Basically the setup should be symmetrical to the `BufferStage` setup,
which implies to install the `LocalMemPool` as a thread-local »slice«.
However, in addition we need a mechanism to trigger an automatic clean-up
of the local pool whenever a worker-thread terminates, with the additional
complication that thread-local variables are effectively static.
Since `LocalMemPool` is written such as to abstract from the setup environment,
we need to inject a shutdown function from the setup within `LocalSlice`,
thereby directly accessing the `EngineBufferManager` via dependency-injection.
Most relevant parts of the `LocalBufferStore` are already available
and just need to be connected. One point poses a challenge however:
when and how to trigger the heuristic clean-up of the local pool,
so that ''on average neither too much memory is wasted, nor
to much unnecessary allocations'' are performed...
Signed-off-by: Ichthyostega <prg@ichthyostega.de>
This is a collection with unique-ownership,
and the ability to create all content objects in ''RAII style''
from the constructor call, using a ''population functor''.
It is often used to maintain a collection of resource handles,
and especially to start a series of threads (for testing).
With the variadic λ-captures from C++20, it is possible to rewrite
the most frequently used population functors to be easier to use.
Notably we can now bind constructor arguments, that will be passed
to each of the created elements — a prominent example being the
thread function when launching several threads. Furthermore,
I added a constructor variant that consumes a Lumiera Iterator.
Adding this kind of generic variants is possible now with C++20,
since overload resolution takes concepts into account.
Well, finally.
Lumiera Iterators were conceived as a concept from the very beginning,
and this idea dates back even before C++11. Furthermore, over time some
''trait templates'' have been developed, to detect if some template argument
might match the known pattern of functions associated with a Lumiera Iterator.
Meanwhile this concept is used pervasively in the whole code base, and also
the traits templates proved to be reliable. It therefore seems that the time is ripe
to gradually introduce a genuine **concept** `lum_iter`
I'll take a defensive approach here, since any mistake would lead to
a massive breakage of compilation. The new concept is defined simply
by delegating to to the tried-and-tested trait template, which is
also renamed `can_IterForEach` ⟼ `can_LumieraIter`
One immediate benefit is that the compiler is now able to select
overloads based on that concept, which allows to drop the tricky
selection of variants based on the SFINAE technique and `enable_if`.
For validation, one very heavily used case of that pattern was switched
in `lib/format-util.hpp', which notably lies in the code-path used by `util::join`
The test performs a massively concurrent scenario
that presses with 100 threads and 1 MiB allocation for each.
This highlighted my still limited understanding of the situation
and leads to a much better solution, where, on-demand, some worker
performs all the accrued allocation and management work.
So effectively we do not even need to use the synchronous API
Implementing the `EngineBufferManager` the desired way,
with the ability to post allocation requests asynchronously,
raises the difficult question of who will ultimately perform
the allocation / deallocation and further management work.
The point is that this management effort takes up capacity
originally assigned for the render workers themselves, either
by implicitly pushing aside those threads, or as a dedicated
processing assigned to some worker.
In the current stage of development it seems impossible to decide
upon the proper mechanism to schedule this excess work explicitly.
The solution proposed hereby however circumvents this dilemma,
insofar it allows to implement the synchronous, fallback processing
in terms of the asynchronous processing, just performed exclusively
within the current thread. Unless some processing will be scheduled
explicitly (which can be added as performance optimisation later on),
the handling of a direct allocation request will ''fold down'' into
the current worker eventually, thereby using it's processing capability
without the need of any explicit scheduling.
- use a base class with the lock-free queue as building block
- both `LocalMemPool` and `EngineBufferManager` are built on top,
allowing them to exchange `Alloc` entries asynchronously
- ''Allocation requests'' include a receiver callback,
so that a newly allocated memory block can be dispatched
While all of this aggregate can be considered ''deep implementation'',
the structure of dependencies bears danger to get confusing. Even more so,
since an elaborate pooling memory allocator will be added to and interconnected
with this structure.
Now considering the following route:
- `engine-buffer-allocator.hpp` has elements of a common service / interface
- but `engine-buffer-allocator.cpp `is also the most deeply nested implementation backend
- `local-buffer-store.hpp` is a junction point
- `local-mem-pool.hpp` is considered an insolated implementation detail
- ultimately, this will be included and wired in `render-enviroment.cpp`
This implementation and the accompanying test might seem pointless,
yet the point is that the interface is aligned to the emerging
concepts for memory allocators in Lumiera, while also directly
delegating to the std::allocator for the time being.
Before we're able to observe the actual allocator trend and behaviour,
it is pretty much moot to engage into any kind of optimisation;
the generic heap allocator performs quite well on recent C++ runtimes,
and, furthermore, the most important performance tweak is to use
a thread-local pool of recent allocations, which should allow
our jobs to run without impeding each other.
Seems to work now roughly as expected, so that the next layers
of functionality can be built on top. It seems impossible
to determine the viability of this whole approach without
taking the full buffer memory management into use with
a real-world load scenario.
At the begin of each Render Node `pull()`, the render job is expected
to ''announce'' a maximum expected allocation requirement to the `BufferProvider`,
which information is passed-on to the `LocalMemPool` to determine
to which degree it can be satisfied with allocations from that local pool.
Without actual empiric observation, it is impossible to determine
the relevance of this step, how it should be resolved heuristically
against the known contents of the pool, and how much effort to spend
on that step.
Thus I'll settle for the ''simplest thing that can possibly work'' —
which is to check only for close matches and rather rely on additional
allocations, forwarded asynchronously, assuming that the clean-up
heuristics will deal with excess allocations quickly.
- take into account how much memory will be wasted when
using some allocation to satisfy a (smaller) allocation request
- furthermore prefer allocations used frequently in the past
- when an allocation is used, tag it with a score that also
reflects this degree of waste
- select buffers with low score first for clean-up
- use boost::lockfree to implement the in-queue with new allocations
- use a std::list with an internal administrative record
- automatically ingest the in-queue prior to accessing the pool
After several fruitless attempts, I have decided now
to settle for self-contained ''memory pool'' component,
that can be operated **thread-locally** and employs a
**lock-free** message queue for the allotment of memory.
Seemingly the question of how this relates to `BufferMetadata`
must remain unresolved for the time being, until the entire
buffer memory subsystem is available for empirical investigation.
As staring point, it is clear that the `BufferStore` side of the
implementation will rely on a similar structure with a thread-local
base service — which in this case will have to maintain a small pool
of recently used buffer allocations...
...which happens when a buffer type is referred for the first time
from within a new worker thread, while it is already stored in the
global registry; the full key-chain is down-synced then.
...using the first usage situation as guideline,
and verify proper operation through unit-test
- the LocalBufferStage gets a set of thread-local BufferMetadata
- furthermore, it accesses the central hub via dependency-injection
- after each type registration ''from a worker thread'' data is published
As an asside: ensure most metadata registration can ''move'' the TypeHandler
This test setup defines one of the problematic situations
where a synchronisation is necessary.
__TODO__: actually add locking and that synchronisation to LocalBufferStage
What I am building step by step right now is the first draft
of a production-grade render buffer management scheme.
For this I need suitable implementations of `BufferStag` (and `BufferStore`)
These will include a `LocalSlice` to hold thread-local »satelitte services«,
which in turn need a connection to some central coordination hub.
The thread-local instances can access the hub via our `lib::Depend` (I have just
verified that this works as intended, with on-demand creation of the
central instance on first access from some thread. This is important,
since very likely we need to install a dedicated and speficically wired
service instance, which must happen prior to first dependency-request
via `lib::Depend`. And since the worker-threads are created by the Engine,
we get ample opportunity to do so before starting those threads.
Included in this changeset is a first rough draft of the central registry
`EngineBufferMetadata` — I have modelled it analogous to the `SimpleBufferStateRegistry`,
which is the test-level implementation of the `BufferStage` API.
However, we do not want to expose direct buffer allocation operations
on the central hub, and thus I have left out the respecive parts of the API.
Furthermore, I have added an `abandon()` operation, without there beeing
a need for such an operation right now. But since this operation is tricky
and potentially dangerous, I prefer implementing it the correct way,
since I have a clear understanding what this entails at the moment.
The recursive synchronisation up the parent-chain can in most cases be skipped,
by relying on the fact that inconsistent data is never stored in the BufferMetadata
table. If thus an entry is locally present and matches the entry from the
other side, then it can be concluded that its parent must also be already in sync.
Since the intention is to run several thread-local instances of `BufferMetadata`
in a concurrent setup, we need a way to publish and exchange chains of
registered type-keys and metadata entries.
The first step is to develop a mechanic for data exchange,
and to define the rules when such synchronisation is (not) allowed.
Since thread_local implies a static setup, we need a way how to
build a new service instance. Usually this would be a default construction,
yet there might be services with special init requirements.
Using an explicit template parameter for this setup configuration
allows to make the primary template parameter just the service itself,
which makes the default usage much clearer.
- Use a ''monostate'' as access point to thread-local service instances
- rely on a configuration-context to setup the actual service
- investigate runtime overhead using a multithreaded microbenchmark
During the last days, I started with analysis to find a way
to operate the `BufferProvider`, as implemented thus far,
within a massively concurrent environment.
Since I intend to create thread-local caches,
the initialisation of the `BufferMetadata` store must be static.
It turns out that the root-ID, which acts like a mandator ID,
is only used at a single code path, when creating or accessing
an abstract type-Key. It is thus easy to extract and relocate
this distinguishing ID into the actual `BufferStage` implementation.
Looking through all my work notes, it seems that all the
problematic aspects are now solved in a satisfactory way.
* it is clear how to dependency-inject a `BufferProvider`
* the internals of the `BufferProvider` front-End have been restructured
* there are now two ''pillars'' of the implementation: `BufferStage` and `BufferStore`
* both can be subclassed independently
* `BufferProviderSetup` exposes a proxying-mechanism (used for `DiagnosticBufferProvider`)
* the `BufferMetadata` have been reviewed and some minor details were rectified
* after a re-Design of `OutputSlot`, `BuffHandle` can now be used as unified front-End
* both `DiagnosticBufferProvider` and `DiagnosticOutputSlot` have been rewritten
* they are now independent, self-contained specialisations, layered on top
Overall, what might seem like a detour, was an important step towards
well-balanced structures and a better understanding of the Render Engine.
The actual mystery, that could ''not'' be clarified,
is why that test did pass previously. Because it was
always broken, insofar it pretended that the output buffer
does hold a valid test frame, while in fact the buffer is
just newly allocated.
The deeper reason behind this mismatch lies in the concept of
''in-place processing'' — which is a very obvious, nullary optimisation,
and thus we want our engine to be ''in-place capable'' in the long run.
Which explains why I did write the dummy-test operations with in-place
processing in mind. Yet the actual version of the render engine leaves
out a lot of future improvements, amongst others the in-place processing.
This implies that — at least at the moment — it is clearly the responsibility
of the processing function to prepare the output buffer, in cases where such
preparation is necessary (and this applies in the case of `TestFrame`,
which needs to be primed with a valid header mark)
This change touches on another unsolved problem in the `TestDomainOntology`:
we need a mechanism to allow the generic builder front-end to set
specific configuration for some kind of processing. Generally speaking,
such things are possible in C++, yet somewhat tricky — and thus I postponed
that aspect and just included a configuration variant, hard-coded for the moment.
After an extended round of analysis and cascading reworks,
the long-standing discrepancy between `OutputSlot` and the
''Buffer Provider Protocol'' (#1395) has finally been resolved.
The foundation for this success was laid by means of a restructured
implementation within `BufferProvider` – which has now been
successfully completed and validated.
__note__: regarding hash-standard.hpp see Ticket #1422
Mostly this hack worked -- but it is very likely not needed any more.
Today I had to re-wire a partial specialisation for hashing pointers,
that was added somewhere in the past into the STL.
Now able to reap the rewards: since all that convoluted and tangled
old ''state tracking'' has been resolved and replaced by dedicated
and self-contained diagnostic subclasses, the old dummy BufferProvider
implementation can be detached, using a similar yet very simple
fake implementation.
For context: the `HeapMemBufferStore` is part of the default configuration
and thus automatically used whenever a unit test needs a `BufferProvider`.
This default implementation is not optimised for speed, and actually
never releases any memory, which allows to verify proper handling
of buffer memory allocation.
''This is a breakthrough...''
Using the realigned implementation structure in `BufferProvider`,
and especially the ability to establish an overarching context by
passing an connection-ID through the `LocalTag` in the `BufferMetadata`,
it was possible to resolve the longstanding mismatch buried deep within
the `OutputSlot` implementation.
Based on the new internal APIs in `BufferProvider`, it was thus possible
to setup an `OututBufferProxy` that acts as a hub to connect the
`DataSink` handles with the actual `OutputSlot::Connection` implementation.
This change allowed to rectify the convoluted control structure in
`OutputSlot` and finally to manage the Lifecycle properly. Furthermore,
one level of unnecessary implementation-inheritance could be removed,
so that it is no longer necessary to subclass `OutputSlot` in order to
configure the link to a specific output backend.
Using this redesigned internal structure, it was now fairly simple to
re-implement a tracking `DiagnosticOutputSlot` and then to verify proper
integration by passing the `OutputSlotProtocol_test` again.
This is a design study to find out what would be required
to build an opaque function wrapper, similar to `std::function`,
yet without using any heap memory ever.
Such a building block would allow for some interesting options
for a well structured design of elaborate data structures.
This changeset provides an implementation draft.
__Two important findings to note__:
* it is impossible to get away with just a single function pointer
* because at the very least we need to invoke the destructor of the
embedded delegate. Thus something akin to a "manager" is required.
* this implies that the component would not be so efficient as I hoped,
since it requires two pointers and requires to emit code for each
type of delegate.
* by making the component **NonCopyable**, I avoided the complexities
of virtual copy support. In case the latter is desired, it would be
much simpler to switch to a classical OO approach with a VTable
and use our `PolymorphicValue` wrapper to handle copy and cloning.
This commit is the aggregated result of extended research
that is only loosely related to current development topics.
One theme was the construction of nested data structures using tuples,
which faces some difficulties due to tuple manipulations being limited
almost entirely to compile-time processing. However, the barrier between
compile-time and runtime can be overcome with the help of a dispatcher
table -- which leads to the secondary problem to handle the diverse
tuple elements with a common, generic functor. Such a functor can be
constructed, if and only if those element types are ''reconcilable''
The STDLIB provides two metafunctions, `std::common_type` and
`std::common_refernce`, which however pose some challenges for the
task at hand, since they include additional processing. In the end,
a custom version of the ''basic type reconciliation'' was implemented
and tested, and then used as foundation to build a dispatcher table
for runtime-indexed access and iteration over tuples.
As an aside, several deeply hidden bugs in our library code were resolved,
and the tuple-helpers, tuple string rendering and associated utilities
were regrouped and modernised.
As result of this research work, the following components are provided:
- meta/common-result
- meta/fun-trampoline
- meta/tuple-index-adaptor
The detailed analysis of corner cases indicated a lot of hidden dangers
when stretching the `IterExplorer` design beyond its limits.
By deliberate choice, this design was kept open so that advanced cohesive
implementation structures can be built. It is even possible ''and tolerated''
to rely on ''references'' to some degree -- assuming however that the
referred memory location remains stable.
As it turned out, these basic assumptions can be broken by a combination
of read-ahead processing (as is necessary for grouping and filtering),
together with adaptor objects embedded as value, especially when
the resulting pipeline object is then moved or copied during evaluation.
While my general attitude towards such extremely clever implementation
constructs is "don't shoot yourself into your foot unless you need to" --
it seems prudent to clean-up some crappy code from the first attempts
towards an iterator pipeline -- notably some of the »Itertools«.
This expands the processing scheme developed thus far, so that
an arbitrary nested iterable can be handled, as long as it can be
adapted by the generic mechanisms provided by IterExplorer.
Notably there is now a type pattern match to distinguish various cases,
even spanning the ''rather esoteric'' possibility that an underlying
iterator produces STL collections by-value.
By examining these special cases, the limitations of the `IteratorExplorer`
design were highlighted, and a number of tricky bugs could be fixed.
Since the new `flatten()` operation was implemented in the `IterExplorer` framework,
some non-obvious corner cases must be considered; notably the source iterator
might yield something that is not by itself a Lumiera Iterator, yet could be
adapted by applying those generic mechanisms already in place to handle such cases;
however there is the special twist in the case at hand that we do not have to adapt
the ''source iterator'', yet rather the ''product'' of the source iterator —
a requirement that, once understood, can be easily adopted with
yet another iterator-decorator layered below.
This is a feature I wasn’t sure for a long time how best to tackle.
It’s worth noting that an initial attempt at `IterExplorer` was based
on the monad pattern, which turned out to be a mistake, since monads
are more of an anti-pattern, than being genuinely useful.
The actual use-case that made me reconsider such a feature it
to build an seamlessly chained sequence from several iterators.
It would be simple to develop a suitable utility as a stand-alone helper,
yet it seemed worth the effort to integrate this into the `IterExplorer`
ecosystem, due to the flexible automatic adaptation to iterable source
pipelines provided therein.
Going this route incurred some cost however, unveiling several defects
related to handling nested iterable types in the generic processing framework.
On occasion of the recent improvements, I noticed that somewhere
in the past I had misunderstood the meaning of the typedefs in iterators:
* `reference` is the central definition; this is what the iterator yields
* `value_type` is secondary; this is the type you use to store results
So at times I defined only a `value_type`, wrongly assuming this to be
the primary definition, and all the other ones to be ephemeral.
Consequently the trait to detect a »Lumiera Forward Iterator« should be sharpened...
...which unveils a long-standing bug in our ''Duck Detector''
(it is not possible to define a pointer to a reference, causing
a check for the `reference` typedef erroneously to fail.
- use uniform naming scheme
- extract type-binding from RangeIter into a trait-template
- use the C++20 iterator concept and trait to detect a STL iterator
After some investigation, I seem to understand the problem better now.
This builds upon the Introduction of our `ValueTraits´ two years ago:
e176e54004
The root cause to these ongoing sporadic problems with type deduction
seems to be that we encounter a **clash of concepts**: What Lumiera calls
an »Iterator« is essentially different to the order of terms used by the STL.
In the STL, an iterator can be seen as an ''abstracted pointer'', and is
thus only meaningful in conjunction with a backing container. It is also
this backing STL container that provides the value-type definitions.
In Lumiera however, an ''Iterator is a promise to retrieve a sequence of evaluations''.
For that reason, the Iterator itself provides value-type definitions,
and the notion of "iterable" can largely be conflated with "Lumiera Iterator".
Now, the `lib::RangeIter` can act as a bridge between both usage patterns,
and thus it is possible to create a `RangeIter` both from a Lumiera Iterator
or from a STL container. (As an aside, this is a similar idea as promoted
by the "Range TS" in C++20 -- yet the Lumiera iterator concept is much more
abstract and allows evaluations like tree expansion and search algorithms,
which would be never possible with the STL based notion of iteration)...
Thus the value type binding in `RangeIter` is the only correct place
where to mitigate this ''clash of concepts:''
* if the `IT` parameter is a Lumiera iterator, we need to retrieve
the type bindings defined therein
* otherwise, we can assume that `IT` is an STL iterator and thus
essentially a pointer, and we must strip that pointer and work from there.
...to emplace some object into the buffer space.
I wouldn't expect that to be necessary for the real usage,
but it comes in handy for tests, where we often want to place
some sentinel value into a buffer.
- visible Signature of TestFrame changed due to being relocated into the Vault
- attempt to make the WorkForce_test more robust;
frequently fails on testsuite run after full build;
so better ''wait'' for the workers to have started working,
instead of using a fixed delay (that must be extremely short
and is hard to predict)
Straight-forward:
- add a map keyed by observed frame numbers
- individual connection needs to know its own seqNr :-(
- also need to store and recall the frameNr in the buffer-Index
Fix bug with DataSink constructor
(value constructor shadows copy constructor => Stackoverflow)
''Complete construction and wiring sequence works as intended''
- can now throw away the old convoluted `TrackingInMemoryBlockSequence`
- instead use a simplistic implementation of the Connection interface
- do not perform any kind of validation, allocate new buffer for each frame
- use the `NaiveBufferSetup` (`BufferProvider`) as implementation backend.
This will be the first ''reality test'' for the new `OutputSlot` internal design:
Is it possible to implement a tracking test implementation in a clean
and straight-forward way, avoiding all the tangling and indirection
that plagued the old prototype implementation from 2012?
After resolving the inner tensions in the old prototype from 2012,
it was now possible to resolve all issues in a single strike
* `DataSink` becomes a functor, yet embeds a shared ref-count
* in the λ-closure a ''specific Buffer Descriptor'' is prepared for each feed
* building upon the prototyping work, a dedicated `OutputBufferProxy`
exposes a `BuffHandle` to the client, with `emit()` and `release()` calls
* the implementation ''delegates directly'' to the `OutputSlot::Connection`
* an this is where all actual output technology is encapsulated
