Nikita Popov [Sat, 5 Jan 2019 17:36:05 +0000 (17:36 +0000)]
[InstCombine] Improve cttz/ctlz + icmp tests; NFC
Change part of the tests to use vectors (I'm using scalar for ugt
and vector for ult), add multiuse variations, rename %lz to %tz
for the cttz tests.
llvm-svn: 350471
Mike Spertus [Sat, 5 Jan 2019 17:01:34 +0000 (17:01 +0000)]
Improve MSVC type visualizations
Display TypeBits in a single line.
Fix bit rot in template visualizations
Rudimentary support for deduced types
llvm-svn: 350470
Mike Spertus [Sat, 5 Jan 2019 16:59:27 +0000 (16:59 +0000)]
Enhance MSVC visualization of PointerUnions
Add a "deref" view that displays the pointed to objects since
other visualizers often need to display data reference by internal
PointerUnions
llvm-svn: 350469
Nikita Popov [Sat, 5 Jan 2019 15:51:59 +0000 (15:51 +0000)]
[InstCombine] Add cttz/ctlz + icmp ugt/ult tests; NFC
llvm-svn: 350468
Serge Guelton [Sat, 5 Jan 2019 12:07:36 +0000 (12:07 +0000)]
[python] Make the collections import future-proof
On Python 3.7 the old code raises a warning:
DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working
class ArgumentsIterator(collections.Sequence):
On Python 3.8 it wouldn't work anymore.
Commited on behalf of Jakub Stasiak.
Differential Revision: https://reviews.llvm.org/D56341
llvm-svn: 350467
Jonas Toth [Sat, 5 Jan 2019 11:40:05 +0000 (11:40 +0000)]
[Documentation] fix order of checks in checks/list.rst
llvm-svn: 350466
Martin Storsjo [Sat, 5 Jan 2019 10:44:03 +0000 (10:44 +0000)]
[docs] Add a known limitation to the comment about lld supporting GCC/MinGW object files. NFC.
Differential Revision: https://reviews.llvm.org/D56335
llvm-svn: 350465
Martin Storsjo [Sat, 5 Jan 2019 10:43:59 +0000 (10:43 +0000)]
[MinGW] Expand comment for MinGW driver. NFC.
Originally authored by Rui Ueyama.
Differential Revision: https://reviews.llvm.org/D53031
llvm-svn: 350464
Nikita Popov [Sat, 5 Jan 2019 09:48:16 +0000 (09:48 +0000)]
[InstCombine] Relax cttz/ctlz with select on zero
The cttz/ctlz intrinsics have a parameter specifying whether the
result is undefined for zero. cttz(x, false) can be relaxed to
cttz(x, true) if x is known non-zero, and in fact such an optimization
is already performed. However, this currently doesn't work if x is
non-zero as a result of a select rather than an explicit branch.
This patch adds handling for this case, thus allowing
x != 0 ? cttz(x, false) : y to simplify to x != 0 ? cttz(x, true) : y.
Differential Revision: https://reviews.llvm.org/D55786
llvm-svn: 350463
Nikita Popov [Sat, 5 Jan 2019 09:48:05 +0000 (09:48 +0000)]
[InstCombine] Add vector tests for select + ctlz/cttz; NFC
llvm-svn: 350462
Petr Hosek [Sat, 5 Jan 2019 07:57:53 +0000 (07:57 +0000)]
[CMake][Fuchsia] Enable experimental new pass manager by default
This change enableds experimental new pass manager.
Differential Revision: https://reviews.llvm.org/D56350
llvm-svn: 350461
Petr Hosek [Sat, 5 Jan 2019 07:57:46 +0000 (07:57 +0000)]
[CMake][Fuchsia] Enable x86 relaxation by default
This enables x86 relaxation by default. This depends on a linker new
enough to support the new reloc types but since we default to lld we
don't worry about host system linkers that might be too old to support
the new reloc types.
Differential Revision: https://reviews.llvm.org/D56349
llvm-svn: 350460
Petr Hosek [Sat, 5 Jan 2019 07:57:38 +0000 (07:57 +0000)]
[CMake][Fuchsia] Enable --build-id linker flag by default
This enables passing --build-id to linker by default.
Differential Revision: https://reviews.llvm.org/D56348
llvm-svn: 350459
Nico Weber [Sat, 5 Jan 2019 05:01:20 +0000 (05:01 +0000)]
[gn build] Merge r350423
llvm-svn: 350458
Nico Weber [Sat, 5 Jan 2019 04:05:25 +0000 (04:05 +0000)]
[gn build] Add build files for unittests under llvm/unittests/ExecutionEngine
Differential Revision: https://reviews.llvm.org/D56328
llvm-svn: 350457
Easwaran Raman [Sat, 5 Jan 2019 02:26:29 +0000 (02:26 +0000)]
[Inliner] Optimize shouldBeDeferred
This has some minor optimizations to shouldBeDeferred. This is not
strictly NFC because the early exit inside the loop assumes
TotalSecondaryCost is monotonically non-decreasing, which is not true if
the threshold used by CostAnalyzer is negative. AFAICT the thresholds do
not go below 0 for the default values of the various options we use.
llvm-svn: 350456
Craig Topper [Sat, 5 Jan 2019 01:40:29 +0000 (01:40 +0000)]
[X86] Require second operand of X86vshiftuniform to be an integer. NFC
We don't need to require the first operand to be an integer because we already said it was the same type as the result which we also constrained to an integer.
llvm-svn: 350455
Nico Weber [Sat, 5 Jan 2019 01:39:18 +0000 (01:39 +0000)]
[gn build] Add build files for unittests that load shared libraries
This is slightly ugly for three reasons:
- The shlib needs to go next to the binary to be found on all platforms, so the
build files refer to target_out_dir
- The explicit -fPIC flag needed on the shared lib side, and the -rdynamic flag
needed on the host side, on Linux
- Plugins that refer to LLVM code and assume that the host will resolve them
don't work on Windows -- PluginsTests won't test anything on Windows (but
DynamicLibraryTests will, since the dll here doesn't call LLVM code)
If we get lots more of these plugin / plugin host targets it might make sense
to add a template for them. But for now, these are the last ones we need.
(We're at 6 plugin hosts, 2 of them tests, and at 6 shared libraries, 2 of them
tests as well. clang is a plugin host by default in the CMake build but not
(yet?) in the GN build.)
Differential Revision: https://reviews.llvm.org/D56330
llvm-svn: 350454
Nico Weber [Sat, 5 Jan 2019 01:19:14 +0000 (01:19 +0000)]
Let new test from r350340 still pass even after r350451.
llvm-svn: 350453
Alexandre Ganea [Sat, 5 Jan 2019 01:16:24 +0000 (01:16 +0000)]
[LLD][COFF] PDB: Parallel sort publics
Saves up to 1.3 sec on large PDBs.
Figures below are for the "Globals Stream Layout" pass:
Before This patch
Large EXE (PDB is ~2 GB) 3330 ms 2022 ms
Large EXE (PDB is ~2 GB) 2680 ms 1608 ms
Large DLL (PDB is ~1 GB) 1455 ms 938 ms
Large DLL (PDB is ~800 MB) 1215 ms 800 ms
Small DLL (PDB is ~200 MB) 224 ms 146 ms
Differential Revision: https://reviews.llvm.org/D56334
llvm-svn: 350452
Nico Weber [Sat, 5 Jan 2019 01:10:20 +0000 (01:10 +0000)]
Move -add-plugin validation after -load was executed.
Moves the code added in r350340 around a bit, to hopefully make the existing
plugin tests pass when clang is built with examples enabled.
llvm-svn: 350451
Alexandre Ganea [Sat, 5 Jan 2019 01:08:10 +0000 (01:08 +0000)]
[LLD][COFF] Fix namespace compilation issue with a upcoming patch. NFC
llvm-svn: 350450
Evgeniy Stepanov [Sat, 5 Jan 2019 00:45:14 +0000 (00:45 +0000)]
Revert "Revert "Switch Android from TLS_SLOT_TSAN(8) to TLS_SLOT_SANITIZER(6)""
This reapplies commit r348984.
llvm-svn: 350449
Evgeniy Stepanov [Sat, 5 Jan 2019 00:44:58 +0000 (00:44 +0000)]
Revert "Revert "[hwasan] Android: Switch from TLS_SLOT_TSAN(8) to TLS_SLOT_SANITIZER(6)""
This reapplies commit r348983.
llvm-svn: 350448
Nico Weber [Sat, 5 Jan 2019 00:14:37 +0000 (00:14 +0000)]
[gn build] Add build file for DebugInfoPDBTests
I'm pretty unhappy this patch: DebugInfoPDBTests uses an API that requires some
magic txt file to be next to the unit test executable that stores the absolute
path to the LLVM source root.
The choices here are:
1. Don't use the unittest() template for DebugInfoPDBTests and set output_dir
for unit tests in two places (the gni file for every test but this one, and the
BUILD.gn file for this specific test).
2. Add another unittest_foo() template variation for this one test.
I went with the former, and added a comment to the template to look out for
this.
(The CMake build has the same issue.)
Differential Revision: https://reviews.llvm.org/D56324
llvm-svn: 350447
Jonas Devlieghere [Sat, 5 Jan 2019 00:01:04 +0000 (00:01 +0000)]
[Driver] Some more cleanup. NFC
llvm-svn: 350446
Peter Collingbourne [Fri, 4 Jan 2019 23:24:02 +0000 (23:24 +0000)]
Bring back the pthread_create interceptor, but only on non-aarch64.
We still need the interceptor on non-aarch64 to untag the pthread_t
and pthread_attr_t pointers and disable tagging on allocations done
internally by glibc.
llvm-svn: 350445
Peter Collingbourne [Fri, 4 Jan 2019 23:10:47 +0000 (23:10 +0000)]
Disable system-allocator-fallback.cc test on Android O and earlier.
The dynamic loader on Android O appears to have a bug where it crashes
when dlopening DF_1_GLOBAL libraries.
llvm-svn: 350444
Evgeniy Stepanov [Fri, 4 Jan 2019 22:55:04 +0000 (22:55 +0000)]
[sanitizer] Reduce stack depot size on Android.
Summary:
The default setting kTabSizeLog=20 results in an 8Mb global hash table,
almost all of it in private pages. That is not a sane setting in a
mobile, system-wide use case: with ~150 concurrent processes stack
depot will account for more than 1Gb of RAM.
Reviewers: kcc, pcc
Subscribers: srhines, kubamracek, llvm-commits
Differential Revision: https://reviews.llvm.org/D56333
llvm-svn: 350443
Rong Xu [Fri, 4 Jan 2019 22:54:03 +0000 (22:54 +0000)]
[PGO] Use SourceFileName rather module name in PGOFuncName
In LTO or Thin-lto mode (though linker plugin), the module
names are of temp file names which are different for
different compilations. Using SourceFileName avoids the issue.
This should not change any functionality for current PGO as
all the current callers of getPGOFuncName() is before LTO.
llvm-svn: 350442
Joel E. Denny [Fri, 4 Jan 2019 22:12:13 +0000 (22:12 +0000)]
[OpenMP] Refactor const restriction for linear
As discussed in D56113, this patch refactors the implementation of the
const restriction for linear to reuse a function introduced by D56113.
A side effect is that, if a variable has mutable members, this
diagnostic is now skipped, and the diagnostic for the variable not
being an integer or pointer is reported instead.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56299
llvm-svn: 350441
Joel E. Denny [Fri, 4 Jan 2019 22:11:56 +0000 (22:11 +0000)]
[OpenMP] Refactor const restriction for reductions
As discussed in D56113, this patch refactors the implementation of the
const restriction for reductions to reuse a function introduced by
D56113. A side effect is that diagnostics sometimes now say
"variable" instead of "list item" when a list item is a variable.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56298
llvm-svn: 350440
Joel E. Denny [Fri, 4 Jan 2019 22:11:31 +0000 (22:11 +0000)]
[OpenMP] Replace predetermined shared for const variable
The following appears in OpenMP 3.1 sec. 2.9.1.1 as a predetermined
data-sharing attribute:
> Variables with const-qualified type having no mutable member are
> shared.
It does not appear in OpenmP 4.0, 4.5, or 5.0. This patch removes the
implementation of that attribute when the requested OpenMP version is
greater than 3.1.
One effect of that removal is that `default(none)` affects const
variables without mutable members.
Also, without this patch, if a const variable without mutable members
was explicitly lastprivate or private, it was an error because it was
predetermined shared. Now, clang instead complains that it's const
without mutable fields, which is a more intelligible diagnostic. That
should be fine for all of the above versions because they all have
something like the following, which is quoted from OpenMP 5.0
sec. 2.19.3:
> A variable that is privatized must not have a const-qualified type
> unless it is of class type with a mutable member. This restriction does
> not apply to the firstprivate clause.
reduction and linear clauses already have separate checks for const
variables. Future patches will merge the implementations.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56113
llvm-svn: 350439
Alexandre Ganea [Fri, 4 Jan 2019 21:49:22 +0000 (21:49 +0000)]
[LLD][COFF] Fix file/line retrieval when a undefined symbol is to be printed
Differential Revision: https://reviews.llvm.org/D55951
llvm-svn: 350438
Nikita Popov [Fri, 4 Jan 2019 21:41:35 +0000 (21:41 +0000)]
[X86] Fix warning; NFC
llvm-svn: 350437
Vyacheslav Zakharin [Fri, 4 Jan 2019 21:25:01 +0000 (21:25 +0000)]
Update the pr_datasz of .note.gnu.property section.
Patch by Xiang Zhang.
Differential Revision: https://reviews.llvm.org/D56080
llvm-svn: 350436
Nikita Popov [Fri, 4 Jan 2019 21:21:43 +0000 (21:21 +0000)]
[BDCE] Remove dead uses of arguments
In addition to finding dead uses of instructions, also find dead uses
of function arguments, and replace them with zero as well.
I'm changing the way the known bits are computed here to remove the
coupling between the transfer function and the algorithm. It previously
relied on the first op being visited first and computing known bits --
unless the first op is not an instruction, in which case they're computed
on the second op. I could have adjusted this to check for "instruction
or argument", but I think it's better to avoid the repeated calculation
with an explicit flag.
Differential Revision: https://reviews.llvm.org/D56247
llvm-svn: 350435
Evandro Menezes [Fri, 4 Jan 2019 21:02:25 +0000 (21:02 +0000)]
[AArch64] Adjust the cost model for Exynos M3
Improve the modeling of ASIMD loads and stores.
llvm-svn: 350434
Peter Collingbourne [Fri, 4 Jan 2019 20:51:54 +0000 (20:51 +0000)]
Fix default-function-attr.c so that it works on Windows.
llvm-svn: 350433
Craig Topper [Fri, 4 Jan 2019 20:50:59 +0000 (20:50 +0000)]
[X86] Add INSERT_SUBVECTOR to ComputeNumSignBits
This adds support for calculating sign bits of insert_subvector. I based it on the computeKnownBits.
My motivating case is propagating sign bits information across basic blocks on AVX targets where concatenating using insert_subvector is common.
Differential Revision: https://reviews.llvm.org/D56283
llvm-svn: 350432
Alexey Bataev [Fri, 4 Jan 2019 20:16:54 +0000 (20:16 +0000)]
[OPENMP][NVPTX]General formatting/code improvement, NFC.
Summary: Formatting.
Reviewers: gtbercea, grokos, kkwli0
Subscribers: guansong, openmp-commits, caomhin
Differential Revision: https://reviews.llvm.org/D56290
llvm-svn: 350431
Sanjay Patel [Fri, 4 Jan 2019 20:14:53 +0000 (20:14 +0000)]
[x86] add tests for potential horizontal vector ops; NFC
These are modified versions of the FP tests from rL349923.
llvm-svn: 350430
Peter Collingbourne [Fri, 4 Jan 2019 19:27:04 +0000 (19:27 +0000)]
hwasan: Implement lazy thread initialization for the interceptor ABI.
The problem is similar to D55986 but for threads: a process with the
interceptor hwasan library loaded might have some threads started by
instrumented libraries and some by uninstrumented libraries, and we
need to be able to run instrumented code on the latter.
The solution is to perform per-thread initialization lazily. If a
function needs to access shadow memory or add itself to the per-thread
ring buffer its prologue checks to see whether the value in the
sanitizer TLS slot is null, and if so it calls __hwasan_thread_enter
and reloads from the TLS slot. The runtime does the same thing if it
needs to access this data structure.
This change means that the code generator needs to know whether we
are targeting the interceptor runtime, since we don't want to pay
the cost of lazy initialization when targeting a platform with native
hwasan support. A flag -fsanitize-hwaddress-abi={interceptor,platform}
has been introduced for selecting the runtime ABI to target. The
default ABI is set to interceptor since it's assumed that it will
be more common that users will be compiling application code than
platform code.
Because we can no longer assume that the TLS slot is initialized,
the pthread_create interceptor is no longer necessary, so it has
been removed.
Ideally, lazy initialization should only cost one instruction in the
hot path, but at present the call may cause us to spill arguments
to the stack, which means more instructions in the hot path (or
theoretically in the cold path if the spills are moved with shrink
wrapping). With an appropriately chosen calling convention for
the per-thread initialization function (TODO) the hot path should
always need just one instruction and the cold path should need two
instructions with no spilling required.
Differential Revision: https://reviews.llvm.org/D56038
llvm-svn: 350429
Davide Italiano [Fri, 4 Jan 2019 19:23:52 +0000 (19:23 +0000)]
[Scalar] Simplify comparison operators and add coverage.
llvm-svn: 350428
Peter Collingbourne [Fri, 4 Jan 2019 19:21:51 +0000 (19:21 +0000)]
hwasan: Use system allocator to realloc and free untagged pointers in interceptor mode.
The Android dynamic loader has a non-standard feature that allows
libraries such as the hwasan runtime to interpose symbols even after
the symbol already has a value. The new value of the symbol is used to
relocate libraries loaded after the interposing library, but existing
libraries keep the old value. This behaviour is activated by the
DF_1_GLOBAL flag in DT_FLAGS_1, which is set by passing -z global to
the linker, which is what we already do to link the hwasan runtime.
What this means in practice is that if we have .so files that depend
on interceptor-mode hwasan without the main executable depending on
it, some of the libraries in the process will be using the hwasan
allocator and some will be using the system allocator, and these
allocators need to interact somehow. For example, if an instrumented
library calls a function such as strdup that allocates memory on
behalf of the caller, the instrumented library can reasonably expect
to be able to call free to deallocate the memory.
We can handle that relatively easily with hwasan by using tag 0 to
represent allocations from the system allocator. If hwasan's realloc
or free functions are passed a pointer with tag 0, the system allocator
is called.
One limitation is that this scheme doesn't work in reverse: if an
instrumented library allocates memory, it must free the memory itself
and cannot pass ownership to a system library. In a future change,
we may want to expose an API for calling the system allocator so
that instrumented libraries can safely transfer ownership of memory
to system libraries.
Differential Revision: https://reviews.llvm.org/D55986
llvm-svn: 350427
Aaron Enye Shi [Fri, 4 Jan 2019 19:09:20 +0000 (19:09 +0000)]
[HIP][DRIVER][OFFLOAD] Do not unbundle unsupported file types
The offload bundler action should not unbundle the input file types that does not match the action type. This fixes an issue where .so files are unbundled when the action type is object files.
Reviewers: yaxunl
Differential Revision: https://reviews.llvm.org/D56321
llvm-svn: 350426
Aaron Enye Shi [Fri, 4 Jan 2019 19:05:41 +0000 (19:05 +0000)]
[HIP][DRIVER][OFFLOAD] Do not unbundle unsupported file types
The offload bundler action should not unbundle the input file types that does not match the action type. This fixes an issue where .so files are unbundled when the action type is object files.
llvm-svn: 350425
Teresa Johnson [Fri, 4 Jan 2019 19:05:01 +0000 (19:05 +0000)]
[ThinLTO] Clang changes to utilize new pass to handle chains of aliases
Summary:
As with NameAnonGlobals, invoke the new CanonicalizeAliases via clang
when using the new PM.
Depends on D54507.
Reviewers: pcc, davidxl
Subscribers: mehdi_amini, inglorion, steven_wu, dexonsmith, cfe-commits
Differential Revision: https://reviews.llvm.org/D55620
llvm-svn: 350424
Teresa Johnson [Fri, 4 Jan 2019 19:04:54 +0000 (19:04 +0000)]
[ThinLTO] Handle chains of aliases
At -O0, globalopt is not run during the compile step, and we can have a
chain of an alias having an immediate aliasee of another alias. The
summaries are constructed assuming aliases in a canonical form
(flattened chains), and as a result only the base object but no
intermediate aliases were preserved.
Fix by adding a pass that canonicalize aliases, which ensures each
alias is a direct alias of the base object.
Reviewers: pcc, davidxl
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, arphaman, llvm-commits
Differential Revision: https://reviews.llvm.org/D54507
llvm-svn: 350423
Erik Pilkington [Fri, 4 Jan 2019 18:33:06 +0000 (18:33 +0000)]
[ObjCARC] Add an new attribute, objc_externally_retained
This attribute, called "objc_externally_retained", exposes clang's
notion of pseudo-__strong variables in ARC. Pseudo-strong variables
"borrow" their initializer, meaning that they don't retain/release
it, instead assuming that someone else is keeping their value alive.
If a function is annotated with this attribute, implicitly strong
parameters of that function aren't implicitly retained/released in
the function body, and are implicitly const. This is useful to expose
for performance reasons, most functions don't need the extra safety
of the retain/release, so programmers can opt out as needed.
This attribute can also apply to declarations of local variables,
with similar effect.
Differential revision: https://reviews.llvm.org/D55865
llvm-svn: 350422
Sanjay Patel [Fri, 4 Jan 2019 17:48:13 +0000 (17:48 +0000)]
[x86] lower extracted fadd/fsub to horizontal vector math; 2nd try
The 1st try for this was at rL350369, but it caused IR-level diffs because
our cost models differentiate custom vs. legal/promote lowering. So that was
reverted at rL350373. The cost models were fixed independently at rL350403,
so this is effectively the same patch as last time.
Original commit message:
This would show up if we fix horizontal reductions to narrow as they go along,
but it's an improvement for size and/or Jaguar (fast-hops) independent of that.
We need to do this late to not interfere with other pattern matching of larger
horizontal sequences.
We can extend this to integer ops in a follow-up patch.
Differential Revision: https://reviews.llvm.org/D56011
llvm-svn: 350421
Vedant Kumar [Fri, 4 Jan 2019 17:43:22 +0000 (17:43 +0000)]
[CodeExtractor] Do not extract unsafe lifetime markers
Lifetime markers which reference inputs to the extraction region are not
safe to extract. Example ('rhs' will be extracted):
```
entry:
+------------+
| x = alloca |
| y = alloca |
+------------+
/ \
lhs: rhs:
+-------------------+ +-------------------+
| lifetime_start(x) | | lifetime_start(x) |
| use(x) | | lifetime_start(y) |
| lifetime_end(x) | | use(x, y) |
| lifetime_start(y) | | lifetime_end(y) |
| use(y) | | lifetime_end(x) |
| lifetime_end(y) | +-------------------+
+-------------------+
```
Prior to extraction, the stack coloring pass sees that the slots for 'x'
and 'y' are in-use at the same time. After extraction, the coloring pass
infers that 'x' and 'y' are *not* in-use concurrently, because markers
from 'rhs' are no longer available to help decide otherwise.
This leads to a miscompile, because the stack slots actually are in-use
concurrently in the extracted function.
Fix this by moving lifetime start/end markers for memory regions defined
in the calling function around the call to the extracted function.
Fixes llvm.org/PR39671 (rdar://
45939472).
Differential Revision: https://reviews.llvm.org/D55967
llvm-svn: 350420
Sanjay Patel [Fri, 4 Jan 2019 17:38:12 +0000 (17:38 +0000)]
[InstCombine] reduce raw IR narrowing rotate patterns to funnel shift
Similar to rL350199 - there are no known analysis/codegen holes for
funnel shift intrinsics now, so we can canonicalize the 6+ regular
instructions to funnel shift to improve vectorization, inlining,
unrolling, etc.
llvm-svn: 350419
Nico Weber [Fri, 4 Jan 2019 17:32:28 +0000 (17:32 +0000)]
[gn build] Merge r350351
llvm-svn: 350418
Nico Weber [Fri, 4 Jan 2019 17:26:05 +0000 (17:26 +0000)]
[gn build] Commit change that should have been in r350410.
llvm-svn: 350416
Alexey Bataev [Fri, 4 Jan 2019 17:25:09 +0000 (17:25 +0000)]
[OPENMP][NVPTX]Use new functions from the runtime library.
Updated codegen to use the new functions from the runtime library.
llvm-svn: 350415
Aaron Ballman [Fri, 4 Jan 2019 17:20:00 +0000 (17:20 +0000)]
Add two new pragmas for controlling software pipelining optimizations.
This patch adds #pragma clang loop pipeline and #pragma clang loop pipeline_initiation_interval for debugging or reducing compile time purposes. It is possible to disable SWP for concrete loops to save compilation time or to find bugs by not doing SWP to certain loops. It is possible to set value of initiation interval to concrete number to save compilation time by not doing extra pipeliner passes or to check created schedule for specific initiation interval.
Patch by Alexey Lapshin.
llvm-svn: 350414
Nico Weber [Fri, 4 Jan 2019 17:16:21 +0000 (17:16 +0000)]
[gn build] Add even more build files for LLVM unittests
Another random assortment of easy build files.
Differential Revision: https://reviews.llvm.org/D56217
llvm-svn: 350413
Nico Weber [Fri, 4 Jan 2019 17:15:38 +0000 (17:15 +0000)]
[gn build] Add more build files for LLVM unittests
A fairly random assortment of build files that are easy.
Differential Revision: https://reviews.llvm.org/D56213
llvm-svn: 350412
Nico Weber [Fri, 4 Jan 2019 17:14:55 +0000 (17:14 +0000)]
[gn build] Start adding build files for LLVM unittests
Adds build files for //llvm/unittest/[A-D].
Also teach sync_source_lists_from_cmake.py to not complain about missing
BUILD.gn files for CMakeLists.txt files that just call add_subdirectory()
without calling add_.+_unittest, like e.g.
llvm/unittests/Target/CMakeLists.txt.
(Omits CodeGen/GlobalISel and DebugInfo/PDB because their build files are somewhat interesting, and this patch is already on the larger side.)
Differential Revision: https://reviews.llvm.org/D56212
llvm-svn: 350411
Nico Weber [Fri, 4 Jan 2019 17:13:33 +0000 (17:13 +0000)]
[gn build] Add check-llvm target and make it work
With this, check-llvm runs and passes all of clang's lit tests. It doesn't run
any of its unit tests yet.
This is the only change in the GN build patch series that needs a change to a
file outside of llvm/utils/gn: llvm/test/tools/llvm-config/booleans.test checks
the result of llvm-config --build-system for some reason, so I'm updating the
test to accept "gn" as valid output in addition to "cmake". (The alternative
would be to let the gn build self-identify as cmake, which seems worse.)
Like with check-clang and check-lld, running just ninja -C out/gn will build
all prerequisites needed to run tests, but it won't run the tests (so that the
build becomes clean after one build). Running ninja -C out/gn check-llvm will
build prerequisites if needed and run the tests. The check-llvm target never
becomes clean and runs tests every time.
Differential Revision: https://reviews.llvm.org/D56195
llvm-svn: 350410
Nico Weber [Fri, 4 Jan 2019 17:12:25 +0000 (17:12 +0000)]
[gn build] Add build file for libLTO.dylib
Not used by anything yet, but will be needed to make check-llvm run ld64's libLTO plugin tests.
Differential Revision: https://reviews.llvm.org/D56317
llvm-svn: 350409
John Brawn [Fri, 4 Jan 2019 17:12:09 +0000 (17:12 +0000)]
[LICM] Adjust how moving the re-hoist point works
In some cases the order that we hoist instructions in means that when rehoisting
(which uses the same order as hoisting) we can rehoist to a block A, then a
block B, then block A again. This currently causes an assertion failure as it
expects that when changing the hoist point it only ever moves to a block that
dominates the hoist point being moved from.
Fix this by moving the re-hoist point when it doesn't dominate the dominator of
hoisted instruction, or in other words when it wouldn't dominate the uses of
the instruction being rehoisted.
Differential Revision: https://reviews.llvm.org/D55266
llvm-svn: 350408
Nico Weber [Fri, 4 Jan 2019 17:11:46 +0000 (17:11 +0000)]
[gn build] Add build files for llvm/lib/{LineEditor,Testing/Support,TextAPI}
Nothing pulls them in yet, but they will be needed for check-llvm.
LineEditor depends on libedit, so create a gn/build/lib for it, following the
usual pattern.
Differential Revision: https://reviews.llvm.org/D56316
llvm-svn: 350407
Nirav Dave [Fri, 4 Jan 2019 17:11:15 +0000 (17:11 +0000)]
Undo r350355 "[X86] Remove terrible DX Register parsing hack in parse operand. NFCI."
Add missing test case and update comments.
llvm-svn: 350406
Alexey Bataev [Fri, 4 Jan 2019 17:09:12 +0000 (17:09 +0000)]
[OPENMP][NVPTX]Improve performance + reduce number of used registers.
Summary:
Reduced number of the used register + improved performance propagating
the information about current execution/data sharing mode directly from
the compiler, where it is possible.
In some cases, it requires new/reworked interfaces of the runtime
external functions. Old functions are marked as deprecated.
Reviewers: grokos, gtbercea, kkwli0
Subscribers: guansong, jfb, openmp-commits, caomhin
Differential Revision: https://reviews.llvm.org/D56278
llvm-svn: 350405
Aaron Ballman [Fri, 4 Jan 2019 16:58:14 +0000 (16:58 +0000)]
Refactor the way we handle diagnosing unused expression results.
Rather than sprinkle calls to DiagnoseUnusedExprResult() around in places where we want diagnostics, we now diagnose unused expression statements and full expressions in a more generic way when acting on the final expression statement. This results in more appropriate diagnostics for [[nodiscard]] where we were previously lacking them, such as when the body of a for loop is not a compound statement.
This patch fixes PR39837.
llvm-svn: 350404
Simon Pilgrim [Fri, 4 Jan 2019 16:55:57 +0000 (16:55 +0000)]
[CostModel][X86] Fix SSE1 FADD/FSUB costs
Noticed in D56011 - handle the case that scalar fp ops are quicker on P3 than P4
Add the other costs so that we're not relying on the default "is legal/custom" cost logic.
llvm-svn: 350403
Ranjeet Singh [Fri, 4 Jan 2019 16:39:10 +0000 (16:39 +0000)]
Revert patches 348835 and 348571 because they're
causing code size performance regressions.
llvm-svn: 350402
Simon Pilgrim [Fri, 4 Jan 2019 16:37:01 +0000 (16:37 +0000)]
[CostModel][X86] Add SSE1 fp cost tests
llvm-svn: 350401
Mark Searles [Fri, 4 Jan 2019 16:35:01 +0000 (16:35 +0000)]
Fix typo: "with he MODULE" -> "with the MODULE"
Differential Revision: https://reviews.llvm.org/D56302
llvm-svn: 350400
Simon Pilgrim [Fri, 4 Jan 2019 15:43:43 +0000 (15:43 +0000)]
[X86] Add VPSLLI/VPSRLI ((X >>u C1) << C2) SimplifyDemandedBits combine
Repeat of the generic SimplifyDemandedBits shift combine
llvm-svn: 350399
Erich Keane [Fri, 4 Jan 2019 15:24:06 +0000 (15:24 +0000)]
Prevent unreachable when checking invalid multiversion decls.
CPUSpecifc/CPUDispatch call resolution assumed that all declarations
that would be passed are valid, however this was an invalid assumption.
This patch deals with those situations by making the valid version take
priority. Note that the checked ordering is arbitrary, since both are
replaced by calls to the resolver later.
Change-Id: I7ff2ec88c55a721d51bc1f39ea1a1fe242b4e45f
llvm-svn: 350398
Andrea Di Biagio [Fri, 4 Jan 2019 15:08:38 +0000 (15:08 +0000)]
[MCA] Improved handling of in-order issue/dispatch resources.
Added field 'MustIssueImmediately' to the instruction descriptor of instructions
that only consume in-order issue/dispatch processor resources.
This speeds up queries from the hardware Scheduler, and gives an average ~5%
speedup on a release build.
No functional change intended.
llvm-svn: 350397
Simon Pilgrim [Fri, 4 Jan 2019 14:56:10 +0000 (14:56 +0000)]
[X86] Split immediate shifts tests. NFCI.
A future patch will combine logical shifts more aggressively.
llvm-svn: 350396
Florian Hahn [Fri, 4 Jan 2019 14:53:22 +0000 (14:53 +0000)]
[ValueTracking] Fix a misuse of APInt in GetPointerBaseWithConstantOffset
GetPointerBaseWithConstantOffset include this code, where ByteOffset
and GEPOffset are both of type llvm::APInt :
ByteOffset += GEPOffset.getSExtValue();
The problem with this line is that getSExtValue() returns an int64_t, but
the += matches an overload for uint64_t. The problem is that the resulting
APInt is no longer considered to be signed. That in turn causes assertion
failures later on if the relevant pointer type is > 64 bits in width and
the GEPOffset was negative.
Changing it to
ByteOffset += GEPOffset.sextOrTrunc(ByteOffset.getBitWidth());
resolves the issue and explicitly performs the sign-extending
or truncation. Additionally, instead of asserting later if the result
is > 64 bits, it breaks out of the loop in that case.
See also
https://reviews.llvm.org/D24729
https://reviews.llvm.org/D24772
This commit must be merged after D38662 in order for the test to pass.
Patch by Michael Ferguson <mpfergu@gmail.com>.
Reviewers: reames, sanjoy, hfinkel
Reviewed By: hfinkel
Differential Revision: https://reviews.llvm.org/D38501
llvm-svn: 350395
Nico Weber [Fri, 4 Jan 2019 13:48:58 +0000 (13:48 +0000)]
[gn build] Make write_cmake_config.py check that each key passed is unique
I got that wrong once while locally while working on check-llvm.
Reviewed as part of https://reviews.llvm.org/D56195
llvm-svn: 350394
Stefan Granitz [Fri, 4 Jan 2019 12:47:02 +0000 (12:47 +0000)]
[CMake] Python bindings generation polishing
Summary:
Simplify SWIG invocation and handling of generated files.
The `swig_wrapper` target can generate `LLDBWrapPython.cpp` and `lldb.py` in its own binary directory, so we can get rid of a few global variables and their logic. We can use the swig_wrapper's BINARY_DIR target property to refer to it and liblldb's LIBRARY_OUTPUT_DIRECTORY to refer to the framework/shared object output directory.
Reviewers: JDevlieghere, aprantl, stella.stamenova, beanz, zturner, xiaobai
Reviewed By: aprantl
Subscribers: mgorny, lldb-commits, #lldb
Differential Revision: https://reviews.llvm.org/D55332
llvm-svn: 350393
Stefan Granitz [Fri, 4 Jan 2019 12:46:57 +0000 (12:46 +0000)]
[CMake] Revised RPATH handling
Summary:
If we build LLDB.framework, dependant tools need appropriate RPATHs in both locations, the build-tree (for testing) and the install-tree (for deployment). Luckily, CMake can handle it for us: https://gitlab.kitware.com/cmake/community/wikis/doc/cmake/RPATH-handling.
* In the build-tree, tools use the absolute path to the framework's actual output location.
* In the install-tree, tools get a list of RPATHs to look for the framework when deployed.
`LLDB_FRAMEWORK_INSTALL_DIR` is added to the `CMAKE_INSTALL_PREFIX` to change the relative location of LLDB.framework in the install-tree.
If it is not empty, it will be added as an additional RPATH to all dependant tools (so they are functional in the install-tree).
If it is empty, LLDB.framework goes to the root and tools will not be functional in the directory structure of the LLVM install-tree.
For historical reasons `LLDB_FRAMEWORK_INSTALL_DIR` defaults to "Library/Frameworks".
Reviewers: xiaobai, JDevlieghere, aprantl, clayborg
Reviewed By: JDevlieghere
Subscribers: ki.stfu, mgorny, lldb-commits, #lldb
Differential Revision: https://reviews.llvm.org/D55330
llvm-svn: 350392
Stefan Granitz [Fri, 4 Jan 2019 12:46:50 +0000 (12:46 +0000)]
[CMake] Revised LLDB.framework builds
Summary:
Add features to LLDB CMake builds that have so far only been available in Xcode. Clean up a few inconveniences and prepare further improvements.
Options:
* `LLDB_FRAMEWORK_BUILD_DIR` determines target directory (in build-tree)
* `LLDB_FRAMEWORK_INSTALL_DIR` **only** determines target directory in install-tree
* `LLVM_EXTERNALIZE_DEBUGINFO` allows externalized debug info (dSYM on Darwin, emitted to `bin`)
* `LLDB_FRAMEWORK_TOOLS` determines which executables will be copied to the framework's Resources (dropped symlinking, removed INCLUDE_IN_SUITE, removed dummy targets)
Other changes:
* clean up `add_lldb_executable()`
* include `LLDBFramework.cmake` from `source/API/CMakeLists.txt`
* use `*.plist.in` files, which are typical for CMake and independent from Xcode
* add clang headers to the framework bundle
Reviewers: xiaobai, JDevlieghere, aprantl, davide, beanz, stella.stamenova, clayborg, labath
Reviewed By: aprantl
Subscribers: friss, mgorny, lldb-commits, #lldb
Differential Revision: https://reviews.llvm.org/D55328
llvm-svn: 350391
Stefan Granitz [Fri, 4 Jan 2019 12:46:38 +0000 (12:46 +0000)]
[CMake] Move debugserver options to separate debugserverConfig.cmake
Summary:
One place for debugserver options, analog to LLDBConfig for LLDB options (see D55317). It was discussed in earlier reviews already, e.g. D55013.
Reviewers: JDevlieghere, aprantl, xiaobai
Reviewed By: aprantl, xiaobai
Subscribers: mgorny, lldb-commits, #lldb
Differential Revision: https://reviews.llvm.org/D55320
llvm-svn: 350390
Stefan Granitz [Fri, 4 Jan 2019 12:46:34 +0000 (12:46 +0000)]
[CMake] Aggregate options for LLDB in LLDBConfig.cmake
Summary: In preparation for LLDB.framework changes, collect options for LLDB in LLDBConfig.cmake (used for both, standalone and in-tree builds of LLDB).
Reviewers: JDevlieghere, aprantl, xiaobai
Reviewed By: aprantl
Subscribers: srhines, mgorny, lldb-commits, #lldb
Differential Revision: https://reviews.llvm.org/D55317
llvm-svn: 350389
Stefan Granitz [Fri, 4 Jan 2019 12:46:30 +0000 (12:46 +0000)]
[CMake] Streamline code signing for debugserver #2
Summary:
Major fixes after D54476 (use Diff1 as base for comparison to see only recent changes):
* In standalone builds target directory for debugserver must be LLDB's bin, not LLVM's bin
* Default identity for code signing must not force-override LLVM_CODESIGNING_IDENTITY globally
We have a lot of cases, make them explicit:
* ID used for code signing (debugserver and in tests):
** `LLDB_CODESIGN_IDENTITY` if set explicitly, or otherwise
** `LLVM_CODESIGNING_IDENTITY` if set explicitly, or otherwise
** `lldb_codesign` as the default
* On Darwin we have a debugserver target that:
* On other systems, the debugserver target is not defined, which is equivalent to **[3A]**
Common configurations on Darwin:
* **[1A]** `cmake -GNinja ../llvm` builds debugserver from source and signs with `lldb_codesign`, no code signing for other binaries (prints status: //lldb debugserver: /path/to/bin/debugserver//)
* **[1A]** `cmake -GNinja -DLLVM_CODESIGNING_IDENTITY=- -DLLDB_CODESIGN_IDENTITY=lldb_codesign ../llvm` builds debugserver from source and signs with `lldb_codesign`, ad-hoc code signing for other binaries (prints status: //lldb debugserver: /path/to/bin/debugserver//)
* **[2A]** `cmake -GNinja -DLLVM_CODESIGNING_IDENTITY=- -DLLDB_USE_SYSTEM_DEBUGSERVER=ON ../llvm` copies debugserver from system, ad-hoc code signing for other binaries (prints status: //Copy system debugserver from: /path/to/system/debugserver//)
* **[2B]** `cmake -GNinja -DLLVM_CODESIGNING_IDENTITY=- ../llvm` same, but prints additional warning: //Cannot code sign debugserver with identity '-'. Will fall back to system's debugserver. Pass -DLLDB_CODESIGN_IDENTITY=lldb_codesign to override the LLVM value for debugserver.//
* **[3A]** `cmake -GNinja -DLLVM_CODESIGNING_IDENTITY=- -DLLDB_NO_DEBUGSERVER=ON ../llvm` debugserver not available (prints status: //lldb debugserver will not be available)//
Reviewers: JDevlieghere, beanz, davide, vsk, aprantl, labath
Reviewed By: JDevlieghere, labath
Subscribers: mgorny, #lldb, lldb-commits
Differential Revision: https://reviews.llvm.org/D55013
llvm-svn: 350388
Andrea Di Biagio [Fri, 4 Jan 2019 12:31:14 +0000 (12:31 +0000)]
[MCA] Store extra information about processor resources in the ResourceManager.
Method ResourceManager::use() is responsible for updating the internal state of
used processor resources, as well as notifying resource groups that contain used
resources.
Before this patch, method 'use()' didn't know how to quickly obtain the set of
groups that contain a particular resource unit. It had to discover groups by
perform a potentially slow search (done by iterating over the set of processor
resource descriptors).
With this patch, the relationship between resource units and groups is stored in
the ResourceManager. That means, method 'use()' no longer has to search for
groups. This gives an average speedup of ~4-5% on a release build.
This patch also adds extra code comments in ResourceManager.h to better describe
the resource mask layout, and how resouce indices are computed from resource
masks.
llvm-svn: 350387
Anastasia Stulova [Fri, 4 Jan 2019 11:50:36 +0000 (11:50 +0000)]
[Basic] Extend DiagnosticEngine to store and format Qualifiers.
Qualifiers can now be streamed into the DiagnosticEngine using
regular << operator. If Qualifiers are empty 'unqualified' will
be printed in the diagnostic otherwise regular qual syntax is
used.
Differential Revision: https://reviews.llvm.org/D56198
llvm-svn: 350386
Diogo N. Sampaio [Fri, 4 Jan 2019 11:04:18 +0000 (11:04 +0000)]
[AArch64] Add command-line option predres
Prediction control instructions are only
mandatory from v8.5a onwards but is optional
from Armv8.0-A. This patch adds a command
line option to enable it by it's own.
Differential Revision: https://reviews.llvm.org/D56007
llvm-svn: 350385
Pavel Labath [Fri, 4 Jan 2019 10:11:25 +0000 (10:11 +0000)]
Symtab: Remove one copy of symbol size computation code
Summary:
The implementation in CalculateSymbolSizes has been made redundant in
D19004, as this patch added another copy of size computation code into
InitAddressIndexes (which is called by CalculateSymbolSizes).
Reviewers: clayborg, jasonmolenda, tberghammer
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D56132
llvm-svn: 350384
Stefan Granitz [Fri, 4 Jan 2019 09:22:32 +0000 (09:22 +0000)]
[CMake] Use XCODE_ATTRIBUTE properties for code signing and entitlements in Xcode
Summary: A post-commit comment to D55116 amended that this was the correct way for code signing in Xcode.
Reviewers: beanz
Reviewed By: beanz
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D55816
llvm-svn: 350383
Pavel Labath [Fri, 4 Jan 2019 07:14:17 +0000 (07:14 +0000)]
RangeMap.h: merge RangeDataArray and RangeDataVector
Summary:
The main difference between the classes was supposed to be the fact that
one is backed by llvm::SmallVector, and the other by std::vector.
However, over the years, they have accumulated various other differences
too.
This essentially removes the std::vector version, as that is pretty much
identical to llvm::SmallVector<T, 0>, and combines their interfaces. It
does not attempt to do a more significant refactoring, even though there
is still a lot of duplication in this file, as it is hard to tell which
quirk of some API is depended on by somebody (and, a previous, more
ambitious attempt at this in D16769 has failed).
I also add some tests, including one which demonstrates one of the
quirks/bugs of the API I have noticed in the process.
Reviewers: clayborg, teemperor, tberghammer
Subscribers: mgorny, JDevlieghere, lldb-commits
Differential Revision: https://reviews.llvm.org/D56170
llvm-svn: 350380
Richard Trieu [Fri, 4 Jan 2019 06:49:24 +0000 (06:49 +0000)]
[WebAssembly] Split the checking from the sorting logic.
Move the check for -1 and identical values outside the vector sorting code.
Compare functions need to be able to compare identical elements to be
conforming.
llvm-svn: 350379
Xin Tong [Fri, 4 Jan 2019 02:13:22 +0000 (02:13 +0000)]
[memcpyopt] Remove a few unnecessary isVolatile() checks. NFC
We already checked for isSimple() on the store.
llvm-svn: 350378
Joel E. Denny [Fri, 4 Jan 2019 02:07:13 +0000 (02:07 +0000)]
[OpenMP] Fix nvidia-cuda-toolkit detection on Debian/Ubuntu
The OpenMP runtime's cmake scripts do not correctly locate the
libdevice that the Debian/Ubuntu package nvidia-cuda-toolkit currently
includes, at least on my Ubuntu 18.04.1 installation. This patch
fixes that for me.
This problem was discussed at length in D55269. D40453 added a
similar adjustment in clang, but reviewers of D55269 concluded that,
for the OpenMP runtime, the right place to address this problem is in
cmake's CUDA support. However, it was also suggested we could add a
workaround to OpenMP's cmake scripts now. This patch contains such a
workaround, which I've tried to design so that it will have no harmful
effect if cmake improves in the future.
nvidia-cuda-toolkit also needs improvements because its intended
monolithic CUDA tree shim, /usr/lib/cuda, has many empty directories,
such as bin. I reported that at:
<https://bugs.launchpad.net/ubuntu/+source/nvidia-cuda-toolkit/+bug/1808999>
Reviewed By: grokos
Differential Revision: https://reviews.llvm.org/D55588
llvm-svn: 350377
Kuba Mracek [Fri, 4 Jan 2019 00:25:08 +0000 (00:25 +0000)]
[lldb] Fix ObjCExceptionRecognizedStackFrame to populate the list of recognized arguments
Differential Revision: https://reviews.llvm.org/D56027
llvm-svn: 350376
Kuba Mracek [Fri, 4 Jan 2019 00:20:52 +0000 (00:20 +0000)]
[lldb] Check SafeToCallFunctions before calling functions in GetExceptionObjectForThread
Differential Revision: https://reviews.llvm.org/D56115
llvm-svn: 350375
Craig Topper [Fri, 4 Jan 2019 00:10:58 +0000 (00:10 +0000)]
[X86] Add post-isel peephole to fold KAND+KORTEST into KTEST if only the zero flag is used.
Doing this late so we will prefer to fold the AND into a masked comparison first. That can be better for the live range of the mask register.
Differential Revision: https://reviews.llvm.org/D56246
llvm-svn: 350374
Sanjay Patel [Fri, 4 Jan 2019 00:02:02 +0000 (00:02 +0000)]
revert r350369: [x86] lower extracted fadd/fsub to horizontal vector math
There are non-codegen tests that need to be updated with this code change.
llvm-svn: 350373
Daniel Dunbar [Thu, 3 Jan 2019 23:24:50 +0000 (23:24 +0000)]
Adopt SwiftABIInfo for WebAssembly.
Summary:
- This adopts SwiftABIInfo as the base class for WebAssemblyABIInfo, which is in keeping with what is done for other targets for which Swift is supported.
- This is a minimal patch to unblock exploration of WASM support for Swift (https://bugs.swift.org/browse/SR-9307)
Reviewers: rjmccall, sunfish
Reviewed By: rjmccall
Subscribers: ahti, dschuff, sbc100, jgravelle-google, aheejin, cfe-commits
Differential Revision: https://reviews.llvm.org/D56188
llvm-svn: 350372
Mike Spertus [Thu, 3 Jan 2019 23:24:39 +0000 (23:24 +0000)]
Fix MSVC visualizations for ActionResult and OpaquePtr
Mainly just fixing buggy code. Also removed unnecessary visualizers
for specializations of OpaquePtr
llvm-svn: 350371
Evgeniy Stepanov [Thu, 3 Jan 2019 23:19:02 +0000 (23:19 +0000)]
[hwasan] Switch to 64 allocator with a dense size class map.
Summary:
Replace the 32-bit allocator with a 64-bit one with a non-constant
base address, and reduce both the number of size classes and the maximum
size of per-thread caches.
As measured on [1], this reduces average weighted memory overhead
(MaxRSS) from 26% to 12% over stock android allocator. These numbers
include overhead from code instrumentation and hwasan shadow (i.e. not a
pure allocator benchmark).
This switch also enables release-to-OS functionality, which is not
implemented in the 32-bit allocator. I have not seen any effect from
that on the benchmark.
[1] https://android.googlesource.com/platform/system/extras/+/master/memory_replay/
Reviewers: vitalybuka, kcc
Subscribers: kubamracek, cryptoad, llvm-commits
Differential Revision: https://reviews.llvm.org/D56239
llvm-svn: 350370
Sanjay Patel [Thu, 3 Jan 2019 23:16:19 +0000 (23:16 +0000)]
[x86] lower extracted fadd/fsub to horizontal vector math
This would show up if we fix horizontal reductions to narrow as they go along,
but it's an improvement for size and/or Jaguar (fast-hops) independent of that.
We need to do this late to not interfere with other pattern matching of larger
horizontal sequences.
We can extend this to integer ops in a follow-up patch.
Differential Revision: https://reviews.llvm.org/D56011
llvm-svn: 350369