Started at 349,880,038 allocations and 42s
- Switched to making `Symbol` in GOOS be a "fixed type", just a wrapper
around a `const char*` pointing to the string in the symbol table. This
is a step toward making a lot of things better, but by itself not a huge
improvement. Some things may be worse due to more temp `std::string`
allocations, but one day all these can be removed. On linux it saved
allocations (347,685,429), and saved a second or two (41 s).
- cache `#t` and `#f` in interpreter, better lookup for special
forms/builtins (hashtable of pointers instead of strings, vector for the
small special form list). Dropped time to 38s.
- special-case in quasiquote when splicing is the last thing in a list.
Allocation dropped to 340,603,082
- custom hash table for environment lookups (lexical vars). Dropped to
36s and 314,637,194
- less allocation in `read_list` 311,613,616. Time about the same.
- `let` and `let*` in Interpreter.cpp 191,988,083, time down to 28s.
This sets up the extractor for jak 2. I was expecting that I'd have to
make some more significant changes to the decompiler/compiler path
stuff, but this was not the case!
The only real change is that you can now provide multiple ISO hashes for
an entry in `ISOMetadata`. This is needed for the two different NTSC
versions, which have the same configs, serials, and ELF hashes, but
slightly different contents.
I also didn't add the korean version because I don't have the info for
it.
---------
Co-authored-by: ManDude <7569514+ManDude@users.noreply.github.com>
This change adds a few new features:
- Decompiler automatically knows the type of `find-parent-method` use in
jak 1 and jak2 when used in a method or virtual state handler.
- Decompiler inserts a call to `call-parent-method` or
`find-parent-state`
- Removed most casts related to these functions
There are still a few minor issues around this:
- There are still some casts needed when using `post` methods, as `post`
is just a `function`, and needs a cast to `(function none)` or similar.
It didn't seem easy to change the type of `post`, so I'm not going to
worry about it for this PR. It only shows up in like 3 places in jak 2.
(and 0 in jak 1)
- If "call the handler if it's not #f" logic should probably be another
macro.
Fixes#805
- state handlers that are not inlined lambdas have smarter type
checking, getting rid of 99.9% of the casts emitted (they were not
useful)
- art groups were not being properly linked to their "master" groups.
- `max` in `ja` in Jak 2 was not being detected.
Another huge PR...
Previously, `object` and `none` were both top-level types. This made
decompilation rather messy as they have no LCA and resulted in a lot of
variables coming out as type `none` which is very very wrong and
additionally there were plenty of casts to `object`. This changes it so
`none` becomes a child of `object` (it is still represented by
`NullType` which remains unusable in compilation).
This change makes `object` the sole top-level type, and the type that
can represent *any* GOAL object. I believe this matches the original
GOAL built-in type structure. A function that has a return type of
`object` can now return an integer or a `none` at the same time.
However, keep in mind that the return value of `(none)` is still
undefined, just as before. This also makes a cast to `object`
meaningless in 90% of the situations it showed up in (as every single
thing is already an `object`) and the decompiler will no longer emit
them. Casts to `none` are also reduced. Yay!
Additionally, state handlers also don't get the final `(none)` printed
out anymore. The return type of a state handler is completely
meaningless outside the event handler (which is return type `object`
anyway) so there are no limitations on what the last form needs to be. I
did this instead of making them return `object` to trick the decompiler
into not trying to output a variable to be used as a return value
(internally, in the decompiler they still have return type `none`, but
they have `object` elsewhere).
Fixes#1703Fixes#830Fixes#928
There are potentially still some minor issues with the resulting files.
Some of them appear to have minor artifacts that playing through the
actual game do not -- but this is a much better starting point for
someone to iterate from if they are interested in improving things.
Some general improvements for the texture animator:
- Clouds are special cased, saving about 1 ms per frame
- Adjusting the amount of clouds now actually works.
- Fixed an issue with the brightness of clouds, and the way that they
fade out around the edges.
Rotates the log files with a timestamp instead of copying all files and
incrementing an integer. Increases the amount of info you have when
looking at user's log files (ie. when looking at all the files, the file
creation dates are accurate).
Also simplifies the API for setting the log file, and `gk` logs are now
game specific with `jak1` or `jak2`. Which should be useful going
forward.
Lastly, added a flag to all CLIs to disable ansi colors for people that
want to do so. Though at the same time, there is finally a workaround in
jenkins to fix ANSI colors in the truncated log view -- so I'm not sure
why anyone would want to get rid of the color information. You can even
setup text editors to display the color info making log parsing much
easier. Fixes#1917
---------
Co-authored-by: ManDude <7569514+ManDude@users.noreply.github.com>
- Add security wall animation
- Add waterfall animations
- Add lava animations
- Update layer values from the game to fix the security wall
- Remove leftover debug in `level.gc` that would break level-specific
animations on the second time you visited the level
- Optionally load animated slot textures to the pool so generic can use
them (fixes skull gems in UI)
Added framework to do texture animations entirely in C++. Currently only
works on relatively simple ones, and doesn't handle updating all
parameters - only the speeds.
Connected texture animations to merc and tfrag for skull gems, dark
bomb, and scrolling conveyors.
Cleaned up Tfragment/Tfrag3, which used to be two classes. This was one
of the first C++ renderers, so it had a weird design.
I havn't tested it yet, but I can almost guarantee that atleast `goalc`
will not work in the slightest!
But the project is atleast fully compiling. My hope is to start
translating some AVX to NEON next / get `goalc` working...eventually.
The main thing that was done here was to slightly modify the new
subtitle-v2 JSON schema to be more similar to the existing one so that
it can properly be used in Crowdin.
Draft while I double-check the diff myself
Along the way the following was also done (among other things):
- got rid of as much duplication as was feasible in the serialization
and editor code
- separated the text serialization code from the subtitle code for
better organization
- simplified "base language" in the editor. The new subtitle format has
built-in support for defining a base language so the editor doesn't have
to be used as a crutch. Also, cutscenes only defined in the base come
first in the list now as that is generally the order you'd work from
(what you havn't done first)
- got rid of the GOAL subtitle format code completely
- switched jak 2 text translations to the JSON format as well
- found a few mistakes in the jak 1 subtitle metadata files
- added a couple minor features to the editor
- consolidate and removed complexity, ie. recently all jak 1 hints were
forced to the `named` type, so I got rid of the two types as there isn't
a need anymore.
- removed subtitle editor groups for jak 1, the only reason they existed
was so when the GOAL file was manually written out they were somewhat
organized, the editor has a decent filter control, there's no need for
them.
- removed the GOAL -> JSON python script helper, it's been a month or so
and no one has come forward with existing translations that they need
help with migrating. If they do need it, the script will be in the git
history.
I did some reasonably through testing in Jak1/Jak 2 and everything
seemed to work. But more testing is always a good idea.
---------
Co-authored-by: ManDude <7569514+ManDude@users.noreply.github.com>
The current event-based approach is very difficult to get right, and it
depends on no events ever being missed. This changes the keyboard/mouse
handling code to a polling-based approach.
Other fixes:
- an issue where modifier keys were not able to be successfully bound
(like Left Shift to `X`)
- improves cursor hiding (except when you use the start menu, this seems
like an SDL issue, see comment)
- Better discarding of kb/mouse inputs when imgui intercepts input
- properly swap bindings when an already set key is assigned, even if it
crosses the distinction of an analog vs normal button
Fixes#2800
Fixes empty boxed arrays of strings breaking some decomp
(`ctywide-speech` and `race-info`).
Adds `decomp-as` tag to decompiler types so that the static data
decompiler can use macros like `meters` and `seconds` on fields that
aren't of type `meters` or `time-frame`.
Adds `override` tag to decompiler types which overrides the type of
field with that name. The type must be a child type of the original
field's type (or the same type, but why would you do this?).
Fixes the camera being offset for `drillmtn` after loading `palout`
once.
This is a huge refactor sadly.
Trying to make up for some of the startup speed lost in the SDL
transition. This saves about 1s from start (from ~3s), and about 500 MB
of RAM.
- Faster TIE unpack by merging matrix groups, more efficient vertex
transforms, and skipping normal transforms on groups with no normals.
- Refactor generic merc and merc to use a single renderer with multiple
interfaces, rather than many renderers. Removed "LightningRenderer" as a
special thing, but Warp is still special
- Add more profiling stuff to startup and the loader.
- Remove `SDL_INIT_HAPTIC` - this turned out to be needed for
force-feedback steering wheels, and not needed for controller vibration
- Switched `vag-player` to use quicksort instead of the default GOAL
sort (very slow)
