One or more non-volatile computer readable storage media storing computer readable instructions, wherein the computer readable instructions, when executed by one or more processors, are configured to cause the one or more processors to:
- acquire network description data, describing network layer tasks corresponding to each of network layers in a target neural network;
- determine a dependency relationship between the network layer tasks on the basis of the network description data;
- detect an idle register group on an artificial intelligence accelerated computing component, each function module on the artificial intelligence accelerated computing component corresponding to at least two register groups, wherein the at least two register groups are configured to receive a configuration information of the each function module and a hardware controller controls an operation of the each function module according to the configuration information, wherein the idle register group comprises a register group without data or a register group storing register group data corresponding to a network layer task that has been completely executed;
- in response to the detection of the idle register group, generate a target register configuration on the basis of the dependency relationship, the network description data, and a type of the idle register group, and write the target register configuration into the idle register group, wherein in order to generate a target register configuration on the basis of the dependency relationship, the network description data, and a type of the idle register group, the computer readable instructions cause the one or more processors to:
- in response to the detection of the idle register group, determine a candidate network layer task matching the type of the idle register group on the basis of the dependency relationship;
- in response to the determination of the candidate network layer task, determine candidate description sub-data corresponding to the candidate network layer task in the network description data;
- and
- in response to the determination of the candidate description sub-data, generate the target register configuration using the candidate description sub-data and a hardware architecture description data corresponding to the artificial intelligence accelerated computing component;
- and
- in response to detection of a target network layer task completion notification, determine a next task of a target network layer task corresponding to the target network layer task completion notification on the basis of the dependency relationship, and start the next task on a corresponding first function module on the artificial intelligence accelerated computing component so as to execute the next task;
- wherein the detection of the idle register group on the artificial intelligence accelerated computing component comprises:
- in response to being in an initialization phase, determine all register groups on the artificial intelligence accelerated computing component to be the idle register groups;
- in response to not being in the initialization phase and detection of the target network layer task completion notification, determine a second function module corresponding to the target network layer task;
- update an idle register group identification corresponding to a type of the second function module in the network description data on the basis of the type of the second function module;
- and
- determine the idle register group corresponding to the second function module according to the updated idle register group identification, wherein the idle register group is one of two ping-pong register groups corresponding to the second function module.