Code generator and method for generating extra code for a virtual machine

A method of generating code for a virtual machine, the method comprising:

• executing bytecode at an interpreter implemented by a processor;
• collecting tracking information at an information collecting unit, wherein the tracking information indicates a logical flow of execution of the bytecode by the interpreter, and a number of executions of the bytecode by the interpreter, and wherein the information collecting unit is implemented separately from both the interpreter and a compiler implemented by the processor;
• determining whether the compiler is idle,
• in response to the compiler being idle
• dividing the bytecode into bytecode segments having a predefined length, based on the logical flow of execution of the bytecode;
• and
• determining a bytecode trace among the bytecode segments, based on the number of executions of the bytecode, the bytecode trace having the predefined length, wherein the bytecode trace comprises a starting point that is a bytecode segment that has been executed by the interpreter more than a threshold number of executions;
• in response to the compiler not being idle,
• collecting tracking information at an information collecting unit;
• storing the bytecode trace in a work queue;
• and
• transmitting the stored bytecode trace to the compiler, in order to be compiled into machine code,
• and defining the predefined length of the bytecode trace based on a capacity of the work queue of the compiler and a capacity of a code cache configured to store machine code compiled from the bytecode executed by the interpreter.

A method of generating code for a virtual machine comprising bytecode, the method comprising:

• receiving tracking information at an information collecting unit implemented in a processor, that indicates a logical flow by which bytecode is executed by an interpreter wherein the information collecting unit is implemented separately from both the interpreter and a compiler implemented by the processor;
• determining whether the compiler is idle,
• in response to the compiler being idle dividing the bytecode into bytecode segments having a predefined length, based on the logical flow of execution of the bytecode;
• checking, at a code determining unit, whether the interpreter has determined a bytecode trace, and in response to the bytecode trace not being determined by the interpreter, determining, by the code determining unit, a new bytecode trace among the bytecode segments, based on the number of executions of the bytecode, and storing the new bytecode trace in a work queue;
• and
• in response to the compiler not being idle,
• receiving tracking information at an information collecting unit;
• compiling into machine code, at the compiler, when the compiler exits a standby state, the new bytecode trace determined by the code determining unit and stored in a work queue;
• and
• defining the predefined length of the new bytecode trace based on a capacity of the work queue of the compiler and a capacity of a code cache configured to store machine code compiled from the bytecode executed by the interpreter,
• wherein the new bytecode trace is a segment of code having the predefined length beginning at a starting point, and the starting point is a bytecode segment that has been executed by the interpreter more than a threshold number of executions.

A terminal, the terminal comprising:

• a processor;
• an interpreter implemented by the processor and configured to execute bytecode of a virtual machine, and determine a bytecode trace, to determine machine code to be generated;
• an information collecting unit implemented in the processor and configured to collect tracking information indicating a logical flow of the bytecode currently being executed by the interpreter, collect a number of executions of the bytecode by the interpreter;
• determining whether a compiler implemented by the processor is idle,
• wherein in response to the compiler being idle
• a code generator implemented by the processor, and configured to divide the bytecode into bytecode segments having a predefined length, based on the logical flow of execution of the bytecode;
• check whether the interpreter has determined the bytecode trace, and in response to the bytecode trace not being determined by the interpreter,
• determine a new bytecode trace among the bytecode segments, based on the number of executions of the bytecode, wherein the new bytecode trace is a segment of code having the predefined length beginning at a starting point, the starting point being a bytecode segment that has been executed by the interpreter more than a threshold number of executions;
• and
• in response to the compiler not being idle, collect tracking information;
• wherein the compiler is configured to compile the new bytecode trace into machine code, and the compiler includes a work queue configured to store bytecode sections to be translated into machine code,
• wherein the information collecting unit is implemented separately from both the interpreter and the compiler, and
• wherein the predefined length of the new bytecode trace is defined based on a capacity of the work queue and a capacity of a code cache configured to store machine code compiled from the bytecode executed by the interpreter.