Abstract: A channel based network is provided that allows one or more hosts to communicate with one or more remote fabric attached I/O units. A split-model network driver includes a host module driver and I/O unit module driver. The host module driver and the I/O unit module driver each includes a messaging layer that allows the hosts and I/O units to communicate over the switched fabric using a push-push messaging protocol. For a host to send data, the host either initiates a RDMA write to a pre-registered buffer or initiates a message Send to a pre-posted buffer on the target. For the RDMA case, the initiator would have to send the target some form of transfer indication specifying where the data has been written. This notification can be done with either a separate message or more preferably with immediate data that is included with the RDMA write.

Abstract: An asynchronous programming environment is disclosed. In one embodiment, the environment comprises a dynamic object storage scheme, a dynamic dispatch scheme, and an object recognition scheme. The dynamic object storage scheme stores a plurality of objects. The dynamic dispatch scheme is presence one of a certain number objects. The object recognition scheme describes the plurality of objects.

Abstract: One embodiment of the present invention provides a system that facilitates automated isolation of native code within a computer program that has been compiled to a platform-independent code. The system operates by receiving a library containing a native code sub-routine that provides a service to the computer program. The system analyzes the library to determine the symbol name for the native code sub-routine. A proxy sub-routine is generated for each native code sub-routine exported by the native library that forms a link to the native code sub-routine. This proxy sub-routine is placed into a new library using the original name of the native code sub-routine. The system runs the native code sub-routine in one process, and executes the platform-independent code in a separate process. The system invokes the native code sub-routine in the first process by calling the proxy sub-routine from the platform-independent code in the second process.

Abstract: Disclosed is a method for executing a remote method. Each argument of the remote method is wrapped when the argument is a remote object (e.g. the argument's declared class implements a remote marker interface). However, each argument of the remote method is copied when an argument is not a remote object. The remote method is invoked using the wrapped or copied argument(s) (e.g. the wrapped or copied arguments are passed to the invoked method). A result of the invoked remote method is then wrapped (e.g., prior to being passed to the process that is receiving the result) when the result is a remote object (e.g., the result's declared class does not implement the remote marker interface). In contrast, the result of the remote method is copied (e.g., the copy is then passed to the process that is receiving the result) when the result is are remote object.

Abstract: A system for sharing information between separately executable programs is provided. The data sharing system includes a control unit (11) having a processor (12) and a memory (16) coupled to the processor (12). The memory (16) stores information pertaining to a first program that was previously executed by the processor (12). The data sharing system also includes a data storage medium (14) that stores a second program executable by the processor (12). When executed by the processor, the second program retrieves information pertaining to the first program from the memory and utilizes the information pertaining to the first program with the execution of the second program.

Abstract: A method and implementing computer system is provided for the creation of large numbers of threads in a computer system. An exemplary embodiment supports up to sixteen segments in memory of thread private data for each process or application program running on the system. Each segment contains support for 2K threads. These segments are identified in process' user structure which is located in the process private data segment of memory allowing cleanup collection on a per-segment basis. The thread's private data is composed of two parts, viz. its private kernel thread stack (96K) and uthread data structure. The uthread contains the individual data fields that are referenced only by the thread, including the register save area for the thread.