Repository navigation

Script Studio: LSL optimizer size work, LL's compiler as a reference, and SLua converter QA - #404

Merged

RyeMutt merged 59 commits into

developfrom

rye/script-optim

Oct 5, 2026

Merged

Script Studio: LSL optimizer size work, LL's compiler as a reference, and SLua converter QA#404
RyeMutt merged 59 commits into
developfrom
rye/script-optim

Commits on Oct 4, 2026
- The optimizer keeps a pure call made again in a local, where that is no larger
  
  RyeMutt
  and
  claude
  committed
  6b5cf17
- Two vectors' dot and cross products fold to what the VM gives
  
  RyeMutt
  and
  claude
  committed
  1533532
- A local set to a constant is that constant where it is read before it is set again
  
  RyeMutt
  and
  claude
  committed
  d8f8440
- A comparison the ranges of its sides settle folds: a count is never below nought
  
  RyeMutt
  and
  claude
  committed
  c2864ac
- Convert to SLua makes bit32's answers LSL's signed integers
  
  RyeMutt
  and
  claude
  committed
  42c0059
- Convert to SLua carries the script's comments over
  
  RyeMutt
  and
  claude
  committed
  040bb77
- The optimizer's float folding is built without the viewer's fast-math flags
  
  RyeMutt
  and
  claude
  committed
  817b3cf
- The registry's baseline takes Tailslide's vector products and slua's jump direction
  
  RyeMutt
  and
  claude
  committed
  df255ac
- Bit tests of one value merge into one (L18)
  
  RyeMutt
  and
  claude
  committed
  b389ff7
- The column selection's commands are registered as ones no menu item gives
  
  RyeMutt
  and
  claude
  committed
  8f07605
- Convert to SLua's carrying of the script's comments has a preference
  
  RyeMutt
  and
  claude
  committed
  3bdffa5
- More of the library folds where the two VMs cannot part (L17)
  
  RyeMutt
  and
  claude
  committed
  558e96b
- A jump on LSL on Luau costs what it costs, every run
  
  RyeMutt
  and
  claude
  committed
  3dbbfc4
- Convert to SLua signs bit32's answers with SLua's own bit32.s32
  
  RyeMutt
  and
  claude
  committed
  aade1e1
- LL's LSL compiler, restored as it was (L1)
  
  RyeMutt
  and
  claude
  committed
  f644eaa
- LL's compiler's entry points, restored as they were (L1)
  
  RyeMutt
  and
  claude
  committed
  cd58672
- LL's compiler takes what follows from the definitions from lsl-definitions (L1)
  
  RyeMutt
  and
  claude
  committed
  7e0e818
- LL's compiler builds, for the tests (L1)
  
  RyeMutt
  and
  claude
  committed
  d696196
- LL's compiler frees what it made once, and finds a name in an outer scope (L1)
  
  RyeMutt
  and
  claude
  committed
  a2dee42
- LL's compiler boxes a list in a list as the grid's release build does (L1)
  
  RyeMutt
  and
  claude
  committed
  376514d
- LL's compiler called on a text, to LSO or to CIL, with tests (L2)
  
  RyeMutt
  and
  claude
  committed
  c069d87
- What a text view draws before its rows stays within its text
  
  RyeMutt
  and
  claude
  committed
  ed8357f
- Convert to SLua writes its script into the prim the LSL is in
  
  RyeMutt
  and
  claude
  committed
  cbd838a
- A line diff lined up at lines known to stand for each other
  
  RyeMutt
  and
  claude
  committed
  38b7688
- A comparison lines up at anchors, and what is typed in it goes to its source
  
  RyeMutt
  and
  claude
  committed
  157e359
- Convert to SLua says which line of SLua each line of LSL became
  
  RyeMutt
  and
  claude
  committed
  265e652
- Script Studio's comparisons say unsaved only while it is, and give way to typing
  
  RyeMutt
  and
  claude
  committed
  fa7e779
- LL's LSO VM, restored as it was (L5)
  
  RyeMutt
  and
  claude
  committed
  eb88323
- LL's LSO VM builds, and runs a compiled image (L5)
  
  RyeMutt
  and
  claude
  committed
  5d327ea
- The Mono weigher reads any compiler's CIL, and counts a member by its metadata name
  
  RyeMutt
  and
  claude
  committed
  29729bf
- Tailslide's LSO and CIL held against LL's compiler's, and the Mono weigher too (L3, L4)
  
  RyeMutt
  and
  claude
  committed
  d232ba7
- The optimizer's size shapes leave two numbers to the folder
  
  RyeMutt
  and
  claude
  committed
  50e5b84
- Tailslide 1.5.14, compiling as LL's compiler does, and Mono weighed as the grid keeps it
  
  RyeMutt
  and
  claude
  committed
  d1b45c9
- The optimizer writes ++x, leaves a default out, and casts a negative number, where each is smaller
  
  RyeMutt
  and
  claude
  committed
  123bc62
- The optimizer's passes each in a file of their own
  
  RyeMutt
  and
  claude
  committed
  27ed957
- alscript in folders: core, lint, lsl, lsl/optimizer, luau, preprocessor
  
  RyeMutt
  and
  claude
  committed
  21323d1
- The optimizer's operators, branches, loops and keys in fewer bytes
  
  RyeMutt
  and
  claude
  committed
  cc99f91
- The optimizer's functions, scratch globals and dead stores in fewer bytes
  
  RyeMutt
  and
  claude
  committed
  6d8771a
- The optimizer keeps globals' reads and repeated constants once, and charges Mono's locals their frames
  
  RyeMutt
  and
  claude
  committed
  fa1b0f0
- LSO joins string literals, and a loop that never ends runs by a jump back
  
  RyeMutt
  and
  claude
  committed
  309b70b
- Clean up duplicated cmake
  RyeMutt
  committed
  2c5b9a7
- Convert to SLua takes a state's field handlers off with LLEvents:off
  
  RyeMutt
  and
  claude
  committed
  03fd58b
- A copy from the compare view leaves out the rows that line the sides up
  
  RyeMutt
  and
  claude
  committed
  c349bc5
- Convert to SLua in fewer calls: bit32.btest, table.append, NULL_KEY
  
  RyeMutt
  and
  claude
  committed
  771d495
- Convert to SLua counts by a constant's name, and declares first only what is called early
  
  RyeMutt
  and
  claude
  committed
  a0fda9b
- Convert to SLua keeps the script's comments where it laid them out
  
  RyeMutt
  and
  claude
  committed
  2bf3525
- Convert to SLua lines up again what the LSL lined up
  
  RyeMutt
  and
  claude
  committed
  91b94b5
- Vim's registers are shared by every editor of the studio, in every window
  
  RyeMutt
  and
  claude
  committed
  5255b63
- Convert to SLua keeps hexadecimal past 0x7FFFFFFF where bit32 takes it
  
  RyeMutt
  and
  claude
  committed
  1b3d094
- Integer literals past 32 bits read as the grid reads them: -1
  
  RyeMutt
  and
  claude
  committed
  df130d6
- Convert to SLua keeps a mask's hexadecimal where only bit32 reads it
  
  RyeMutt
  and
  claude
  committed
  56f8a78
- SlIntegerPast32Bits: a whole number past 0xFFFFFFFF, which LSL reads as -1
  
  RyeMutt
  and
  claude
  committed
  d19dd22
- Convert to SLua says once what hexadecimal past 0x7FFFFFFF and bit32.s32 are
  
  RyeMutt
  and
  claude
  committed
  671891c
- [pre-commit.ci] auto fixes from pre-commit.com hooks
  pre-commit-ci[bot]
  committed
  2e9f9f8
- The optimizer casts a string to an integer as the grid's 32-bit hosts do
  
  RyeMutt
  and
  claude
  committed
  81d4ee6
- Tailslide folds a string cast to an integer as the grid's 32-bit hosts do
  
  RyeMutt
  and
  claude
  committed
  30f6b2e
- The library's float functions fold only where two ways of working them out agree
  
  RyeMutt
  and
  claude
  committed
  40ff804
- Fix pre-commit failure due to tabs
  RyeMutt
  committed
  c66c793
- The analyzers test stands in for the viewer's RLV name, as the other tests that link llui do
  
  RyeMutt
  and
  claude
  committed
  a2817eb

Commits

Commits on Oct 4, 2026