Forward dataflow analysis includes reaching definitions and available expressions. The analysis should be implemented in the directory src/oqd-core/analysis/.
Using lattice.py and dataflow.py in oqd-compiler-infrastructure, forward dataflow analysis can be implemented for both analog and atomic interfaces. After the program is passed through the frontend, and the typed AST is obtained, a control flow graph (CFG) is built from it. The type checker verifies the types in the CFG.
Dataflow analysis needs a symbol table. The CFG can be modified to build and hold a symbol table, or the symbol table can be built from the typed AST separately. This is a design choice which depends on how the different analyses will use the CFG and the symbol table.
The result of the analysis should be used to build analysis passes (or something similar) to update the CFG or the typed AST. Ideally, since analysis passes can be run sequentially, they should update the CFG. After running all the passes, the updated CFG should be used to modify the typed AST. This is necessary as the current requirement of the compiler passes is a typed AST.
If SSA is implemented in the future, it will simplify dataflow analysis. However, it would require a refactor of the compiler passes, and lowering the code into ARTIQ.
Forward dataflow analysis includes reaching definitions and available expressions. The analysis should be implemented in the directory
src/oqd-core/analysis/.Using
lattice.pyanddataflow.pyinoqd-compiler-infrastructure, forward dataflow analysis can be implemented for both analog and atomic interfaces. After the program is passed through the frontend, and the typed AST is obtained, a control flow graph (CFG) is built from it. The type checker verifies the types in the CFG.Dataflow analysis needs a symbol table. The CFG can be modified to build and hold a symbol table, or the symbol table can be built from the typed AST separately. This is a design choice which depends on how the different analyses will use the CFG and the symbol table.
The result of the analysis should be used to build analysis passes (or something similar) to update the CFG or the typed AST. Ideally, since analysis passes can be run sequentially, they should update the CFG. After running all the passes, the updated CFG should be used to modify the typed AST. This is necessary as the current requirement of the compiler passes is a typed AST.
If SSA is implemented in the future, it will simplify dataflow analysis. However, it would require a refactor of the compiler passes, and lowering the code into ARTIQ.