chenglin.bi [Mon, 21 Nov 2022 14:10:55 +0000 (22:10 +0800)]
[DAGCombiner] fold or (and x, y), x --> x
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D138398
Nico Weber [Fri, 18 Nov 2022 15:47:34 +0000 (10:47 -0500)]
[llvm,polly,clang] Stop setting config.enable_shared in most places
Clang's lit.cfg.py reads this to add an "enable-shared" feature that
three of clang's lit tests use. Nothing else reads enable_shared, so
remove it from most lit.site.cfg.py.in files.
Differential Revision: https://reviews.llvm.org/D138301
Benjamin Maxwell [Mon, 21 Nov 2022 13:22:03 +0000 (13:22 +0000)]
[AArch64][SVE] Precommit tests for VSCALE/SVE CNTx demanded bits
Louis Dionne [Mon, 8 Aug 2022 18:06:40 +0000 (14:06 -0400)]
[libc++] Remove workarounds for systems that used to require __need_XXX macros
Libc++ tried accomodating systems that need to be able to define various
__need_FOO macros before including C library headers, however it does not
appear to be needed anymore in most cases. Indeed, glibc used to use that
system to conditionally provide definitions, however almost all instances
of these macros have been removed from glibc years ago.
I think the next step would be to also fix Clang's own builtin headers
to stop needing these macros.
Differential Revision: https://reviews.llvm.org/D131425
Aliia Khasanova [Fri, 18 Nov 2022 18:00:10 +0000 (18:00 +0000)]
Merge kDynamicSize and kDynamicSentinel into one constant.
resolve conflicts
Differential Revision: https://reviews.llvm.org/D138282
Shengchen Kan [Mon, 21 Nov 2022 12:59:22 +0000 (20:59 +0800)]
[X86][NFC] Minor improvement in X86InstrInfo::optimizeCompareInstr
Before this patch, the code enumerated `getCondFromBranch`, `getCondFromSETCC` and `getCondFromFromCMov` to get the condition code of a `MachineInstr`, and assigned the result to variable `OldCC` when `MI || IsSwapped || ImmDelta != 0` was satisfiled.
After this patch, the `if-else` structure is eliminated by using `getCondFromMI`. Since `OldCC` is only used when `MI || IsSwapped || ImmDelta != 0` is true, it is initialized with `getCondFromMI` directly outside the scope of `if` now.
Reviewed By: pengfei
Differential Revision: https://reviews.llvm.org/D138349
Max Kazantsev [Mon, 21 Nov 2022 12:18:53 +0000 (19:18 +0700)]
[NFC][IndVars] Add LLVM_DEBUG printout to replaceExitCond
Pavel Labath [Mon, 21 Nov 2022 12:13:34 +0000 (13:13 +0100)]
[lldb] Enable TestTargetVar on non-darwin
Björn Schäpers [Sun, 20 Nov 2022 14:08:28 +0000 (15:08 +0100)]
[clang-format][NFC] Removed unused include
Björn Schäpers [Sat, 19 Nov 2022 16:22:09 +0000 (17:22 +0100)]
[clang-format][NFC] Don't add a load of 0es
Shift is 0 for all non comments, maybe even for comments. Don't add the
0 up to three times.
Differential Revision: https://reviews.llvm.org/D138357
Björn Schäpers [Thu, 17 Nov 2022 20:07:31 +0000 (21:07 +0100)]
[clang-format][NFC] Format lambda in conditional
This sort of amends
cdbe296853b1b3fc6415236f05770360e23f0d39
Differential Revision: https://reviews.llvm.org/D138356
Björn Schäpers [Thu, 17 Nov 2022 20:01:50 +0000 (21:01 +0100)]
[clang-format][NFC] Remove unneeded braces
Done by clang-format itself.
Differential Revision: https://reviews.llvm.org/D138354
Oleg Shyshkov [Mon, 21 Nov 2022 12:15:50 +0000 (13:15 +0100)]
[mlir][linalg] Change linalg.broadcast `dimensions` attribute to represent added dimensions.
Original [RFC](discourse.llvm.org/t/rfc-primitive-ops-add-broadcastop-to-linalg/66313) defined `dimensions` as a map from input to init, but a discussion in reviews.llvm.org/D138291 concluded that it's more natural for `dimensions` to represent added dims. Also this way is more consistent with `linalg.reduce`.
Differential Revision: https://reviews.llvm.org/D138408
Benjamin Kramer [Mon, 21 Nov 2022 10:46:02 +0000 (11:46 +0100)]
Mark test using -debug-only as requiring assertions
Anton Sidorenko [Fri, 18 Nov 2022 09:33:02 +0000 (12:33 +0300)]
[MachineTraceMetrics] Pick the trace successor for an entry block
We generate erroneous trace for a basic block if it does not have at least one
predecessor when MinInstr strategy is used. Currently only this strategy is
implemented, so we always have a wrong trace for any entry block. This results in
wrong instructions heights calculation and also leads to wrong critical path.
The described behavior is demonstrated on a simple test. It shows that early
if-conv pass makes wrong decisions due to incorrectly calculated critical path
lenght.
Differential Revision: https://reviews.llvm.org/D138272
Anton Sidorenko [Fri, 18 Nov 2022 09:31:28 +0000 (12:31 +0300)]
[MachineTraceMetrics] Precommit test for D138272
Guillaume Chatelet [Mon, 21 Nov 2022 09:38:42 +0000 (09:38 +0000)]
[libc][test][NFC] bump mpfr version to 4.1.1
Alex Zinenko [Mon, 21 Nov 2022 09:20:19 +0000 (10:20 +0100)]
[mlir] bazel for D137997
Max Kazantsev [Mon, 21 Nov 2022 09:19:01 +0000 (16:19 +0700)]
[Test] Add some test showing limitations of makeIVComparisonInvariant
The transform doesn't work if argument isn't immediately a Phi.
Jean Perier [Mon, 21 Nov 2022 09:04:28 +0000 (10:04 +0100)]
[flang] Add hlfir.designate codegen for Fortran ArrayRef
Differential Revision: https://reviews.llvm.org/D138198
Valentin Clement [Mon, 21 Nov 2022 09:04:58 +0000 (10:04 +0100)]
[flang][NFC] Update TODOs list
Fangrui Song [Mon, 21 Nov 2022 09:02:04 +0000 (09:02 +0000)]
[ELF] InputFile: change symbols to unique_ptr<Symbol *[]>
We may try a custom allocator in the future.
Valentin Clement [Mon, 21 Nov 2022 09:00:17 +0000 (10:00 +0100)]
[flang] Add fir.select_type conversion to if-then-else ladder
Convert fir.select_type operation to an if-then-else ladder.
The type guards are sorted before the conversion so it follows the
execution of SELECT TYPE construct as mentioned in 11.1.11.2 point 4
of the Fortran standard.
Depends on D138279
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D138280
wanglei [Mon, 21 Nov 2022 07:36:28 +0000 (15:36 +0800)]
[LoongArch] AsmParser support for the li.[wd] pseudo instructions
The `li.[wd]` pseudo instructions are used to load an immediate value
into a GPR. These expand directly during asm parsing. As the result,
only real MC instructions are emitted to the MCStreamer. The actual
expansion to real instructions is similar to the expansion performed by
the GAS.
Note: The `li.w` always treats the imm operand as a 32-bit signed value.
Reviewed By: SixWeining
Differential Revision: https://reviews.llvm.org/D138086
Mel Chen [Mon, 14 Nov 2022 13:54:37 +0000 (05:54 -0800)]
[RISCV] Enable reduction pattern SelectICmp and SelectFCmp.
Reviewed By: david-arm
Differential Revision: https://reviews.llvm.org/D137940
Kazu Hirata [Mon, 21 Nov 2022 08:15:35 +0000 (00:15 -0800)]
[llvm] Use Optional::operator-> (NFC)
Since std::optional does not offer getPointer(), this patch replaces
X.getPointer()->foo with X->foo to make the migration from
llvm::Optional to std::optional easier.
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716
Max Kazantsev [Mon, 21 Nov 2022 08:14:28 +0000 (15:14 +0700)]
[Test] One more test for IndVars, showing imperfect work with context
Andrzej Warzynski [Mon, 21 Nov 2022 07:15:26 +0000 (12:45 +0530)]
[mlir][test] Fix how the number of flops is calculated
Make sure that the number of repetitions is correctly incorporated when
calculating the number of floating point operations.
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D138382
Amy Wang [Sat, 19 Nov 2022 13:27:54 +0000 (14:27 +0100)]
[MLIR][Transform] Transform dialect support for affine
Create the directory and file structures to support transform ops for
the affine dialect in the future. Enable affine.unroll and affine.get_parent_for
transform operations in this first batch of check in.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D137997
wanglei [Mon, 21 Nov 2022 07:13:10 +0000 (15:13 +0800)]
[LoongArch] Support parsing la* pseudo instructions
This patch makes `IAS` compatible with `GAS`. It accepts `la*` pseudo
instructions, and expands `la{,.local,.global}` into different
instructions according to different features.
```
Default:
la = la.global = la.got
la.local = la.pcrel
With feature "+la-global-with-pcrel":
la = la.global = la.pcrel
With feature "+la-global-with-abs":
la = la.global = la.abs
With feature "+la-local-with-abs":
la.local = la.abs
With features "+la-global-with-pcrel,+la-global-with-abs"(disorder):
la = la.global = la.pcrel
```
Note: To keep consistent with `GAS` behavior, the "la" can only have
one register operand.
Reviewed By: SixWeining
Differential Revision: https://reviews.llvm.org/D138021
Max Kazantsev [Mon, 21 Nov 2022 06:14:13 +0000 (13:14 +0700)]
[Test] Add some examples where IndVars fails to turn condition into invariant
Kazu Hirata [Mon, 21 Nov 2022 06:38:43 +0000 (22:38 -0800)]
[Scalar] Teach matchExpandedRem to return std::optional (NFC)
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716
Paul Scoropan [Mon, 21 Nov 2022 05:16:26 +0000 (00:16 -0500)]
[PowerPC] XCOFF exception section support on the integrated assembler path
Continuation of https://reviews.llvm.org/D132146 (direct assembly path support, needs to merge first). Adds support to the integrated assembler path for emitting XCOFF exception sections. Both features need https://reviews.llvm.org/D133030 to merge first
Reviewed By: shchenz, DiggerLin
Differential Revision: https://reviews.llvm.org/D134195
Fangrui Song [Mon, 21 Nov 2022 06:04:18 +0000 (06:04 +0000)]
[ELF] Simplify postScanRelocations with in.got
Christopher Di Bella [Mon, 21 Nov 2022 04:36:14 +0000 (04:36 +0000)]
Revert "Revert "[clang-tblgen][NFC] renames Diagnostic.Text to Diagnostic.Summary""
This reverts commit
196edb9f3f4a036371af0013a3d8b857fe0924d5.
Christopher Di Bella [Mon, 21 Nov 2022 04:34:56 +0000 (04:34 +0000)]
Revert "[clang-tblgen][NFC] renames Diagnostic.Text to Diagnostic.Summary"
This reverts commit
eb3f7880a272b818940e2bd2510ae04b0e32873f.
Fangrui Song [Mon, 21 Nov 2022 04:12:03 +0000 (04:12 +0000)]
[ELF] Add InputSectionBase::{addRelocs,relocs} and GotSection::addConstant to add/access relocations
to prepare for changing `relocations` from a SmallVector to a pointer.
Also change the `isec` parameter in `addAddendOnlyRelocIfNonPreemptible` to `GotSection &`.
Mehdi Amini [Thu, 17 Nov 2022 09:43:07 +0000 (09:43 +0000)]
Apply clang-tidy fixes for readability-identifier-naming in TosaValidation.cpp (NFC)
Mehdi Amini [Thu, 17 Nov 2022 09:42:32 +0000 (09:42 +0000)]
Apply clang-tidy fixes for modernize-use-equals-default in TosaValidation.cpp (NFC)
Christopher Di Bella [Mon, 21 Nov 2022 03:44:11 +0000 (03:44 +0000)]
[clang-tblgen][NFC] renames Diagnostic.Text to Diagnostic.Summary
The [Improving Clang's Diagnostics RFC][1] identifies eight broad fields
for Clang to surface, two of which are text-based. Since the current
diagnostics more closely map to the diagnostic summary (or headline), we
should rename them to ensure that there's no confusion when
Diagnostic.Reason is introduced in the near future.
[1]: https://discourse.llvm.org/t/rfc-improving-clang-s-diagnostics/62584
Reviewed By: aaron.ballman, erichkeane
Differential Revision: https://reviews.llvm.org/D135820
gonglingqin [Mon, 21 Nov 2022 02:51:58 +0000 (10:51 +0800)]
[LoongArch] Make function name in error message consistent with the user input. NFC
Zi Xuan Wu (Zeson) [Mon, 21 Nov 2022 01:33:09 +0000 (09:33 +0800)]
[CSKY] Lower ISD::ConstantPool node to support getting the address of ConstantPool entry
When there is not GRS or MOVIH/ORI instruction, we can not get the address of
ConstantPool entry directly. So we need put the address into ConstantPool to leverage CSKY::LRW instruction.
gonglingqin [Mon, 21 Nov 2022 01:12:18 +0000 (09:12 +0800)]
[LoongArch] Add intrinsics for ibar, break and syscall
Diagnostics for intrinsic input parameters have also been added.
Differential Revision: https://reviews.llvm.org/D138094
Fangrui Song [Mon, 21 Nov 2022 00:52:03 +0000 (00:52 +0000)]
[ELF] Don't ignore --detect-odr-violations
We don't implement the gold specific feature and it's inappropriate to ignore it.
The option is used by some projects under gold configure checks.
Fangrui Song [Mon, 21 Nov 2022 00:37:32 +0000 (00:37 +0000)]
[CMake][OpenMP] Remove -Wl,--warn-shared-textrel
This option is legacy and is removed from GNU ld's doc (many binutils
distributions are configured with --enable-textrel-check=). lld ignores the
option.
Fangrui Song [Mon, 21 Nov 2022 00:02:36 +0000 (00:02 +0000)]
[ELF] Simplify InputSectionBase::getSize
Fangrui Song [Sun, 20 Nov 2022 23:22:32 +0000 (23:22 +0000)]
[ELF] Replace rawData+size with content_+size+compressedSize
compressedSize resides in an existing union. sizeof(InputSection) decreases from
160 to 152 on ELF64 systems.
Fangrui Song [Sun, 20 Nov 2022 22:43:22 +0000 (22:43 +0000)]
[ELF] Change rawData to content() and data() to contentMaybeDecompress()
Clarify data() which may trigger decompression and make it feasible to refactor
the member variable rawData.
Kazu Hirata [Sun, 20 Nov 2022 22:32:17 +0000 (14:32 -0800)]
[clang] Remove unused forward declarations (NFC)
Adrian Vogelsgesang [Fri, 26 Aug 2022 13:00:43 +0000 (06:00 -0700)]
[LLDB] Do not dereference promise pointer in `coroutine_handle` pretty printer
So far, the pretty printer for `std::coroutine_handle` internally
dereferenced the contained frame pointer displayed the `promise`
as a sub-value. As noticed in https://reviews.llvm.org/D132624
by @labath, this can lead to an endless loop in lldb during printing
if the coroutine frame pointers form a cycle.
This commit breaks the cycle by exposing the `promise` as a pointer
type instead of a value type. The depth to which the `frame variable`
and the `expression` commands dereference those pointers can be
controlled using the `--ptr-depth` argument.
Differential Revision: https://reviews.llvm.org/D132815
Fangrui Song [Sun, 20 Nov 2022 21:56:13 +0000 (21:56 +0000)]
[ELF] InputSectionBase: add bool compressed to avoid overloading size with compressed semantics
Rename uncompressedSize to size, to prepare for shrinking rawData+size from 3 words to 2 words.
Kazu Hirata [Sun, 20 Nov 2022 21:30:19 +0000 (13:30 -0800)]
[Analysis] Remove getInverseMinMaxPred
The last uses of getInverseMinMaxPred were removed on February 24,
2022 in commit
5423b0a52567b2fba8cbae3afee760f53c3f14d2.
Fangrui Song [Sun, 20 Nov 2022 21:23:30 +0000 (21:23 +0000)]
[ELF] Shuffle SectionBase fields. NFC
Move frequently used fields (e.g. sectionKind, partition) to the top. My x86-64 lld executable is 600+B smaller.
Thomas Preud'homme [Thu, 17 Nov 2022 23:35:18 +0000 (23:35 +0000)]
Add version to all LLVM cmake package
Add a version to non-LLVM cmake package so that users needing an exact
version match can use the version parameter to find_package. Also adjust
the find_package(LLVM) to use an exact version match as well.
Reviewed By: arsenm, stellaraccident
Differential Revision: https://reviews.llvm.org/D138274
Greg Clayton [Sun, 20 Nov 2022 17:40:23 +0000 (09:40 -0800)]
Add a new top level statistic that tracks how many modules have variable errors.
We have a statistic on each module named "debugInfoHadVariableErrors" which tracks when we have debug info, but an error prevented the variables from being displayed. This patch adds a new top level statistic named "totalModuleCountWithVariableErrors" which is a count of the modules that have "debugInfoHadVariableErrors" set to true.
Differential Revision: https://reviews.llvm.org/D138383
Kazu Hirata [Sun, 20 Nov 2022 20:24:16 +0000 (12:24 -0800)]
[ADT] Remove deperecated methods in OptionalStorage
Note that I deprecated these methods on August 7, 2022 in commit
commit
b5f8d42efe3e246d582d4a1a328fac915e4ce8dc.
Kazu Hirata [Sun, 20 Nov 2022 20:06:19 +0000 (12:06 -0800)]
[ADT] Remove llvm::empty
Note that I deprecated llvm::empty on September 18, 2022 in commit
6b49f30fca04ea3e4d57aef40e4bb6716a378870.
Bjorn Pettersson [Sat, 19 Nov 2022 00:40:37 +0000 (01:40 +0100)]
[LV] Clean up LoopVectorizationCostModel::calculateRegisterUsage. NFC
Minor refactoring in LoopVectorizationCostModel::calculateRegisterUsage.
Also adding some FIXME:s related to what appears to be some short
comings related to how the register usage is calculated.
Differential Revision: https://reviews.llvm.org/D138342
Bjorn Pettersson [Sat, 19 Nov 2022 00:16:11 +0000 (01:16 +0100)]
[TableGen] Fix SubRegIndex size computation for concatenated subregs of unknown size
When calculating the size of concatenated subregisters, and at least
one of the subregisters involved has an unknown size (-1), then the
concatenated size should be set to -1 as well.
This bug was found for an out-of-tree target.
Looking at lib/Target the only in-tree target that has a subregister
with unknown size is X86:
X86RegisterInfo.td: def sub_mask_0 : SubRegIndex<-1>;
But it looks like sub_mask_0 don't result in any concatenated subreg
index with faulty size if looking at X86SubRegIdxRanges[].
Differential Revision: https://reviews.llvm.org/D138341
Bjorn Pettersson [Sat, 19 Nov 2022 00:00:57 +0000 (01:00 +0100)]
[TableGen] Let -register-info-debug dump Offset/Size for each SubRegIndex
This patch adds dumping of the Offset and Size info for each
SubRegIndex printed when using
llvm-tblgen -gen-register-info -register-info-debug
It also updates the ConcatenatedSubregs.td to check those printouts,
including some new subreg definitions that show short-comings in
how the size is calculated when concatenating subregisters and at
least one has an incomplete size (-1). Today TableGen will just add
sizes together, resulting in MCRegisterInfo::getSubRegIdxSize()
returning a value that isn't -1 even if the combined subregister size
is unknown.
Differential Revision: https://reviews.llvm.org/D138340
Kazu Hirata [Sun, 20 Nov 2022 19:52:04 +0000 (11:52 -0800)]
[Support] Use std::is_convertible_v (NFC)
Kazu Hirata [Sun, 20 Nov 2022 19:32:31 +0000 (11:32 -0800)]
[ADT] Use std::is_copy_constructible_v (NFC)
Simon Pilgrim [Sun, 20 Nov 2022 19:28:27 +0000 (19:28 +0000)]
[X86] IceLakeServer - PACKS instructions take latency 3cy
This appears to be a slow down vs Skylake (which the model was copied off) - confirmed with uops.info / instlatx64
Noticed as D138359 was reporting that many of the PACKS overrides were redundant, but were in fact incorrect
Adrian Vogelsgesang [Thu, 25 Aug 2022 01:12:08 +0000 (18:12 -0700)]
[LLDB] Recognize `std::noop_coroutine()` in `std::coroutine_handle` pretty printer
With this commit, the `std::coroutine_handle` pretty printer now
recognizes `std::noop_coroutine()` handles. For noop coroutine handles,
we identify use the summary string `noop_coroutine` and we don't print
children
Instead of
```
(std::coroutine_handle<void>) $3 = coro frame = 0x555555559058 {
resume = 0x00005555555564f0 (a.out`std::__1::coroutine_handle<std::__1::noop_coroutine_promise>::__noop_coroutine_frame_ty_::__dummy_resume_destroy_func() at noop_coroutine_handle.h:79)
destroy = 0x00005555555564f0 (a.out`std::__1::coroutine_handle<std::__1::noop_coroutine_promise>::__noop_coroutine_frame_ty_::__dummy_resume_destroy_func() at noop_coroutine_handle.h:79)
}
```
we now print
```
(std::coroutine_handle<void>) $3 = noop_coroutine
```
Differential Revision: https://reviews.llvm.org/D132735
Louis Dionne [Fri, 18 Nov 2022 21:17:10 +0000 (16:17 -0500)]
[libc++][NFC] Refactor run-buildbot to make it closer to our buildkite pipeline
Differential Revision: https://reviews.llvm.org/D138380
Joe Loser [Sun, 20 Nov 2022 05:01:29 +0000 (22:01 -0700)]
[ADT] Remove deprecated `llvm::array_lengthof`
`llvm::array_lengthof` was marked deprecated a bit over two months ago in
5758c824dab4851fd061201483987b338c051417. Now, finally remove it. Since this
function was the only one in `STLArrayExtras.h`, get rid of the whole header.
D118342 (2022-01) moved the function to `STLArrayExtras.h`. This function should
have very few users now (MaskRay made some downstream inspection in
https://reviews.llvm.org/D133502#3777981). D133502 deprecated the function.
Differential Revision: https://reviews.llvm.org/D138374
Joe Loser [Sun, 20 Nov 2022 05:11:18 +0000 (22:11 -0700)]
[ADT] Simplify determining size for joining items
`join_items_size` uses recursive function templates for determining the total
items size. Instead, use fold expressions since we have C++17 available now.
Differential Revision: https://reviews.llvm.org/D138375
Adrian Vogelsgesang [Sun, 20 Nov 2022 14:52:15 +0000 (06:52 -0800)]
[lldb] Fix test expectation in `TestCoroutineHandle.py` for 32-bit systems
Kazu Hirata [Sun, 20 Nov 2022 17:59:36 +0000 (09:59 -0800)]
[llvm] Remove unused forward declarations (NFC)
Corentin Jabot [Sun, 20 Nov 2022 17:51:52 +0000 (18:51 +0100)]
[Clang] Add papers approved in Kona by WG21 to the C++ status page
Approved core issues will be updated separately through an update
of the defect report list when a new core issue list is
published.
Simon Pilgrim [Sun, 20 Nov 2022 17:19:50 +0000 (17:19 +0000)]
[MCA][X86] Add missing test coverage for BWI instructions
Simon Pilgrim [Sun, 20 Nov 2022 16:00:26 +0000 (16:00 +0000)]
[MCA][X86] Add test coverage for IFMA instructions
Nico Weber [Sun, 20 Nov 2022 17:02:33 +0000 (12:02 -0500)]
[gn build] port
af8c49dc1ec4
Louis Dionne [Fri, 18 Nov 2022 20:01:33 +0000 (15:01 -0500)]
[libc++] Introduce a setting to remove fstream from the library
This allows porting the library to platforms that are able to support
<iostream> but that do not have a notion of a filesystem, and where it
hence doesn't make sense to support std::fstream (and never will).
Also, remove reliance on <fstream> in various tests that didn't
actually need it.
Differential Revision: https://reviews.llvm.org/D138327
Alexandre Ganea [Sun, 20 Nov 2022 15:24:02 +0000 (10:24 -0500)]
[LLD][COFF] Survive empty and invalid PCH signature
Solve two issues that showed up when using LLD with Unreal Engine & FASTBuild:
1. It seems the S_OBJNAME record doesn't always record the "precomp signature". We were relying on that to match the PCH.OBJ with their dependent-OBJ.
2. MSVC link.exe is able to link a PCH.OBJ when the "precomp signatureÈ doesn't match, but LLD was failing. This was occuring since the Unreal Engine Build Tool was compiling the PCH.OBJ, but the dependent-OBJ were compiled & cached through FASTBuild. Upon a clean rebuild, the PCH.OBJs were recompiled by the Unreal Build Tool, thus the "precomp signatures" were changing; however the OBJs were already cached by FASTBuild, thus having an old "precomp signatures".
We now ignore "precomp signatures" and properly fallback to cmd-line name lookup, like MSVC link.exe does, and only fail if the PCH.OBJ type stream doesn't match the count expected by the dependent-OBJ.
Differential Revision: https://reviews.llvm.org/D136762
Benjamin Kramer [Sun, 20 Nov 2022 14:24:14 +0000 (15:24 +0100)]
[X86] Fix atomic rmw intrinsic expansion for non-opaque pointers
This is a bit annoying, but there are still users out there that got
broken by this (this time it was numba). We need to keep some barebones
support around until non-opaque pointers are completely gone.
Adrian Vogelsgesang [Wed, 24 Aug 2022 03:53:00 +0000 (20:53 -0700)]
Reapply "[LLDB] Devirtualize coroutine promise types for `std::coroutine_handle`"
The original commit was missing a `ClangASTImporter::CopyType` call.
Original commit message:
This commit teaches the `std::coroutine_handle` pretty-printer to
devirtualize type-erased promise types. This is particularly useful to
resonstruct call stacks, either of asynchronous control flow or of
recursive invocations of `std::generator`. For the example recently
introduced by https://reviews.llvm.org/D132451, printing the `__promise`
variable now shows
```
(std::__coroutine_traits_sfinae<task, void>::promise_type) __promise = {
continuation = coro frame = 0x555555562430 {
resume = 0x0000555555556310 (a.out`task detail::chain_fn<1>() at llvm-nested-example.cpp:66)
destroy = 0x0000555555556700 (a.out`task detail::chain_fn<1>() at llvm-nested-example.cpp:66)
promise = {
continuation = coro frame = 0x5555555623e0 {
resume = 0x0000555555557070 (a.out`task detail::chain_fn<2>() at llvm-nested-example.cpp:66)
destroy = 0x0000555555557460 (a.out`task detail::chain_fn<2>() at llvm-nested-example.cpp:66)
promise = {
...
}
}
result = 0
}
}
result = 0
}
```
(shortened to keep the commit message readable) instead of
```
(std::__coroutine_traits_sfinae<task, void>::promise_type) __promise = {
continuation = coro frame = 0x555555562430 {
resume = 0x0000555555556310 (a.out`task detail::chain_fn<1>() at llvm-nested-example.cpp:66)
destroy = 0x0000555555556700 (a.out`task detail::chain_fn<1>() at llvm-nested-example.cpp:66)
}
result = 0
}
```
Note how the new debug output reveals the complete asynchronous call
stack: our own function resumes `chain_fn<1>` which in turn will resume
`chain_fn<2>` and so on. Thereby this change allows users of lldb to
inspect the logical coroutine call stack without using any custom debug
scripts (although the display is still a bit clumsy. It would be nicer
to also integrate this into lldb's backtrace feature, but I don't know
how to do so)
The devirtualization currently works by introspecting the function
pointed to by the `destroy` pointer. (The `resume` pointer is not worth
much, given that for the final suspend point `resume` is set to a
nullptr. We have to use the `destroy` pointer instead.) We then look
for a `__promise` variable inside the `destroy` function. This
`__promise` variable is synthetically generated by LLVM, and looking at
its type reveals the type-erased promise_type.
This approach only works for clang-generated code, though. While gcc
also adds a `_Coro_promise` variable to the `resume` function, it does
not do so for the `destroy` function. However, we can't use the `resume`
function, as it will be reset to a nullptr at the final suspension
point. For the time being, I am happy with de-virtualization only working
for clang. A follow-up commit will further improve devirtualization and
also expose the variables spilled to the coroutine frame. As part of
this, I will also revisit gcc support.
Differential Revision: https://reviews.llvm.org/D132624
Simon Pilgrim [Sun, 20 Nov 2022 14:17:44 +0000 (14:17 +0000)]
[X86] Remove unnecessary SHLD32rri8/SHRD16rri8 instruction override from bdver2 model
Reported by D138359 - the override matches the WriteSHDrri base sched def
Simon Pilgrim [Sun, 20 Nov 2022 14:05:05 +0000 (14:05 +0000)]
[X86] Remove unnecessary XGETBV instruction overrides from znver1/znver2 models
Reported by D138359 - znver models already treats all WriteSystem sched instructions as microcoded
Simon Pilgrim [Sun, 20 Nov 2022 13:56:04 +0000 (13:56 +0000)]
[MCA][X86] Add test coverage for XSAVE instructions
Simon Pilgrim [Sun, 20 Nov 2022 13:19:49 +0000 (13:19 +0000)]
[X86] Remove unnecessary RDPMC/RDTSC instruction overrides from znver1/znver2 models
Reported by D138359
Simon Pilgrim [Sun, 20 Nov 2022 12:57:44 +0000 (12:57 +0000)]
[X86] Remove unnecessary string instruction overrides from znver1/znver2 models
Reported by D138359 - they were being overridden as WriteMicrocoded despite already being declared WriteMicrocoded
It also fixes a rather funny instregex mismatch that was matching the movsldup shuffle by mistake
Simon Pilgrim [Sun, 20 Nov 2022 12:22:11 +0000 (12:22 +0000)]
[X86] Remove unnecessary bit test instruction overrides from znver2 model
Reported by D138359 and confirmed with AMD SoG - matches znver1 model
Simon Pilgrim [Sun, 20 Nov 2022 12:13:02 +0000 (12:13 +0000)]
[X86] Improve LOOP/LOOPE/LOOPNE schedule on SandyBridge model
D138359 was reporting that this override was superfluous, but it had never been setup - I took the numbers from uops.info (I couldn't find an estimate in Intel docs).
Simon Pilgrim [Sun, 20 Nov 2022 11:35:21 +0000 (11:35 +0000)]
[MCA][X86] Add test coverage for LOOP/LOOPE/LOOPNE instructions
These were missed for some reason - only noticed this while investigating a FIXME in the SandyBridge model
Also sync the znver2/znver3 tests which had been missed when LOCK test coverage was added
Simon Pilgrim [Sun, 20 Nov 2022 10:48:27 +0000 (10:48 +0000)]
[X86] Remove unnecessary zmm shuffle instruction overrides from IceLake model
Reported by D138359 and confirmed with Intel AoM, Agner + uops.info
Simon Pilgrim [Sat, 19 Nov 2022 18:53:02 +0000 (18:53 +0000)]
[X86] znver2 - (V)EXTRACTPSrr takes 2 uops
D138359 was reporting that the EXTRACTPSrr override was unnecessary, however the AMD SoG and Agner both confirm that both the rr and rm versions take 2uops (matching znver1)
Fangrui Song [Sun, 20 Nov 2022 09:07:03 +0000 (09:07 +0000)]
[ADT] Remove None=1 MSVC workaround
This workaround from
0cd22f9540c0591132ec991c51103cf800cf4e24 (2017-01) is very
likely unneeded.
Kazu Hirata [Sun, 20 Nov 2022 08:24:40 +0000 (00:24 -0800)]
Use None consistently (NFC)
This patch replaces NoneType() and NoneType::None with None in
preparation for migration from llvm::Optional to std::optional.
In the std::optional world, we are not guranteed to be able to
default-construct std::nullopt_t or peek what's inside it, so neither
NoneType() nor NoneType::None has a corresponding expression in the
std::optional world.
Once we consistently use None, we should even be able to replace the
contents of llvm/include/llvm/ADT/None.h with something like:
using NoneType = std::nullopt_t;
inline constexpr std::nullopt_t None = std::nullopt;
to ease the migration from llvm::Optional to std::optional.
Differential Revision: https://reviews.llvm.org/D138376
Noah Goldstein [Sat, 19 Nov 2022 09:06:37 +0000 (01:06 -0800)]
[clang-format] Don't use PPIndentWidth inside multi-line macros
Differential Revision: https://reviews.llvm.org/D137181
Kazu Hirata [Sun, 20 Nov 2022 07:22:48 +0000 (23:22 -0800)]
[lldb] Use Optional::has_value instead of Optional::hasValue (NFC)
Kazu Hirata [Sun, 20 Nov 2022 05:11:10 +0000 (21:11 -0800)]
[Analysis] Use llvm::Optional::value_or (NFC)
Phoebe Wang [Sun, 20 Nov 2022 02:42:26 +0000 (10:42 +0800)]
[X86] Use lock or/and/xor for cases that we only care about the EFLAGS
This is a follow up of D137711 to fix the reset of #58685.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D138294
Vitaly Buka [Sun, 20 Nov 2022 01:12:32 +0000 (17:12 -0800)]
[test][asan] Rename variables for less confusion
Kazu Hirata [Sat, 19 Nov 2022 23:36:50 +0000 (15:36 -0800)]
[Analysis] Teach getOptionalIntLoopAttribute to return std::optional (NFC)
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716
Kazu Hirata [Sat, 19 Nov 2022 23:00:19 +0000 (15:00 -0800)]
[SelectionDAG] Teach getRegistersForValue to return std::optional (NFC)
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716/11
Fangrui Song [Sat, 19 Nov 2022 22:09:05 +0000 (22:09 +0000)]
Remove unused llvm/IRPrinter/IRPrintingPasses.h or reorder #include after D137768
Simon Pilgrim [Sat, 19 Nov 2022 21:46:14 +0000 (21:46 +0000)]
[MCA][X86] Add test coverage for BF16 instructions
Simon Pilgrim [Sat, 19 Nov 2022 21:20:31 +0000 (21:20 +0000)]
[MCA][X86] Add test coverage for VP2INTERSECT instructions
NOTE: For IceLakeServer we actually test TigerLake as that's the only target that supports it (we do something similar for F16C on IvyBridge in the SandyBridge tests).
Kazu Hirata [Sat, 19 Nov 2022 21:45:27 +0000 (13:45 -0800)]
[ADT] Remove deprecated Any::hasValue
Note that I deprecated Any::hasValue on August 20, 2022 in commit
abb6271d809b37d3c3318894d697d5820032bec8.