Abstract: A system allocates scratchpad memory (SPM) to heterogeneous devices for neural network computing. The system executes the operations of a global optimization manager. The global optimization manager receives compilation states from compilers, which compile corresponding subgraphs of a neural network model into corresponding subcommands that run on the heterogeneous devices. The global optimization manager unifies records of a same object across different ones of the compilation states, and allocates the SPM to the subgraphs according to the unified records of the compilation states.

Abstract: The present invention relates to a chip tray positioning device, which mainly comprises a frame body, a tray conveying module, a pulling module, a pushing module and a controller. The tray conveying module is disposed on the frame body, electrically connected to the controller and controlled to convey a chip tray from the start area to the end area. The pulling module and the pushing module are disposed on the frame body, electrically connected to the controller and controlled to cause the chip tray to be abutted against the end wall and the lateral wall of the frame body, thereby realizing the positioning of the chip tray and eliminating an error formed in the transfer process of the chip tray. In addition, the controller also controls the pushing module to knock the chip tray at a specific frequency so that the chip tray is vibrated.

Abstract: A system allocates scratchpad memory (SPM) to heterogeneous devices for neural network computing. The system executes the operations of a global optimization manager. The global optimization manager receives compilation states from compilers, which compile corresponding subgraphs of a neural network model into corresponding subcommands that run on the heterogeneous devices. The global optimization manager unifies records of a same object across different ones of the compilation states, and allocates the SPM to the subgraphs according to the unified records of the compilation states.

Abstract: The present invention relates to a chip tray positioning device, which mainly comprises a frame body, a tray conveying module, a pulling module, a pushing module and a controller. The tray conveying module is disposed on the frame body, electrically connected to the controller and controlled to convey a chip tray from the start area to the end area. The pulling module and the pushing module are disposed on the frame body, electrically connected to the controller and controlled to cause the chip tray to be abutted against the end wall and the lateral wall of the frame body, thereby realizing the positioning of the chip tray and eliminating an error formed in the transfer process of the chip tray. In addition, the controller also controls the pushing module to knock the chip tray at a specific frequency so that the chip tray is vibrated.

Abstract: An automatic probe construction system and the method thereof are provided. The automatic probe construction system includes a data dereference analysis module, a probe selection module, and a probe verification module. The data dereference analysis module dereferences a plurality of target data of an electronic apparatus according to a plurality of pointers, and constructs a plurality of candidate probes. The probe selection module constructs a control flow graph according to the candidate probes and an instruction code of an executable image file, to select via the control flow graph from the candidate probes at least one dedicated probe. The probe verification module searches the executable image file for an instruction code corresponding to the dedicated probe, to verify the dedicated probe. Therefore, the dedicated probe can be constructed automatically, and the number of the dedicated probes is thus limited.

Abstract: A method for determining whether a to-be-tested program contains malicious behavior is disclosed. The method includes steps of providing an emulator having a kernel and a plurality of installed hook points, wherein the kernel has a plurality of in-kernel functions; executing the to-be-tested program in the emulator dynamically to invoke the plurality of installed hook points to obtain a specific in-kernel function set from the plurality of in-kernel functions; and determining whether the to-be-tested program contains instructions for malicious behavior based on an invocation sequence of the specific in-kernel function set.

Abstract: An automatic probe construction system and the method thereof are provided. The automatic probe construction system includes a data dereference analysis module, a probe selection module, and a probe verification module. The data dereference analysis module dereferences a plurality of target data of an electronic apparatus according to a plurality of pointers, and constructs a plurality of candidate probes. The probe selection module constructs a control flow graph according to the candidate probes and an instruction code of an executable image file, to select via the control flow graph from the candidate probes at least one dedicated probe. The probe verification module searches the executable image file for an instruction code corresponding to the dedicated probe, to verify the dedicated probe. Therefore, the dedicated probe can be constructed automatically, and the number of the dedicated probes is thus limited.

Abstract: A computer system and a method for tracking information flow are provided. The computer system divides an information flow tracking task into two decoupled tasks executed by two procedures. The first procedure emulates execution of instructions and divides the instructions into code blocks according to an instruction executing sequence. The first procedure translates the instructions of the code blocks into information flow codes and transmits them to the second procedure. The first procedure further translates the instructions into dynamic emulation instructions and executes the dynamic emulation instructions to generate addressing results of the dynamic addressing instructions. The second procedure executes the information flow codes according to the addressing results to emulate the instructions of the code blocks. Moreover, the method also tries to reduce the amount of data transmission between the two procedures when the first procedure executes the emulation task.

Abstract: A method for determining whether a to-be-tested program contains malicious behavior is disclosed. The method includes steps of providing an emulator having a kernel and a plurality of installed hook points, wherein the kernel has a plurality of in-kernel functions; executing the to-be-tested program in the emulator dynamically to invoke the plurality of installed hook points to obtain a specific in-kernel function set from the plurality of in-kernel functions; and determining whether the to-be-tested program contains instructions for malicious behavior based on an invocation sequence of the specific in-kernel function set.

Abstract: The present invention relates to compound semiconductor ESD protection devices of three types. The device comprises a multi-gate enhancement mode PET (E-PET). For the type I compound semiconductor ESD protection device, the source electrode is connected to the plural gate electrodes through at least one first resistor, and the drain electrode is connected to the plural gate electrodes through at least one second resistor. For the type II compound semiconductor ESD protection device, at least one of the plural gate electrodes are connected to at least one of the inter-gate regions between two adjacent gate electrodes through at least one fourth resistor. For the type compound semiconductor ESD protection device, the plural gate electrodes are connected to the source or drain electrodes through at least one seventh resistor. Any two gate electrodes in the three types of compound semiconductor ESD protection devices can be connected by a resistor.

Abstract: The present invention relates to compound semiconductor ESD protection devices using plural compound semiconductor E-FETs or compound semiconductor multi-gate E-FETs. The device comprises plural compound semiconductor E-FETs or multi-gate E-FETs, in which each of the gates is DC-connected to the source, drain, or an inter-gate region between two adjacent gates in the multi-gate E-FET through at least one first resistor, and at least one of the gates is AC-connected to the source, drain, or an inter-gate region between two adjacent gates in the multi-gate E-FET through a gate capacitor.

Abstract: The present invention relates to compound semiconductor ESD protection devices using plural compound semiconductor E-FETs or compound semiconductor multi-gate E-FETs. The device comprises plural compound semiconductor E-FETs or multi-gate E-FETs, in which each of the gates is DC-connected to the source, drain, or an inter-gate region between two adjacent gates in the multi-gate E-FET through at least one first resister, and at least one of the gates is AC-connected to the source, drain, or an inter-gate region between two adjacent gates in the multi-gate E-FET through a gate capacitor.

Abstract: The present invention relates to compound semiconductor ESD protection devices of three types. The device comprises a multi-gate enhancement mode FET (E-FET). For the type I compound semiconductor ESD protection device, the source electrode is connected to the plural gate electrodes through at least one first resistor, and the drain electrode is connected to the plural gate electrodes through at least one second resistor. For the type II compound semiconductor ESD protection device, the plural gate electrodes are connected to at least one of the inter-gate regions between two adjacent gate electrodes through at least one fourth resistor. For the type III compound semiconductor ESD protection device, the plural gate electrodes are connected to the source or drain electrodes through at least one seventh resistor. Any two gate electrodes in the three types of compound semiconductor ESD protection devices can be connected by a resistor.

Abstract: A tracing device for detecting whether a specific attribute datum has a possibility of being stolen is provided. The tracing device includes a label map and a first processing device, wherein the label map has a specific label attached on the specific attribute datum and a buffer region, and the first processing device is coupled to the label map and determines whether there is the specific label in the buffer region.

Abstract: A method for authenticating a message in a network is provided. The method includes a step of generating, in a sink device, a private key and a public key. The private key includes a plurality of sub-private keys. The method further includes a step of generating, in the sink device, a signature for the message. The signature includes a sub-private key and an authentication path associated with the sub-private key in a hash tree. The hash tree is constructed during the generation of the sub-public keys.

Abstract: In one embodiment of the present invention, a system for dynamic content distribution and grouping includes a server for providing a plurality of content channels, at least one tuning/decoding means for receiving/decoding at least one of the plurality of content channels, and at least one respective display means for displaying the at least one received/decoded content channel. In such embodiments of the present invention, the server is configured to dynamically configure the at least one tuning/decoding means to receive/decode at least one of the plurality of content channels. More specifically, in embodiments of the present invention, the server dynamically configures the at least one tuning/decoding means by communicating a channel change command to the at least one tuning/decoding means over a dedicated radio-frequency channel or, in an alternate embodiment, using an internet protocol.

Abstract: A computer system and a method for tracking information flow are provided. The computer system divides an information flow tracking task into two decoupled tasks executed by two procedures. The first procedure emulates execution of instructions and divides the instructions into code blocks according to an instruction executing sequence. The first procedure translates the instructions of the code blocks into information flow codes and transmits them to the second procedure. The first procedure further translates the instructions into dynamic emulation instructions and executes the dynamic emulation instructions to generate addressing results of the dynamic addressing instructions. The second procedure executes the information flow codes according to the addressing results to emulate the instructions of the code blocks. Moreover, the method also tries to reduce the amount of data transmission between the two procedures when the first procedure executes the emulation task.

Abstract: A method for authenticating a message in a network is provided. The method includes a step of generating, in a sink device, a private key and a public key. The private key includes a plurality of sub-private keys. The method further includes a step of generating, in the sink device, a signature for the message. The signature includes a sub-private key and an authentication path associated with the sub-private key in a hash tree. The hash tree is constructed during the generation of the sub-public keys.

Abstract: In one embodiment of the present invention, a system for dynamic content distribution and grouping includes a server for providing a plurality of content channels, at least one tuning/decoding means for receiving/decoding at least one of the plurality of content channels, and at least one respective display means for displaying the at least one received/decoded content channel. In such embodiments of the present invention, the server is configured to dynamically configure the at least one tuning/decoding means to receive/decode at least one of the plurality of content channels. More specifically, in embodiments of the present invention, the server dynamically configures the at least one tuning/decoding means by communicating a channel change command to the at least one tuning/decoding means over a dedicated radio-frequency channel or, in an alternate embodiment, using an internet protocol.

Abstract: A fan apparatus has a frame, a motor and at least one fan blade part. The frame has a through chamber where the motor and the fan blade assembly are configured. The fan blade assembly has a stationary blade and a moving blade. The stationary blade is fixed inside the chamber to go around the motor and has multiple first radial blade parts. The moving blade is coupled to a shaft of the motor and has multiple second radial blade parts. When each first blade part is corresponding to the corresponding second blade part, a first included angle is formed between each first blade part and the corresponding second blade part. Therefore, a pair of the stationary blade and the moving blade enhances the wind pressure of the fan apparatus.