Commit Graph

1403 Commits

Author SHA1 Message Date
Georgii Rymar 76e0ab23f6 [FileCheck] - Refactor the code related to string arrays. NFCI.
There are few `std::vector<std::string>` members in
`FileCheckRequest`. This patch changes these arrays to `std::vector<StringRef>`
and refactors the code related to cleanup/improve/simplify it.

Differential revision: https://reviews.llvm.org/D78202
2020-04-20 14:54:49 +03:00
Yi-Hong Lyu 54cfc6944e [CommandLine] Fix cl::ConsumeAfter support with more than one positional argument
Summary:
Currently, cl::ConsumeAfter only works for the case that has exactly one
positional argument. Without the fix, it skip fulfilling first positional
argument and put that additional positional argument in interpreter arguments.

Reviewers: bkramer, Mordante, rnk, lattner, beanz, craig.topper

Reviewed By: rnk

Subscribers: JosephTremoulet, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77242
2020-04-17 02:12:54 -07:00
Chris Lattner 953a814aae Remove the llvm/Support/StringPool.h file and related support now that it has no clients. A plain old StringSet<> is a better replacement.
Subscribers: mgorny, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D78336
2020-04-16 17:57:39 -07:00
Sergej Jaskiewicz 5cef31074f Introduce llvm::sys::Process::getProcessId() and adopt it
Differential Revision: https://reviews.llvm.org/D78022
2020-04-16 15:05:37 +03:00
Michael Spencer bca47ef80e [LLVM][SupportTests] Ask the OS how large the page size is instead of guessing.
PPC64 had a larger than expected page size. Instead of guessing just
use the same API that `MemoryBuffer` does to determine the page size.
2020-04-15 15:12:28 -07:00
Michael Spencer 18ee0fca8c [LLVM][SupportTests] Fix Windows build breakage
`MemoryBuffer::getOpenFile` take an OS file handle, not an int.
2020-04-15 14:47:27 -07:00
Michael Spencer 92e8af0ecb [Clang] Expose RequiresNullTerminator in FileManager.
This is needed to fix the reason
0a2be46cfd (Modules: Invalidate out-of-date PCMs as they're
discovered) and 5b44a4b07fc1d ([modules] Do not cache invalid state for
modules that we attempted to load.) were reverted.

These patches changed Clang to use `isVolatile` when loading modules.
This had the side effect of not using mmap when loading modules, and
thus greatly increased memory usage.

The reason it wasn't using mmap is because `MemoryBuffer` plays some
games with file size when you request null termination, and it has to
disable these when `isVolatile` is set as the size may change by the
time it's mmapped. Clang by default passes
`RequiresNullTerminator = true`, and `shouldUseMmap` ignored if
`RequiresNullTerminator` was even requested.

This patch adds `RequiresNullTerminator` to the `FileManager` interface
so Clang can use it when loading modules, and changes `shouldUseMmap` to
only take volatility into account if `RequiresNullTerminator` is true.
This is fine as both `mmap` and a `read` loop are vulnerable to
modifying the file while reading, but are immune to the rename Clang
does when replacing a module file.

Differential Revision: https://reviews.llvm.org/D77772
2020-04-15 14:17:51 -07:00
Thomas Preud'homme 9743123af8 [FileCheck] Better diagnostic for format conflict
Summary:
Improve error message in case of conflict between several implicit
format to mention the operand that conflict.

Reviewers: jhenderson, jdenny, probinson, grimar, arichardson, rnk

Reviewed By: jdenny

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77741
2020-04-15 13:46:45 +01:00
River Riddle 204c3b5516 [llvm][STLExtras] Move various iterator/range utilities from MLIR to LLVM
This revision moves the various range utilities present in MLIR to LLVM to enable greater reuse. This revision moves the following utilities:

* indexed_accessor_*
This is set of utility iterator/range base classes that allow for building a range class where the iterators are represented by an object+index pair.

* make_second_range
Given a range of pairs, returns a range iterating over the `second` elements.

* hasSingleElement
Returns if the given range has 1 element. size() == 1 checks end up being very common, but size() is not always O(1) (e.g., ilist). This method provides O(1) checks for those cases.

Differential Revision: https://reviews.llvm.org/D78064
2020-04-14 15:14:40 -07:00
Georgii Rymar 1647ff6e27 [ADT/STLExtras.h] - Add llvm::is_sorted wrapper and update callers.
It can be used to avoid passing the begin and end of a range.
This makes the code shorter and it is consistent with another
wrappers we already have.

Differential revision: https://reviews.llvm.org/D78016
2020-04-14 14:11:02 +03:00
Lang Hames e823068306 [Support] Add support RTTI support for open class hierarchies.
This patch extracts the RTTI part of llvm::ErrorInfo into its own class
(RTTIExtends) so that it can be used in other non-error hierarchies, and makes
it compatible with the existing LLVM RTTI function templates (isa, cast,
dyn_cast, dyn_cast_or_null) by adding the classof method.

Differential Revision: https://reviews.llvm.org/D39111
2020-04-13 12:52:44 -07:00
John McCall 8423a6f363 Rename OptimalLayout to OptimizedStructLayout at Chris's request. 2020-04-10 00:14:20 -04:00
Jay Foad c63aed890e [KnownBits] Move AND, OR and XOR logic into KnownBits
Summary:
There are at least three clients for KnownBits calculations:
ValueTracking, SelectionDAG and GlobalISel. To reduce duplication the
common logic should be moved out of these clients and into KnownBits
itself.

This patch does this for AND, OR and XOR calculations by implementing
and using appropriate operator overloads KnownBits::operator& etc.

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74060
2020-04-09 10:10:37 +01:00
Pavel Labath 8f1233699b [llvm/Support] Don't crash on empty nullptr ranges when decoding LEBs
Summary:
If the decoding functions are called with both start and end pointers
being nullptr, the function will crash due to a nullptr dereference.
This happens because the function does not recognise nullptr as a valid
end pointer.

Obviously, nobody is going to pass null pointers here deliberately, but
it can happen indirectly (as it did for me), when calling these
functions on an ArrayRef, as a default-initialized empty ArrayRef will
have both begin() and end() pointers equal to nullptr.

The fix is to simply remove the nullptr check. Passing nullptr for "end"
with a valid "begin" pointer will still work, as one cannot reach
nullptr by incrementing a valid pointer without triggerring UB.

Reviewers: dblaikie

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77304
2020-04-06 14:14:11 +02:00
Pavel Labath 9154a6398e [llvm/Support] Make more DataExtractor methods error-aware
Summary:
This patch adds the optional Error argument, and the Cursor variants to
more DataExtractor methods. The functions now behave the same way as
other error-aware functions (they set the error when they fail, and
don't do anything if the error is already set).

I have merged the LEB128 implementations via a template (similarly to
how fixed-size functions are handled) to reduce code duplication.

Depends on D77304.

Reviewers: dblaikie, aprantl

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77306
2020-04-06 14:14:11 +02:00
Pavel Labath a16fffa3f6 [Support] Make DataExtractor string functions error-aware
Summary:
This patch adds an optional Error argument to DataExtractor functions
for string extraction, and makes them behave like other DataExtractor
functions (set the error if extraction fails, don't do anything if the
error is already set).

I have merged the StringRef and C string versions of the functions to
reduce code duplication.

Reviewers: dblaikie, MaskRay

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D77307
2020-04-06 14:14:11 +02:00
Sylvain Audi e4ae0a2e97 [Support/Path] sys::path::replace_path_prefix fix and simplifications
Added unit tests for 2 scenarios that were failing.
Made replace_path_prefix back to 3 parameters instead of 5, simplifying the implementation. The other 2 were always used with the default value.

This commit is intended to be the first of 3:
1) simplify/fix replace_path_prefix.
2) use it in the context of -fdebug-prefix-map and -fmacro-prefix-map (see D76869).
3) Make Windows version of replace_path_prefix insensitive to both case and separators (slash vs backslash).

Differential Revision: https://reviews.llvm.org/D77223
2020-04-03 13:50:23 -04:00
Kai Wang 501522b5b2 [RISCV] Support RISC-V ELF attributes sections in llvm-readobj.
Enable llvm-readobj to handle RISC-V ELF attribute sections.

Differential Revision: https://reviews.llvm.org/D75833
2020-04-01 21:50:11 +08:00
Kai Wang 581ba35291 [RISCV] ELF attribute section for RISC-V.
Leverage ARM ELF build attribute section to create ELF attribute section
for RISC-V. Extract the common part of parsing logic for this section
into ELFAttributeParser.[cpp|h] and ELFAttributes.[cpp|h].

Differential Revision: https://reviews.llvm.org/D74023
2020-03-31 16:16:19 +08:00
Jonas Devlieghere 4151f2d04a Re-land "[FileCollector] Add a method to add a whole directory and it contents."
Extend the FileCollector's API with addDirectory which adds a directory
and its contents to the VFS mapping.

Differential revision: https://reviews.llvm.org/D76671
2020-03-30 13:19:18 -07:00
Jonas Devlieghere 190df4a5bc Revert "[FileCollector] Add a method to add a whole directory and it contents."
This reverts commit 8913769e35 because the
unit test is failing on the Windows bot.
2020-03-27 19:21:48 -07:00
Jonas Devlieghere 8913769e35 [FileCollector] Add a method to add a whole directory and it contents.
Extend the FileCollector's API with addDirectory which adds a directory
and its contents to the VFS mapping.

Differential revision: https://reviews.llvm.org/D76671
2020-03-27 17:38:24 -07:00
Jonas Devlieghere 3ef33e69de [VirtualFileSystem] Support directory entries in the YAMLVFSWriter
The current implementation of the JSONWriter does not support writing
out directory entries. Earlier today I added a unit test to illustrate
the problem. When an entry is added to the YAMLVFSWriter and the path is
a directory, it will incorrectly emit the directory as a file, and any
files inside that directory will not be found by the VFS.

It's possible to partially work around the issue by only adding "leaf
nodes" (files) to the YAMLVFSWriter. However, this doesn't work for
representing empty directories. This is a problem for clients of the VFS
that want to iterate over a directory. The directory not being there is
not the same as the directory being empty.

This is not just a hypothetical problem. The FileCollector for example
does not differentiate between file and directory paths. I temporarily
worked around the issue for LLDB by ignoring directories, but I suspect
this will prove problematic sooner rather than later.

This patch fixes the issue by extending the JSONWriter to support
writing out directory entries. We store whether an entry should be
emitted as a file or directory.

Differential revision: https://reviews.llvm.org/D76670
2020-03-27 15:16:52 -07:00
Ties Stuij 71ae267d1f [PATCH] [ARM] ARMv8.6-a command-line + BFloat16 Asm Support
Summary:
This patch introduces command-line support for the Armv8.6-a architecture and assembly support for BFloat16. Details can be found
https://community.arm.com/developer/ip-products/processors/b/processors-ip-blog/posts/arm-architecture-developments-armv8-6-a

in addition to the GCC patch for the 8..6-a CLI:
https://gcc.gnu.org/legacy-ml/gcc-patches/2019-11/msg02647.html

In detail this patch

- march options for armv8.6-a
- BFloat16 assembly

This is part of a patch series, starting with command-line and Bfloat16
assembly support. The subsequent patches will upstream intrinsics
support for BFloat16, followed by Matrix Multiplication and the
remaining Virtualization features of the armv8.6-a architecture.

Based on work by:
- labrinea
- MarkMurrayARM
- Luke Cheeseman
- Javed Asbar
- Mikhail Maltsev
- Luke Geeson

Reviewers: SjoerdMeijer, craig.topper, rjmccall, jfb, LukeGeeson

Reviewed By: SjoerdMeijer

Subscribers: stuij, kristof.beyls, hiraditya, dexonsmith, danielkiss, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D76062
2020-03-26 09:17:20 +00:00
John McCall 49e5a97ec3 Add an algorithm for performing "optimal" layout of a struct.
The algorithm supports both assigning a fixed offset to a field prior to
layout and allowing fields to have sizes that aren't multiples of their
required alignments.  This means that the well-known algorithm of sorting
by decreasing alignment isn't always good enough.  Still, we start with
that, and only if that leaves padding around do we fall back on a greedy
padding-minimizing algorithm.

There is no known efficient algorithm for producing a guaranteed-minimal
layout in all cases.  In fact, allowing arbitrary fixed-offset fields means
there's a straightforward reduction from bin-packing, making this NP-hard.
But as usual with such problems, we can still efficiently produce adequate
solutions to the cases that matter most to us.

I intend to use this in coroutine frame layout, where the retcon lowerings
very badly want to minimize total space usage, and where the switch lowering
can indeed produce a header with interior padding if the promise field is
highly-aligned.  But it may be useful in a much wider variety of situations.
2020-03-23 23:24:48 -04:00
Jonas Devlieghere 42df3e2985 [VirtualFileSystem] Add unit test for vfs::YAMLVFSWriter
Add a unit test for vfs::YAMLVFSWriter.

This patch exposes an issue in the writer: when we call addFileMapping
with a directory, the VFS writer will emit it as a regular file, causing
any of the nested files or directories to not be found.
2020-03-23 18:49:06 -07:00
Alexandre Ganea c1f8595fe5 [Support] Silence warning in Path unittests when compiling with clang-cl
warning: comparison of integers of different signs: 'const unsigned long long' and 'const int' [-Wsign-compare]
2020-03-23 12:20:18 -04:00
Andrew Ng e6f6c55121 [Support] Improve Windows widenPath and add support for long UNC paths
Check the path length limit against the length of the UTF-16 version of
the input rather than the UTF-8 equivalent, as the UTF-16 length may be
shorter. Move widenPath from the llvm::sys::path namespace in Path.h to
the llvm::sys::windows namespace in WindowsSupport.h. Only use the
reduced path length limit for create directory. Canonicalize using
sys::path::remove_dots().

Differential Revision: https://reviews.llvm.org/D75372
2020-03-19 13:00:21 +00:00
Guillaume Chatelet a060a1782e [Alignment] Add alignTo with skew parameter
Summary: courbet

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D76274
2020-03-17 15:54:47 +01:00
Reid Kleckner 213aea4c58 Remove unused Endian.h includes, NFC
Mainly avoids including Host.h everywhere:

$ diff -u <(sort thedeps-before.txt) <(sort thedeps-after.txt) \
    | grep '^[-+] ' | sort | uniq -c | sort -nr
   3141 - /usr/local/google/home/rnk/llvm-project/llvm/include/llvm/Support/Host.h
2020-03-11 15:45:34 -07:00
KAWASHIMA Takahiro c8cd1a994d [AArch64] Add support for Fujitsu A64FX
A64FX is an Armv8.2-A CPU used in FUJITSU Supercomputer
PRIMEHPC FX1000, PRIMEHPC FX700, and supercomputer Fugaku.

https://www.fujitsu.com/global/products/computing/servers/supercomputer/specifications/

Differential Revision: https://reviews.llvm.org/D75594
2020-03-09 19:15:09 +09:00
Fangrui Song 791efb148f [ARM] Rewrite ARMAttributeParser
* Delete boilerplate
* Change functions to return `Error`
* Test parsing errors
* Update callers of ARMAttributeParser::parse() to check the `Error` return value.

Since this patch touches nearly everything in the file, I apply
http://llvm.org/docs/Proposals/VariableNames.html and change variable
names to lower case.

Reviewed By: compnerd

Differential Revision: https://reviews.llvm.org/D75015
2020-03-05 10:57:27 -08:00
serge-sans-paille 1454c27b60 Syndicate, test and fix base64 implementation
llvm/Support/Base64, fix its implementation and provide a decent test suite.

Previous implementation code was using + operator instead of | to combine

results, which is a problem when shifting signed values. (0xFF << 16) is
implicitly converted to a (signed) int, and thus results in 0xffff0000,
h is
negative. Combining negative numbers with a + in that context is not what we
want to do.

This is a recommit of 5a1958f267 with UB removved.

This fixes https://github.com/llvm/llvm-project/issues/149.

Differential Revision: https://reviews.llvm.org/D75057
2020-03-03 12:17:53 +01:00
Hans Wennborg 916be8fd6a Revert abb00753 "build: reduce CMake handling for zlib" (PR44780)
and follow-ups:
a2ca1c2d "build: disable zlib by default on Windows"
2181bf40 "[CMake] Link against ZLIB::ZLIB"
1079c68a "Attempt to fix ZLIB CMake logic on Windows"

This changed the output of llvm-config --system-libs, and more
importantly it broke stand-alone builds. Instead of piling on more fix
attempts, let's revert this to reduce the risk of more breakages.
2020-03-03 11:03:09 +01:00
Mitch Phillips 49684f9db5 Revert "Syndicate, test and fix base64 implementation"
This reverts commit 5a1958f267.

This change broke the UBSan build bots. See
https://reviews.llvm.org/D75057 for more information.
2020-03-02 09:33:22 -08:00
Mitch Phillips 21e83244cf Revert "Fix Base64Test - for StringRef size"
This reverts commit b52355f8a1.

The change this patch depends on
(5a1958f267) broke the UBSan buildbots.
See https://reviews.llvm.org/D75057 for more information.
2020-03-02 09:33:21 -08:00
Luke Geeson 7d594cf003 [ARM] Add Cortex-M55 Support for clang and llvm
This patch upstreams support for the ARM Armv8.1m cpu Cortex-M55.

In detail adding support for:

 - mcpu option in clang
 - Arm Target Features in clang
 - llvm Arm TargetParser definitions

details of the CPU can be found here:
https://developer.arm.com/ip-products/processors/cortex-m/cortex-m55

Reviewers: chill

Reviewed By: chill

Subscribers: dmgreen, kristof.beyls, hiraditya, cfe-commits,
llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74966
2020-03-02 11:42:26 +00:00
serge-sans-paille b52355f8a1 Fix Base64Test - for StringRef size
Original failures: http://lab.llvm.org:8011/builders/clang-with-lto-ubuntu/builds/15975/steps/test-stage1-compiler/logs/stdio
2020-03-02 11:13:30 +01:00
serge-sans-paille 5a1958f267 Syndicate, test and fix base64 implementation
Move Base64 implementation from clangd/SemanticHighlighting to
llvm/Support/Base64, fix its implementation and provide a decent test suite.

Previous implementation code was using + operator instead of | to combine some
results, which is a problem when shifting signed values. (0xFF << 16) is
implicitly converted to a (signed) int, and thus results in 0xffff0000, which is
negative. Combining negative numbers with a + in that context is not what we
want to do.

This fixes https://github.com/llvm/llvm-project/issues/149.

Differential Revision: https://reviews.llvm.org/D75057
2020-03-02 10:02:25 +01:00
Reid Kleckner af450eabb9 Avoid including FileSystem.h from MemoryBuffer.h
Lots of headers pass around MemoryBuffer objects, but very few open
them. Let those that do include FileSystem.h.

Saves ~250 includes of Chrono.h & FileSystem.h:

$ diff -u thedeps-before.txt thedeps-after.txt | grep '^[-+] ' | sort | uniq -c | sort -nr
    254 -    ../llvm/include/llvm/Support/FileSystem.h
    253 -    ../llvm/include/llvm/Support/Chrono.h
    237 -    ../llvm/include/llvm/Support/NativeFormatting.h
    237 -    ../llvm/include/llvm/Support/FormatProviders.h
    192 -    ../llvm/include/llvm/ADT/StringSwitch.h
    190 -    ../llvm/include/llvm/Support/FormatVariadicDetails.h
...

This requires duplicating the file_t typedef, which is unfortunate. I
sunk the choice of mapping mode down into the cpp file using variable
template specializations instead of class members in headers.
2020-02-29 12:30:23 -08:00
Reid Kleckner e3a9b0f359 [Support] Remove byte swapping from MathExtras.h
MathExtras.h was just wrapping SwapByteOrder.h functionality, so have
the callers use it directly.  Use the MathExtras.h name (ByteSwap_NN) as
the standard naming, since it appears to be the most popular.
2020-02-27 17:23:48 -08:00
Greg Clayton df8dda67ed Add methods to data extractor for extracting bytes and fixed length C strings.
Summary:
These modificaitons will be used in D74883.

Fixed length C strings can have trailing NULLs or sometimes spaces (BSD archive files), so the fixed length C string defaults to stripping trailing NULLs, but can have the arguments specify to remove one or more kinds of spaces if needed. This is used to extract fixed length C strings from ELF NOTEs in D74883.

Reviewers: labath, dblaikie, aprantl

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74991
2020-02-24 14:17:43 -08:00
Fangrui Song 73d8d83a6d [ARM] Change ARMAttributeParser::Parse to use support::endianness and simplify 2020-02-21 11:05:33 -08:00
Pavel Labath b55c58a2d5 [Error/unittests] Add a FailedWithMessage gtest matcher
Summary:
We already have a "Failed" matcher, which can be used to check any
property of the Error object. However, most frequently one just wants to
check the error message, and while this is possible with the "Failed"
matcher, it is also very convoluted
(Failed<ErrorInfoBase>(testing::Property(&ErrorInfoBase::message, "the
message"))).

Now, one can just write: FailedWithMessage("the message"). I expect that
most of the usages will remain this simple, but the argument of the
matcher is not limited to simple strings -- the argument of the matcher
can be any other matcher, so one can write more complicated assertions
if needed (FailedWithMessage(ContainsRegex("foo|bar"))). If one wants to
match multiple error messages, he can pass multiple arguments to the
matcher.

If one wants to match the message list as a whole (perhaps to check the
message count), I've also included a FailedWithMessageArray matcher,
which takes a single matcher receiving a vector of error message
strings.

Reviewers: sammccall, dblaikie, jhenderson

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74898
2020-02-21 15:29:48 +01:00
Pavel Labath 9746b3ae07 ErrorTest: Break up "ErrorMatchers" test
This test was getting a bit long. Before adding more checks, group the
existing checks according to the matcher used, and break it up into
smaller tests.
2020-02-19 14:19:48 +01:00
Luke Geeson 4518aab289 [AArch64] Add Cortex-A34 Support for clang and llvm
This patch upstreams support for the AArch64 Armv8-A cpu Cortex-A34.

In detail adding support for:
 - mcpu option in clang
 - AArch64 Target Features in clang
 - llvm AArch64 TargetParser definitions

details of the cpu can be found here:
https://developer.arm.com/ip-products/processors/cortex-a/cortex-a34

Reviewers: SjoerdMeijer

Reviewed By: SjoerdMeijer

Subscribers: SjoerdMeijer, kristof.beyls, hiraditya, cfe-commits,
llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74483

Change-Id: Ida101fc544ca183a0a0e61a1277c8957855fde0b
2020-02-18 14:56:16 +00:00
Alexandre Ganea b25fc4123c [Support] In tests, fix warning: variable ‘Threads’ set but not used 2020-02-15 09:05:01 -05:00
Amy Huang cb36bfa3de Fix 01b02a73de to use correct macro spelling and fix unit tests. 2020-02-14 15:58:36 -08:00
Alexandre Ganea 8404aeb56a [Support] On Windows, ensure hardware_concurrency() extends to all CPU sockets and all NUMA groups
The goal of this patch is to maximize CPU utilization on multi-socket or high core count systems, so that parallel computations such as LLD/ThinLTO can use all hardware threads in the system. Before this patch, on Windows, a maximum of 64 hardware threads could be used at most, in some cases dispatched only on one CPU socket.

== Background ==
Windows doesn't have a flat cpu_set_t like Linux. Instead, it projects hardware CPUs (or NUMA nodes) to applications through a concept of "processor groups". A "processor" is the smallest unit of execution on a CPU, that is, an hyper-thread if SMT is active; a core otherwise. There's a limit of 32-bit processors on older 32-bit versions of Windows, which later was raised to 64-processors with 64-bit versions of Windows. This limit comes from the affinity mask, which historically is represented by the sizeof(void*). Consequently, the concept of "processor groups" was introduced for dealing with systems with more than 64 hyper-threads.

By default, the Windows OS assigns only one "processor group" to each starting application, in a round-robin manner. If the application wants to use more processors, it needs to programmatically enable it, by assigning threads to other "processor groups". This also means that affinity cannot cross "processor group" boundaries; one can only specify a "preferred" group on start-up, but the application is free to allocate more groups if it wants to.

This creates a peculiar situation, where newer CPUs like the AMD EPYC 7702P (64-cores, 128-hyperthreads) are projected by the OS as two (2) "processor groups". This means that by default, an application can only use half of the cores. This situation could only get worse in the years to come, as dies with more cores will appear on the market.

== The problem ==
The heavyweight_hardware_concurrency() API was introduced so that only *one hardware thread per core* was used. Once that API returns, that original intention is lost, only the number of threads is retained. Consider a situation, on Windows, where the system has 2 CPU sockets, 18 cores each, each core having 2 hyper-threads, for a total of 72 hyper-threads. Both heavyweight_hardware_concurrency() and hardware_concurrency() currently return 36, because on Windows they are simply wrappers over std:🧵:hardware_concurrency() -- which can only return processors from the current "processor group".

== The changes in this patch ==
To solve this situation, we capture (and retain) the initial intention until the point of usage, through a new ThreadPoolStrategy class. The number of threads to use is deferred as late as possible, until the moment where the std::threads are created (ThreadPool in the case of ThinLTO).

When using hardware_concurrency(), setting ThreadCount to 0 now means to use all the possible hardware CPU (SMT) threads. Providing a ThreadCount above to the maximum number of threads will have no effect, the maximum will be used instead.
The heavyweight_hardware_concurrency() is similar to hardware_concurrency(), except that only one thread per hardware *core* will be used.

When LLVM_ENABLE_THREADS is OFF, the threading APIs will always return 1, to ensure any caller loops will be exercised at least once.

Differential Revision: https://reviews.llvm.org/D71775
2020-02-14 10:24:22 -05:00
Justin Lebar fb45968e62 Use C++14-style return type deduction in LLVM.
Summary:
Simplifies the C++11-style "-> decltype(...)" return-type deduction.

Note that you have to be careful about whether the function return type
is `auto` or `decltype(auto)`.  The difference is that bare `auto`
strips const and reference, just like lambda return type deduction.  In
some cases that's what we want (or more likely, we know that the return
type is a value type), but whenever we're wrapping a templated function
which might return a reference, we need to be sure that the return type
is decltype(auto).

No functional change.

Subscribers: dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74383
2020-02-11 07:38:42 -08:00