luaV_execute should compute 'ra' only when the instruction uses it.
Computing an illegal address is undefined behavior even if the address
is never dereferenced.
taken from raptorjit
we can check the hash first to skip quite a bunch of memcmps which should be faster in most cases (hopefully but quick timedemo tests confirmed that)
the string hashing was the most hit and most spent in part of the whole game in alot of cases
replacing it with luajits algo absolutely massively improves lua performance (10-15fps in my bench timedemos!)
however idk how all this stuff exactly works so this needs alot of testing to make sure nothing breaks
ive rewritten the unaligned ptr access slightly to use memcpy for safety with some platforms that require stricter alignment
This reverts commit f72475b557, reversing
changes made to 0c89c85aae.
Sorry Generic this branch is really cool but it needs more work. I shouldn't have merged it too early...
- Add read-only `actionname` field to Lua states.
- Resolves to function name definition of global Lua functions.
- Resolves to all-uppercase name for C actions.
- Resolves to "NULL" string if nil.
Pretty much removed everything except for time operations. File operations steps on the toes of the IO library, and the others were useless or invasive. The only one I wasn't sure about was `os.tmpname()`, but I think that's something that could be implemented for IO library instead if necessary.
The main reason I want this is because os.time provides an OK way to initially seed RNG on game first boot. But it would also be cute to let Lua modders be able to re-implement Christmas mode / day&night cycles themselves :p
This is a squashed commit of the following commits from
upstream:
James R <justsomejames2@gmail.com> on 2021-05-01
Commit 44d217807f71d15bd7ca69dad226458e7060ccab
> Collect makefiles
James R <justsomejames2@gmail.com> on 2021-05-04
Commit f637e28d0c1877c44acd7b01f7130625d5ea4099
> Remove bin, objs and dep directories
James R <justsomejames2@gmail.com> on 2021-05-02
Commit b31056c7d977fe3c2a5e0637589fd9521f1f6c04
> Rewrite Makefile to be modular as well as more automated
>
> Some key points for programmers:
> - Source code files are mostly listed in a 'Sourcefile'.
> So you no longer directly edit the object list. There
> can be multiple Sourcefiles and they can even live in
> subdirectories--the directory name will be prepended to
> every filename in the list. Of course, the Makefile
> still needs to be edited to read from each Sourcefile.
> - Different rules are no longer required for source code
> files that live in subdirectories (such as sdl/ or
> hardware/). Subdirectories Just Work so go ham!
>
> In addition to those points, another important change is
> that the bin directory is no longer divided into platform
> subdirectories (Linux64, Mingw, etc). Executables now go
> directly into bin. If you use DEBUGMODE or target 64-bit,
> then subdirectories for 'debug' and '64' will be made
> though.
>
> Oh by the way, I don't think make clean actually removed
> files before on Windows. It should now. I also fixed as
> many little inconsistencies like that as I noticed.
>
> And now just an overview of the technical aspects that
> shouldn't affect anyone who doesn't REALLY care about the
> Makefile...
>
> objs and dep directories have been moved to a make
> directory. Makefile.cfg and its variants have been moved
> out of their various subdirectories to src/Makefile.d
> make distclean removes the bin and make directories
> entirely, but make clean and cleandep still only affect
> the current build target.
>
> When I say automation, I mean that a lot of copy pasting
> in the Makefile has been reduced.
James R <justsomejames2@gmail.com> on 2021-05-04
Commit 53d1cbe8264196e46911c86bc4be13d44aa35dfe
> Appveyor: update executable directory
James R <justsomejames2@gmail.com> on 2021-05-04
Commit 888073d64d1885cae0ad8b5d54bfec31e9793e0a
> Fix make clean printing header
James R <justsomejames2@gmail.com> on 2021-05-04
Commit 3d7205d4942d1e1b2627fdf5acf9d9c82437a97b
> Fix minor errors with Windows ECHO, DEL, MD
>
> - Quotes were not removed by ECHO.
> - DEL would print an error on nonexistent file.
> - MD would do this plus return a nonzero exit code.
James R <justsomejames2@gmail.com> on 2021-05-05
Commit f9813844e7f0dfa6229062b1ebd83e4e8150c120
> Update CMakeLists.txt to use Sourcefiles
>
> This establishes (near) parity of source code file lists
> between the Makefile and CMakeLists.txt
>
> To make that change I messed around CMakeLists.txt a bit.
> It now uses target_sources and target_compile_definitions.
> I also removed some MSVC stuff since we don't actually
> care about MSVC--it made things easier.
>
> CMake minimum version 3.0 -> 3.13 for target_sources.
James R <justsomejames2@gmail.com> on 2021-05-05
Commit ec8b63d6759f0c124693ea6d5a521e8d82efd566
> Makefile: remove last of unused flags
James R <justsomejames2@gmail.com> on 2021-05-07
Commit c06817d0085ee77b0027917bca91a24e60c7deba
> Makefile: fix mingw/64 swapped with 32-bit
James R <justsomejames2@gmail.com> on 2021-05-07
Commit 9e7d80c2c44199126fa451c59fca0f98d23ad14d
> Makefile: suppress DEL error
James R <justsomejames2@gmail.com> on 2021-05-07
Commit 87afa7655a71539c7cbc7db12df18c4253187fbc
> CMake: fix ASM compile
>
> - target_sources from correct directory
> - enable_language must be used in add_executable directory
James R <justsomejames2@gmail.com> on 2021-05-08
Commit 210c9419e4bddd1fe8e9be7cb66de31119fa0d68
> Ignore -Wtrigraphs
James R <justsomejames2@gmail.com> on 2021-06-16
Commit 3a044e71cc7cf6c80d9836189511c1a58c108370
> Disable EXE disassembly and compression by default
>
> NOOBJDUMP=1 and NOUPX=1 have been removed. Make 'dump'
> target to disassemble. UPX=upx to compress executable.
>
> Setting UPX used to cause it to fail. That has been fixed.
James R <justsomejames2@gmail.com> on 2021-06-16
Commit c142b3241ddb8f1ea9b5090d20dffa43a17129ab
> Makefile: alert full path of final executable
>
> With added textual contrast.
James R <justsomejames2@gmail.com> on 2021-06-16
Commit b04c79d8a7ba4ae8f888ab94d394b6714699e107
> Say 'at'
James R <justsomejames2@gmail.com> on 2021-06-21
Commit 2d7a8c3c571bcc5335437e9db074c75af00408ce
> Makefile: use shell commands to read in Sourcefile
>
> File function is not supported < Make 4.2.
James R <justsomejames2@gmail.com> on 2021-06-21
Commit d5146945a69ebf30e6fc1fbb64558e456e0b7459
> Makefile: don't automatically set WINDOWSHELL unless PATH matches Windows norms
>
> This is for MSYS2, which requires unix shell commands.
James R <justsomejames2@gmail.com> on 2021-06-22
Commit 5f4d7e3c5b013afe03ae2a441c82ad79bae7c66c
> Makefile: fail if old build directories exist
>
> After a checkout from before revision, old directories
> such as bin/Linux64 only remain if untracked files exist
> within. This may be confusing to the user. They may even
> use an outdated executable if it is one of those untracked
> files.
James R <justsomejames2@gmail.com> on 2021-06-22
Commit c3ad5de912fccfdd64c510f5684883afceaa7d36
> Makefile: let variables be defined on Make line
>
> If a variable is defined as in 'make CC=gcc-10', then that
> definition overrides anything other definition in the
> Makefile.
James R <justsomejames2@gmail.com> on 2021-06-22
Commit ed85e994a46ace07cf022b38a9abac1698a8667e
> Remove misplaced parentheses
James R <justsomejames2@gmail.com> on 2021-06-29
Commit c1ecfa306f6a0b168d4f6cbc3224ebebd4deec51
> Makefile: 'dep' not 'deps'
James R <justsomejames2@gmail.com> on 2021-06-29
Commit 6c03f9b5b3590e01773f05bc1c31cc67a2d88fc2
> fuck
James R <justsomejames2@gmail.com> on 2021-06-29
Commit abdf5c101cbe80657f210c0b90800e2c4c6f0c05
> Makefile: report SHELL
James R <justsomejames2@gmail.com> on 2021-06-29
Commit 22ab611daa17bd23225f3afeeba39147c0d0fe17
> Makefile: do not automatically set WINDOWSHELL
>
> According to this answer--
> https://stackoverflow.com/a/45952425
> --Make will always prefer a unix shell, even on Windows,
> if one can be found in Path. So we can't check PATH to
> determine if it's a Windows shell... this is just too much
> bother.
James R <justsomejames2@gmail.com> on 2021-06-29
Commit f79e0ee540564fcec5da7c0dc4006906492a7f30
> Appveyor: remove 64-bit target
James R <justsomejames2@gmail.com> on 2021-06-29
Commit faee657572f23cd6f8b68c077873e01c4278414e
> Appveyor: update for new Makefile
James R <justsomejames2@gmail.com> on 2021-06-29
Commit 44b82dea58612e6f07830647e2b08555f54a7caa
> Appveyor: remove redundant CFLAGS
>
> Also changed -Wno-error=implicit-fallthrough to
> -Wno-implicit-fallthrough. For some reason Appveyor's
> version of GCC is triggering these warnings despite the
> comments, so just shut it up.
James R <justsomejames2@gmail.com> on 2021-06-29
Commit db919accd22350c087c1fd0e5122d8e490e2f171
> Appveyor: suppress real time file names
James R <justsomejames2@gmail.com> on 2021-07-12
Commit 58fa44e8dc0444eecbe701f31872fad9fa563a6a
> CMP0115
SteelT <steeltitanium1@gmail.com> on 2021-07-16
Commit 71f905f95bd4fe62dbca87208cf71b7309f12de1
> Fix gme support being effectively disabled
>
> I found this easier than trying to adjust the makefile, as it uses the same thing to automatically generate the various NO* compile options.
James R <justsomejames2@gmail.com> on 2021-08-14
Commit 824b1ab28cdddfd1c018454a9af14dabaab49fcd
> Makefile: use full stem in dependency generation
>
> Previously took only the filename, so the directory
> component was stripped. This broke dependencies for
> basically every file.
GoldenTails <milestailsprower101n2@gmail.com> on 2021-08-28
Commit 1a8ec7975c20e1d652cbbdb7a9e366192e41c05f
> Makefile: Improve gcc detection
>
> Wasn't working for me until I fixed it.
> Turns out gcc really doesn't like giving its name out.
> Most of the time it reads argv[0].
The global "tags" can be iterated upon for every unique tag which is set in the
level. If a tag is set on a sector/line/thing, it will be included. Taking the
length of "tags" will give you the number of these unique tags. (If a tag is
set on multiple sectors/lines/things, it will only be counted once though.)
For sectors, lines and mapthings, call the field "tagged". This function takes
one argument, which is the tag. The return value can be iterated over for all
the sectors/lines/things with that tag. The length can also be taken for the
number of such objects. If no argument is given, the global tag is default.
* create new file lua_polyobjlib.c
* made a stub LUA_PolyObjLib function
* added META_POLYOBJ to lua_libs.h
* updated makefile, CMake and MSVC project files for lua_polyobjlib.c
