Tracing the Meta-Level: PyPy's Tracing JIT Compiler

More documents

Recommendations

Info

Applying a Tracing JIT to an Interpreter When to check for a Closing Loop ◮ program counter of language interpreter can only be the same a second time if it is set to an earlier version by an instruction ◮ this only happens at a backward jump in the language interpreter ◮ jit needs another hint to recognize a backward jump
Applying a Tracing JIT to an Interpreter When to use Generated Code ◮ problem: all pieces of assembler code correspond to the same hot loop, the bytecode dispatch loop ◮ however: they correspond to dierent paths through that loop ◮ need to check the program counter to pick the correct machine code ◮ if program counter corresponds to the position key of one of the machine codes, this code can be executed ◮ this only needs to be checked at a backward branch
Page 1 and 2: Tracing the Meta-Level: PyPy's Trac
Page 3 and 4: Introduction What We want to do ◮
Page 5 and 6: The PyPy project Advantages for wri
Page 7 and 8: Tracing JIT compilers ◮ initially
Page 9 and 10: Tracing JIT compilers What should b
Page 11 and 12: Tracing JIT compilers Tracing JIT P
Page 13 and 14: Applying a Tracing JIT to an Interp
Page 17: Applying a Tracing JIT to an Interp
Page 23 and 24: Improving the Result II ◮ if oper
Page 25 and 26: Implementation issues II ◮ on fai
Page 27 and 28: Evaluation PyPy ◮ PyPy with JIT i
Page 29 and 30: Related Work Piumarta and Riccardi
Page 31 and 32: Related Work Partial Evaluation ◮
Page 33 and 34: Related Work Dynamic Partial Evalua

Tracing the Meta-Level: PyPy's Tracing JIT Compiler

Create successful ePaper yourself

Delete template?

Save as template?