OSDN Git Service
Juneyoung Lee [Mon, 20 Jul 2020 00:21:28 +0000 (09:21 +0900)]
[ValueTracking] Let isGuaranteedNotToBeUndefOrPoison use canCreateUndefOrPoison
This patch adds support more operations.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D83926
sstefan1 [Sun, 19 Jul 2020 18:45:24 +0000 (20:45 +0200)]
[Utils][Fix] remove unnecessary ; at the end
sstefan1 [Sat, 11 Jul 2020 19:53:50 +0000 (21:53 +0200)]
[Utils] Check function attributes in update_test_checks
Summary:
This introduces new flag to the update_test_checks and
update_cc_test_checks that allows for function attributes
to be checked in a check-line. If the flag is not set,
the behavior should remain the same.
Reviewers: jdoerfert
Subscribers: arichardson, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83629
Lang Hames [Sun, 19 Jul 2020 03:43:13 +0000 (20:43 -0700)]
[JITLink][MachO] Fix handling of non-extern UNSIGNED pair of SUBTRACTOR relocs.
When processing a MachO SUBTRACTOR/UNSIGNED pair, if the UNSIGNED target
is non-extern then check the r_symbolnum field of the relocation to find
the targeted section and use the section's address to find 'ToSymbol'.
Previously 'ToSymbol' was found by loading the initial value stored at
the fixup location and treating this as an address to search for. This
is incorrect, however: the initial value includes the addend and will
point to the wrong block if the addend is less than zero or greater than
the block size.
rdar://
65756694
Juneyoung Lee [Sun, 19 Jul 2020 17:21:52 +0000 (02:21 +0900)]
Fix ValueTrackingTest.cpp to use pair instead of tuple
Jameson Nash [Wed, 15 Jul 2020 13:17:42 +0000 (09:17 -0400)]
[ConstantFolding] check applicability of AllOnes constant creation first
The getAllOnesValue can only handle things that are bitcast from a
ConstantInt, while here we bitcast through a pointer, so we may see more
complex objects (like Array or Struct).
Differential Revision: https://reviews.llvm.org/D83870
Alex Richardson [Sun, 19 Jul 2020 16:24:57 +0000 (17:24 +0100)]
[llvm-reduce] Fix incorrect indices in argument reduction pass
The function extractArgumentsFromModule() was passing a one-based index to,
but replaceFunctionCalls() was expecting a zero-based argument index. This
resulted in assertion errors when reducing function call arguments with
different types. Additionally, the
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D84099
Logan Smith [Sun, 19 Jul 2020 16:35:29 +0000 (09:35 -0700)]
[llvm][NFC] Add missing 'override'
Juneyoung Lee [Sun, 19 Jul 2020 16:23:58 +0000 (01:23 +0900)]
[ValueTracking] Add canCreateUndefOrPoison & let canCreatePoison use Operator
This patch
- adds `canCreateUndefOrPoison`
- refactors `canCreatePoison` so it can deal with constantexprs
`canCreateUndefOrPoison` will be used at D83926.
Reviewed By: nikic, jdoerfert
Differential Revision: https://reviews.llvm.org/D84007
Bruno Ricci [Sun, 19 Jul 2020 16:08:17 +0000 (17:08 +0100)]
[clang] Fix the warning for a non-void consteval function without a return value to actually say "consteval".
This warning was modified in
796ed03b8412 to use the term "consteval"
for consteval functions. However the warning has never worked as
intended since the diagnostic's arguments are used in the wrong order.
This was unfortunately missed by
796ed03b8412 since no test did exercise
this specific warning.
Additionally send the NamedDecl* into the diagnostic instead of just the
IdentifierInfo* to correctly work with special names and template
arguments.
Bruno Ricci [Sun, 19 Jul 2020 15:53:59 +0000 (16:53 +0100)]
[clang] Disable a few formatting options for test/
Hopefully this will make the bot a little less noisy. Rationale for each:
AlignTrailingComments: We don't want to force-align the various expected-error
and friends.
CommentPragmas: Tell clang-format to leave the "// CHECK:" and the
"// expected-" alone.
AlwaysBreakTemplateDeclarations: Templates in tests often have no break between
the template-head and the declaration.
Differential Revision: https://reviews.llvm.org/D83901
Wenlei He [Sun, 19 Jul 2020 15:49:04 +0000 (08:49 -0700)]
Revert "[InlineAdvisor] New inliner advisor to replay inlining from optimization remarks"
This reverts commit
2d6ecfa168c2d36ac88efc854f19b05d1c540ded.
Wenlei He [Fri, 17 Jul 2020 23:35:56 +0000 (16:35 -0700)]
[InlineAdvisor] New inliner advisor to replay inlining from optimization remarks
Summary:
This change added a new inline advisor that takes optimization remarks from previous inlining as input, and provides the decision as advice so current inlining can replay inline decisions of a different compilation. Dwarf inline stack with line and discriminator is used as anchor for call sites including call context. The change can be useful for Inliner tuning as it provides a channel to allow external input for tweaking inline decisions. Existing alternatives like alwaysinline attribute is per-function, not per-callsite. Per-callsite inline intrinsic can be another solution (not yet existing), but it's intrusive to implement and also does not differentiate call context.
A switch -sample-profile-inline-replay=<inline_remarks_file> is added to hook up the new inline advisor with SampleProfileLoader's inline decision for replay. Since SampleProfileLoader does top-down inlining, inline decision can be specialized for each call context, hence we should be able to replay inlining accurately. However with a bottom-up inliner like CGSCC inlining, the replay can be limited due to lack of specialization for different call context. Apart from that limitation, the new inline advisor can still be used by regular CGSCC inliner later if needed for tuning purpose.
Subscribers: mgorny, aprantl, hiraditya, llvm-commits
Tags: #llvm
Resubmit for https://reviews.llvm.org/D84086
Aaron Ballman [Sun, 19 Jul 2020 15:19:48 +0000 (11:19 -0400)]
Silence a "logical operation on address of string constant" via CMake instead.
Aaron Ballman [Sun, 19 Jul 2020 15:12:39 +0000 (11:12 -0400)]
Aaron Ballman [Sun, 19 Jul 2020 14:50:51 +0000 (10:50 -0400)]
Silencing some 'logical operation on address of string constant diagnostics; NFC
Sanjay Patel [Sun, 19 Jul 2020 14:03:55 +0000 (10:03 -0400)]
[x86] split FMA with fast-math-flags to avoid libcall
fma reassoc A, B, C --> fadd (fmul A, B), C (when target has no FMA hardware)
C/C++ code may use explicit fma() calls (which become LLVM fma
intrinsics in IR) but then gets compiled with -ffast-math or similar.
For targets that do not have FMA hardware, we don't want to go out to
the math library for a precise but slow FMA result.
I tried this as a generic DAGCombine, but it caused infinite looping
on more than 1 other target, so there's likely some over-reaching fma
formation happening.
There's also a potential intersection of strict FP with fast-math here.
Deferring to current behavior for that case (assuming that strict-ness
overrides fast-ness).
Differential Revision: https://reviews.llvm.org/D83981
Roman Lebedev [Sun, 19 Jul 2020 13:37:03 +0000 (16:37 +0300)]
Reland "[InstCombine] Lower infinite combine loop detection thresholds"
This reverts commit
4500db8c59621a31c622862a2946457fdee481ce,
which was reverted because lower thresholds exposed a new issue (PR46680).
Now that it was resolved by
d12ec0f752e7f2c7f7252539da2d124264ec33f7,
we can reinstate lower limits and wait for a new bugreport before
reverting this again...
Nikita Popov [Thu, 9 Jul 2020 20:52:50 +0000 (22:52 +0200)]
[PredicateInfo] Add a method to interpret predicate as cmp constraint
Both users of predicteinfo (NewGVN and SCCP) are interested in
getting a cmp constraint on the predicated value. They currently
implement separate logic for this. This patch adds a common method
for this in PredicateBase.
This enables a missing bit of PredicateInfo handling in SCCP: Now
the predicate on the condition itself is also used. For switches
it means we know that the switched-on value is the same as the case
value. For assumes/branches we know that the condition is true or
false.
Differential Revision: https://reviews.llvm.org/D83640
Roman Lebedev [Sun, 19 Jul 2020 13:09:46 +0000 (16:09 +0300)]
[NFCI][GVN] Make IsValueFullyAvailableInBlock() readable - use enum class instead of magic numbers
This does not change any logic, it only wraps the magic 0/1/2/3 constants
into an enum class.
Sanjay Patel [Sun, 19 Jul 2020 13:24:03 +0000 (09:24 -0400)]
[InstSimplify] fold fcmp with infinity constant using isKnownNeverInfinity
This is a step towards trying to remove unnecessary FP compares
with infinity when compiling with -ffinite-math-only or similar.
I'm intentionally not checking FMF on the fcmp itself because
I'm assuming that will go away eventually.
The analysis part of this was added with rGcd481136 for use with
isKnownNeverNaN. Similarly, that could be an enhancement here to
get predicates like 'one' and 'ueq'.
Differential Revision: https://reviews.llvm.org/D84035
Nikita Popov [Sat, 18 Jul 2020 20:22:41 +0000 (22:22 +0200)]
[InstCombine] Fix store merge worklist management (PR46680)
Fixes https://bugs.llvm.org/show_bug.cgi?id=46680.
Just like insertions through IRBuilder, InsertNewInstBefore()
should be using the deferred worklist mechanism, so that processing
of newly added instructions is prioritized.
There's one side-effect of the worklist order change which could be
classified as a regression. An add op gets pushed through a select
that at the time is not a umax. We could add a reverse transform
that tries to push adds in the reverse direction to restore a min/max,
but that seems like a sure way of getting infinite loops... Seems
like something that should best wait on min/max intrinsics.
Differential Revision: https://reviews.llvm.org/D84109
David Green [Thu, 16 Jul 2020 09:36:23 +0000 (10:36 +0100)]
[ARM] Don't mark vctp as having sideeffects
As far as I can tell, it should not be necessary for VCTP to be
unpredictable in tail predicated loops. Either it has a a valid loop
counter as a operand which will naturally keep it in the right loop, or
it doesn't and it won't be converted to a tail predicated loop. Not
marking it as having side effects allows it to be scheduled more cleanly
for cases where it is not expected to become a tail predicate loop.
Differential Revision: https://reviews.llvm.org/D83907
Fangrui Song [Sun, 19 Jul 2020 07:26:01 +0000 (00:26 -0700)]
[llvm-cov gcov] Don't require NUL terminator when reading files
.gcno, .gcda and source files can be modified while we are reading them. If the
concurrent modification of a file being read nullifies the NUL terminator
assumption, llvm-cov can trip over an assertion failure in MemoryBuffer::init.
This is not so rare - the source files can be in an editor and .gcda can be
written by an running process (if the process forks, when .gcda gets written is
probably more unpredictable).
There is no accompanying test because an assertion failure requires data
races with some involved setting.
Kang Zhang [Sun, 19 Jul 2020 07:01:45 +0000 (07:01 +0000)]
[PowerPC] Remove the redundant implicit operands in ppc-early-ret pass
Summary:
In the `ppc-early-ret` pass, we have use `BuildMI` and `copyImplicitOps` when the branch instructions can do the early return. But the two functions will add implicit operands twice, this is not correct.
This patch is to remove the redundant implicit operands in `ppc-early-ret pass`.
Reviewed By: jsji
Differential Revision: https://reviews.llvm.org/D76042
Yuanfang Chen [Sun, 19 Jul 2020 05:23:07 +0000 (22:23 -0700)]
[NewPM] Allow passes to never be skipped
A pass declares itself unskippable by defining a method `static bool isRequired()`.
Also, this patch makes pass managers and adaptor passes required (unskippable).
PassInstrumentation before-pass-callbacks could be used to skip passes by returning false.
However, some passes should not be skipped at all. Especially so for special-purpose passes such as pass managers and adaptor passes since if they are skipped for any reason, the passes contained by them would also be skipped ignoring contained passes's return value of `isRequired()`.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D82344
Yuanfang Chen [Sun, 19 Jul 2020 05:26:37 +0000 (22:26 -0700)]
[NewPM] make parsePassPipeline parse adaptor-wrapped user passes
Currently, when parsing text pipeline, different kinds of passes always
introduce nested pass managers. This makes it impossible to test the
adaptor-wrapped user passes from the text pipeline interface which is needed
by D82344 test cases. This also seems useful in general. See comments above
`parsePassPipeline`.
The syntax would be like mixing passes of different types, but it is
not the same as inferring the correct pass type and then adding the
matching nested pass managers. Strictly speaking, the resulted pipelines
are different.
Reviewed By: asbirlea, aeubanks
Differential Revision: https://reviews.llvm.org/D82698
Fangrui Song [Sun, 19 Jul 2020 04:01:41 +0000 (21:01 -0700)]
[Driver] --print-search-dirs: print -B options and COMPILER_PATH
Fangrui Song [Sun, 19 Jul 2020 03:26:01 +0000 (20:26 -0700)]
[Driver] -B: don't search for target triple prefixes
To match GCC (either crossing or not), which doesn't prepend target triple prefixes to `exec_prefixes`.
As an example, powerpc64le-linux-gnu-gcc does not search "powerpc64le-linux-gnu-${name}" in a -B path.
Mehdi Amini [Sat, 18 Jul 2020 22:21:29 +0000 (22:21 +0000)]
Fix invalid link in the MLIR Standard Dialect www page (2nd attempt)
Fangrui Song [Sat, 18 Jul 2020 22:07:46 +0000 (15:07 -0700)]
[gcov] Add __gcov_dump/__gcov_reset and delete __gcov_flush
GCC r187297 (2012-05) introduced `__gcov_dump` and `__gcov_reset`.
`__gcov_flush = __gcov_dump + __gcov_reset`
The resolution to https://gcc.gnu.org/PR93623 ("No need to dump gcdas when forking" target GCC 11.0) removed the unuseful and undocumented __gcov_flush.
Close PR38064.
Reviewed By: calixte, serge-sans-paille
Differential Revision: https://reviews.llvm.org/D83149
Fangrui Song [Sun, 12 Jul 2020 00:17:21 +0000 (17:17 -0700)]
[gcov][test] Simplify instrprof-dlopen-dlclose-gcov.test
Nikita Popov [Sat, 18 Jul 2020 21:36:42 +0000 (23:36 +0200)]
[InstCombine] Add test for PR46680 (NFC)
Bruno Ricci [Sat, 18 Jul 2020 19:44:06 +0000 (20:44 +0100)]
[clang][NFC] Tests showing the problems with some uses of NamedDecl::getDeclName in diagnostics, SemaOverload.cpp+SemaStmt.cpp part
Bruno Ricci [Sat, 18 Jul 2020 19:39:16 +0000 (20:39 +0100)]
[clang][NFC] Tests showing the problems with some uses of NamedDecl::getDeclName in diagnostics, SemaExpr.cpp part
Bruno Ricci [Sat, 18 Jul 2020 19:35:16 +0000 (20:35 +0100)]
[clang] Provide a more specific diagnostic for a misplaced lambda capture-default.
Currently a capture-default which is not the first element in the lambda-capture
is diagnosed with a generic expected variable name or 'this' in lambda capture
list, which is true but not very helpful.
If we don't have already parsed a capture-default then a lone "&" or "=" is
likely to be a misplaced capture-default, so diagnose it as such.
Differential Revision: https://reviews.llvm.org/D83681
Reviewed By: aaron.ballman
peter klausler [Fri, 17 Jul 2020 18:17:39 +0000 (11:17 -0700)]
[flang] Improve output from a STOP statement
Add a missing newline to IEEE FP flag formatting, and
don't neglect to emit STOP when there's no code number.
Reviewed By: tskeith
Differential Revision: https://reviews.llvm.org/D84060
Fangrui Song [Sat, 18 Jul 2020 18:10:18 +0000 (11:10 -0700)]
[gcov][test] Call wait() to make gcov-fork.c reliable
If the parent exit before the child, the line counts might be 1.
next:18'0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
26: 1: 21: if (fork() == -1) return 1; // CHECK-NEXT: 1: [[#@LINE]]:
next:18'0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
27: 1: 22: func2(); // CHECK-NEXT: 2: [[#@LINE]]:
next:18'0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
28: 1: 23: return 0; // CHECK-NEXT: 2: [[#@LINE]]:
next:18'0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
peter klausler [Fri, 17 Jul 2020 18:14:28 +0000 (11:14 -0700)]
[flang] Insert leading blanks in LOGICAL formatted output fields
Summary: For Lw output editing, emit (w-1) blanks before the T or the F.
Reviewed By: sscalpone, PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D84059
Artem Dergachev [Sat, 18 Jul 2020 02:42:20 +0000 (19:42 -0700)]
[analyzer] scan-build: Fix silencing multiple core checkers.
It was only silencing one checker because -analyzer-config flags
can only carry one value at a time.
Joseph Huber [Sat, 18 Jul 2020 16:00:08 +0000 (12:00 -0400)]
[OpenMP] Add Additional Function Attribute Information to OMPKinds.def
Summary:
This patch adds more function attribute information to the runtime function definitions in OMPKinds.def. The goal is to provide sufficient information about OpenMP runtime functions to perform more optimizations on OpenMP code.
Reviewers: jdoerfert
Subscribers: aaron.ballman cfe-commits yaxunl guansong sstefan1 llvm-commits
Tags: #OpenMP #clang #LLVM
Differential Revision: https://reviews.llvm.org/D81031
Roman Lebedev [Sat, 18 Jul 2020 16:41:21 +0000 (19:41 +0300)]
[NFC][CVP] processSDiv(): pacify gcc compilers
Mehdi Amini [Sat, 18 Jul 2020 16:22:31 +0000 (16:22 +0000)]
Fix dead link on MLIR website
Mehdi Amini [Sat, 18 Jul 2020 16:12:41 +0000 (16:12 +0000)]
Fix Markdown format for lists in the Standard Dialect documentation
This affects the rendering on the website.
Fangrui Song [Sat, 18 Jul 2020 16:07:40 +0000 (09:07 -0700)]
[ELF][test] Update test after DW_EH_PE_sdata4/DW_EH_PE_sdata8 change
Anders Waldenborg [Sat, 18 Jul 2020 15:26:06 +0000 (17:26 +0200)]
[clang-format] Make sure rst documentation matches comments
clang/docs/tools/dump_format_style.py is used to read the comments
from clang/include/clang/Format/Format.h and update the contents of
clang/docs/ClangFormatStyleOptions.rst
Recent changes made these out of date. This commit syncs them by
folding the improved wording back to the comments and then
regenerating the rst file.
Differential Revision: https://reviews.llvm.org/D84103
Fangrui Song [Sat, 18 Jul 2020 15:59:05 +0000 (08:59 -0700)]
[DebugInfo] Respect relocations when decoding DW_EH_PE_sdata4 & DW_EH_PE_sdata8 and support R_ARM_REL32
The addresses in llvm-dwarfdump --eh-frame output for object files are closer to readelf -wf output now.
Florian Hahn [Sat, 18 Jul 2020 14:59:51 +0000 (15:59 +0100)]
[PredicateInfo] Fold PredicateWithCondition into PredicateBase (NFC).
Each concrete instance of a predicate has a condition (also noted in the
original PredicateBase comment) and to me it seems like there is no
clear benefit of having both PredicateBase and PredicateWithCondition
and they can be folded together.
Reviewers: nikic, efriedma
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D84089
Roman Lebedev [Sat, 18 Jul 2020 10:23:26 +0000 (13:23 +0300)]
[CVP] Soften SDiv into a UDiv as long as we know domains of both of the operands.
Yes, if operands are non-positive this comes at the extra cost
of two extra negations. But a. division is already just
ridiculously costly, two more subtractions can't hurt much :)
and b. we have better/more analyzes/folds for an unsigned division,
we could end up narrowing it's bitwidth, converting it to lshr, etc.
This is essentially a take two on
0fdcca07ad2c0bdc2cdd40ba638109926f4f513b,
which didn't fix the potential regression i was seeing,
because ValueTracking's computeKnownBits() doesn't make use
of dominating conditions in it's analysis.
While i could teach it that, this seems like the more general fix.
This big hammer actually does catch said potential regression.
Over vanilla test-suite + RawSpeed + darktable
(10M IR instrs, 1M IR BB, 1M X86 ASM instrs), this fires/converts 5 more
(+2%) SDiv's, the total instruction count at the end of middle-end pipeline
is only +6, so out of +10 extra negations, ~half are folded away,
and asm instr count is only +1, so practically speaking all extra
negations are folded away and are therefore free.
Sadly, all these new UDiv's remained, none folded away.
But there are two less basic blocks.
https://rise4fun.com/Alive/VS6
Name: v0
Pre: C0 >= 0 && C1 >= 0
%r = sdiv i8 C0, C1
=>
%r = udiv i8 C0, C1
Name: v1
Pre: C0 <= 0 && C1 >= 0
%r = sdiv i8 C0, C1
=>
%t0 = udiv i8 -C0, C1
%r = sub i8 0, %t0
Name: v2
Pre: C0 >= 0 && C1 <= 0
%r = sdiv i8 C0, C1
=>
%t0 = udiv i8 C0, -C1
%r = sub i8 0, %t0
Name: v3
Pre: C0 <= 0 && C1 <= 0
%r = sdiv i8 C0, C1
=>
%r = udiv i8 -C0, -C1
Roman Lebedev [Sat, 18 Jul 2020 10:13:01 +0000 (13:13 +0300)]
[NFC][CVP] Rename predicates - s/positive/non negative/ to better note that zero is ok
Roman Lebedev [Sat, 18 Jul 2020 10:07:13 +0000 (13:07 +0300)]
[NFC][CVP] Refactor isPositive() out of hasPositiveOperands()
Roman Lebedev [Sat, 18 Jul 2020 14:30:42 +0000 (17:30 +0300)]
[NFC][CVP] Add tests for possible sdiv->udiv where operands are not non-negative
Currently that fold requires both operands to be non-negative,
but the only real requirement for the fold is that we must know
the domains of the operands.
Matt Arsenault [Sat, 18 Jul 2020 14:35:40 +0000 (10:35 -0400)]
AMDGPU/GlobalISel: Address some test fixmes that don't fail now
Matt Arsenault [Sat, 18 Jul 2020 14:08:17 +0000 (10:08 -0400)]
AMDGPU/GlobalISel: Fix test copy paste error
David Green [Fri, 17 Jul 2020 07:35:04 +0000 (08:35 +0100)]
[LV] Add additional InLoop redution tests. NFC
Evgeny Leviant [Sat, 18 Jul 2020 11:11:40 +0000 (14:11 +0300)]
[CodeGen][TargetPassConfig] Add TargetTransformInfo pass correctly
Patch adds tti pass directly enforcing its execution with correctly set
TargetTransformInfo.
Differential revision: https://reviews.llvm.org/D84047
Vitaly Buka [Sat, 18 Jul 2020 09:16:17 +0000 (02:16 -0700)]
[sanitizer] Disable test on incompatible platforms
Yash Jain [Sat, 18 Jul 2020 07:39:30 +0000 (13:09 +0530)]
[MLIR] Add lowering for affine.parallel to scf.parallel
Add lowering conversion from affine.parallel to scf.parallel.
Differential Revision: https://reviews.llvm.org/D83239
Fangrui Song [Sat, 18 Jul 2020 06:49:15 +0000 (23:49 -0700)]
[RelocationResolver] Support R_AARCH64_PREL32
Code from D83800 by Yichao Yu
Fangrui Song [Sat, 18 Jul 2020 06:29:50 +0000 (23:29 -0700)]
[RelocationResolver] Support R_PPC_REL32 & R_PPC64_REL{32,64}
This suppresses `failed to compute relocation: R_PPC_REL32, Invalid data was encountered while parsing the file`
and its 64-bit variants when running llvm-dwarfdump on a PowerPC object file with .eh_frame
Unfortunately it is difficult to test the computation:
DWARFDataExtractor::getEncodedPointer does not use the relocated value
and even if it does, we need to teach llvm-dwarfdump --eh-frame to do
some linker job to report a reasonable address.
Gui Andrade [Sat, 18 Jul 2020 03:49:17 +0000 (03:49 +0000)]
[MSAN tests] Require android for sigandorset.cpp
Gui Andrade [Sat, 18 Jul 2020 03:53:00 +0000 (03:53 +0000)]
Revert "update libatomic instrumentation"
This was committed mistakenly.
This reverts commit
1f29171ae77f81cacea32808b67d7ae62da23e0c.
Gui Andrade [Sat, 18 Jul 2020 03:39:14 +0000 (03:39 +0000)]
update libatomic instrumentation
Gui Andrade [Sat, 18 Jul 2020 03:18:48 +0000 (03:18 +0000)]
[LLVM] Add libatomic load/store functions to TargetLibraryInfo
This allows treating these functions like libcalls.
This patch is a prerequisite to instrumenting them in MSAN: https://reviews.llvm.org/D83337
Differential Revision: https://reviews.llvm.org/D83361
Gui Andrade [Sat, 18 Jul 2020 03:09:39 +0000 (03:09 +0000)]
sanitizers: Implement sig{and,or}set interceptors
Also adds a sanitizers-wide test, and a msan test, for these functions.
Chen Zheng [Sat, 18 Jul 2020 02:51:00 +0000 (22:51 -0400)]
[PowerPC]add testcase for adding store (load float*) pattern, nfc
Chen Zheng [Wed, 15 Jul 2020 08:03:35 +0000 (04:03 -0400)]
[SCEV][IndVarSimplify] insert point should not be block front.
Recommit after removing the unused cast instructions.
Differential Revision: https://reviews.llvm.org/D80975
Joel E. Denny [Sat, 18 Jul 2020 01:35:21 +0000 (21:35 -0400)]
[OpenMP] Fix map clause for unused var: don't ignore it
For example, without this patch:
```
$ cat test.c
int main() {
int x[3];
#pragma omp target map(tofrom:x[0:3])
#ifdef USE
x[0] = 1
#endif
;
return 0;
}
$ clang -fopenmp -fopenmp-targets=nvptx64-nvidia-cuda -S -emit-llvm test.c
$ grep '^@.offload_maptypes' test.ll
$ echo $?
1
$ clang -fopenmp -fopenmp-targets=nvptx64-nvidia-cuda -S -emit-llvm test.c \
-DUSE
$ grep '^@.offload_maptypes' test.ll
@.offload_maptypes = private unnamed_addr constant [1 x i64] [i64 35]
```
With this patch, both greps produce the same result.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D83922
Vitaly Buka [Sat, 18 Jul 2020 01:29:47 +0000 (18:29 -0700)]
Corrected __libfuzzer_is_present symbol inclusion for MSVC x86 32-bit
The incorrect symbol will cause linking failures for 32-bit targets:
clang_rt.fuzzer-i386.lib(FuzzerDriver.obj) : error LNK2001: unresolved external symbol __libfuzzer_is_present
Verified no longer fails to link with this change for 32-bit and still succeeds for 64-bit MSVC.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D83594
Arthur Eubanks [Sat, 18 Jul 2020 00:49:46 +0000 (17:49 -0700)]
[NewPM][ASan] Make ASan tests work under NPM
Under NPM, the asan-globals-md analysis is required but cannot be run
within the asan function pass due to module analyses not being able to
run from a function pass. So this pins all tests using "-asan" to the
legacy PM and adds a corresponding RUN line with
-passes='require<asan-globals-md>,function(asan)'.
Now all tests in Instrumentation/AddressSanitizer pass when
-enable-new-pm is by default on.
Tests were automatically converted using the following python script and
failures were manually fixed up.
import sys
for i in sys.argv:
with open(i, 'r') as f:
s = f.read()
with open(i, 'w') as f:
for l in s.splitlines():
if "RUN:" in l and ' -asan -asan-module ' in l and '\\' not in l:
f.write(l.replace(' -asan -asan-module ', ' -asan -asan-module -enable-new-pm=0 '))
f.write('\n')
f.write(l.replace(' -asan -asan-module ', " -passes='require<asan-globals-md>,function(asan),module(asan-module)' "))
f.write('\n')
elif "RUN:" in l and ' -asan ' in l and '\\' not in l:
f.write(l.replace(' -asan ', ' -asan -enable-new-pm=0 '))
f.write('\n')
f.write(l.replace(' -asan ', " -passes='require<asan-globals-md>,function(asan)' "))
f.write('\n')
else:
f.write(l)
f.write('\n')
See https://bugs.llvm.org/show_bug.cgi?id=46611.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D83921
Lang Hames [Fri, 17 Jul 2020 21:56:51 +0000 (14:56 -0700)]
[llvm-jitlink] Add suppport for testing GOT entries and stubs for ELF.
This enables regression testing of GOT and stub handling with
llvm-jitlink.
Kuba Mracek [Sat, 18 Jul 2020 00:39:43 +0000 (17:39 -0700)]
[asan] Use dynamic shadow memory position on Apple Silicon macOS
This is needed because macOS on Apple Silicon has some reserved pages inside the "regular" shadow memory location, and mapping over that location fails.
Differential Revision: https://reviews.llvm.org/D82912
Logan Smith [Fri, 17 Jul 2020 03:39:45 +0000 (20:39 -0700)]
[clang][NFC] Add a missing 'override'
Logan Smith [Fri, 17 Jul 2020 03:38:41 +0000 (20:38 -0700)]
[llvm][NFC] Add missing 'override's
Logan Smith [Fri, 17 Jul 2020 03:36:46 +0000 (20:36 -0700)]
[llvm][NFC] Add missing 'override's in unittests/
Vitaly Buka [Sat, 18 Jul 2020 00:29:20 +0000 (17:29 -0700)]
[sanitizer] Another attempt to fix protoent test
Now we are going to pick name and index based on output of getprotoent_r.
peter klausler [Sat, 18 Jul 2020 00:02:47 +0000 (17:02 -0700)]
[flang] Prevent bogus runtime I/O error message
The runtime was requiring that STATUS='OLD' be explicitly specified
on an OPEN statement for a connected unit. There error should issue
only if a STATUS= other than 'OLD' is specified; an OPEN with no
STATUS= specifier is okay.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D84079
peter klausler [Fri, 17 Jul 2020 23:40:49 +0000 (16:40 -0700)]
[flang] Treat tabs like spaces in formatted input.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D84078
peter klausler [Fri, 17 Jul 2020 23:36:59 +0000 (16:36 -0700)]
[flang] Support <name>=<integer> syntax in compiler directives
Accept name=value as part of a !DIR$ compiler directive. These
are currently ignored in semantics, but we should recognize
more directive forms to facilitate testing. In due course,
these placeholding directive parsers will be replaced.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D84077
Arthur Eubanks [Fri, 17 Jul 2020 23:47:41 +0000 (16:47 -0700)]
Revert "[PGO][PGSO] Add profile guided size optimization to loop vectorization legality."
This reverts commit
30c382a7c6607a7d898730f8d288768110cdf1d2.
See https://crbug.com/
1106813.
Michele Scandale [Fri, 17 Jul 2020 23:43:05 +0000 (16:43 -0700)]
[CMake] Make `intrinsics_gen` dependency unconditional.
The `intrinsics_gen` target exists in the CMake exports since r309389
(see LLVMConfig.cmake.in), hence projects can depend on `intrinsics_gen`
even it they are built separately from LLVM.
Reviewed By: MaskRay, JDevlieghere
Differential Revision: https://reviews.llvm.org/D83454
Aditya Nandakumar [Fri, 17 Jul 2020 23:03:20 +0000 (16:03 -0700)]
[GISel: Add support for CSEing SrcOps which are immediates
https://reviews.llvm.org/D84072
Add G_EXTRACT to CSEConfigFull and add unit test as well.
Leonard Chan [Fri, 17 Jul 2020 22:29:50 +0000 (15:29 -0700)]
Revert "[AddressSanitizer] Don't use weak linkage for __{start,stop}_asan_globals"
This reverts commit
d76e62fdb7a93d9a33f642b6b528f2562cc3c3f4.
Reverting since this can lead to linker errors:
```
ld.lld: error: undefined hidden symbol: __start_asan_globals
```
when using --gc-sections. The linker can discard __start_asan_globals
once there are no more `asan_globals` sections left, which can lead to
this error if we have external linkages to them.
peter klausler [Fri, 17 Jul 2020 18:21:08 +0000 (11:21 -0700)]
[flang] Allow ! and // comments after some preprocessing directives
Old-style C /*comments*/ are omitted from preprocessor directive
token sequences by the prescanner, but line-ending C++ and Fortran
free-form comments are not since their handling might depend on
the directive. Add code to skip these line-ending comments as
appropriate in place of existing code that just skipped blanks.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D84061
AndreyChurbanov [Fri, 17 Jul 2020 22:15:37 +0000 (01:15 +0300)]
[OpenMP] libomp cleanup: check presence of hwloc objects CORE, PACKAGE
hwloc documentation guarantees the only object that is always present
in the topology is PU. We can check the presence of other objects
in the topology, just in case.
Differential Revision: https://reviews.llvm.org/D84065
Michael Spencer [Fri, 17 Jul 2020 22:12:18 +0000 (16:12 -0600)]
[clangd] Fix the build with clang <3.9.
In clang <3.9 the `unique_ptr` constructor that is supposed to allow
for Derived to Base conversion does not work. Remove this if we drop
support for such configurations.
This is the same fix as in
fda901a987ddd, and it updates the comments
to better reflect the actual issue. The same thing reproduces with
libc++ with older clangs.
Eric Christopher [Fri, 17 Jul 2020 22:05:42 +0000 (15:05 -0700)]
Temporarily Revert "[OpenMP] Add Additional Function Attribute Information to OMPKinds.def"
as it's causing a few unused variable warnings via the macro instantiation:
sources/llvm-project/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def:649:17: error: unused variable 'InaccessibleOnlyAttrs' [-Werror,-Wunused-variable]
__OMP_ATTRS_SET(InaccessibleOnlyAttrs,
^
This reverts commit
09fe0c5ab9ca5846d6cb625e9be47fdcfe49b93c.
Eric Christopher [Fri, 17 Jul 2020 21:58:01 +0000 (14:58 -0700)]
Temporarily Revert "[InlineAdvisor] New inliner advisor to replay inlining from optimization remarks"
as it is failing the inline-replay.ll test as well as sanitizers/Werror
from returning a stack local variable.
This reverts commit
029946b112684c27b27f7c2d7554f22b33ae1e0b.
Joseph Huber [Fri, 17 Jul 2020 21:54:01 +0000 (17:54 -0400)]
[OpenMP] Add Additional Function Attribute Information to OMPKinds.def
Summary:
This patch adds more function attribute information to the runtime function definitions in OMPKinds.def. The goal is to provide sufficient information about OpenMP runtime functions to perform more optimizations on OpenMP code.
Reviewers: jdoerfert
Subscribers: aaron.ballman cfe-commits yaxunl guansong sstefan1 llvm-commits
Tags: #OpenMP #clang #llvm
Differential Revision: https://reviews.llvm.org/D81031
Teresa Johnson [Fri, 17 Jul 2020 21:48:28 +0000 (14:48 -0700)]
[compiler-rt][asan][hwasan] Fix Fuchsia build
Fix build failure in Fuchsia build from refactoring in
5d2be1a18845c528d3e86f7efcc59872e4a757c3
Guard the moved versions of ReserveShadowMemoryRange and ProtectGap
the same way they were in the asan code originally (not for Fuchsia or
RTEMS). Otherwise we end up with unsats as they invoke functions not
defined there.
peter klausler [Fri, 17 Jul 2020 18:24:29 +0000 (11:24 -0700)]
[flang] Rework read/write permission management for runtime file opening
Anonymous Fortran unit files (e.g., "./fort.7") need to be created
O_RDWR so that they can be written, rewound, and read. Other
files opened with no ACTION= specifier need to set read/write
permissions based on the file, if it exists.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D84063
Sjoerd Meijer [Fri, 17 Jul 2020 21:40:41 +0000 (22:40 +0100)]
Remove clang matrix lowering test for now as it is still failing under the NPM.
Wenlei He [Tue, 14 Jul 2020 05:12:28 +0000 (22:12 -0700)]
[InlineAdvisor] New inliner advisor to replay inlining from optimization remarks
Summary:
This change added a new inline advisor that takes optimization remarks for previous inlining as input, and provide the decision as advice so current inlining can replay inline decision of a different compilation. Dwarf inline stack with line and discriminator is used as anchor for call sites. The change can be useful for Inliner tuning.
A switch -sample-profile-inline-replay=<inline_remarks_file> is added to hook up the new inliner advisor with SampleProfileLoader's inline decision for replay. The new inline advisor can also be used by regular CGSCC inliner later if needed.
Reviewers: davidxl, mtrofin, wmi, hoy
Subscribers: aprantl, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83743
Xiangling Liao [Thu, 16 Jul 2020 18:29:13 +0000 (14:29 -0400)]
[AIX] report_fatal_error on `-fregister_global_dtors_with_atexit` for static init
On AIX, the semantic of global_dtors contains __sterm functions associated with C++
cleanup actions and user-declared __attribute__((destructor)) functions. We should
never merely register __sterm with atexit(), so currently
-fregister_global_dtors_with_atexit does not work well on AIX: It would cause
finalization actions to not occur when unloading shared libraries. We need to figure
out a way to handle that when we start supporting user-declared
__attribute__((destructor)) functions.
Currently we report_fatal_error on this option temporarily.
Differential Revision: https://reviews.llvm.org/D83974
Mitch Phillips [Fri, 17 Jul 2020 19:59:09 +0000 (12:59 -0700)]
[GWP-ASan] Split the unwinder into segv/non-segv.
Summary:
Splits the unwinder into a non-segv (for allocation/deallocation traces) and a
segv unwinder. This ensures that implementations can select an accurate, slower
unwinder in the segv handler (if they choose to use the GWP-ASan provided one).
This is important as fast frame-pointer unwinders (like the sanitizer unwinder)
don't like unwinding through signal handlers.
Reviewers: morehouse, cryptoad
Reviewed By: morehouse, cryptoad
Subscribers: cryptoad, mgorny, eugenis, pcc, #sanitizers
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D83994
Xinan Jiang [Fri, 17 Jul 2020 19:57:08 +0000 (20:57 +0100)]
[InstCombine] Fix typo in comment.
Reviewers: fhahn
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D83951
Roman Lebedev [Fri, 17 Jul 2020 19:35:21 +0000 (22:35 +0300)]
[InstCombine] Fold X sdiv (-1 << C) -> -(X u>> Y) iff X is non-negative
This is the one i'm seeing as missed optimization,
although there are likely other possibilities, as usual.
There are 4 variants of a general sdiv->udiv fold:
https://rise4fun.com/Alive/VS6
Name: v0
Pre: C0 >= 0 && C1 >= 0
%r = sdiv i8 C0, C1
=>
%r = udiv i8 C0, C1
Name: v1
Pre: C0 <= 0 && C1 >= 0
%r = sdiv i8 C0, C1
=>
%t0 = udiv i8 -C0, C1
%r = sub i8 0, %t0
Name: v2
Pre: C0 >= 0 && C1 <= 0
%r = sdiv i8 C0, C1
=>
%t0 = udiv i8 C0, -C1
%r = sub i8 0, %t0
Name: v3
Pre: C0 <= 0 && C1 <= 0
%r = sdiv i8 C0, C1
=>
%r = udiv i8 -C0, -C1
If we really don't like sdiv (more than udiv that is),
and are okay with increasing instruction count (2 new negations),
and we ensure that we don't undo the fold,
then we could just implement these..
Roman Lebedev [Fri, 17 Jul 2020 19:28:36 +0000 (22:28 +0300)]
[NFC][InstCombine] Add some tests with sdiv-by-negative-power-of-two
peter klausler [Fri, 17 Jul 2020 18:27:40 +0000 (11:27 -0700)]
[flang] Add entry points for internal scalar list-directed I/O
BeginInternalListInput and BeginInternalListOutput were missing
from the I/O API implementation; add them.
Reviewed By: PeteSteinfeld
Differential Revision: https://reviews.llvm.org/D84066
Michael Spencer [Fri, 17 Jul 2020 19:32:12 +0000 (13:32 -0600)]
[Clang] Fix building with Clang < 3.9.
This is a workaround for a bug in older versions of Clang when. The
constructor that is supposed to allow for Derived to Base conversion
does not work. Remove this if we drop support for such configurations.