Would it be wise to use LLVM (with its full-fledged JIT and optimizers) to run interpreted language code such as Matlab?
What are the specific parts of the execution that could be improved using the abstract compiler optimizer instead of Matlab current modus operandi? (I understand that it probably has a simple LLVM-like optimizer, but I don't think it is as powerful as LLVM)
The reason I thought about this is because, using runtime information, a typical Matlab calculation has (argument size, etc.), I think that a full-fledged optimizer can increase the speed of execution in very simple scenarios, where Matlab now requires special measures to ensure optimal performance (for example, explicit predefined variables, etc.)
I understand that there is currently no support for any of this, but I'm just wondering what impact such an interpreter will have.
source share