| Age | Commit message (Collapse) | Author |
|---|
|
modules/checkers w/fill_module=$module requires a consistent
mapping of cpu to fragnum since it creates a per-cpu
til_module_context_t for the fill_module.
The existing implementation for threaded rendering maximizes
performance by letting *any* scheduled to run thread advance
fragnum atomically and render the acquired fragnum
indiscriminately. A side effect of this is any given frame, even
rendered by the same module, will have a random mapping of
cpus/threads to fragnums.
With this change, the simple til_module_t.prepare_frame() API of
returning a bare fragmenter function is changed to instead return
a "frame plan" in til_frame_plan_t. Right now til_frame_plan_t
just contains the same fragmenter as before, but also has a
.cpu_affinity member for setting if the frame requires a stable
relationship of cpu/thread to fragnum.
Setting .cpu_affinity should be avoided if unnecessary, and that
is the default if you don't mention .cpu_affinity at all when
initializing the plan in the ergonomic manner w/designated
initializers. This is because the way .cpu_affinity is
implemented will leave threads spinning while they poll for
*their* next fragnum using atomic intrinsics. There's probably
some room for improvement here, but this is good enough for now
to get things working and correct.
|
|
Also wire this up to the til_module_context_new() helper and
all its callers.
This is in preparation for modules doing more correct delta-T
derived animation.
|
|
- modules now allocate their contexts using
til_module_context_new() instead of [cm]alloc().
- modules simply embed til_module_context_t at the start of their
respective private context structs, if they do anything with
contexts
- modules that do nothing with contexts (lack a create_context()
method), will now *always* get a til_module_context_t supplied
to their other methods regardless of their create_context()
presence. So even if you don't have a create_context(), your
prepare_frame() and/or render_fragment() methods can still
access seed and n_cpus from within the til_module_context_t
passed in as context, *always*.
- modules that *do* have a create_context() method, implying they
have their own private context type, will have to cast the
til_module_context_t supplied to the other methods to their
private context type. By embedding the til_module_context_t at
the *start* of their private context struct, a simple cast is
all that's needed. If it's placed somewhere else, more
annoying container_of() style macros are needed - this is
strongly discouraged, just put it at the start of struct.
- til_module_create_context() now takes n_cpus, which may be set
to 0 for automatically assigning the number of threads in its
place. Any non-zero value is treated as an explicit n_cpus,
primarily intended for setting it to 1 for single-threaded
contexts necessary when embedded within an already-threaded
composite module.
- modules like montage which open-coded a single-threaded render
are now using the same til_module_render_fragment() as
everything else, since til_module_create_context() is accepting
n_cpus.
- til_module_create_context() now produces a real type, not void
*, that is til_module_context_t *. All the other module
context functions now operate on this type, and since
til_module_context_t.module tracks the module this context
relates to, those functions no longer require both the module
and context be passed in. This is especially helpful for
compositing modules which do a lot of module context creation
and destruction; the module handle is now only needed to create
the contexts. Everything else operating on that context only
needs the single context pointer, not module+context pairs,
which was unnecessarily annoying.
- if your module's context can be destroyed with a simple free(),
without any deeper knowledge or freeing of nested pointers, you
can now simply omit destroy_context() altogether. When
destroy_context() is missing, til_module_context_free() will
automatically use libc's free() on the pointer returned from
your create_context() (or on the pointer that was automatically
created if you omitted create_context() too, for the
bare til_module_context_t that got created on your behalf
anyways).
For the most part, these changes don't affect module creation.
In some ways this eases module creation by making it more
convenient access seed and n_cpus if you had no further
requirement for a context struct.
In other ways it's slightly annoying to have to do type-casts
when you're working with your own context type, since before it
was all void* and didn't require casts when assigning to your
typed context variables.
The elimination for requiring a destroy_context() method in
simple free() of private context scenarios removes some
boilerplate in simple cases.
I think it's a wash for module writers, or maybe a slight win for
the simple cases.
|
|
This is a mostly mechanical change of using rand_r() in place of
rand(), using the provided seed as the seed state.
There's some outstanding rand()s outside of create_context()
which should probably get switched over, with the seed being
stowed in the context struct. I didn't bother going deeper on
this at the moment in the interests of getting to sleep soon.
|
|
In the recent surge of ADD-style rtv+compose focused development,
a bunch of modules were changed to randomize initial states at
context_create() so they wouldn't be so repetitive.
But the way this was done in a way that made it impossible to
suppress the randomized initial state, which sometimes may be
desirable in compositions. Imagine for instance something like
the checkers module, rendering one module in the odd cells, and
another module into the even cells. Imagine if these modules are
actually the same, but if checkers used one seed for all the odd
cells and another seed for all the even cells. If the modules
used actually utilized the seed provided, checkers would be able
to differentiate the odd from even by seeding them differently
even when the modules are the same.
This commit is a step in that direction, but rototiller and all
the composite modules (rtv,compose,montage) are simply passing
rand() as the seeds. Also none of the modules have yet been
modified to actually make use of these seeds.
Subsequent commits will update modules to seed their
pseudo-randomized initial state from the seed value rather than
always calling things like rand() themselves.
|
|
Just adds TIL_FB_DRAW_FLAG_TEXTURABLE so callers can granularly
inhibit texturing if desired.
|
|
Just one case, modules/submit, was using 32x32 tiles and is now
using 64x64. I don't expect it to make any difference.
While here I fixed up the num_cpus/n_cpus naming inconsistencies,
normalizing on n_cpus.
|
|
Fragmenting is often dimensioned according to the number of cpus,
and by not supplying this to the fragmenter it was made rather
common for module contexts to plumb this themselves - in some
cases incorporating a context type/create/destroy rigamarole
for the n_cpus circuit alone.
So just plumb it in libtil, and the prepare_frame functions can
choose to ignore it if they have something more desirable onhand.
Future commits will remove a bunch of n_cpus from module contexts
in favor of this.
|
|
This brings something resembling an actual type to the private
objects returrned in *res_setup. Internally libtil/rototiller
wants this to be a til_setup_t, and it's up to the private users
of what's returned in *res_setup to embed this appropriately and
either use container_of() or casting when simply embedded at the
start to go between til_setup_t and their private containing
struct.
Everywhere *res_setup was previously allocated using calloc() is
now using til_setup_new() with a free_func, which til_setup_new()
will initialize appropriately. There's still some remaining work
to do with the supplied free_func in some modules, where free()
isn't quite appropriate.
Setup freeing isn't actually being performed yet, but this sets
the foundation for that to happen in a subsequent commit that
cleans up the setup leaks.
Many modules use a static default setup for when no setup has
been provided. In those cases, the free_func would be NULL,
which til_setup_new() refuses to do. When setup freeing actually
starts happening, it'll simply skip freeing when
til_setup_t.free_func is NULL.
|
|
This is a preparatory commit for cleaning up the existing sloppy
global-ish application of settings during the iterative _setup()
call sequences.
Due to how this has evolved from a very rudimentary thing
enjoying many assumptions about there ever only being a single
module instance being configured by the settings, there's a lot
of weirdness and inconsistency surrounding module setup WRT
changes being applied instantaneously to /all/ existing and
future context's renderings of a given module vs. requiring a new
context be created to realize changes.
This commit doesn't actually change any of that, but puts the
plumbing in place for the setup methods to allocate and
initialize a private struct encapsulating the parsed and
validated setup once the settings are complete. This opaque
setup pointer will then be provided to the associated
create_context() method as the setup pointer. Then the created
context can configure itself using the provided setup when
non-NULL, or simply use defaults when NULL.
A future commit will update the setup methods to allocate and
populate their respective setup structs, adding the structs as
needed, as well as updating their create_context() methods to
utilize those setups.
One consequence of these changes when fully realized will be that
every setting change will require a new context be created from
the changed settings for the change to be realized.
For settings appropriately manipulated at runtime the concept of
knobs was introduced but never finished. That will have to be
finished in the future to enable more immediate/interactive
changing of settings-like values appropriate for interactive
manipulation
|
|
Originally the thinking was that rototiller modules would become
dlopen()ed shared objects, and that it would make sense to let
them be licensed differently.
At this time only some modules I have written were gplv3, Phil's
modules are all gplv2, and I'm not inclined to pivot towards a
dlopen model.
So this commit drops the license field from til_module_t,
relicenses my v3 code to v2, and adds a gplv2 LICENSE file to the
source root dir. As of now rototiller+libtil and all its modules
are simply gplv2, and anything linking in libtil must use a gplv2
compatible license - the expectation is that you just use gplv2.
|
|
Largely mechanical rename of librototiller -> libtil, but
introducing a til_ prefix to all librototiller (now libtil)
functions and types where a rototiller prefix was absent.
This is just a step towards a more libized librototiller, and til
is just a nicer to type/read prefix than rototiller_.
|
|
This is a first approximation of separating the core modules and
threaded rendering from the cli-centric rototiller program and
its sdl+drm video backends.
Unfortunately this seemed to require switching over to libtool
archives (.la) to permit consolidating the per-lib and
per-module .a files into the librototiller.a and linking just
with librototiller.a to depend on the aggregate of
libs+modules+librototiller-glue in a simple fashion.
If an alternative to .la comes up I will switch over to it,
using libtool really slows down the build process.
Those are implementation/build system details though. What's
important in these changes is establishing something resembling a
librototiller API boundary, enabling creating alternative
frontends which vendor this tree as a submodule and link just to
librototiller.{la,a} for all the modules+threaded rendering of
them, while providing their own fb_ops_t for outputting into, and
their own settings applicators for driving the modules setup.
|
|
My current version of mingw in debian 9.11 doesn't seem to have
rand_r(), so let's just use plain rand() and lose the improved
parallelization on win builds.
|
|
Most modules find themselves wanting some kind of "t" value increasing
with time or frames rendered. It's common for them to create and
maintain this variable locally, incrementing it with every frame
rendered.
It may be interesting to introduce a global notion of ticks since
rototiller started, and have all modules derive their "t" value from
this instead of having their own private versions of it.
In future modules and general innovations it seems likely that playing
with time, like jumping it forwards and backwards to achieve some
visual effects, will be desirable. This isn't applicable to all
modules, but for many their entire visible state is derived from their
"t" value, making them entirely reversible.
This commit doesn't change any modules functionally, it only adds the
plumbing to pull a ticks value down to the modules from the core.
A ticks offset has also been introduced in preparation for supporting
dynamic shifting of the ticks value, though no API is added for doing
so yet.
It also seems likely an API will be needed for disabling the
time-based ticks advancement, with functions for explicitly setting
its value. If modules are created for incorporating external
sequencers and music coordination, they will almost certainly need to
manage the ticks value explicitly. When a sequencer jumps
forwards/backwards in the creative process, the module glue
responsible will need to keep ticks synchronized with the
sequencer/editor tool.
Before any of this can happen, we need ticks as a first-class core
thing shared by all modules.
Future commits will have to modify existing modules to use the ticks
appropriately, replacing their bespoke variants.
|
|
Since snow_context_t needs another member anyways, stick n_cpus in there
to inform the fragmenter of precisely how many fragments to make.
This renderer doesn't benefit from tiling or any such locality, so it uses
the slice fragmenter and really only benefits from as many fragments as there
are CPUs. Any additional fragments is just wasted fragmenting overhead.
|
|
Snow was already threaded, but used a global seed with rand_r() meaning
the CPUs were hammering on the same address. There wasn't any locking or
atomics, as it isn't terribly critical when generating white noise if the
seed access is racy. But the writes still caused cache lines to ping-pong.
This commit gives a ~15.5% speedup in my measurements on an i7-2640M.
Note without the padded union, using just an array of ints, zero gain
is realized. I used a padding of 256 just to have some headroom, x86
is 64 but other CPUs vary, POWER9 is 128 for example.
|
|
Mechanical change removing abbreviation for consistency
|
|
Mostly mechanical change, though threads.c needed some jiggering to
make the logical cpu id available to the worker threads.
Now render_fragment() can easily addresss per-cpu data created by
create_context().
|
|
I wanted to add some noise to the rtv module and figured why not
just add a snow module and make rtv pass through it briefly when
switching modules.
It's not interesting by itself, but as more composite/meta modules
like rtv get made it might be handy beyond rtv.
|
