Remove mutex around LLVM calls from booster

Previously the LLVM backend was not thread-safe so the simplification calls had to be sequentialised for concurrent RPC requests.

Since thread-local storage is now used, the calls can be executed concurrently/in parallel using multiple cores. This PR removes the mutex that sequentialised the calls before.

Author: jberthold

booster/library/Booster/LLVM/Internal.hs

@@ -104,21 +104,19 @@ data API = API
     , simplifyBool :: KorePatternPtr -> IO (Either LlvmError Bool)
     , simplify :: KorePatternPtr -> KoreSortPtr -> IO (Either LlvmError ByteString)
     , collect :: IO ()
-    , mutex :: MVar ()
     }
 
 newtype LLVM a = LLVM (ReaderT API IO a)
     deriving newtype (Functor, Applicative, Monad, MonadIO, MonadThrow, MonadCatch, MonadMask)
 
-{- | Uses dlopen to load a .so/.dylib C library at runtime. For doucmentation of flags such as `RTL_LAZY`, consult e.g.
+{- | Uses dlopen to load a .so/.dylib C library at runtime. For documentation of flags such as `RTL_LAZY`, consult e.g.
      https://man7.org/linux/man-pages/man3/dlopen.3.html
 -}
 withDLib :: FilePath -> (Linker.DL -> IO a) -> IO a
 withDLib dlib = Linker.withDL dlib [Linker.RTLD_LAZY]
 
 runLLVM :: API -> LLVM a -> IO a
-runLLVM api (LLVM m) =
-    withMVar api.mutex $ const $ runReaderT m api
+runLLVM api (LLVM m) = runReaderT m api
 
 mkAPI :: Linker.DL -> IO API
 mkAPI dlib = flip runReaderT dlib $ do
@@ -282,8 +280,7 @@ mkAPI dlib = flip runReaderT dlib $ do
                 stderr
                 "[Warn] Using an LLVM backend compiled with --llvm-mutable-bytes (unsound byte array semantics)"
 
-    mutex <- liftIO $ newMVar ()
-    pure API{patt, symbol, sort, simplifyBool, simplify, collect, mutex}
+    pure API{patt, symbol, sort, simplifyBool, simplify, collect}
 
 ask :: LLVM API
 ask = LLVM Reader.ask