From 85c654a102701958d3748e82ecac9c1bc4dbbcba Mon Sep 17 00:00:00 2001 From: dec05eba Date: Tue, 16 Jul 2019 00:27:53 +0200 Subject: Start on real bytecode & doc parsing --- doc/CompilerFlow.md | 6 ++++++ 1 file changed, 6 insertions(+) create mode 100644 doc/CompilerFlow.md (limited to 'doc/CompilerFlow.md') diff --git a/doc/CompilerFlow.md b/doc/CompilerFlow.md new file mode 100644 index 0000000..e42b94d --- /dev/null +++ b/doc/CompilerFlow.md @@ -0,0 +1,6 @@ +# Compiler flow +(Tokenize&parse -> Resolve AST -> Generate SSA -> Generate bytecode) -> Generate program\ +Each step except the last is done using multiple threads in parallel and the output of each step is used +in the next step. The last step is not done in parallel because the last step is combining all bytecode +and writing it to a file, which is an IO bottlenecked operation and it won't benefit from multithreading +and may even lose performance because of it. \ No newline at end of file -- cgit v1.2.3