +0100)]
[flang] Fix controlSuccessor chain for select type construct
Represent the select type statement + type guard statement
the same way the select case statement and case statement are represented.
controlSuccessor was not correctly attributed to the next type guard stmt.
Reviewed By: PeteSteinfeld, vdonaldson
Differential Revision: https://reviews.llvm.org/D137460
Simon Pilgrim [Sat, 5 Nov 2022 13:57:07 +0000 (13:57 +0000)]
[X86] Fix override for CVTPD2PS/CVTPD2DQ/CVTTPD2DQ AVX variants
These were lost when they were converted from instregex to instrs
Simon Pilgrim [Sat, 5 Nov 2022 12:41:45 +0000 (12:41 +0000)]
[X86] Add SchedWriteVecTruncate scheduler per-width wrapper
Replaces hard coded uses of WriteVPMOV256 for all the vector truncations instructions
We still need to work out how to fix folded stores (see Issue #36236)
Simon Pilgrim [Sat, 5 Nov 2022 11:31:53 +0000 (11:31 +0000)]
[X86] Add missing ReadAfterFold attributes from CVTSD2SSrm/CVTSS2SDrm to match the AVX + _Int equivalents
This was never added in D52886, probably because the defs were already missing the old ReadAfterLd attribute, but they do exhibit the same behaviour of reading the dst reg after the load.
Simon Pilgrim [Sat, 5 Nov 2022 11:19:36 +0000 (11:19 +0000)]
[X86] Remove SandyBridge CVTSS2SIrm/CVTSD2SIrm overrides
Just use the default WriteCvtSS2I/WriteCvtSD2I folded variants (already used by the VCVTSS2SI/VCVTSD2SI AVX variants).
Confirmed with agner and uops.info - there should be no difference between the SSE/AVX1 variants in folded load latency
Simon Pilgrim [Sat, 5 Nov 2022 11:27:46 +0000 (11:27 +0000)]
[X86] Replace unnecessary int2double overrides with a better WriteCvtI2PD def
Broadwell, Haswell and SkylakeClient were completely overriding the WriteCvtI2PD defs - we can remove those overrides entirely by replacing the unused WriteCvtI2PD values
There's plenty more of these in the scheduler models - I'm looking at improving warnings in llvm-tblgen to catch them all
Ivan Butygin [Fri, 4 Nov 2022 22:26:02 +0000 (23:26 +0100)]
[mlir] Add `populateFunctionOpInterfaceTypeConversionPattern` version which operates on any `FunctionOpInterface`
Exisitng version is always limited to some specific op.
Differential Revision: https://reviews.llvm.org/D137469
Siva Chandra Reddy [Sat, 5 Nov 2022 07:58:16 +0000 (00:58 -0700)]
[libc][docs] Add a build and test document.
Siva Chandra Reddy [Sat, 5 Nov 2022 07:21:09 +0000 (00:21 -0700)]
[libc][docs] Fix typos.
Siva Chandra Reddy [Fri, 4 Nov 2022 19:53:24 +0000 (12:53 -0700)]
[libc][docs] Move links to internal developer guides out of the main side bar.
The links are all now from a "Developer Guides" page which is linked from the
main side bar.
Uday Bondhugula [Sat, 5 Nov 2022 05:38:21 +0000 (11:08 +0530)]
MLIR. NFC. Remove trailing whitespaces in MemRefOps.td
Remove trailing whitespaces in MemRefOps.td. NFC.
Med Ismail Bennani [Sat, 5 Nov 2022 04:58:51 +0000 (21:58 -0700)]
[lldb/crashlog] Standardize file path key in the ScriptedProcess Dictionary
This patch replaces the backing file path key to "file_path" to keep it
consistent.
rdar://
101652618
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Sat, 5 Nov 2022 04:57:31 +0000 (21:57 -0700)]
[lldb/test] Fix StructuredDataTest::GetDescriptionEmpty warning (NFC)
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Med Ismail Bennani [Fri, 4 Nov 2022 21:49:52 +0000 (14:49 -0700)]
[lldb/crashlog] Fix frame parser regex for when there is no source info
It can happen that the originator of a crash report doesn't have access
to certain images. When that's the case, ReportCrash won't show the
source info in the crash report stack frames, but only the stack address
and image name.
This patch fixes a bug in the crashlog stackframe parser regular
expression to optionally match the source info group.
rdar://
101934135
Differential Revision: https://reviews.llvm.org/D137466
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
chenglin.bi [Sat, 5 Nov 2022 04:58:14 +0000 (12:58 +0800)]
[InstCombine] fold `sub + and` pattern with specific const value
`C1 - ((C3 - X) & C2) --> (X & C2) + (C1 - (C2 & C3))`
when:
(C3 - ((C2 & C3) - 1)) is pow2 &&
((C2 + C3) & ((C2 & C3) - 1)) == ((C2 & C3) - 1) &&
C2 is negative pow2 || (C3 - X) is nuw
https://alive2.llvm.org/ce/z/HXQJV-
Fix: #58523
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D136582
Tue Ly [Fri, 4 Nov 2022 18:22:35 +0000 (14:22 -0400)]
[libc] Add add_with_carry to builtin wrapper.
Add add_with_carry to builtin wrapper to be used by UInt class.
Reviewed By: orex
Differential Revision: https://reviews.llvm.org/D137453
River Riddle [Sat, 5 Nov 2022 01:53:50 +0000 (18:53 -0700)]
[mlir] Attempt to fix windows build after D137374
River Riddle [Fri, 4 Nov 2022 02:12:53 +0000 (19:12 -0700)]
[mlir:LLVM] Fix translation of DISubRangeAttr to LLVM
We were missing DISubRangeAttr in the type switch. This
also updates a few of the range parameters to be optional.
River Riddle [Thu, 3 Nov 2022 23:06:59 +0000 (16:06 -0700)]
[mlir] Infer SubElementInterface implementations using the storage KeyTy
The KeyTy of attribute/type storage classes provide enough information for
automatically implementing the necessary sub element interface methods. This
removes the need for derived classes to do it themselves, which is both much
nicer and easier to handle certain invariants (e.g. null handling). In cases where
explicitly handling for parameter types is necessary, they can provide an implementation
of `AttrTypeSubElementHandler` to opt-in to support.
This tickles a few things alias wise, which annoyingly messes with tests that hard
code specific affine map numbers.
Differential Revision: https://reviews.llvm.org/D137374
Quentin Colombet [Thu, 3 Nov 2022 19:14:52 +0000 (19:14 +0000)]
[mlir][MemRef] Fix the lowering of extract_strided_metadata
The first result of the extract_strided_metadata operation is a MemRef,
not a naked pointer.
This patch fixes the lowering of this operation in MemRefToLLVM so that
we properly materialize the full MemRef structure and not just the base,
naked, pointer.
Differential Revision: https://reviews.llvm.org/D137364
Craig Topper [Sat, 5 Nov 2022 00:49:08 +0000 (17:49 -0700)]
[RISCV] Add isel pattern for (and X, 0xffffffff << C) with Zba.
We can use SRLI by C to clear LSBs followed by a SLLI_UW.
Craig Topper [Sat, 5 Nov 2022 00:26:16 +0000 (17:26 -0700)]
[RISCV] Remove duplicate SDNodeXForm. NFC
Jennifer Yu [Sat, 5 Nov 2022 00:15:14 +0000 (17:15 -0700)]
Back out test that failed.
But I can not reproduce the problem on my local machine. My local machine run:
222 0x5a6780
222 0x7fffbef9400e
222 0x5a677e 0x5a6780 0x7fffbef936c8
222 0x376f8e 0x376f90 0x7fffbef94008
222 0x281f20
222 0x7fffbef9400e
PASSED
Matthew Voss [Sat, 5 Nov 2022 00:08:54 +0000 (17:08 -0700)]
[llvm-profdata] Check for all duplicate entries in MemOpSize table
Previously, we only checked for duplicate zero entries when merging a
MemOPSize table (see D92074), but a user recently provided a reproducer
demonstrating that other entries can also be duplicated. As demonstrated
by the test in this patch, PGOMemOPSizeOpt can potentially generate
invalid IR for non-zero, non-consecutive duplicate entries. This seems
to be a rare case, since the duplicate entry is often below the
threshold, but possible. This patch extends the existing warning to
check for any duplicate values in the table, both in the optimization
and in llvm-profdata.
Differential Revision: https://reviews.llvm.org/D136211
Craig Topper [Fri, 4 Nov 2022 19:59:55 +0000 (12:59 -0700)]
[RISCV] Rename isAllUsesReadW to hasAllWUsers in RISCVSExtWRemoval. NFC
Peiming Liu [Fri, 4 Nov 2022 19:35:29 +0000 (19:35 +0000)]
[mlir][sparse] fix sparse tensor rewriting patterns that do not propagate sparse tensor SSA properly.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D137468
David Blaikie [Fri, 4 Nov 2022 23:35:05 +0000 (23:35 +0000)]
Peiming Liu [Fri, 4 Nov 2022 17:52:21 +0000 (17:52 +0000)]
[mlir][sparse] extend foreach operation to accept reduction arguments.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D137463
Matt Arsenault [Thu, 3 Nov 2022 21:05:37 +0000 (14:05 -0700)]
R600: Remove broken atomicrmw patterns
inc/dec are not add/sub of 1.
Mircea Trofin [Fri, 4 Nov 2022 23:05:10 +0000 (16:05 -0700)]
[MLGO][NFC] Use std::map instead of DenseMap to avoid use after free
In `MLInlineAdvisor::getAdviceImpl`, we call `getCachedFPI` twice, once
for the caller, once for the callee, so the second may invalidate the
reference obtained by the first because the underlying implementation of
the cache is a `DenseMap`. `std::map` doesn't have that problem.
Ryan Prichard [Fri, 4 Nov 2022 22:51:44 +0000 (15:51 -0700)]
[libc++abi] Use std::nullptr_t instead of declaring it manually
Sometimes libc++'s stddef.h wrapper gets included, which defines
::nullptr_t. This test is compiled with -Wshadow -Werror, so shadowing
::nullptr_t with a nullptr_t in main is an error. Include cstddef,
which is guaranteed to define std::nullptr_t in C++11 and forward.
Reviewed By: ldionne, #libc_abi
Differential Revision: https://reviews.llvm.org/D137127
Matt Arsenault [Mon, 17 Oct 2022 04:35:07 +0000 (21:35 -0700)]
AMDGPU: Fold undef rcp to qnan
This matches the behavior in instcombine, and for fdiv.
Matt Arsenault [Fri, 4 Nov 2022 22:49:15 +0000 (15:49 -0700)]
AMDGPU: Fix test failure
Peiming Liu [Tue, 18 Oct 2022 16:41:03 +0000 (16:41 +0000)]
[mlir][sparse] support Parallel for/reduction.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D135927
George Hu [Fri, 4 Nov 2022 22:37:18 +0000 (15:37 -0700)]
Revert "[lldb-vscode] Send Statistics Dump in terminated event"
This reverts commit
e3ccbae309273900a42e30b606c15c873d57f1ea.
There is a bug which is failing the test running on mac.
Jennifer Yu [Wed, 2 Nov 2022 22:43:26 +0000 (15:43 -0700)]
Remove redundant loads.
It is caused by regenerate captured var value when processing the
has_device_addr, the captured var value has been generated in
GenerateOpenMPCapturedVars and passed as Arg in generateInfoForCapture.
The fix just use Arg instead regenerated just same as is_device_ptr
Matt Arsenault [Tue, 25 Oct 2022 05:41:41 +0000 (22:41 -0700)]
AMDGPU: Fix assert when trying to overextend liverange
This was trying to add segments beyond the new and use,
so skip additional segments.
This would hit (S < E && "Cannot create empty or backwards segment").
Wanyi Ye [Fri, 4 Nov 2022 19:01:00 +0000 (12:01 -0700)]
Fix test TestVSCode_terminatedEvent.py
This test is broken due to the flaky encoding of top-level JSON key 'memory'
When I run locally (linux) the test passed. However, it failed the build bot: https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/48111/
I will find a way to repro before I can actually fix this issue correctly.
https://reviews.llvm.org/D137455
Jakub Kuderski [Fri, 4 Nov 2022 21:37:18 +0000 (17:37 -0400)]
[mlir][spirv] Fold noop `BitcastsOp`s
This allows for bitcast conversion to roundtrip.
Fixes: https://github.com/llvm/llvm-project/issues/58801
Reviewed By: antiagainst, Hardcode84, mravishankar
Differential Revision: https://reviews.llvm.org/D137459
Xiang Li [Fri, 4 Nov 2022 19:37:07 +0000 (12:37 -0700)]
[CMake] Only set LLVM_DEFAULT_TARGET_TRIPLE to LLVM_HOST_TRIPLE when native target is enabled
This is for case when native target like X86 is not in LLVM_TARGETS_TO_BUILD.
Right now LLVM_DEFAULT_TARGET_TRIPLE is set to LLVM_HOST_TRIPLE even when native target is not enabled,
As a result, many lit tests will fail because default_triple is set for lit test but not enabled when build LLVM.
Reviewed By: smeenai
Differential Revision: https://reviews.llvm.org/D134972
Corentin Jabot [Fri, 4 Nov 2022 21:09:58 +0000 (22:09 +0100)]
Hanhan Wang [Fri, 4 Nov 2022 20:58:59 +0000 (13:58 -0700)]
[mlir][tiling] Relax tiling to accept generating multiple operations.
Some operations need to generate multiple operations when implementing
the tiling interface. Here is a sound example in IREE, see
https://github.com/iree-org/iree/pull/10905 for more details.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D137300
Renaud-K [Wed, 2 Nov 2022 22:39:58 +0000 (15:39 -0700)]
Alias Analysis infra in Flang
Differential revision: https://reviews.llvm.org/D136889
Rafael Auler [Sat, 29 Oct 2022 01:34:48 +0000 (18:34 -0700)]
[BOLT] Always move JTs in jump-table=move
We should always move jump tables when requested. Previously,
we were not moving jump tables of non-simple functions in relocation
mode. That caused a bug detailed in the attached test case: in PIC
jump tables, we force jump tables to be moved, but if they are not
moved because the function is not simple, we could incorrectly update
original entries in .rodata, corrupting it under special circumstances
(see testcase).
Reviewed By: #bolt, maksfb
Differential Revision: https://reviews.llvm.org/D137357
Rafael Auler [Fri, 4 Nov 2022 20:08:31 +0000 (13:08 -0700)]
[BOLT][Hugify] Fix apple builds
Fix placement of ifdefs in hugify.cpp after D129107 landed.
Mike Rice [Thu, 3 Nov 2022 15:50:08 +0000 (08:50 -0700)]
Some uses of the preprocessor can result in multiple target regions on the
same line. Cases such as those in the associated lit tests, can now be
supported.
This adds a 'Count' field to TargetRegionEntryInfo to differentiate
regions with the same source position.
The OffloadEntriesInfoManager routines are updated to maintain a count of
regions seen at a location. The registration of regions proceeds that same as
before, but now the next available count is always determined and used in the
offload entry.
Fixes: https://github.com/llvm/llvm-project/issues/52707
Differential Revision: https://reviews.llvm.org/D134816
Hanhan Wang [Thu, 3 Nov 2022 01:09:46 +0000 (18:09 -0700)]
[mlir] Move the assertion to a valid place.
The defining Op may live in an unlinked block so its parent Op may be
null. Only assert it when the parent Op is not null.
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D137306
Jakub Kuderski [Fri, 4 Nov 2022 19:10:58 +0000 (15:10 -0400)]
[mlir][arith][spirv] Remove duplicate test cases
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D137166
Jakub Kuderski [Fri, 4 Nov 2022 19:10:18 +0000 (15:10 -0400)]
[mlir][spirv] Do not truncate i/f64 -> i/f32 in SPIRVConversion
This truncation can be unexpected and break program behavior.
Dedicated emulation passes should be used instead.
Also rename pass options to "emulate-lt-32-bit-scalar-types".
Fixes: https://github.com/llvm/llvm-project/issues/57917
Reviewed By: antiagainst
Differential Revision: https://reviews.llvm.org/D137115
Florian Hahn [Fri, 4 Nov 2022 18:42:06 +0000 (18:42 +0000)]
[IndVars] Forget SCEV for replaced PHI.
Additional SCEV verification highlighted a case where the cached loop
dispositions where incorrect after simplifying a phi node in IndVars.
Fix it by invalidating the phi before replacing it.
Fixes #58750
Shilei Tian [Fri, 4 Nov 2022 18:10:54 +0000 (14:10 -0400)]
[LLVM][AMDGPU] Specialize 32-bit atomic fadd instruction for generic address space
The 32-bit floating-point atomic add instructions on AMDGPUs does not support a
"flat" or "generic" address space. So, if the address space cannot be determined
statically, the AMDGPU backend will fall back to a CAS loop (which does support
"flat" addressing). Instead, this patch emits runtime address-space checks to
allow native FP atomic add instructions for global and LDS memory (and non-atomic
FP add instructions for private/scratch memory).
In order to do that, this patch introduces a new interface function
`emitExpandAtomicRMW`. It is expected to be called when a common atomic expand
doesn't work for a specific target, such as the case we discussed here.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D129690
Benjamin Kramer [Fri, 4 Nov 2022 18:06:44 +0000 (19:06 +0100)]
[bazel] Port
589764a38264
Benjamin Kramer [Fri, 4 Nov 2022 17:58:43 +0000 (18:58 +0100)]
Fix format specifier warning in EmulateInstructionRISCV more
Yes, the portable macro is still the only way to do this.
Slava Zakharin [Wed, 2 Nov 2022 01:27:04 +0000 (18:27 -0700)]
[mlir][math] Simplify pow(x, 0.75) into sqrt(sqrt(x)) * sqrt(x).
Trivial simplification for CPU2017/503.bwaves resulting in 3.89%
speed-up on icelake.
Differential Revision: https://reviews.llvm.org/D137351
Jonas Devlieghere [Fri, 4 Nov 2022 17:46:21 +0000 (10:46 -0700)]
[lldb] Fix format specifier warning in EmulateInstructionRISCV
Fixes warning: format specifies type 'unsigned long' but the argument
has type 'lldb::addr_t' (aka 'unsigned long long') [-Wformat]
Slava Zakharin [Fri, 4 Nov 2022 16:52:47 +0000 (09:52 -0700)]
[mlir][math] Initial support for fastmath flag attributes for Math dialect.
Added arith::FastMathAttr and ArithFastMathInterface support for Math dialect
floating point operations.
This change-set creates ArithCommon conversion utils that currently
provide classes and methods to aid with arith::FastMathAttr conversion
into LLVM::FastmathFlags. These utils are used in ArithToLLVM and
MathToLLVM convertors, but may eventually be used by other converters
that need to convert fast math attributes.
Since Math dialect operations use arith::FastMathAttr, MathOps.td now
has to include enum and attributes definitions from Arith dialect.
To minimize the amount of TD code included from Arith dialect,
I moved FastMathAttr definition into ArithBase.td.
Differential Revision: https://reviews.llvm.org/D136312
Tom Eccles [Wed, 19 Oct 2022 10:51:31 +0000 (10:51 +0000)]
[flang] add -f[no-]reciprocal-math
Only add the option processing and store the result. No attributes are
added to FIR yet.
Differential Revision: https://reviews.llvm.org/D137330
Tom Eccles [Wed, 19 Oct 2022 10:33:12 +0000 (10:33 +0000)]
[flang] Add -f[no-]associative-math and -mreassociate
Only add the option processing and store the result. No attributes are
added to FIR yet.
Clang only forwards -mreassociate
if (AssociativeMath && !SignedZeros && !TrappingMath)
Flang doesn't have -f[no-]trapping-math, so this part of the condition
has been omitted. !TrappingMath is the default.
Differential Revision: https://reviews.llvm.org/D137329
Tom Eccles [Tue, 18 Oct 2022 17:59:03 +0000 (17:59 +0000)]
[flang] Add -f[no-]signed-zeros
Only add the option processing and store the result. No attributes are
added to FIR yet.
Differential Revision: https://reviews.llvm.org/D137328
Tom Eccles [Tue, 18 Oct 2022 16:14:52 +0000 (16:14 +0000)]
[flang] Add -f[no-]approx-func
Only add the option processing and store the result. No attributes are
added to FIR yet.
Differential Revision: https://reviews.llvm.org/D137326
Tom Eccles [Mon, 17 Oct 2022 17:45:15 +0000 (17:45 +0000)]
[flang] Add -f[no-]honor-nans and -menable-no-nans
Only add the option processing and store the result. No attributes are
added to FIR yet.
Differential Revision: https://reviews.llvm.org/D137325
Zequan Wu [Wed, 2 Nov 2022 21:25:48 +0000 (14:25 -0700)]
[Test] Fix CHECK typo.
Differential Revision: https://reviews.llvm.org/D137287
Philip Reames [Fri, 4 Nov 2022 16:59:40 +0000 (09:59 -0700)]
[RISCV] Define custom-N opcodes
Differential Revision: https://reviews.llvm.org/D137355