Matt Arsenault [Thu, 5 Dec 2019 09:37:27 +0000 (15:07 +0530)]
Set some fast math attributes in setFunctionAttributes
This will provide a more consistent view to codegen for these
attributes. The current system is somewhat awkward, and the fields in
TargetOptions are reset based on the command line flag if the
attribute isn't set. By forcing these attributes with the flag, there
can never be an inconsistency in the behavior if code directly
inspects the attribute on the function without considering the command
line flags.
Jonas Devlieghere [Thu, 16 Jan 2020 03:12:50 +0000 (19:12 -0800)]
[lldb/Reproducers] Extract function for reading environment override (NFC)
Create a helper function for reading reproducer overrides from
environment variables.
Richard Smith [Thu, 16 Jan 2020 02:37:32 +0000 (18:37 -0800)]
PR42694 Support explicit(bool) in older language modes as an extension.
This needs somewhat careful disambiguation, as C++2a explicit(bool) is a
breaking change. We only enable it in cases where the source construct
could not possibly be anything else.
Wei Mi [Thu, 16 Jan 2020 02:23:36 +0000 (18:23 -0800)]
[SampleFDO] Fix invalid branch profile generated by indirect call promotion.
Suppose an inline instance has hot total sample count but 0 entry count, and
it is an indirect call target. If the indirect call has no other call target
and inline instance associated with it and it is promoted, currently the
conditional branch generated by indirect call promotion will have invalid
branch profile which is !{!"branch_weights", i32 0, i32 0} -- because the
entry count of the promoted target is 0 and the total entry count of all
targets is also 0. This caused a SEGV in Control Height Reduction and may
cause problem in other passes.
Function entry count of an inline instance is computed by a heuristic --
using either the sample of the starting line or starting inner inline
instance. The patch changes the heuristic a little bit so that when total
sample count is larger than 0, the computed entry count will be at least 1.
Then the new branch profile will be !{!"branch_weights", i32 1, i32 0}.
Differential Revision: https://reviews.llvm.org/D72790
Craig Topper [Thu, 16 Jan 2020 02:07:15 +0000 (18:07 -0800)]
[X86] When handling i64->f32 sint_to_fp on 32-bit targets only bitcast to f64 if sse2 is enabled.
The code is trying to copy the i64 value to an xmm register to
use a 64-bit store so that the 64-bit fild can benefit from
store forwarding.
But this trick only works if f64 is going to be stored in an
XMM register. If we only have SSE1 then only float is in xmm
register. So this trick just causes 2 stores i32 stores, an f64
load into the x87, an f64 from x87, and a 64-bit fild. So we end
up with an extra stack temporary and still didn't get store forwarding.
We might be able to use v2f32 here instead, but I didn't check. I
just wanted the code to make sense.
Found by inspection as I continue to stare too hard at our
int_to_fp conversions.
Craig Topper [Thu, 16 Jan 2020 02:04:46 +0000 (18:04 -0800)]
[X86] Add 32-bit mode sse1 command line to scalar-int-to-fp.ll. NFC
Jim Ingham [Thu, 16 Jan 2020 02:10:31 +0000 (18:10 -0800)]
Fix the macos build after D71575.
size_t and uint64_t are spelled slightly differently on macOS, which was
causing the compiler to error out calling std::min - since the two types have
to be the same.
I fixed this by casting the uint64_t computation to a size_t. That's probably
not the cleanest solution, but it gets us back to building.
Matt Arsenault [Mon, 9 Dec 2019 13:09:58 +0000 (18:39 +0530)]
llc: Don't overwrite frame-pointer attribute
Continue making command line flags with matching attribute behavior
consistent.
Yuanfang Chen [Thu, 16 Jan 2020 01:46:18 +0000 (17:46 -0800)]
Revert "[Support] make report_fatal_error `abort` instead of `exit`"
This reverts commit
647c3f4e47de8a850ffcaa897db68702d8d2459a.
Got bots failure from sanitizer-windows and maybe others.
Vedant Kumar [Thu, 16 Jan 2020 00:57:45 +0000 (16:57 -0800)]
debugserver: Cut dependency on intrinsics_gen
debugserver does not depend on intrinsics_gen or on llvm.
Yuanfang Chen [Wed, 15 Jan 2020 06:43:51 +0000 (22:43 -0800)]
[Support] make report_fatal_error `abort` instead of `exit`
Summary:
This patch could be treated as a rebase of D33960. It also fixes PR35547.
A fix for `llvm/test/Other/close-stderr.ll` is proposed in D68164. Seems
the consensus is that the test is passing by chance and I'm not
sure how important it is for us. So it is removed like in D33960 for now.
The rest of the test fixes are just adding `--crash` flag to `not` tool.
** The reason it fixes PR35547 is
`exit` does cleanup including calling class destructor whereas `abort`
does not do any cleanup. In multithreading environment such as ThinLTO or JIT,
threads may share states which mostly are ManagedStatic<>. If faulting thread
tearing down a class when another thread is using it, there are chances of
memory corruption. This is bad 1. It will stop error reporting like pretty
stack printer; 2. The memory corruption is distracting and nondeterministic in
terms of error message, and corruption type (depending one the timing, it
could be double free, heap free after use, etc.).
Reviewers: rnk, chandlerc, zturner, sepavloff, MaskRay, espindola
Reviewed By: rnk, MaskRay
Subscribers: wuzish, jholewinski, qcolombet, dschuff, jyknight, emaste, sdardis, nemanjai, jvesely, nhaehnle, sbc100, arichardson, jgravelle-google, aheejin, kbarton, fedor.sergeev, asb, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, jrtc27, zzheng, edward-jones, atanasyan, rogfer01, MartinMosbeck, brucehoult, the_o, PkmX, jocewei, jsji, lenary, s.egerton, pzheng, cfe-commits, MaskRay, filcab, davide, MatzeB, mehdi_amini, hiraditya, steven_wu, dexonsmith, rupprecht, seiya, llvm-commits
Tags: #llvm, #clang
Differential Revision: https://reviews.llvm.org/D67847
Jonas Devlieghere [Thu, 16 Jan 2020 00:49:33 +0000 (16:49 -0800)]
[lldb/Tools] Remove lldb-mi.exports
lldb-mi was removed from the repo a while ago.
Matt Arsenault [Mon, 9 Dec 2019 12:33:25 +0000 (18:03 +0530)]
llc: Change behavior of -mattr with existing attribute
Append this to the existing target-features attribute on the function.
Some flags ignore existing attributes, and some overwrite them. Move
towards consistently respecting existing attributes if present. Since
target features act as a state machine on their own, append to the
function attribute. The backend default added feature list, function
attributes, and -mattr will all be appended together, and the later
features can individually toggle the earlier settings.
Paolo Severini [Wed, 15 Jan 2020 23:29:24 +0000 (15:29 -0800)]
[LLDB] Add ObjectFileWasm plugin for WebAssembly debugging
Summary:
This is the first in a series of patches to enable LLDB debugging of
WebAssembly targets.
Current versions of Clang emit (partial) DWARF debug information in WebAssembly
modules and we can leverage this debug information to give LLDB the ability to
do source-level debugging of Wasm code that runs in a WebAssembly engine.
A way to do this could be to use the remote debugging functionalities provided
by LLDB via the GDB-remote protocol. Remote debugging can indeed be useful not
only to connect a debugger to a process running on a remote machine, but also to
connect the debugger to a managed VM or script engine that runs locally,
provided that the engine implements a GDB-remote stub that offers the ability to
access the engine runtime internal state.
To make this work, the GDB-remote protocol would need to be extended with a few
Wasm-specific custom query commands, used to access aspects of the Wasm engine
state (like the Wasm memory, Wasm local and global variables, and so on).
Furthermore, the DWARF format would need to be enriched with a few Wasm-specific
extensions, here detailed: https://yurydelendik.github.io/webassembly-dwarf.
This CL introduce classes **ObjectFileWasm**, a file plugin to represent a Wasm
module loaded in a debuggee process. It knows how to parse Wasm modules and
store the Code section and the DWARF-specific sections.
Reviewers: jasonmolenda, clayborg, labath
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D71575
Richard Smith [Wed, 15 Jan 2020 21:14:13 +0000 (13:14 -0800)]
Fix pack deduction to only deduce the arity of packs that are actually
expanded by the deduced pack.
We recently started also deducing the arity of separately-expanded packs
that are merely mentioned within the pack in question, which is
incorrect.
Jonas Devlieghere [Thu, 16 Jan 2020 00:01:00 +0000 (16:01 -0800)]
[lldb/Utils] Patch all variables used by lit (3/3)
Instead of passing all the arguments for dotest.py as a single CMake
variable, lit now uses separate variables for the different test
binaries. Before this change they'd all get patched as part of the
LLDB_DOTEST_ARGS. We need to patch the new variables as well.
Jonas Devlieghere [Wed, 15 Jan 2020 23:55:51 +0000 (15:55 -0800)]
[lldb/Utils] Patch all variables used by lldb-dotest (2/2)
Instead of passing all the arguments for dotest.py as a single CMake
variable, lldb-dotest now uses separate variables for the different test
binaries. Before this change they'd all get patched as part of the
LLDB_DOTEST_ARGS. We need to patch the new variables as well.
Stanislav Mekhanoshin [Wed, 15 Jan 2020 17:38:08 +0000 (09:38 -0800)]
Process BUNDLE in tail duplication
When tail duplication estimates a size of tail it uses instruction
count. Account for a number of instrictions in a bundle too.
Differential Revision: https://reviews.llvm.org/D72783
Amy Huang [Wed, 15 Jan 2020 23:27:58 +0000 (15:27 -0800)]
Revert "Further implement CWG 2292"
This reverts commit
ee0f1f1edc3ec0d4e698d50cc3180217448802b7 because it
causes an error on valid code.
See https://reviews.llvm.org/rGee0f1f1edc3ec0d4e698d50cc3180217448802b7.
Alexey Bataev [Wed, 15 Jan 2020 22:37:12 +0000 (17:37 -0500)]
[OPENMP]Use regular processing of vtable used when TU is a prefix.
If current kind of the translation unit is TU_Prefix and it is not
complete, cannot decide what to do with virtual members/table at that
time, need to delay it to later stages.
Vedant Kumar [Wed, 15 Jan 2020 19:26:34 +0000 (11:26 -0800)]
[CodeExtractor] Transfer debug info to extracted function
After extracting, fix up debug info in both the old and new functions by
1) Pointing line locations and debug intrinsics to the new subprogram
scope, and
2) Deleting intrinsics which point to values outside of the new
function.
Depends on https://reviews.llvm.org/D72795.
Testing: check-llvm, check-clang, a build of LNT in the `-Os -g` config
with "-mllvm -hot-cold-split=1" set, and end-to-end debugging of a toy
program which undergoes splitting to verify that lldb can find
variables, single step, etc. in extracted code.
rdar://
45507940
Differential Revision: https://reviews.llvm.org/D72801
Matt Arsenault [Sun, 5 Jan 2020 15:51:54 +0000 (10:51 -0500)]
AMDGPU/GlobalISel: Select exp with patterns
This does produce slightly different code. Now a unique IMPLICIT_DEF
is emitted for each of the implicit_def operands, rather than reusing
the same one.
Matt Arsenault [Sun, 5 Jan 2020 00:49:17 +0000 (19:49 -0500)]
AMDGPU: Remove custom node for exports
I'm mildly worried about potentially reordering exp/exp_done with
IntrWriteMem on the intrinsic.
Requires hacking out the illegal type on SI, so manually select that
case during lowering.
Matt Arsenault [Fri, 10 Jan 2020 15:07:24 +0000 (10:07 -0500)]
GlobalISel: Handle more cases of G_SEXT narrowing
This now develops the same problem G_ZEXT/G_ANYEXT have where the
requested type is assumed to be the source type. This will be fixed
separately by creating intermediate merges.
Brian Gesiak [Wed, 15 Jan 2020 21:39:39 +0000 (16:39 -0500)]
[IR] Module's NamedMD table needn't be 'void *'
Summary:
In July 21 2010 `llvm::NamedMDNode` was refactored such that it would no
longer subclass `llvm::Value`:
https://github.com/llvm/llvm-project/commit/
2637cc1a38d7336ea30caf
As part of this change, a map type from metadata names to their named
metadata, `llvm::MDSymbolTable`, was deleted. In its place, the type
of member `llvm::Module::NamedMDSymTab` was changed, from
`llvm::MDSymbolTable` to `void *`. The underlying memory allocations
for this pointer were changed to `new StringMap<NamedMDNode *>()`.
However, as far as I can tell, there's no need for obscuring the
underlying type being pointed to by the `void *`, and no need for
static casts from `void *` to `StringMap`. In fact, I don't think
there's a need for explicit calls to `new` and `delete` at all.
This commit changes `NamedMDSymTab` from a pointer to a reference, which
automatically couples its lifetime with the lifetime of its owning
`llvm::Module` instance, thus removing the explicit calls to `new` and
`delete` in the `llvm::Module` constructor and destructor. It also
changes the type from `void *` to a newly defined `NamedMDSymTabType`,
and removes the static casts.
Test Plan:
An ASAN-enabled build and run of `check-all` succeeds with this change
(aside from some tests that always fail for me in ASAN for some reason,
such as `check-clang` `SemaTemplate/stack-exhaustion.cpp`).
Reviewers: aprantl, dblaikie, chandlerc, pcc, echristo
Reviewed By: dblaikie
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72812
Jonas Devlieghere [Wed, 15 Jan 2020 23:15:51 +0000 (15:15 -0800)]
[lldb/Utils] Patch all variables used by lldb-dotest
Instead of passing all the arguments for dotest.py as a single CMake
variable, lldb-dotest now uses separate variables for the different test
binaries. Before this change they'd all get patched as part of the
LLDB_DOTEST_ARGS. We need to patch the new variables as well.
Petr Hosek [Wed, 15 Jan 2020 21:58:29 +0000 (13:58 -0800)]
[libcxx] Use mtx_plain | mtx_recursive following C11 API
The C11 API specifies that to initialize a recursive mutex,
mtx_plain | mtx_recursive should be used with mtx_init.
Differential Revision: https://reviews.llvm.org/D72809
Reid Kleckner [Wed, 15 Jan 2020 23:09:25 +0000 (15:09 -0800)]
[COFF] Warn that LLD does not support /PDBSTRIPPED:
Doesn't really fix PR44491, but it avoids treating it as an input.
Amy Huang [Wed, 15 Jan 2020 22:34:19 +0000 (14:34 -0800)]
Revert "Allow system header to provide their own implementation of some builtin"
This reverts commit
921f871ac438175ca8fcfcafdfcfac4d7ddf3905 because it
causes libc++ code to trigger __warn_memset_zero_len.
See https://reviews.llvm.org/D71082.
Jonas Devlieghere [Wed, 15 Jan 2020 22:56:28 +0000 (14:56 -0800)]
[lldb/Debugger] Rename IO handler methods to be more meaningful (NFC)
Make it clear form the method names whether they are synchronous or
asynchronous.
Alexey Bataev [Wed, 15 Jan 2020 22:41:15 +0000 (17:41 -0500)]
Revert "[OPENMP]Do not use RTTI by default for NVPTX devices."
This reverts commit
23058f9dd4d7e18239fd63b6da52549514b45fda. It breaks
builds of cuda code somehow in some cases.
Eric Fiselier [Wed, 15 Jan 2020 22:29:55 +0000 (17:29 -0500)]
[libc++] Fix parsing <string> in C++03.
Specifically, add a space between >> when closing templates.
Eric Fiselier [Wed, 15 Jan 2020 22:27:10 +0000 (17:27 -0500)]
[libc++] Optimize basic_string::operator=(const basic_string&) for SSO assignments
This change optimizes the operator=() assignment for short strings by direcly
copying the raw data from the source into the current instance. This creates an
optimized / inlined mempcy up to over 2X faster for short string assignments.
With inlining enabled for operator=, performance is up to 6X faster.
Benchmarks 'as is':
name old time/op new time/op delta
BM_StringAssignStr_Empty_Opaque 6.05ns ± 2% 3.59ns ± 0% -40.67%
BM_StringAssignStr_Empty_Transparent 5.15ns ± 0% 3.08ns ± 0% -40.12%
BM_StringAssignStr_Small_Opaque 7.71ns ± 0% 3.59ns ± 0% -53.45%
BM_StringAssignStr_Small_Transparent 7.66ns ± 0% 3.09ns ± 0% -59.66%
BM_StringAssignStr_Large_Opaque 24.1ns ± 0% 24.9ns ± 0% +3.22%
BM_StringAssignStr_Large_Transparent 22.2ns ± 0% 22.8ns ± 0% +2.77%
BM_StringAssignStr_Huge_Opaque 315ns ± 6% 320ns ± 5% ~
BM_StringAssignStr_Huge_Transparent 318ns ± 5% 321ns ± 4% ~
Benchmarks with partial inlining operator=():
name old time/op new time/op delta
BM_StringAssignStr_Empty_Opaque 5.94ns ± 2% 1.95ns ± 0% -67.21%
BM_StringAssignStr_Empty_Transparent 5.14ns ± 0% 1.04ns ± 1% -79.73%
BM_StringAssignStr_Small_Opaque 7.69ns ± 0% 1.96ns ± 0% -74.48%
BM_StringAssignStr_Small_Transparent 7.65ns ± 0% 1.04ns ± 0% -86.40%
BM_StringAssignStr_Large_Opaque 24.1ns ± 0% 24.5ns ± 0% +1.61%
BM_StringAssignStr_Large_Transparent 22.2ns ± 0% 21.1ns ± 0% -4.70%
BM_StringAssignStr_Huge_Opaque 317ns ± 5% 323ns ± 4% ~
BM_StringAssignStr_Huge_Transparent 318ns ± 5% 320ns ± 5% ~
Patch by Martijn Vels (mvels@google.com)
Reviewed as https://reviews.llvm.org/D72704
Vedant Kumar [Wed, 15 Jan 2020 22:19:53 +0000 (14:19 -0800)]
[test] Move call-site-entry-linking.test into test/tools/dsymutil/X86
This should fix a failure on the clang-cmake-armv7-quick bot.
Vedant Kumar [Wed, 15 Jan 2020 22:15:45 +0000 (14:15 -0800)]
DWARF: Simplify the way the return PC is attached to call site tags, NFC
This cleanup was suggested by Djordje in D72489.
Fedor Sergeev [Wed, 15 Jan 2020 21:57:34 +0000 (00:57 +0300)]
[BasicBlock] add helper getPostdominatingDeoptimizeCall
It appears to be rather useful when analyzing Loops with multiple
deoptimizing exits, perhaps merged ones.
For now it is used in LoopPredication, will be adding more uses
in other loop passes.
Reviewers: asbirlea, fhahn, skatkov, spatel, reames
Reviewed By: reames
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72754
Eric Fiselier [Wed, 15 Jan 2020 22:12:09 +0000 (17:12 -0500)]
[libc++] Explicitly enumerate std::string external instantiations - Attempt 2
The GCC build failures have been addressed, and the LLDB failures were
fixed by LLDB.
I have also verified that the apple-clang 9.0 segfault no longer
occurs.
Original Message:
The external instantiation of std::string is a problem for libc++.
Additions and removals of inline functions in string can cause ABI
breakages, including introducing new symbols.
This patch aims to:
(1) Make clear which functions are explicitly instatiated.
(2) Prevent new functions from being accidentally instantiated.
(3) Allow a migration path for adding or removing functions from the
explicit instantiation over time.
Although this new formulation is uglier, it is preferable from a
maintainability and readability standpoint because it explicitly
enumerates the functions we've chosen to expose in our ABI. Changing
this list is non-trivial and requires thought and planning.
(3) is achieved by making it possible to control the extern template declaration
separately from it's definition. Meaning we could add a new definition to
the dylib, wait for it to roll out, then add the extern template
declaration to the header. Similarly, we could remove existing extern
template declarations while still keeping the definition to prevent ABI
breakages.
Eric Fiselier [Wed, 15 Jan 2020 22:02:17 +0000 (17:02 -0500)]
[libc++] Explicitly mark basic_string<...>::npos with default
visibility.
This ensures that the version compiled into the library isn't
accidentally hidden.
Eric Fiselier [Wed, 15 Jan 2020 21:57:08 +0000 (16:57 -0500)]
[libc++] Make SFINAE'd member functions in string mutually exclusive.
This patch is needed in order to work around a GCC bug that fails to
explicitly instantiate a non-template function of a class template when
there is another overload that's a function template.
(See https://godbolt.org/z/4bUQ_b)
This patch SFINAE's away the function templates when the argument is
a basic_string.
Vedant Kumar [Wed, 15 Jan 2020 21:59:12 +0000 (13:59 -0800)]
lldb: Run TestCrossDSOTailCalls.py and TestCrossObjectTailCalls.py on Darwin only
See https://bugs.llvm.org/show_bug.cgi?id=44561, these tests are failing
on an aarch64/Linux bot:
http://lab.llvm.org:8011/builders/lldb-aarch64-ubuntu/builds/655
For some reason the backtrace the tests are expecting to find is
incomplete.
Jinsong Ji [Wed, 15 Jan 2020 21:19:34 +0000 (21:19 +0000)]
[MachineScheduler][NFC] Don't swap when we can't cluster
https://reviews.llvm.org/D72706 tried to reduce reordering due to mem op
clustering. This patch avoid doing the swap when we can't cluster.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D72800
Mircea Trofin [Wed, 15 Jan 2020 21:33:58 +0000 (13:33 -0800)]
[NFC] Refactor InlineResult for readability
Summary:
InlineResult is used both in APIs assessing whether a call site is
inlinable (e.g. llvm::isInlineViable) as well as in the function
inlining utility (llvm::InlineFunction). It means slightly different
things (can/should inlining happen, vs did it happen), and the
implicit casting may introduce ambiguity (casting from 'false' in
InlineFunction will default a message about hight costs,
which is incorrect here).
The change renames the type to a more generic name, and disables
implicit constructors.
Reviewers: eraman, davidxl
Reviewed By: davidxl
Subscribers: kerbowa, arsenm, jvesely, nhaehnle, eraman, hiraditya, haicheng, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72744
Jacques Pienaar [Thu, 9 Jan 2020 02:48:38 +0000 (18:48 -0800)]
[mlir] Add shaped container component type interface
Summary:
* Add shaped container type interface which allows infering the shape, element
type and attribute of shaped container type separately. Show usage by way of
tensor type inference trait which combines the shape & element type in
infering a tensor type;
- All components need not be specified;
- Attribute is added to allow for layout attribute that was previously
discussed;
* Expand the test driver to make it easier to test new creation instances
(adding new operands or ops with attributes or regions would trigger build
functions/type inference methods);
- The verification part will be moved out of the test and to verify method
instead of ops implementing the type inference interface in a follow up;
* Add MLIRContext as arg to possible to create type for ops without arguments,
region or location;
* Also move out the section in OpDefinitions doc to separate ShapeInference doc
where the shape function requirements can be captured;
- Part of this would move to the shape dialect and/or shape dialect ops be
included as subsection of this doc;
* Update ODS's variable usage to match camelBack format for builder,
state and arg variables;
- I could have split this out, but I had to make some changes around
these and the inconsistency bugged me :)
Differential Revision: https://reviews.llvm.org/D72432
Zhongduo Lin [Wed, 15 Jan 2020 21:22:47 +0000 (16:22 -0500)]
[NFC][IndVarSimplify] remove duplicate code in widenWithVariantLoadUseCodegen.
Summary: Duplicate code in widenWithVariantLoadUseCodegen is removed and also use assert to check unknown extension type as it should be filtered out by the pre condition check before calling this function.
Reviewers: az, sanjoy, sebpop, efriedma, javed.absar, sanjoy.google
Reviewed By: efriedma
Subscribers: hiraditya, llvm-commits, amehsan
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72652
Richard Smith [Thu, 9 May 2019 06:30:37 +0000 (23:30 -0700)]
PR17164: Change clang's default behavior from -flax-vector-conversions=all to -flax-vector-conversions=integer.
Summary:
See proposal on cfe-dev:
http://lists.llvm.org/pipermail/cfe-dev/2019-April/062030.html
Reviewers: SjoerdMeijer, eli.friedman
Subscribers: kristof.beyls, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D67678
Richard Smith [Tue, 17 Sep 2019 03:36:55 +0000 (20:36 -0700)]
Work around PR43337: don't try to use the vec_sel overloads for vector long long, since clang's <altivec.h> doesn't provide it yet!
Vedant Kumar [Wed, 15 Jan 2020 19:22:06 +0000 (11:22 -0800)]
DebugInfo: Factor out logic to update locations in MD_loop metadata, NFC
Factor out the logic needed to update debug locations contained within
MD_loop metadata.
This refactor is preparation for a future change that also needs to
rewrite MD_loop metadata.
rdar://
45507940
Vedant Kumar [Fri, 10 Jan 2020 02:00:33 +0000 (18:00 -0800)]
[DWARF] Emit DW_AT_call_return_pc as an address
This reverts D53469, which changed llvm's DWARF emission to emit
DW_AT_call_return_pc as a function-local offset. Such an encoding is not
compatible with post-link block re-ordering tools and isn't standards-
compliant.
In addition to reverting back to the original DW_AT_call_return_pc
encoding, teach lldb how to fix up DW_AT_call_return_pc when the address
comes from an object file pointed-to by a debug map. While doing this I
noticed that lldb's support for tail calls that cross a DSO/object file
boundary wasn't covered, so I added tests for that. This latter case
exercises the newly added return PC fixup.
The dsymutil changes in this patch were originally included in D49887:
the associated test should be sufficient to test DW_AT_call_return_pc
encoding purely on the llvm side.
Differential Revision: https://reviews.llvm.org/D72489
Lang Hames [Wed, 15 Jan 2020 19:30:04 +0000 (11:30 -0800)]
[docs][ORC] Update the laziness section of the ORCv2 design doc.
This updates the discussion of lazy reexports, fixes a TBD for a usage example,
and adds a reference to the fully worked lazy reexports example that was added
in
e9e26c01cd865da678b1af6ba5f417c713956a66.
Nico Weber [Wed, 15 Jan 2020 18:35:58 +0000 (13:35 -0500)]
[gn build] re-run "gn format" with trunk gn
Nico Weber [Wed, 15 Jan 2020 18:30:57 +0000 (13:30 -0500)]
[gn build] add multi-line forcing comments in more places
Nico Weber [Wed, 15 Jan 2020 18:24:28 +0000 (13:24 -0500)]
[gn build] make "gn format" comment slightly more concise
Craig Topper [Wed, 15 Jan 2020 02:19:54 +0000 (18:19 -0800)]
[Mips] Add FileCheck to a test that just tested for a crash.
I believe the generated code here can suffer from double rounding.
So I wanted to capture the existing codegen so we can make
decisions about how to fix it.
Lang Hames [Wed, 15 Jan 2020 18:20:10 +0000 (10:20 -0800)]
[ORC] Set setCloneToNewContextOnEmit on LLJIT's transform layer when needed.
Based on Don Hinton's patch in https://reviews.llvm.org/D72406. This feature
was accidentally left out of
e9e26c01cd865da678b1af6ba5f417c713956a66, and
would have pessimized concurrent compilation in the default case.
Thanks for spotting this Don!
Amara Emerson [Wed, 15 Jan 2020 16:49:22 +0000 (08:49 -0800)]
Revert "Revert rG6078f2fedcac5797ac39ee5ef3fd7a35ef1202d5 - "[AArch64][GlobalISel]: Support @llvm.{return,frame}address selection.""
The original change wasn't constraining the operand regclasses which broke EXPENSIVE_CHECKS.
Nico Weber [Wed, 15 Jan 2020 17:59:45 +0000 (12:59 -0500)]
[gn build] Reformat all build files
Ran `git ls-files '*.gn' '*.gni' | xargs llvm/utils/gn/gn.py format`.
The motivation is to reformat them with trunk gn again right after.
Trunk gn changed formatting of some single-element lists.
Nico Weber [Wed, 15 Jan 2020 15:45:02 +0000 (10:45 -0500)]
Replace CLANG_SPAWN_CC1 env var with a driver mode flag
Flags are clang's default UI is flags.
We can have an env var in addition to that, but in D69825 nobody has yet
mentioned why this needs an env var, so omit it for now. If someone
needs to set the flag via env var, the existing CCC_OVERRIDE_OPTIONS
mechanism works for it (set CCC_OVERRIDE_OPTIONS=+-fno-integrated-cc1
for example).
Also mention the cc1-in-process change in the release notes.
Also spruce up the test a bit so it actually tests something :)
Differential Revision: https://reviews.llvm.org/D72769
Mark Murray [Mon, 13 Jan 2020 13:21:30 +0000 (13:21 +0000)]
[ARM][MVE][Intrinsics] Add VMINAQ, VMINNMAQ, VMAXAQ, VMAXNMAQ intrinsics.
Summary: Add VMINAQ, VMINNMAQ, VMAXAQ, VMAXNMAQ intrinsics and unit tests.
Reviewers: simon_tatham, miyuki, dmgreen
Subscribers: kristof.beyls, hiraditya, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D72761
Kadir Cetinkaya [Wed, 15 Jan 2020 16:59:02 +0000 (17:59 +0100)]
[clangd] Extract string literals in macro arguments to unbreak gcc buildbots
Benjamin Kramer [Wed, 15 Jan 2020 16:51:51 +0000 (17:51 +0100)]
Revert "[mlir] Create a gpu.module operation for the GPU Dialect."
This reverts commit
4624a1e8ac8a3f69cc887403b976f538f587744a. Causing
problems downstream.
Teresa Johnson [Wed, 15 Jan 2020 16:35:12 +0000 (08:35 -0800)]
Fix bot by adjusting wildcard matching
I noticed one bot failure due to
24a00ef2404104e9ca6fbd7eb523a8a340be9d99 because the wildcard matching
was not working as intended, fixed it to act similar to other checks of
CGSCCToFunctionPassAdaptor.
evgeny [Wed, 15 Jan 2020 16:29:01 +0000 (19:29 +0300)]
[ThinLTO] Always import constants
This patch imports constant variables even when they can't be internalized
(which results in promotion). This offers some extra constant folding
opportunities.
Differential revision: https://reviews.llvm.org/D70404
Arkady Shlykov [Wed, 15 Jan 2020 14:12:34 +0000 (06:12 -0800)]
[Loop Peeling] Add possibility to enable peeling on loop nests.
Summary:
Current peeling implementation bails out in case of loop nests.
The patch introduces a field in TargetTransformInfo structure that
certain targets can use to relax the constraints if it's
profitable (disabled by default).
Also additional option is added to enable peeling manually for
experimenting and testing purposes.
Reviewers: fhahn, lebedev.ri, xbolva00
Reviewed By: xbolva00
Subscribers: xbolva00, hiraditya, zzheng, llvm-commits
Differential Revision: https://reviews.llvm.org/D70304
Sanjay Patel [Wed, 15 Jan 2020 13:23:46 +0000 (08:23 -0500)]
[InstCombine] reassociate fsub+fsub into fsub+fadd
As discussed in the motivating PR44509:
https://bugs.llvm.org/show_bug.cgi?id=44509
...we can end up with worse code using fast-math than without.
This is because the reassociate pass greedily transforms fsub
into fneg/fadd and apparently (based on the regression tests
seen here) expects instcombine to clean that up if it wasn't
profitable. But we were missing this fold:
(X - Y) - Z --> X - (Y + Z)
There's another, more specific case that I think we should
handle as shown in the "fake" fneg test (but missed with a real
fneg), but that's another patch. That may be tricky to get
right without conflicting with existing transforms for fneg.
Differential Revision: https://reviews.llvm.org/D72521
Nicolas Vasilache [Wed, 15 Jan 2020 16:12:53 +0000 (11:12 -0500)]
[mlir][Linalg] NFC - Hotfix for gcc-5 build
Nicolas Vasilache [Wed, 15 Jan 2020 14:48:49 +0000 (09:48 -0500)]
[mlir][Linalg] NFC - Cleanup Linalg Pass locations and namespacing
Summary:
This diff moves the conversion pass declaration closer to its definition
and makes the namespacing of passes consistent with the rest of the
infrastructure (i.e. `mlir::linalg::createXXXPass` -> `mlir::createXXXPass`).
Reviewers: ftynse, jpienaar, mehdi_amini
Subscribers: rriddle, burmako, shauheen, antiagainst, arpith-jacob, mgester, lucyrfox, aartbik, liufengdb, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72766
Lang Hames [Wed, 15 Jan 2020 02:14:00 +0000 (18:14 -0800)]
[ORC] Simplify use of lazyReexports with LLJIT.
This patch makes the target triple available via the LLJIT interface, and moves
the IRTransformLayer from LLLazyJIT down into LLJIT. Together these changes make
it easier to use the lazyReexports utility with LLJIT, and to apply IR
transforms to code as it is compiled in LLJIT (rather than requiring transforms
to be applied manually before code is added). An code example is added in
llvm/examples/LLJITExamples/LLJITWithLazyReexports
Lang Hames [Wed, 15 Jan 2020 01:09:02 +0000 (17:09 -0800)]
[ORC] Update lazyReexports to support aliases with different symbol names.
A bug in the existing implementation meant that lazyReexports would not work if
the aliased name differed from the alias's name, i.e. all lazy reexports had to
be of the form (lib1, name) -> (lib2, name). This patch fixes the issue by
capturing the alias's name in the NotifyResolved callback. To simplify this
capture, and the LazyCallThroughManager code in general, the NotifyResolved
callback is updated to use llvm::unique_function rather than a custom class.
No test case yet: This can only be tested at runtime, and the only in-tree
client (lli) always uses aliases with matching names. I will add a new LLJIT
example shortly that will directly test the lazyReexports API and the
non-trivial alias use case.
Hubert Tong [Wed, 15 Jan 2020 15:42:27 +0000 (10:42 -0500)]
DWARFDebugLine.cpp: Format unknown line number standard opcodes
Summary:
This patch implements `formatv()` formatting for `dwarf::LineNumberOps`
and makes use of it for the `llvm-dwarfdump --debug-line` dump.
Previously, unknown line number standard opcodes would lead to undefined
behaviour. The code would attempt to format the data pointer of an empty
`StringRef` (a null pointer) using `%s`. According to the description
for `format()`, use of that interface carries the "risk of `printf`".
Passing a null pointer in place of an array to a C library function
results in undefined behaviour.
Reviewers: jhenderson, daltenty, stevewan
Reviewed By: jhenderson
Subscribers: aprantl, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72369
Hubert Tong [Wed, 15 Jan 2020 15:42:12 +0000 (10:42 -0500)]
[CMake] Enable -qfuncsect when building with IBM XL
Summary:
The IBM XL compiler uses `-qfuncsect` for `-ffunction-sections`.
The comment about sanitizers and `-f[no-]function-sections` is corrected
also, as it is pertinent to this patch.
The sanitizer-related use of `-fno-function-sections` is associated with
powerpc64le, a target for which there is an IBM XL compiler, so that use
is updated in this patch to apply `-qnofuncsect` in case a build using
the XL compiler is viable on that platform.
This patch has been verified with the XL compiler on AIX only.
Reviewers: daltenty, stevewan
Reviewed By: daltenty
Subscribers: mgorny, steven.zhang, #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D72335
Nicolas Vasilache [Wed, 15 Jan 2020 15:36:25 +0000 (10:36 -0500)]
[mlir][Linalg] NFC - Hotfix for gcc-5 build
Georgii Rymar [Wed, 15 Jan 2020 13:17:09 +0000 (16:17 +0300)]
[llvm-readobj][test] - Cleanup SHT_RELR sections testing.
After recent changes (D71872) in yaml2obj, it is possible so cleanup
testing of the SHT_RELR sections.
Differential revision: https://reviews.llvm.org/D71874
Teresa Johnson [Mon, 13 Jan 2020 19:09:13 +0000 (11:09 -0800)]
Restore "[ThinLTO] Add additional ThinLTO pipeline testing with new PM"
This restores
2af97be8027a0823b88d4b6a07fc5eedb440bc1f (reverted at
6288f86e870c7bb7fe47cc138320b9eb34c93941), with all the fixes I had
applied at the time, along with a new fix for non-determinism in the
ordering of a couple of passes due to being accessed as parameters on
the same call.
I've also added --dump-input=fail to the new tests so I can more
thoroughly fix any additional failures.
Kadir Cetinkaya [Wed, 15 Jan 2020 15:22:24 +0000 (16:22 +0100)]
[clangd] Fix windows buildbots
Nico Weber [Wed, 15 Jan 2020 15:05:25 +0000 (10:05 -0500)]
[gn build] find mistakes like the one fixed in
72b5989e0d1 at build time
Simon Tatham [Wed, 15 Jan 2020 14:07:04 +0000 (14:07 +0000)]
[clang] New __attribute__((__clang_arm_mve_strict_polymorphism)).
This is applied to the vector types defined in <arm_mve.h> for use
with the intrinsics for the ARM MVE vector architecture.
Its purpose is to inhibit lax vector conversions, but only in the
context of overload resolution of the MVE polymorphic intrinsic
functions. This solves an ambiguity problem with polymorphic MVE
intrinsics that take a vector and a scalar argument: the scalar
argument can often have the wrong integer type due to default integer
promotions or unsuffixed literals, and therefore, the type of the
vector argument should be considered trustworthy when resolving MVE
polymorphism.
As part of the same change, I've added the new attribute to the
declarations generated by the MveEmitter Tablegen backend (and
corrected a namespace issue with the other attribute while I was
there).
Reviewers: aaron.ballman, dmgreen
Reviewed By: aaron.ballman
Subscribers: kristof.beyls, JDevlieghere, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D72518
Kadir Cetinkaya [Mon, 13 Jan 2020 16:21:31 +0000 (17:21 +0100)]
[clangd] Rearrange type, returntype and parameters in hover card
Summary:
Moves type/returntype into its own line as it is more readable in cases
where the type is long.
Also gives parameter lists a heading, `Parameters:` to make them stand out.
Leaves the `right arrow` instead of `Returns: ` before Return Type to make
output more symmetric.
```
function foo
Returns: ret_type
Parameters:
- int x
```
vs
```
function foo
🡺 ret_type
Parameters:
- int x
```
Reviewers: sammccall, ilya-biryukov
Subscribers: MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D72623
Kadir Cetinkaya [Mon, 13 Jan 2020 16:14:24 +0000 (17:14 +0100)]
[clangd] Add a ruler after header in hover
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D72622
Nico Weber [Wed, 15 Jan 2020 14:49:32 +0000 (09:49 -0500)]
[gn build] (manually) port
b4a99a061f51 better
Nicolas Vasilache [Wed, 15 Jan 2020 14:28:12 +0000 (09:28 -0500)]
[mlir][EDSC] Refactor dependencies involving EDSCs.
Summary: This diff removes the dependency of LinalgOps and VectorOps on EDSCs.
Reviewers: jpienaar, ftynse
Reviewed By: ftynse
Subscribers: merge_guards_bot, mgorny, mehdi_amini, rriddle, burmako, shauheen, antiagainst, csigg, arpith-jacob, mgester, lucyrfox, herhut, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72481
Alex Richardson [Wed, 15 Jan 2020 14:32:14 +0000 (14:32 +0000)]
[ELF] Avoid false-positive assert in getErrPlace()
This assertion was added as part of D70659 but did not account for .bss
input sections. I noticed that this assert was incorrectly triggering
while building FreeBSD for MIPS64. Fixed by relaxing the assert to also
account for SHT_NOBITS input sections and adjust the test
mips-jalr-non-function.s to link a file with a .bss section first.
Reviewed By: MaskRay, grimar
Differential Revision: https://reviews.llvm.org/D72567
Kadir Cetinkaya [Fri, 10 Jan 2020 12:11:09 +0000 (13:11 +0100)]
[clangd] Show hower info for expressions
Summary:
This currently populates only the Name with the expression's type and
Value if expression is evaluatable.
Fixes https://github.com/clangd/clangd/issues/56
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D72500
Ulrich Weigand [Wed, 15 Jan 2020 14:08:35 +0000 (15:08 +0100)]
[FPEnv] Address post-commit review comment for D71467
Remove a bit of code duplication between CreateFCmp and CreateFCmpS
by creating a shared helper function.
Alex Zinenko [Wed, 15 Jan 2020 13:56:25 +0000 (14:56 +0100)]
[mlir] LLVM import: handle function-typed constants
The current implementation of the LLVM-to-MLIR translation could not handle
functions used as constant values in instructions. The handling is added
trivially as `llvm.mlir.constant` can define constants of function type using
SymbolRef attributes, which works even for functions that have not been
declared yet.
Matt Arsenault [Fri, 10 Jan 2020 02:53:28 +0000 (21:53 -0500)]
GlobalISel: Implement lower for G_BITCAST
Bitcast only really applies between scalars and vectors. Implement as
an unmerge and remerge. The test needs to tolerate failure since one
of the unmerges currently fails to legalize.
Matt Arsenault [Tue, 31 Dec 2019 20:28:41 +0000 (15:28 -0500)]
AMDGPU: Partially directly select llvm.amdgcn.interp.p1.f16
The 16 bank LDS case is complicated due to using multiple
instructions. If I attempt to write a pattern for it, the generated
selector incorrectly places the copy to m0 after the first
instruction, so that needs to be separately addressed.
Also fix not gluing the copy to m0 to the second operation in the
second half of the 16 bank lowering.
Matt Arsenault [Fri, 10 Jan 2020 14:47:17 +0000 (09:47 -0500)]
GlobalISel: Fix narrowScalar for G_ANYEXT results
This is nearly the same as G_ZEXT.
Matt Arsenault [Tue, 14 Jan 2020 18:01:46 +0000 (13:01 -0500)]
TableGen: Delete some copy constuctors
Some register related machinery relies on uniqued, static pointers for
register classes and subregisters, so try to make sure these are never
copied.
Matt Arsenault [Tue, 14 Jan 2020 19:09:06 +0000 (14:09 -0500)]
TableGen/GlobalISel: Don't take reference to temporary values
These return temporary Optional<> values which are immediately
destroyed. I'm not sure why no sanitizers seem to have caught this,
but I encountered crashes on these in a future patch.
Matt Arsenault [Tue, 14 Jan 2020 18:48:34 +0000 (13:48 -0500)]
TableGen/GlobalISel: Don't reconstruct CodeGenRegBank
The maps for dealing with the relationships between different register
classes and subregister indexes rely on unique pointers for every
class/index. By constructing a second copy of CodeGenRegBank, two
different pointer values existed for a given subregister depending on
where you were querying.
Use the existing CodeGenRegBank owned by the CodeGenTarget instead of
constructing a second copy. This avoids incorrectly failing map
lookups in a future change.
Luís Marques [Wed, 15 Jan 2020 13:47:23 +0000 (13:47 +0000)]
[RISCV] Fix test for inline asm z constraint modifier
Summary: Use an `i` constraint in the test, to correctly trigger the code for
handling the `z` constraint modifier.
Reviewers: asb, lenary, jrtc27
Reviewed By: lenary, jrtc27
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72134
Soumi Manna [Wed, 15 Jan 2020 13:47:22 +0000 (08:47 -0500)]
Further implement CWG 2292
The core issue is that simple-template-id is ambiguous between class-name
and type-name. This fixes PR43966.
Djordje Todorovic [Wed, 15 Jan 2020 12:00:14 +0000 (13:00 +0100)]
[llvm-locstats] Add the --compare option
Draw a plot showing the difference in debug loc coverage on two
files provided.
Differential Revision: https://reviews.llvm.org/D71870
Nemanja Ivanovic [Wed, 15 Jan 2020 13:00:22 +0000 (07:00 -0600)]
[PowerPC] Legalize saturating vector add/sub
These intrinsics and the corresponding ISD nodes were recently added. PPC has
instructions that do this for vectors. Legalize them and add patterns to emit
the satuarting instructions.
Differential revision: https://reviews.llvm.org/D71940
Hans Wennborg [Wed, 15 Jan 2020 09:02:56 +0000 (10:02 +0100)]
Bump the trunk major version to 11
and clear the release notes.
Simon Pilgrim [Wed, 15 Jan 2020 12:36:11 +0000 (12:36 +0000)]
Revert rG6078f2fedcac5797ac39ee5ef3fd7a35ef1202d5 - "[AArch64][GlobalISel]: Support @llvm.{return,frame}address selection."
These intrinsics expand to a variable number of instructions so just like in
ISelLowering.cpp we use custom code to deal with them.
Committing Tim's original patch.
Differential Revision: https://reviews.llvm.org/D65656
----
Breaks EXPENSIVE_CHECKS builds.
Zakk Chen [Fri, 29 Nov 2019 02:00:54 +0000 (18:00 -0800)]
[RISCV] Support ABI checking with per function target-features
if users don't specific -mattr, the default target-feature come
from IR attribute.
Reviewers: lenary, asb
Reviewed By: lenary, asb
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70837
Zakk Chen [Wed, 15 Jan 2020 12:32:57 +0000 (04:32 -0800)]
Revert "[RISCV] Support ABI checking with per function target-features"
This reverts commit
109e4d12edda07bdec139de36d9fdb6f73399f92.
Simon Pilgrim [Wed, 15 Jan 2020 12:05:30 +0000 (12:05 +0000)]
Fix Wdocumentation warning. NFC.
Simon Pilgrim [Wed, 15 Jan 2020 11:51:55 +0000 (11:51 +0000)]
RegisterClassInfo::computePSetLimit - assert that we actually find a register.
Fixes "pointer is null" clang static analyzer warning.