Printing Newly Created (Allocated) MIR Node

SpiderMonkey
#1

Hi,

Currently, I am pursuing research in JIT compilers, i.e., V8, JSC, and SpiderMonkey.
One task is to identify all the generated IR nodes and extract (print) information about them.
What I currently do is add print statements in the node generator function, e.g., V8 is Node::New and JSC is multiple Node(...) functions under dfg/DFGNode.h.
For SpiderMonkey, I kind of have an idea that I need to search the function(s) under MIR.h file, but it’s not crystal clear which function I need to target for adding the print statements. There may be multiple functions.

Can someone who’s knowledgeable with the optimizer internal give me some advice on it, please?

Thank you!

#2

Hi,

The simplest way would be to use the MDefinition constructor instead of the allocator. Each node is identified with an Opcode, given as argument of the MDefinition constructor, to implement a downcast.

Otherwise, MIR node are created with New and Clone functions, which are present in each MIR node.

#3

Hi,

Thank you for your help.

As recommended, I will try to add prints under the MDefinition constructor as well as New and Clone functions.

Thank you!