Without this patch the file list of the preamble index contains URIs, but other indexes file lists contain file paths.
This makes `indexedFiles()` always returns `IndexContents::None` for the preamble index, because current implementation expects file paths inside the file list of the index.
This patch fixes this problem and also helps to avoid a lot of URI to path conversions during indexes merge.
Reviewed By: kadircet
Differential Revision: https://reviews.llvm.org/D97535
Current indexes merge logic skip data from the static index if the file is in the dynamic index, but sometimes the dynamic index does not contain references (e.g. preamble (dynamic) index vs background (static) index).
This problem is masked with the fact, that the preamble index file list consists of file URI's and other indexes file lists consist of file paths.
This patch introduces the index contents (symbols, references, etc.), which makes indexes merge more flexible and makes it able to use URI's for the index file list.
Reviewed By: sammccall
Differential Revision: https://reviews.llvm.org/D94952
This used to implicitly never happen due to only discovering each CDB
once.
We may want to carefully support reindexing one day, but we need to do
it carefully (tricky tradeoffs) and it would need further support in
background indexer.
Making this explicit here rather than just turning off rebroadcast in
background index for a few reasons:
- allows *new* files in the same CDB to be indexed
- relying on bugs-at-a-distance cancelling each other out is bound to bite us
- gets us closer to actually supporting reindexing, which requires similar tracking
Differential Revision: https://reviews.llvm.org/D94503
This was causing duplicate `symbols` components on the path as both the
edge from an index to filesymbols and filesymbols to symbolslabs were named
symbols.
Differential Revision: https://reviews.llvm.org/D89685
Summary: This will be needed to support call hierarchy
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83536
I've dropped the background context parameter, since we in practice just pass the
current context there, and we now have a different way to specify context too.
While here, clean up a couple of comments.
Reviewed By: kadircet
Differential Revision: https://reviews.llvm.org/D83157
Summary:
We only support Build/Skip for now, but with 'Load' or similar as an
option for future (load existing shards but don't build new ones).
This requires creating the config for each TU on startup. In LLVM, this
is 4000 occurrences for a total of 800ms on my machine.
But together with caching from D83755 it is only 25ms.
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83768
Summary:
ClangdServer owns the question of exactly which config to create, but
TUScheduler/BackgroundIndex control threads and so decide at which point
to inject it.
Reviewers: kadircet
Subscribers: ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83095
Summary:
This confusion was inadvertently introduced in a change to the
heavyweight_hardware_concurrency API: 8404aeb56a
- don't indirect through the rebuilder policy when building the thread pool
- document that rebuilder thresholds are exposed for testing only
- don't use 0 as a sentinel value for "all threads", as we use it as a
sentinel value for "synchronous" (though unsupported for BackgroundIndex)
- rather than pick some new sentinel value, just always use 4 threads for tests
Reviewers: kadircet, aganea
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D82352
Summary:
Clangd uses FSProvider to get threadsafe views into file systems. This
patch changes naming to make that more explicit.
Depends on D81920
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81998
Summary:
We've faced a couple of problems when the returned FS didn't have the
proper working directory. New signature makes the API safer against such
problems.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D81920
Summary:
This makes FileShardedIndex more robust and gets rid of the need for a
URIToFileCache, as it is done by the callers now and it is only once per file,
rather than once per symbol.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D79079
Summary:
This reduces memory usage by dynamic index from more than 400MB to 32MB
when all files in clang-tools-extra/clangd/*.cpp are active in clangd.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D77732
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
This is how it should've been and brings it more in line with
std::string_view. There should be no functional change here.
This is mostly mechanical from a custom clang-tidy check, with a lot of
manual fixups. It uncovers a lot of minor inefficiencies.
This doesn't actually modify StringRef yet, I'll do that in a follow-up.
Summary:
It simply shows the completed/total items on the background queue, e.g.
indexing: 233/1000
The denominator is reset to zero every time the queue goes idle.
The protocol is fairly complicated here (requires creating a remote "progress"
resource before sending updates). We implement the full protocol, but I've added
an extension allowing it to be skipped to reduce the burden on clients - in
particular the lit test takes this shortcut.
The addition of background index progress to DiagnosticConsumer seems ridiculous
at first glance, but I believe that interface is trending in the direction of
"ClangdServer callbacks" anyway. It's due for a rename, but otherwise actually
fits.
Reviewers: kadircet, usaxena95
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, jfb, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D73218
Summary:
we have a few places using `ASTCtx.getLangOpts().IsHeaderFile` to
determine a header file, but it relies on "-x c-header" compiler flag,
if the compilation command doesn't have this flag, we will get a false
positive. We are encountering this issue in bazel build system.
To solve this problem, we infer the file from file name, actual changes will
come in follow-ups.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D70235
This much better reflects what is (now) in this header.
Maybe a rename to ParsedTU would be an improvement, but that's a much
more invasive change and life is too short.
ClangdUnit is dead, long live ClangdUnitTests!
llvm-svn: 370862
Summary:
Those errors are exposed at the first character of a file,
for a lack of a better place.
Previously, all errors were stored inside the AST and report
accordingly. However, errors in command-line argument parsing could
result in failure to produce the AST, so we need an alternative ways to
report those errors.
We take the following approach in this patch:
- buildCompilerInvocation() now requires an explicit DiagnosticConsumer.
- TUScheduler and TestTU now collect the diagnostics produced when
parsing command line arguments.
If pasing of the AST failed, diagnostics are reported via a new
ParsingCallbacks::onFailedAST method.
If parsing of the AST succeeded, any errors produced during
command-line parsing are stored alongside the AST inside the
ParsedAST instance and reported as previously by calling the
ParsingCallbacks::onMainAST method;
- The client code that uses ClangdServer's DiagnosticConsumer
does not need to change, it will receive new diagnostics in the
onDiagnosticsReady() callback
Errors produced when parsing command-line arguments are collected using
the same StoreDiags class that is used to collect all other errors. They
are recognized by their location being invalid. IIUC, the location is
invalid as there is no source manager at this point, it is created at a
later stage.
Although technically we might also get diagnostics that mention the
command-line arguments FileID with after the source manager was created
(and they have valid source locations), we choose to not handle those
and they are dropped as not coming from the main file. AFAICT, those
diagnostics should always be notes, therefore it's safe to drop them
without loosing too much information.
Reviewers: kadircet
Reviewed By: kadircet
Subscribers: nridge, javed.absar, MaskRay, jkorous, arphaman, cfe-commits, gribozavr
Tags: #clang
Differential Revision: https://reviews.llvm.org/D66759
llvm-svn: 370177
Now that we've moved to C++14, we no longer need the llvm::make_unique
implementation from STLExtras.h. This patch is a mechanical replacement
of (hopefully) all the llvm::make_unique instances across the monorepo.
Differential revision: https://reviews.llvm.org/D66259
llvm-svn: 368944
This reverts commit 9c377105da.
[clangd][BackgroundIndexLoader] Directly store DependentTU while loading shard
Summary:
We were deferring the population of DependentTU field in LoadedShard
until BackgroundIndexLoader was consumed. This actually triggers a use after
free since the shards FileToTU was pointing at could've been moved while
consuming the Loader.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64980
llvm-svn: 366559
r366458 is causing test failures. r366467 and r366468 had to be reverted as
they were casuing conflict while reverting r366458.
r366468 [clangd] Remove dead code from BackgroundIndex
r366467 [clangd] BackgroundIndex stores shards to the closest project
r366458 [clangd] Refactor background-index shard loading
llvm-svn: 366551
Summary:
Changes persistance logic to store shards at the directory of closest
CDB. Previously we were storing all shards to directory of the CDB that
triggered indexing, it had its downsides.
For example, if you had two TUs coming from a different CDB but depending on the
same header foo.h, we will store the foo.h only for the first CDB, and it would
be missing for the second and we would never persist it since it was actually
present in the memory and persisted before.
This patch still stores only a single copy of a shard, but makes the directory a
function of the file name. So that the shard place will be unique even with
multiple CDBs accessing the file. This directory is determined as the first
directory containing a CDB in the file's parent directories, if no such
directory exists we make use of the home directory.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64745
llvm-svn: 366467
Summary:
This increases the odds that the boosted file (cpp file matching header)
will be ready. (It always enqueues first, so it'll be present unless
another thread indexes *two* files before the first thread indexes one.)
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64682
llvm-svn: 366199
Summary:
In practice, opening Foo.h will still often result in Foo.cpp making the
second index build instead of the first, as the rebuild policy doesn't
know to wait.
Reviewers: kadircet
Subscribers: ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64575
llvm-svn: 365888
This was reverted in rL365678, the failure was due to YAML parsing of
compile_commands.json.
Converting backslashes to forward slashes to fix the issue in unittest.
llvm-svn: 365748
Summary:
Previously it was rebuilding every 5s by default, which was much too frequent
in the long run - the goal was to provide an early build. There were also some
bugs. There were also some bugs, and a dedicated thread was used in production
but not tested.
- rebuilds are triggered by #TUs built, rather than time. This should scale
more sensibly to fast vs slow machines.
- there are two separate indexed-TU thresholds to trigger index build: 5 TUs
for the first build, 100 for subsequent rebuilds.
- rebuild is always done on the regular indexing threads, and is affected by
blockUntilIdle. This means unit/lit tests run the production configuration.
- fixed a bug where we'd rebuild after attempting to load shards, even if there
were no shards.
- the BackgroundIndexTests don't really test the subtleties of the rebuild
policy (for determinism, we call blockUntilIdle, so rebuild-on-idle is enough
to pass the tests). Instead, we expose the rebuilder as a separate class and
have fine-grained tests for it.
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, jfb, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64291
llvm-svn: 365531
Summary:
Currently SHA1 is about 10% of our CPU, this patch reduces it to ~1%.
xxhash is a well-defined (stable) non-cryptographic hash optimized for
fast checksums (like crc32).
Collisions shouldn't be a problem, despite the reduced length:
- for actual file content (used to invalidate bg index shards), there
are only two versions that can collide (new shard and old shard).
- for file paths in bg index shard filenames, we would need 2^32 files
with the same filename to expect a collision. Imperfect hashing may
reduce this a bit but it's well beyond what's plausible.
This will invalidate shards on disk (as usual; I bumped the version),
but this time the filenames are changing so the old files will stick
around :-( So this is more expensive than the usual bump, but would be
good to land before the v9 branch when everyone will start using bg index.
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D64306
llvm-svn: 365311
Summary:
Clangd currently doesn't cache any indexing failures, which results in
retrying those failed files even if their contents haven't changed.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D63986
llvm-svn: 365120
The bitstream reader handles errors poorly. This has two effects:
* Bugs in file handling (especially modules) manifest as an "unexpected end of
file" crash
* Users of clang as a library end up aborting because the code unconditionally
calls `report_fatal_error`
The bitstream reader should be more resilient and return Expected / Error as
soon as an error is encountered, not way late like it does now. This patch
starts doing so and adopting the error handling where I think it makes sense.
There's plenty more to do: this patch propagates errors to be minimally useful,
and follow-ups will propagate them further and improve diagnostics.
https://bugs.llvm.org/show_bug.cgi?id=42311
<rdar://problem/33159405>
Differential Revision: https://reviews.llvm.org/D63518
llvm-svn: 364464
Summary:
This builds on the relations support added in D59407, D62459,
and D62471 to expose relations in SymbolIndex and its
implementations.
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D62839
llvm-svn: 363481
Summary:
For counting number of references clangd was relying on merging every
duplication of a symbol. Unfortunately this does not apply to FileIndex(and one
of its users' BackgroundIndex), since we get rid of duplication by simply
dropping symbols coming from non-canonical locations. So only one or two(coming
from canonical declaration header and defined source file, if exists)
replications of the same symbol reaches merging step.
This patch changes reference counting logic to rather count number of different
RefSlabs a given SymbolID exists.
Reviewers: ilya-biryukov
Subscribers: ioeric, MaskRay, jkorous, mgrang, arphaman, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D59481
llvm-svn: 360344
Aleksandr Platonov